Compression

Compression in Hearing Aids - Definition Compression in hearing aids refers to a signal processing strategy that automatically and non-linearly adjusts amplification (gain) in response to changing input sound levels, with the primary goal of fitting the wide range of everyday sounds into the narrowed dynamic range of a hearing-impaired listener.

Table summary: The table outlines various clinical applications of compression, focusing on preventing auditory damage, optimizing the remaining dynamic range, improving loudness perception, ensuring comfort, enhancing speech clarity, and mitigating the impact of noise.

Figure 4-5 summary: This figure consists of a conceptual map, a hierarchical flowchart, and a line graph. The content outlines the comprehensive framework of compression in hearing aids, detailing the theoretical foundations, the classification of signal processing into linear and non-linear categories, and the physiological relationship between hearing thresholds and dynamic range. The figure demonstrates that non-linear processing involves complex compression and expansion parameters based on static and dynamic features, while the graph illustrates that hearing impairment results in a shifted and narrower dynamic range compared to normal hearing. It can be concluded that compression is a critical tool used to map a wide range of input sounds into the reduced dynamic range of a hearing-impaired individual to ensure audibility without causing discomfort.

Recent Advances in Hearing Aids

Image summary: This figure is a timeline chart. It illustrates the evolution of hearing aid technology from the sixties to the present day, mapping specific technological advancements and device styles against a chronological axis. The data shows a progression from basic linear amplification and early microphones to the introduction of compression, programmable features, and the eventual transition from analog to digital technology. It can be inferred that hearing aid technology has shifted from simple hardware-based solutions to sophisticated digital systems characterized by increased customization, noise reduction, and a broader variety of fitting options and device forms.

Chapter 1: Introduction

According to the World Health Organization (W.H.O), hearing loss is one of the leading health concerns around the world. At the present time, hearing aids are the most common first step in [re] habilitation. On the surface, this seems fairly straightforward because much is known about the physiology of the auditory system and the psychoacoustics of perception. However, hearing loss impacts an individual in numerous ways, making the fitting of a hearing aid a complex process.

The intact auditory system is capable of perceiving a wide range of sounds, from the quiet pitter-patter of rain to the loud boom of explosives. In Figure 1-1.A, the white bar represents the entire range of sounds, from extremely weak to extremely intense, that may occur in an individual's environment. The weakest sounds that are audible lie at the threshold of hearing sensitivity. At the opposite end is the loudness discomfort level (L.D.L), representing the most intense sounds that can be tolerated without pain. In between these two extremes is the dynamic range of hearing (shown by the blue bar).

For an individual with normal hearing, average conversational speech falls approximately midway within the dynamic range of hearing and coincides with the most comfortable loudness level (M.C.L).

The most common complaint associated with hearing loss is the inability to hear; specifically, the inability to hear soft sounds. Figure 1-1.B depicts a person with sensorineural hearing loss. Once again the white bar is the range of sounds in the environment, while the blue bar represents the individual's dynamic range of hearing. You notice three things immediately.

First, average conversational speech is now barely audible to the individual. Second, weak sounds (for example, gentle rain) are below the threshold of hearing and, therefore, too soft to be heard. And, finally, intense sounds (for example, the boom of an explosion) are still perceived as being loud. As a result of the threshold increasing and the L.D.L remaining the same, the dynamic range of hearing is considerably reduced compared to that of a person with normal hearing.

Human communication is arguably the single most important function of the auditory system. Indeed, reduced ability to hear speech is a major reason for seeking remediation. In addition to a reduced dynamic range, the communication difficulties of a person with hearing impairment are further complicated by the dynamic nature of speech itself.

As shown in Figure 1 to 2, average conversational speech spans a range of 30 decibels. Note that, in general, vowel sounds (for example /a/, /u/, and /i/) are low-pitched, relatively intense, and primarily responsible for making speech audible. On the other hand, consonants (especially unvoiced sounds such as /th/, /f/ and /s/) are high-pitched, relatively weak and carry most of the information that aids in speech understanding.

In light of these considerations, the hearing healthcare professional is faced with squeezing an elephant into a suitcase when fitting hearing aids – soft sounds must be made audible without loud sounds becoming uncomfortable, and speech should remain at a comfortable level. Compression amplification is a means for fitting the world of sound (the elephant) into the dynamic range of the individual with hearing impairment (suitcase).

This handbook is designed to provide the reader with a working knowledge of compression amplification: what it is, how it works, and how it is applied. Also included is some discussion on the principles of fitting compression systems, and troubleshooting problems.

Figure 1-1 summary: This figure consists of two comparative diagrams. The diagrams illustrate how different levels of environmental sound intensity, ranging from weak sounds like rain to intense sounds like explosions, relate to the dynamic range of hearing. The first diagram represents a person with normal hearing, while the second diagram represents a person with sensorineural hearing loss. In the case of normal hearing, the dynamic range is positioned to capture moderate sounds effectively while maintaining boundaries for sounds that are too soft or too loud. In contrast, for the person with sensorineural hearing loss, the dynamic range is shifted upward. This shift indicates that sounds which are typically moderate or weak may be perceived as too soft, while the threshold for sounds becoming too loud is reached more quickly, effectively narrowing the usable range of hearing.

Figure 1-2 summary: This figure is an audiogram plot. It maps various speech sounds across a range of frequencies and hearing levels in decibels. The plot illustrates where different consonants and vowels fall in terms of their spectral frequency and relative intensity during average conversation. The data indicates that speech sounds are distributed across a wide frequency spectrum, with some sounds occurring at lower frequencies and others at higher frequencies. It can be inferred that certain speech sounds require higher hearing sensitivity due to their lower intensity levels, while others are more prominent in the higher frequency range.

Essential Terminology

Before launching into the detailed workings of a compression circuit, it is important to have some general knowledge of amplification. Despite the variety available, all hearing aids have some of the same basic components: a microphone, an amplifier, a receiver, and a battery (Figure 1 to 3). The microphone picks up the incoming acoustic signal and converts it to an electrical signal. The amplifier then magnifies the electrical signal. Like a loudspeaker, the receiver converts the amplified electrical signal back into an acoustic signal that is delivered to the ear. Finally, the battery provides the power for the circuit.

It is essential that the reader understand the following terminology that relates to the signal entering the hearing aid, the amplification, and the sound that is delivered to the ear.

Input

Input refers to the acoustic signal entering the hearing aid. Specifically, the American National Standards Institute defines input level as the sound pressure level (S.P.L) at the microphone opening of a hearing aid. Input level is expressed in decibels S.P.L.

Figure 1-3 summary: This figure is a schematic diagram. It illustrates the fundamental components of a hearing aid, showing the sequential flow from an input signal to an output signal. The process begins with a microphone that captures sound, which is then passed to an amplifier powered by a battery, and finally sent to a receiver. The diagram demonstrates that the system transforms a low-amplitude input wave into a significantly larger output wave, indicating that the primary function of the device is to increase the intensity of sound signals for the user.

Output

Output refers to the amplified signal that is delivered to the ear. The output level is expressed in decibels S.P.L.

Input/Output Function

An input/output (I/O) function is a graphical representation of the output of a hearing aid at various input levels. ansi (2003) defines it as a single-frequency plot of the coupler S.P.L on the ordinate (Y-axis) as a function of the input S.P.L on the abscissa (X-axis) with equal decibel divisions on each axis; a similar definition is used by the International Electrotechnical Commission. Figure 1 to 4 shows a sample I/O function of a hearing aid. In this example, an input of 50 decibels S.P.L results in an output of 80 decibels S.P.L, while an input of 90 decibels S.P.L results in an output of 110 decibels S.P.L. It can also be seen that the output of the hearing aid does not exceed 110 decibels S.P.L. Figure 1 to 5 (page 6) shows sample I/O functions for three hearing aids. Notice that they do not all behave in the same way – inputs of 50 and 90 decibels S.P.L result in different outputs for the three hearing aids.

Gain

Gain refers to the amount of amplification applied to the input signal. Specifically, ansi (2003) defines gain as the difference between the output S.P.L in a coupler and the input S.P.L. Gain is expressed in decibel. The mathematical relationship between input, gain and output is given by the simple formula:

Thus, in Figure 1 to 4, if an input of 50 decibels S.P.L results in an output of 80 decibels S.P.L, the gain of the hearing aid is:

Math summary: This computation calculates the total gain by subtracting the input value from the output value. The process takes an output of eighty decibels and subtracts an input of fifty decibels to produce a final result of thirty decibels.

Similarly, with an output of 105 decibels S.P.L for an input level of 90 decibels S.P.L, Hearing Aid 3 (in Figure 1 to 5) has a gain of 15 decibels.

Figure 1-4 summary: This figure is a line chart. It illustrates the relationship between the input sound level and the corresponding output sound level for a hearing aid, with both axes measured in decibels sound pressure level. The chart shows a linear increase in output as the input increases, which then levels off into a plateau at higher input intensities. From this, it can be inferred that the hearing aid amplifies softer sounds proportionally but implements a compression or limiting function to prevent excessively loud sounds from being output, thereby protecting the user's hearing.

Figure 1-5 summary: This figure is a line chart. It illustrates the relationship between the input sound pressure level and the resulting output sound pressure level for three distinct hearing aid devices. The chart shows that while all three devices increase output as input increases, they exhibit different amplification characteristics. Hearing Aid 3 provides the highest level of gain for lower input levels compared to the other two devices. Hearing Aid 1 demonstrates a steep increase in output before reaching a plateau at higher input levels. In contrast, Hearing Aid 2 shows a more linear and gradual increase in output across the entire input range. Overall, the data indicates that different hearing aids employ varying compression and amplification strategies to manage sound output based on the input intensity.

Input/Gain Function

An input/gain (I/G) function is a graphical representation of the gain of a hearing aid at various input levels. Figure 1 to 6 shows a sample I/G function of a hearing aid. In this example, the hearing aid provides 30 decibels of gain for an input of 50 decibels S.P.L, but only 20 decibels of gain for an input of 90 decibels S.P.L. Figure 1 to 7 shows sample I/G functions for three hearing aids. Notice that they do not all behave in the same way – different amounts of gain are applied to inputs of 50 decibels S.P.L and 90 decibels S.P.L.

Just as gain can be calculated from the input and output, output is calculated by rearranging the formula for calculating gain as follows:

Math summary: This computation calculates the final output value. It is performed by adding the gain to the input values.

Thus, in Figure 1 to 6, if 30 decibels of gain is applied to an input of 50 decibels S.P.L, the output of the hearing aid is:

Math summary: This calculation determines the total output level by adding a gain value to the input level. The process sums an input of fifty decibels with a gain of thirty decibels to produce a final output of eighty decibels.

Similarly, with gain of 18 decibels applied to an input level of 90 decibels S.P.L, Hearing Aid 3 (in Figure 1 to 7) has an output of 108 decibels S.P.L.

An obvious, but frequently overlooked, point is that even though the gain of a device decreases with increasing input level, the output continues to increase. This occurs because the decrease in gain is less than the increase in input level. Figures 1-8.A and 1-8.B, which show the I/O and I/G functions for a hearing aid, illustrate this point. 30 decibels of gain applied to an input of 50 decibels S.P.L results in an output of 80 decibels S.P.L. On the other hand, 20 decibels of gain applied to an input of 90 decibels S.P.L results in an output of 110 decibels S.P.L. In this example, the input level increases by 40 decibels, while the gain decreases only by 10 decibels. Thus, although less gain is applied to the 90 decibels S.P.L input than to the 50 decibels S.P.L input, the output S.P.L is still greater for the 90 decibels S.P.L input.

Figure 1-6 summary: This figure is a line chart. It illustrates the relationship between the input sound level measured in decibels sound pressure level and the resulting gain provided by a hearing aid in decibels. The plot shows a constant gain level for lower and moderate input levels, which then decreases linearly as the input sound level increases beyond a certain threshold. This indicates that the hearing aid provides maximum amplification for softer sounds while reducing the gain for louder sounds to prevent over-amplification and maintain comfort for the user.

Figure 1-7 summary: This figure is a line chart. It displays the relationship between the input sound pressure level and the resulting gain for three distinct hearing aid devices. The chart illustrates how each device modifies its amplification as the input volume increases. Based on the data, Hearing Aid 3 provides the highest initial gain for low-level inputs before decreasing. Hearing Aid 1 maintains a constant gain over a wider range of input levels compared to the others before dropping. All three devices exhibit a trend where gain decreases as input levels reach higher intensities, though they follow different slopes and thresholds for this reduction.

Figure 1-8 summary: This figure consists of two line charts. The first chart illustrates the relationship between the input sound pressure level and the resulting output sound pressure level for a hearing aid, while the second chart shows the relationship between the input sound pressure level and the gain provided by the device. In the input/output function, the output increases linearly with the input until it reaches a maximum plateau. Correspondingly, the input/gain function shows a constant gain for lower input levels, which then decreases as the input increases beyond a certain threshold. These results indicate that the hearing aid provides a steady amount of amplification for soft to moderate sounds but implements compression for louder sounds to prevent the output from exceeding a safe limit.

Frequency Response Curve

A frequency response curve is a graphical representation of the hearing aid output as a function of frequency. Specifically, the I.E.C (1983a) defines it as the S.P.L developed by a hearing aid in the ear simulator expressed as a function of frequency under specified test conditions. The input level and overall gain of the hearing aid are fixed when measuring a frequency response curve. Figure 1 to 9 depicts a sample frequency response curve of a hearing aid for an input of 60 decibels S.P.L. It can be seen that the output of the hearing aid varies across frequencies. Figure 1 to 10 shows sample frequency response curves for speech presented at three different input levels. It is not uncommon for the shape of the curve to change as the input level increases.

A frequency-gain curve is a graph showing the gain of a hearing aid as a function of frequency (Figure 1 to 11) under specified test conditions.

Figure 1-9 summary: This figure is a line chart. It displays the frequency response of a hearing aid, plotting the output sound pressure level against a range of input frequencies for a constant input level. The curve shows that the output remains relatively stable at lower frequencies, increases to a peak in the mid-to-high frequency range, and then drops significantly at the highest frequencies. It can be inferred that the hearing aid provides the most amplification in the mid-high frequency spectrum while offering minimal gain at low frequencies and reduced output at very high frequencies.

Figure 1-10 summary: This figure is a line chart. It displays the frequency response of a hearing aid, plotting the output sound pressure level against a range of frequencies for three distinct input levels. The curves illustrate how the device amplifies speech across the frequency spectrum. The data indicates that as the input level increases, the output level also increases across all frequencies. Additionally, the device shows a peak in output in the mid-to-high frequency range, suggesting targeted amplification for those frequencies, while the output drops off significantly at the highest frequencies.

Figure 1-11 summary: This figure is a line chart. It displays the relationship between input frequency and the resulting gain of a hearing aid when the input level is held constant. The horizontal axis represents frequency in Hertz, while the vertical axis represents gain in decibels. The curve shows that the hearing aid provides a relatively stable amount of gain across low and mid-range frequencies, with a slight peak in the upper-mid frequency range. Beyond a certain threshold, the gain drops sharply as the frequency increases toward the higher end of the spectrum. This indicates that the device is designed to amplify a broad range of audible frequencies but significantly reduces amplification for very high-frequency sounds.

Peak Clipping

In general, the output of a hearing aid increases as the input level increases. However, once the output reaches a certain level, the hearing aid is incapable of producing a louder signal. Maximum output is the highest possible signal that a hearing aid is capable of delivering, regardless of the input level or the gain of the hearing aid.

The maximum output of a hearing aid is determined by the characteristics of the microphone, amplifier and receiver. That is, the maximum output of the hearing aid is only as high as the weakest component in the chain. When the input level and gain exceed the maximum output, the hearing aid is said to be in saturation.

As long as the output of the hearing aid remains below the maximum output, the output signal is similar to the input signal, only larger in amplitude (Figure 1-12.A). When the sum of the input level and gain exceed the maximum output of the hearing aid, however, the peaks of the output signal are clipped at the maximum output (Figure 1-12.B). This is referred to as peak clipping. Note that the shape of the output signal is quite different from that of the input signal once the peaks are clipped.

Peak clipping is one method of controlling or limiting the maximum output of a hearing aid. [Alternate methods of output limiting will be discussed in subsequent chapters.] It is virtually impossible to determine whether or not peak clipping is occurring merely by examining an I/O or I/G function. The hallmark of peak clipping is that it produces an output signal that is distorted, often described as sounding "scratchy."

Distortion

Distortion refers to the presence of frequency components in the output of a hearing aid that were not present in the input signal. There are two types of distortion – harmonic and intermodulation.

Harmonic distortion is said to occur when the output contains frequency components that are integer multiples of the input signal frequency. For example, harmonics of a 500 Hz input signal would occur at 1000 Hz, 1500 Hz, 2000 Hz, 2500 Hz, and so on. Thus, harmonic distortion only occurs at frequencies greater than the input signal frequency. Total harmonic distortion (T.H.D) is the summed power of all the harmonic distortion products relative to the power of the original input signal. T.H.D is typically expressed as a percentage.

Intermodulation distortion is generated by the interaction of at least two signals of different frequencies in the input. It occurs when the frequency components of a complex input signal combine to generate additional frequency components in the output signal. For example, if F.1 and F.2 represent two different frequencies, intermodulation distortion may occur at frequencies corresponding to F.2-F.1, 2F1-F.2, 2F2-F.1, 3F.1-2F.2, and so on. Thus, intermodulation distortion may occur at frequencies above and below the input frequencies.

Both types of distortion may occur simultaneously in a hearing aid. Distortion of any type results in unpleasant sound quality and may adversely affect speech intelligibility.

Now that the reader is familiar with some general hearing aid terminology, the next few chapters will introduce and describe the topic of compression.

Input-Output Functions

Input/output (I/O) functions are the "language" of compression. They are the most common way to explain compression, and therefore it is important for clinicians to become comfortable interpreting them. Ideally, clinicians should be able to examine an I/O function and describe the type of compression being shown.

On all I/O functions:

- The horizontal (10) axis represents input sound pressure levels (S.P.L's).

- The vertical (y) axis represents output S.P.L's.

- The diagonal lines, called functions, show the relationship between corresponding input and output levels.

In other words, the functions illustrate the gain provided by a hearing aid for different input S.P.L's.

A fundamental principle for understanding hearing aid function is:

Input + Gain = Output It is also important to remember that:

- Input and output are described in decibel S.P.L.

- Gain is described in decibel.

bends is called the knee-point, and it is at this point that compression begins.

The knee-point, as measured above the input axis, indicates the input S.P.L at which compression starts.

Throughout compression discussions, the term knee-point refers specifically to the input level where compression begins.

Consider a simple I/O function showing linear gain. Gain remains linear to the left of (or below) the knee-point. This means that any increase in input S.P.L produces an equal increase in output S.P.L.

For example, if a hearing aid provides 60 decibel of gain:

- A 10 decibel S.P.L input produces a 70 decibel S.P.L output.

- A 20 decibel S.P.L input produces an 80 decibel S.P.L output.

- A 30 decibel S.P.L input produces a 90 decibel S.P.L output.

- And so on, until an input level of 60 decibel S.P.L is reached.

Beyond this level, peak clipping is used to ensure that the maximum power output (M.P.O) does not exceed 120 decibel S.P.L.

Figure 1-12 summary: This figure consists of two schematic diagrams illustrating signal transformations. The diagrams show an input wave passing through a boundary to become an output wave. In the first scenario, the input wave maintains its sinusoidal shape as it transitions to the output. In the second scenario, the input wave is transformed into a square-like wave in the output, while a dashed sinusoidal wave is shown as a reference. The comparison indicates that the first process preserves the original waveform, whereas the second process modifies the wave shape, effectively clipping or squaring the signal.

Figure 7-1 summary: This figure is a line chart. It illustrates the relationship between input and output levels, featuring a linear segment that transitions into a shallower slope at a specific bend known as the knee-point. The chart demonstrates how output initially tracks input in a direct ratio before entering a compression phase. It can be inferred that once the input exceeds the knee-point, the system applies compression, meaning that further increases in input result in progressively smaller increases in output. This indicates that the maximum power output is limited by a compression mechanism rather than abrupt peak clipping.

Comparing Linear Gain and Compression

To understand compression more clearly, consider two hearing aids with:

A knee-point at 60 decibel S.P.L.

60 decibel of linear gain below the knee-point.

The differences appear above (or to the right of) the knee-point.

Linear Hearing Aid

In a linear hearing aid with an M.P.O of 120 decibel S.P.L:

- Linear amplification continues until the input reaches 60 decibel S.P.L.

- Once the output reaches 120 decibel S.P.L, the hearing aid enters peak clipping.

- Further increases in input do not increase output.

- The output remains fixed at 120 decibel S.P.L.

Thus, on an I/O function, the line becomes horizontal after reaching M.P.O.

Hearing Aid with Compression

In a compression hearing aid, once the input level exceeds 60 decibel S.P.L, compression rather than peak clipping limits the output.

Instead of becoming horizontal:

- The output continues to increase.

- However, it increases at a slower rate.

- The line rises with a shallower slope.

Thus, compression limits the M.P.O in a gradual manner rather than abruptly stopping output growth.

Understanding Compression

Compression can be thought of as rapids in a river—a countercurrent working against the main flow. The output is still allowed to increase, but not as freely as before.

With compression, gain becomes nonlinear because the amount of gain changes as a function of input S.P.L.

Below the Knee-Point

For input levels below the knee-point:

- Gain remains linear.

- The function maintains a 45 degree angle.

- Equal increases in input produce equal increases in output.

Above the Knee-Point

Once compression begins:

- Input S.P.L continues to increase.

- Output S.P.L also increases.

- However, output increases by a smaller amount than the increase in input.

Consequently, the gain for inputs above the knee-point is less than the gain for inputs below the knee-point.

The slope of the line to the right of the knee-point visually represents the effect of compression on hearing aid gain.

In general:

- Linear gain is represented by a 45 degree slope.

- Compression is represented by a shallower slope.

- The shallower the slope, the greater the amount of compression.

Compression as a Gain-Related Issue

A crucial concept is that compression is fundamentally a gain-related phenomenon.

Compression alters the gain of the hearing aid as input levels increase. Although compression ultimately affects the M.P.O, it does so by changing the gain applied to sounds above the knee-point.

On an I/O function:

- The portion of the graph to the left of the knee-point shows linear gain.

- The portion to the right shows compressed gain.

- The overall height of the line to the right of the knee-point determines the hearing aid's M.P.O.

The M.P.O is therefore influenced by the compression characteristics applied to higher-level inputs.

Interpreting Gain on an I/O Function

To summarize:

- Linear gain is shown by the 45 superscript circle line to the left of the knee-point.

- Compression is shown by the shallower line to the right of the knee-point.

- Linear gain always provides more gain than compressed amplification.

An important point is that I/O functions display only input and output levels. They do not directly display gain.

To determine gain for any input level:

Gain = Output - Input Therefore, one must locate the corresponding output level and subtract the input level from it.

Common Misunderstanding About I/O Functions

Readers should understand that the length of a gain function has virtually nothing to do with the amount of gain provided.

The amount of gain is determined by the position of the function, not by its length.

What matters is whether the function is shifted:

- To the right points to less gain.

- To the left, leads to more gain.

Thus, on an I/O function:

- A rightward shift represents a decrease in gain.

- A leftward shift represents an increase in gain.

Although this may initially seem counterintuitive, it becomes clearer when studying input compression and output compression systems.

Compression Ratios

After understanding knee-points, the next important concept is the compression ratio.

The compression ratio describes the amount of compression provided once compression begins at the knee-point. On an I/O function, the compression ratio can be visualized from the slope of the line to the right of the knee-point. Examples

10:1 Compression Ratio

A 10:1 compression ratio means that:

For every 10 decibel increase in input S.P.L, there is only a 1 decibel increase in output S.P.L.

2:1 Compression Ratio

A 2:1 compression ratio means that:

For every 10 decibel increase in input S.P.L, there is a 5 decibel increase in output S.P.L.

Interpretation of Compression Ratios

The relationship between compression ratio and amount of compression is straightforward:

Higher compression ratios indicate greater compression.

Lower compression ratios indicate less compression.

A useful way to remember the two major compression parameters is:

Knee-point = "When?" compression begins Compression ratio = "How much?" compression occurs The knee-point identifies the input level at which compression starts, while the compression ratio specifies the degree of compression applied beyond that point.

Automatic Gain Control (A.G.C)

In the analog era of hearing aids, compression was commonly referred to as Automatic Gain Control (A.G.C).

This term was used because the hearing aid automatically adjusted its gain in response to changes in input intensity. As input S.P.L changed, the hearing aid controlled and modified gain without requiring manual adjustment.

Thus, A.G.C became an early term used to describe what is now commonly known as hearing aid

Compression Parameters

Because the most common complaint associated with hearing impairment is an inability to hear, it is tempting to alleviate the problem simply by making sounds louder. Figure 2-1.C shows the sensorineural hearing loss to which a fixed amount of gain is applied to all sounds – that is, amplification is linear. This makes weak sounds audible. However, average conversational speech is now loud, and intense sounds are amplified beyond the upper end of the dynamic range making them uncomfortably, or even painfully, loud.

Ouch! In contrast, in Figure 2-1.D, different amounts of gain are applied to weak, moderate and intense sounds – that is, amplification is non-linear. This squeezes the range of environmental sounds to fit within the reduced dynamic range of the person with sensorineural hearing loss. Weak sounds are made audible, moderate sounds are comfortable, while intense sounds are loud without being uncomfortable. The result is that the hearing aid user perceives the world of sounds in much the same way as a person with normal hearing.

The exact manner in which non-linear amplification, or compression, is applied depends on the goal of the hearing aid fitting as well as on the characteristics of the circuit.

Before we begin discussing compression, it may be helpful to lay the groundwork with a brief description of linear amplification, the simplest and most basic amplification system. Figure 2 to 2 shows the I/O function for a linear hearing aid. As expected, the output increases as the input to the hearing aid increases.

Figure 2-2 summary: This figure is a line chart. It displays the relationship between the input sound pressure level and the resulting output sound pressure level for a linear hearing aid. The chart shows a steady increase in output as the input level rises, until it reaches a plateau at the maximum output threshold. The data indicates that the hearing aid provides a consistent amount of gain across a wide range of input levels. Once the input reaches a certain threshold, the output remains constant, demonstrating the device's output limiting capability to prevent excessive sound levels.

More importantly, for every 10 decibels increase in the input level, the output also increases by 10 decibels. For example, when the input level increases from 50 decibels S.P.L to 60 decibels S.P.L, the output level increases from 80 decibels S.P.L to 90 decibels S.P.L. This 1:1 relationship continues until the maximum output of the hearing aid is reached. Notice also, that the gain of the hearing aid is constant at 30 decibels until the maximum output is reached. This is more easily seen in Figure 2 to 3, which shows the I/G function for the same linear hearing aid. The reduction in gain for input levels greater than 80 decibels S.P.L occurs because the maximum output of the hearing aid cannot exceed 110 decibels S.P.L. Thus, the hallmark of a linear hearing aid is that it applies a fixed amount of gain regardless of the level of the input signal, until the maximum output of the hearing aid is reached. In contrast, a non-linear hearing aid applies different amounts of gain to weak, moderate and intense sounds.

Figure 2-3 summary: This figure is a line chart. It illustrates the relationship between the input sound pressure level and the resulting gain for a linear hearing aid. The chart shows that the gain remains constant across a range of lower input levels before decreasing as the input level continues to rise. From this data, it can be inferred that the hearing aid provides a stable level of amplification for quieter sounds, but implements a compression or limiting effect at higher input levels to prevent the output from exceeding a maximum threshold.

Characteristics of a Compressor

The function and application of compression circuits are defined by their static and dynamic features. Static features, such as compression threshold and compression ratio, indicate the behavior of the circuit in response to steady input signals (e.g., a running vacuum cleaner or the constant hub-bub of a noisy restaurant). On the other hand, dynamic features, such as attack time and release time, describe the length of time required for the circuit to respond to a changing input signal (e.g., speech in a one-on-one conversation).

Compression Threshold

Compression threshold (C.T) is defined as the input S.P.L which, when applied to the hearing aid, gives a reduction in the gain of 2 ( plus or minus 0.5 decibel with respect to the gain in the linear mode (I.E.C, 1983b)). In other words, it is the point on the I/O function at which the output level is 2 decibels lower than it would be if no compression had occurred (i.e., if the processing were linear). Thus, in Figure 2 to 4, the C.T is 54 decibels S.P.L. Because it looks like the knee of a bent leg, the point at which the slope of the I/O function changes is referred to as the threshold kneepoint (T.K). The hearing aid shown in Figure 2 to 4 has a T.K of 50 decibels S.P.L. Because the C.T and T.K are within a few decibel of each other, the two terms are often used interchangeably, and will be used as such here.

Depending on the purpose, a compression system may have high or low T.K's. A high T.K, 60 decibels S.P.L or greater, is used to limit the output of a hearing aid so that it does not exceed the individual's loudness discomfort levels and to maximize listening comfort (for example, Hearing Aids 1 and 2, respectively, in Figure 2 to 5). On the other hand, a low T.K, typically set below 60 decibels S.P.L, may be used to improve audibility of the softer components of speech and/or to restore loudness perception (for example, Hearing Aid 3 in Figure 2 to 5). Finally, like Hearing Aid 4 in Figure 2 to 5, a device may have a high and a low T.K to achieve all of these goals.

Figure 2-4 summary: This figure is a line chart. It illustrates the input-output relationship of a hearing aid, mapping the input sound pressure level to the resulting output sound pressure level. The chart identifies two distinct operational regions: a linear region at lower input levels and a compression region at higher input levels, with specific markers for the threshold kneepoint and the compression threshold. The data indicates that for lower input levels, the output increases proportionally with the input. However, once the input reaches the threshold kneepoint, the gain begins to decrease, leading to a compressed output where increases in input result in smaller relative increases in output. This demonstrates the hearing aid's ability to amplify soft sounds more than loud sounds to maintain output within a comfortable listening range.

Figure 2-5 summary: This figure is a line chart. It displays the input-output functions for four different hearing aids, plotting the output sound pressure level against the input sound pressure level. Each line represents a specific hearing aid, with certain points marked as threshold kneepoints. The chart shows that as the input level increases, the output level also rises for all devices, although the rate of increase changes at the identified kneepoints. It can be inferred that different hearing aids employ different compression strategies, as evidenced by the varying locations of their threshold kneepoints and the differing slopes of their output curves. Some devices transition to a lower gain more early than others, indicating diverse levels of output limiting and amplification characteristics across the tested hearing aids.

Compression Ratio

Once the input signal is loud enough to activate compression (i.e., when the input level exceeds the T.K), the compression ratio (C.R) determines how much the signal will be compressed. Specifically, under steady-state conditions, it is the ratio of an input S.P.L difference to the corresponding output S.P.L difference. Thus, C.R relates a change in the input level ( Delta Input) to a change in the output level ( Delta Output). [The symbol Delta is pronounced "delta"] and stands for a change in the target quantity.] C.R is calculated using the formula:

Math summary: This computation calculates the compression ratio. It is performed by dividing the change in input values by the change in output values.

For the hearing aid shown in Figure 2 to 6, increasing the input from 30 to 50 decibels S.P.L ( Delta Input = 20 decibels) increased the output from 60 to 80 decibels S.P.L ( Delta Output = 20 decibels). Using the above formula, the C.R of the hearing aid is:

Math summary: This calculation determines the compression ratio by dividing the change in input values by the change in output values. The process takes a twenty decibel increase in input and divides it by a twenty decibel increase in output to produce a one to one ratio.

Similarly, when the input increases from 70 to 90 decibels S.P.L and the output increases from 90 to 100 decibels S.P.L (Figure 2 to 6), the C.R is 2:1.

C.R's are generally expressed in terms of the number of decibel by which the input must change in order to effect a 1-decibel change in the output. For example, a C.R of 2:1 indicates that a 2-decibel change in input results in a 1-decibel change in output. Because the reference condition is always a 1-decibel change in output, the latter part of the ratio may be dropped. Thus, a C.R of 2:1 can be simply expressed as 2.

Note that, in Figure 2 to 6, 30 decibels of gain is applied at and below the T.K regardless of the input level, resulting in linear amplification. Thus, another way to look at linear amplification is that it has a C.R of 1:1; compression is applied only above the T.K and when the C.R is greater than 1:1.

Depending on the purpose, a compression system may have high or low C.R's. A high C.R, 5.0 or greater, is used to limit the output of a hearing aid so that it does not exceed the individual's loudness discomfort levels (for example, Hearing Aids 1 and 2 in Figure 2 to 7). On the other hand, a low C.R, typically set between 1.0 and 5.0, may be used to improve audibility of the softer components of speech and/or to restore loudness perception (for example, Hearing Aid 3 in Figure 2 to 7). Finally, like Hearing Aid 4 in Figure 2 to 7, a device may have a high and a low C.R to achieve all of these goals. Although necessary for some applications, high C.R's are known to adversely affect the clarity and pleasantness of the amplified sound.

Figure 2-6 summary: This figure is a line graph. It illustrates the relationship between input and output sound pressure levels, demonstrating how compression ratios are calculated based on the change in input relative to the change in output. The graph shows two distinct segments of a function: an initial linear portion where the input and output change at an equal rate, followed by a segment with a shallower slope. It can be inferred that the system operates with no compression at lower input levels and transitions to a state of compression at higher input levels, where the output increases more slowly than the input.

Figure 2-7 summary: This figure is a line chart. It illustrates the input-output functions for four different hearing aids, plotting the output level against the input level in decibels of sound pressure level. Each line represents a different device, with labels indicating various compression ratios across different input ranges. The data shows that while some hearing aids maintain a linear relationship at lower input levels, others begin compressing the signal earlier. As the input level increases, all devices exhibit varying degrees of compression, where the output increases more slowly than the input. Consequently, some hearing aids provide more aggressive compression at high input levels compared to others, resulting in a more constrained output range for loud sounds.

Attack Time and Release Time

When the incoming signal changes abruptly in level from below the T.K to above it, the compressor is unable to change the gain instantaneously. The dynamic characteristics of a compressor refer to the length of time required for the compression circuit to respond to a sudden change in the input. Figure 2 to 8 is a schematic representation of how changes in the input level produce variations in gain and output over time.

Attack time (A.T) is the time delay that occurs between the onset of an input signal loud enough to activate compression (i.e., input signal exceeds the T.K) and the resulting reduction of gain to its target value. Specifically, ansi (2003) defines A.T as the time between the abrupt increase in input level from 55 to 90 decibels S.P.L and the point where the output level has stabilized to within 3 decibels of the steady value for an input of 90 decibels S.P.L. [ defines A.T as the time interval between the moment when the input signal level is increased abruptly by a stated number of decibels and the moment when the output S.P.L from the hearing aid stabilizes at the elevated steady-state level within 2 decibel. The A.T for the normal dynamic range of speech in computed between input levels of 55 and 80 decibels S.P.L, whereas the high-level A.T is computed between 60 and 100 decibels S.P.L.] In Figure 2 to 8, when the input increases to a level above the T.K (Figure 2-8.A), the gain of the hearing aid does not change immediately (Figure 2-8.B). The result is an overshoot in the output (Figure 2-8.C). As the gain approaches its target, so too does the output of the hearing aid – that is, reaches within 3 decibels of its final value.

Release (or recovery) time (R.T) is the time delay that occurs between the offset of an input signal sufficiently loud to activate compression (i.e., input signal falls below the T.K) and the resulting increase of gain to its target value. Specifically, ansi (2003) defines R.T as the interval between the abrupt drop in input level from 90 to 55 decibels S.P.L and the point where the output level has stabilized to within 4 decibels of the steady value for an input of 55 decibels S.P.L. In Figure 2 to 8, the gain of the hearing aid does not change immediately (Figure 2-8.B) when the input decreases to a level below the threshold kneepoint (Figure 2-8.A). The result is an undershoot in the output (Figure 2-8.C), until both the gain and output reach their target – that is, to within 4 decibels of their final values.

Depending on the purpose, a compression system may have fast or slow attack and release times. The faster the A.T, the shorter the duration of the overshoot and the shorter the period of time that the individual has to listen to sounds louder than desired. Indeed, A.T's as fast as 5 ms are especially desirable when compression is used to limit the maximum output of a hearing aid. Although R.T's may also be a few milliseconds in duration, the consequences when combined with a fast A.T may be undesirable because the gain will vary in response to each cycle of the incoming signal resulting in a distorted waveform.

Thus, the R.T is generally longer than the A.T. An R.T of 20 ms is considered fast. The disadvantage of A.T's and R.T's between 100 ms and 2 s is that it causes the compressor to respond to brief sounds, or lack thereof, in the environment. For example, although gain may be reduced in the presence of speech, it will increase during pauses in the utterance.

This results in a pumping sensation where the level of the background (or ambient) noise increases audibly during pauses and decreases when speech is present. This pumping sensation is less problematic for attack and release times faster than 100 ms because the gain changes occur too quickly to be perceived. A fast A.T coupled with a slow R.T (greater than or equal to 2 s) will adversely affect the audibility of speech that follows immediately after the gain reduction in response to a finger snap or the click of a pen. Finally, attack and release times of 2 s or slower respond to changes in the overall level of sound in the environment rather than to individual events.

Figure 2-8 summary: This figure consists of three vertically aligned line charts illustrating the operational behavior of a hearing aid over time. Panel A displays the input level relative to a threshold kneepoint, showing a period where the input signal rises above and then falls below this threshold. Panel B depicts the corresponding gain, which decreases when the input exceeds the threshold and recovers once the input drops. Panel C shows the resulting output level, highlighting the effects of attack time and release time. It can be inferred that the hearing aid employs a compression mechanism where gain is reduced to prevent excessive output when input levels are high. The transition periods, specifically the attack and release times, lead to temporary deviations in the output, resulting in overshoot when the signal first increases and undershoot when the signal decreases.

Some compression circuits incorporate

adaptive or variable release times – that is, the R.T is adjusted based upon the duration of the triggering signal. Figure 2 to 9 (page 16) is a schematic representation of the changes in gain and output associated with changes in the input level of different durations. Thus, for example, if the input is a brief, tran-zee-unt sound such as a door slam, the R.T is fast to have the least possible effect on the audibility of the speech that follows.

On the other hand, an input that is sustained is indicative of a change in the overall level of sound in the environment. In this instance, a slower R.T acts in much the same way as a manual adjustment of the volume control.

Although it is convenient to discuss the static and dynamic characteristics of compression as discrete entities when learning about them, it is important to understand that they interact with each other in systematic ways. For example, C.R's are determined from the response of the hearing aid to relatively steady signals, such as pure tones or speech-shaped noise. For time-varying inputs, such as speech, the effective C.R is significantly affected by the A.T and R.T. When the A.T and R.T are fast – that is, shorter than the duration of a phoneme or syllable – the gain changes sufficiently quickly to amplify softer components more than the louder components.

The result is an effective C.R for speech that is similar to that specified on the basis of steady signals. On the other hand, when the A.T and R.T are slow – that is, longer than the duration of a typical word or utterance – the gain does not change much between softer and louder phonemes. As a result, the effective C.R for speech is much lower than would be expected for steady signals. Figure 2 to 10 shows the effects of fast versus slow A.T and R.T on average conversational speech. Specifically, note that the range between the upper and lower limits of speech is smaller with the fast A.T and R.T, indicating more compression of the signal, than with the slow A.T and R.T.

Another example of the interaction between compression parameters is the observation that fast A.T's and R.T's are more detrimental to the perceived sound quality at high C.R's than at low C.R's.

Figure 2-9 summary: This figure consists of three aligned schematic line graphs. The top panel illustrates an input signal relative to a threshold kneepoint, the middle panel shows the corresponding gain changes, and the bottom panel depicts the resulting output signal over time. When the input signal exceeds the threshold kneepoint, the gain decreases rapidly during the attack time and remains low while the input stays high. Once the input falls below the threshold, the gain gradually increases during the release time. Consequently, the output signal is compressed during periods of high input, with the attack and release times governing how quickly the output level adjusts to changes in the input signal.

Input-Versus Output-Controlled Compression

The term automatic gain control (A.G.C) is often used to describe compression circuits because the amount of gain applied is automatically determined by the signal level. Thus, a level detector is an essential component of any compression circuit. The position of this level detector relative to the volume control influences the operation of the circuit.

With Input-Controlled Compression (A.G.C-I),

the level detector is located before the volume control (Figure 2 to 11) and compression acts on the input to the hearing aid. That is, once the input exceeds the T.K, the compressor is activated and gain is reduced at the pre-amplifier. Therefore, the volume control setting has no impact on the compression parameters.

As shown in Figure 2 to 12 (page 18), rotating the volume control from full-on to -20 decibels (relative to full-on) decreases the gain of the hearing aid above and below the threshold kneepoint. However, neither the T.K nor the C.R changes as a result.

Figure 2-10 summary: This figure is a line chart featuring shaded areas. It illustrates the output levels of a hearing aid across a range of frequencies when utilizing fast versus slow attack and release times in response to a speech signal. The data shows that the slow setting consistently results in higher output levels across the entire frequency spectrum compared to the fast setting. It can be inferred that fast attack and release times provide more aggressive gain reduction, leading to a lower overall output, whereas slow timings allow for higher output levels, particularly in the mid-frequency range.

Figure 2-11 summary: This figure is a schematic block diagram. It illustrates the signal flow of an input-controlled compression circuit, also known as an automatic gain control circuit. The process begins with a microphone that feeds a signal into both a level detector and a pre-amplifier. The level detector provides feedback to the pre-amplifier, which then passes the signal through a volume control stage and an output amplifier before reaching the receiver. The layout indicates that the circuit dynamically manages signal levels by monitoring the input via the level detector to adjust the amplification process, ensuring a consistent output level at the receiver regardless of the initial input strength.

Figure 2-12 summary: This figure is a line chart. It illustrates the relationship between input and output sound pressure levels for an input-controlled compression circuit across different volume control settings, specifically showing the full on position and rotations of negative decibels. The chart identifies specific kneepoints where the slope of the output changes relative to the input. The data indicates that as the volume control is rotated downwards, the overall output level decreases for any given input level. Furthermore, the compression effect begins at a consistent input level regardless of the volume control setting, but the resulting output level at that threshold is lower when the volume is rotated down.

In Output-Controlled Compression (A.G.C-O),

the level detector is located after the volume control (Figure 2 to 13) and compression acts on the output of the hearing aid. That is, the compressor is activated once the output exceeds the T.K. As shown in Figure 2 to 14, rotating the volume control from full-on to -20 decibels (relative to full-on) decreases the gain of the hearing aid. More importantly, however, it results in an increase in the T.K.

It is impossible to determine whether a compression circuit uses A.G.C-I or A.G.C-O simply by examining a single I/O function or frequency response curve. The only way to distinguish between the two types of circuits is by measuring the response of the hearing aid at various settings of the volume control. Although gain changes are seen with both types of circuits, the T.K changes only in A.G.C-O circuits.

Channels and Bands

Although not strictly a “characteristic” of a compressor, a brief discussion of channels and bands is in order. The terms are often used erroneously or interchangeably, but a distinction is made between the two in this book.

Frequency bands are independently

The band is independently controlled areas for gain adjustment only. Thus, increasing or decreasing the gain in a frequency band will equally affect the response to weak, moderate and intense sounds at frequencies within that band. The compression parameters are unaffected. In contrast, compression channels allow separate adjustments for weak and intense input levels; the effect on moderate level inputs depends on the compression architecture. Thus, in addition to gain, changes within a compression channel may affect the C.R and/or the T.K. Depending on the controls provided by the hearing aid manufacturer, it may be possible to make changes across both, frequency bands and compression channels.

A hearing aid may have the flexibility to allow adjustments in multiple frequency bands and/or compression channels. Multiple channels are separated by crossover frequencies, which may or may not be adjustable. The ability to change gain individually in bands and/or channels is useful for two reasons. First, there is a wide variety of audiometric configurations for which a given hearing aid may be useful. The ability to adjust gain and compression in discrete frequency regions permits customization of amplification to the individual's needs.

Another advantage of multichannel compression is that acoustic events in a discrete frequency region do not affect the response of the hearing aid at all frequencies. For example, Figure 2 to 15, shows the response of a 1-channel and 4-channel hearing aid in the presence of a 90 decibels S.P.L tone at 500 Hz. It is clear that the high-frequency response is considerably reduced for the single-channel hearing aid, but unaffected for the multichannel hearing aid. This multichannel advantage may also extend to advanced signal processing features such as noise management.

Although on the surface it appears that a large number of channels would be advantageous, it is unclear as to whether or not this is the case. Trine and Van Tasell (2002) have shown that 3 to 4 channels are adequate to fit the majority of audiometric configurations, with or without the ability to further fine-tune the frequency response within each channel (i.e. in frequency bands), provided that the crossover frequencies are adjustable. One argument against the use of multichannel compression is that it obliterates the relative intensity relationships between various speech sounds, which are an important cue for speech understanding. Finally, the sounds and noises that occur in our environments are usually not discrete in frequency. Thus, gain would be affected in fairly broad frequency regions even for a hearing aid with an infinite number of channels. This problem is further exacerbated when there is considerable overlap across channels.

To underscore this point even further, readers are advised to internalize that on I/O functions, the length of any gain function (showing either linear gain or compression) has precious little (read “nothing”) to do with the amount of gain. It is only the position of the gain functions themselves—right or left—along the horizontal axis that shows an increase or a decrease in gain. In fact, on I/O functions, a rightward shift in gain functions actually shows a decrease in gain, while a leftward shift would show an increase in gain. If this sounds contrary to intuition, the reader is highly encouraged to read on to the next section on “input compression versus output compression”—where this will be explained.

We have talked about compression knee-points, and now it is time to discuss compression ratios. Compression ratios are the amount of compression provided by the hearing aid once compression begins at the knee-point. Compression ratios can be visualized on an I/O function by the slant of the line after (or to the right of) the knee-point.

A 10:1 compression ratio means that for every 10-decibel increase of input S.P.L, there is only a 1-decibel corresponding increase to the output S.P.L. A 2:1 compression ratio means that for every 10-decibel increase of input S.P.L, there is a corresponding 5-decibel increase to the output S.P.L of the hearing aid. Higher compression ratios indicate more compression; lower compression ratios indicate less compression.

In general, one can think of the knee-point as the “when” of compression and the ratio as the “how much” of compression. In the analog era of hearing aids, compression was often referred to as automatic gain control (A.G.C), because the gain of the hearing aid changes as the input intensity S.P.L changes.

Figure 2-13 summary: This figure is a schematic diagram. It illustrates the functional block diagram of an output-controlled compression circuit, showing the signal flow from a microphone through a pre-amplifier, a volume control mechanism, and an output amplifier before reaching the receiver. A level detector monitors the signal after the output amplifier and provides a feedback loop to the pre-amplifier. The arrangement indicates that the system employs a feedback mechanism where the output signal level is used to automatically regulate the gain of the initial amplification stage, ensuring a consistent output level regardless of the input intensity.

Figure 2-14 summary: This figure is a line chart. It illustrates the relationship between input and output sound pressure levels for an output-controlled compression circuit under different volume control settings, specifically when the control is full on and after rotations of negative ten and negative twenty decibels. The chart identifies specific kneepoints where the output behavior transitions from a linear increase to a compressed state. The data indicates that as the volume control is rotated downwards, the input level required to reach the compression threshold increases. Consequently, while the output levels converge at high input intensities, lower volume settings result in a reduced output for a given input level until the compression point is reached.

Figure 2-15 summary: This figure is a line chart. It displays the relationship between frequency and gain for two different hearing aid configurations, specifically a single-channel device and a four-channel device, when exposed to a high-intensity pure tone. The x-axis represents frequency and the y-axis represents gain. The four-channel hearing aid demonstrates a significantly higher peak gain in the higher frequency range compared to the single-channel hearing aid. While both configurations show similar gain levels at lower frequencies, the multi-channel device provides more targeted amplification at higher frequencies before both curves decline at the upper end of the spectrum. This indicates that a higher number of channels allows for more precise and flexible gain adjustment across different frequency bands.

Input Compression Versus Output Compression

Regarding compression in analog hearing aids, this was the first division encountered by the clinician when considering what type of compression to fit. When hearing aids were analog, their compression was either input related or output related. Clinicians thus had to know when to fit which type and, hence, which one to select from a given manufacturer. The digital algorithms in today's hearing aids, however, mathematically calculate and imitate the actual effects of output versus input compression; hence, this is no longer an issue in hearing aid selection.

Basically and in a nutshell, the issue of input versus output compression is one related to how the hearing aid volume control (V.C) operates and what it adjusts. Considering that many of today's hearing aids do not even have a V.C, the issue becomes even more remote. Readers can feel free to skip over this section as it no longer addresses the fitting concerns of today's hearing aids. Otherwise, in the interest of learning how these historically worked in analog hearing aids—as well as for the sake of practicing how to read and interpret I/O functions—do feel free here to "indulge."

In analog hearing aids, the big difference between input and output compression was where the volume control (V.C) sat in the circuit (Figure 7 to 2, bottom). These differences in V.C location consequently affected what the V.C did when manipulated (Figure 7 to 2, top). For output compression hearing aids, the V.C was physically situated “early on” in the circuit, located between the microphone and the amplifier (Figure 7 to 2, bottom left). For input compression hearing aids, the V.C was situated almost dead last in the circuit, located just in front of the receiver that sends sound into the ear (Figure 7 to 2, bottom right).

The two I/O functions in Figure 7 to 2 show the different effects of the V.C with output compression hearing aids (top left) versus input compression hearing aids (top right). With output compression, the V.C adjusted the gain only; that is, it moved the 45° linear gain functions to the right or left. It did not, however, change the height of the line to the right of the knee-point (the M.P.O). With input compression (top right), the V.C adjusted both the gain and the M.P.O together at the same time. Again, in the world of analog hearing aids (where these things could not simply be changed around), if the V.C had a different effect for input than it does for output compression, the clinician had to know what that difference was.

Figure 7-2 summary: This figure consists of two sets of line graphs and corresponding circuit diagrams. The top portion displays input-output functions for output compression and input compression, while the bottom portion shows the relative placement of the volume control within the signal path relative to the amplifier and compression stages. In the output compression model, the volume control is placed before the amplifier, resulting in changes to gain and the threshold knee-point while the maximum power output remains constant. In contrast, the input compression model places the volume control after the amplifier, which alters the gain and the maximum power output while the knee-point remains fixed. Consequently, the figure demonstrates that the position of the volume control determines whether the compression affects the threshold of activation or the final output limit.

Digital hearing aids today are no longer locked into being either input or output compression. The software algorithms mathematically create the effects of different physical V.C locations—and consequently input and output compression—in

Input versus Output Compression: Volume Control Effects

analog circuits. Today, most digital bearing aids incorporate the use of input compression for soft inputs, along with output compression for louder inputs. In other words, for soft inputs, V.C adjustments tend to affect both the gain and M.P.O together. For louder inputs, V.C changes tend to affect the gain only and not the M.P.O. The logic here is that high-intensity inputs plus the choice of an increased gain with a higher V.C position would equal an overly intense output and thus exceed the listener's loudness discomfort levels.

Output Compression on an I/O Function

Look closer now to the left panel of Figure 7 to 2, showing output compression where the V.C affects the gain but not the M.P.O. The three 45° diagonal functions illustrate the effects of three different V.C positions upon the gain of the hearing aid. Note carefully that the right-most line actually shows minimum gain, with the V.C lowered to a minimum position. The left-most line shows maximum gain, with the V.C raised to a maximum position.

To make this clear, it may help to draw some more lines here. On the I/O function showing output compression (left), draw a vertical line down from the right-most knee-point to the input axis. Now from the same knee-point, draw a horizontal line leftward to the output axis. This shows that at this V.C setting, some X amount of input is needed to give some Y amount of output. If similar vertical and horizontal lines are drawn from the left-most knee-point, it becomes clear that for this V.C position, a lot less input is needed to give almost the same amount of output. This would mean that the gain has been increased.

The same I/O function also shows that once “past” or to the right of the knee-point, in the region of compression, there is only one M.P.O line, and it is common to all three diagonal lines. This shows that, for output compression hearing aids, the V.C does not affect the M.P.O. Output compression was thus deemed as very suitable for high-power hearing aids; here, clinicians should be very concerned about providing excessive output that can potentially damage the client's hearing even further. With output compression, no matter what the client does to the V.C position, the M.P.O remains unchanged.

Note also that with output compression, the V.C also changes the compression knee-point. In analog hearing aids, this was because the compression knee-point was adjusted “later on” in the circuit after the V.C (Figure 7 to 2, bottom left). The compressor in the circuit was always set to wait for some steady amount of voltage that would tell it to compress, and the V.C affected this amount.

If it was not sufficient to tell the compressor to compress, then it would not act. Only when the V.C sent the required input signal voltage that the compressor was “waiting for” would the compressor then “do its thing.” Again, in today's digital hearing aids, these actions are mimicked mathematically in the digital software algorithms.

Input Compression on an I/O Function

For input compression hearing aids, the effects of the V.C are completely different (Figure 7 to 2, top right). Here, the V.C affects both the gain and the M.P.O. Again, three diagonal gain lines for three different V.C positions are shown. Once again, the rightmost 45 superscript circle diagonal gain line shows the lowest V.C setting, and the left-most gain line shows the highest or maximum V.C setting. It is obvious that the M.P.O is also affected by the V.C because once “past” or to the right of the knee-point, the height of all three lines also changes.

On analog input compression hearing aids, there was no specifically intended design feature advantage nor is there a real clinical fitting advantage to have the V.C affect the M.P.O. Rather, this was simply a by-product of the V.C placement in the circuit. Fortuitously though, the V.C placed between the amplifier and receiver did allow for another control for adjustment to be placed near the beginning of the circuit (between the microphone and the amplifier). This is what came to be known as the "T.K" control, and it was to be found on the new type of compression known as W.D.R.C, which became extremely popular during the 1990s. The location of this adjustment control explains why W.D.R.C was always found on input (rather than output) compression hearing aids. More on this topic will be described later on in the next section: Output Limiting Compression Versus Wide Dynamic Range Compression.

The vast pool of clients with mild-to-moderate S.N.H.L were the ones for whom W.D.R.C was intended. They have a larger dynamic range than those with severe S.N.H.L and thus can more easily accommodate an M.P.O that rides up and down with V.C adjustments. Input compression hearing aids with W.D.R.C were therefore often moderate-power hearing aids, intended for mild-to-moderate (“sensory”) S.N.H.L.

Note also on Figure 7 to 2 that for input compression hearing aids, the V.C did not affect the knee-point of compression. As Figure 7 to 2, bottom right shows, the compressor was situated before

The V.C. This Meant That The V.C Did Nothing To The Knee-Point Because The Compression Knee-Point Was Already Determined!

I/O functions are not the only way to look at the differences in the V.C effects. The clinician is apt to be familiar with the frequency responses (gain as a function of frequency) seen with ansi testing on a hearing aid test box screen or printout, as shown in Figure 7 to 3. Here, the effects of the V.C on gain and M.P.O are readily apparent for output compression (left) in comparison to input compression (right). For output compression, the V.C increases and decreases the gain; for input compression, the V.C increases and decreases both the gain and the M.P.O together.

In addition to testing each type of compression on a hearing aid test box, one could also hear the effects of output versus input compression. With soft inputs to either hearing aid, adjustments to the V.C would be audible as increases and decreases to the gain. With louder inputs, such as talking loudly into each hearing aid, adjustments to the V.C would be audible mostly.

Input versus Output Compression: Volume Control Effects on Frequency Response

when listening to the input compression hearing aid. This would occur because the loud input plus the gain of the hearing aid would equal the M.P.O, and only on input compression hearing aids did the V.C also adjust the M.P.O.

Figure 7-3 summary: This figure consists of two line graphs comparing frequency responses under different compression settings. The graphs illustrate the relationship between frequency and decibels for both output compression and input compression as the volume control is varied from minimum to maximum. In both scenarios, multiple gain curves are shown. For output compression, a single maximum power output limit is maintained across all volume settings. For input compression, multiple maximum power output limits are shown, shifting in tandem with the gain curves. It can be inferred that for output compression, adjusting the volume control only modifies the gain without affecting the maximum power output. In contrast, for input compression, changing the volume control simultaneously adjusts both the gain and the maximum power output.

Output Limiting Compression Versus Wide Dynamic Range Compression

We have discussed the historic analog hardware differences between input and output compression and their resultant effects upon V.C adjustment. Due to their digital implementation today, these no longer play a part in clinical decisions regarding compression. With this behind us, we now come to another major fork in the road concerning compression—namely, output limiting compression (O.L.C) and W.D.R.C. These are essentially two different compression schemes, and they refer to separate ranges of compression threshold knee-points and compression ratios. The I/O functions of O.L.C versus W.D.R.C, as well as the effects of their respective adjustments, will be compared and contrasted here.

Output Limiting Compression (O.L.C)

The salient features of O.L.C are shown in the I/O function in Figure 7 to 4 (left). Here it can be seen that O.L.C has relatively high compression knee-points and high compression ratios. A high knee-point means that the hearing aid begins to use compression only at high input S.P.L's (e.g., 60 decibel S.P.L or more). Below the knee-point, the O.L.C hearing aid provides linear gain. A high compression ratio is usually defined as being greater than 4:1. Recall also that compression we have discussed so far implies less gain than linear gain.

We have already mentioned that compression ratios are about the amount of compression provided by the hearing aid, once compression begins. As Figure 7 to 4 shows, a 10:1 compression ratio results in an almost completely horizontal line to the right of the knee-point. Compare this to the linear gain shown in Chapter 4, Figure 4 to 3. The two figures look very similar.

Here, it can be seen that O.L.C can be considered to be a “close cousin” to linear gain. The only difference is that with linear gain, the M.P.O

Output Limiting Compression and M.P.O Adjustment

is limited by means of “hard peak clipping.” On the other hand, with O.L.C, the M.P.O is limited with a bit of “give;” in other words, by means of a high compression ratio. Looking at Figures 4 to 3 and 7 to 4 it can also be seen that the linear and O.L.C hearing aids have essentially similar gain characteristics in that they both provide linear gain over a wide range of soft to moderately intense input levels. The numerical values on both figures here are kept the same mainly for illustration and comparison.

Linear hearing aids actually came in all kinds of strengths, with greater and lesser amounts of gain. The thing they all did, however, was to limit the M.P.O by means of peak clipping. O.L.C hearing aids simply utilized a high degree of compression to accomplish the same thing. This introduces a lot less distortion than peak clipping.

By way of analogy, instead of jumping on a bed and hitting one's head against a cement ceiling, it's as if a sponge was attached to the cement ceiling, thus softening the thud. In fact, that is how the first compression came to be; it was a method of limiting the M.P.O without the distortion caused by peak clipping! The first type of compression to emerge in the 1980s was in fact O.L.C.

Along with output compression, then, O.L.C came to be intended to be used by clients with severe hearing loss who would benefit from high-power hearing aids. In fact, in analog hearing aids, O.L.C was typically associated with output compression hearing aids (see Figure 7 to 2), which were also intended for those with severe hearing loss. The combination worked out very well. With output compression, V.C adjustments affected only the gain and not the M.P.O. With O.L.C, the high knee-point and high compression ratio served to “do their work” for loud inputs only.

Notice that Figure 7 to 4 (left) shows that O.L.C provides a strong degree of compression over a narrow range of intense inputs. Below the threshold knee-point, the O.L.C hearing aid provides linear gain for a relatively wide range of soft-to moderate-input S.P.L's. In other words, it “waits” for a fairly high-input S.P.L to go into compression, but once it goes into compression, it really goes into compression. In this way, O.L.C could be said to focus on the “ceiling” of a client's dynamic range.

Regarding compression ratios, we should take a moment to note that there is a law of diminishing returns. As compression ratios increase, there is progressively less and less of an effect; in other words, there is less and less of a reduction in gain. Take, for example, the gain reduction offered by a 2:1 ratio as compared to a 1:1 ratio (linear gain). With a 1:1 ratio, for every 10 decibel of input increase there is a corresponding 10 decibel of output increase.

With a 2:1 ratio, for every 10 decibel of input increase, there is a corresponding 5 decibel of output increase. The gain has gone down here by half. Now consider a 10:1 ratio as compared to a 20:1 ratio. With a 10:1 ratio, for every 10 decibel of input increase, there is a corresponding 1 decibel of output increase. With a 20:1 ratio, for every 10 decibel of input increase, there is a corresponding 0.5 decibel of output increase.

The gain here has also gone down by half, but going down from 10 to 5 is a lot more than going down from 1 to 0.5. In other words, the amount of gain reduction between a 1:1 and a 2:1 ratio is greater than the amount of gain reduction between a 20:1 ratio and a 10:1 ratio. An analogy might help here. Think of standing a meter from a wall. Cut the distance in half and you get that much closer to the wall. Cut the distance in half again, and you get closer still to the wall, but note how the distance has now diminished. By the way, you can keep doing this, but you will never get to the wall.

Figure 7 to 5 shows inputs, outputs, and gain for four different compression ratios. The 1:1 linear gain shown on the left corresponds to the linear example shown in Figure 4 to 3, Chapter 4.

Gain Differences with Different Compression Ratios

Figure 7 to 5. Examples of inputs, outputs, and gain are shown for four different knee-points and ratios. The asterisk in each example shows the knee-point. The underlined values in each example show the results past the knee point.

Linear 1:1 gain with peak clipping is shown in the left-most column. The linear gain here is 60 decibel. Note how with peak clipping, once the input level exceeds the knee-point, the M.P.O does not increase at all. The second column shows O.L.C with a 10:1 compression ratio. The linear gain here is also 60 decibel. Note how with O.L.C, once past the knee-point, the output increases but only by a slight amount.

The third column also shows O.L.C but this time with a 20:1 compression ratio. The linear gain is once again 60 decibel. The right-most column shows an example of W.D.R.C. The linear gain is only 40 decibel and the knee-point is also at 40 decibel. Note the contrast to the previous three examples; once past the knee-point, the outputs and gain decrease gradually as inputs increase.

The second column from the left corresponds to the 10:1 compression ratio of the O.L.C example shown in the left panel of Figure 7 to 5. The third column from the left shows another example of an O.L.C hearing aid; this time, however, the compression ratio is 20:1. Note that the gain reduction here is only very slightly less than that offered with a 10:1 compression ratio. This is an example of the diminishing returns that occur with progressively higher compression ratios. The right-most column shows an example of W.D.R.C, which will be described later.

Adjustment of M.P.O in O.L.C Hearing Aids

We have described how O.L.C was used as an alternative way to limit the M.P.O without the distortion caused by linear peak clipping. Recall from Chapter 4 that the M.P.O in linear hearing aids could be raised or lowered so as to best address the client's loudness tolerance levels. Similarly, the M.P.O could also be raised and lowered in O.L.C hearing aids (see Figure 7 to 4, right panel).

It is important to know that adjustment of O.L.C is completely independent from V.C position. In Figure 7 to 4 (right panel), consider that the V.C is set at a position that is comfortable for the listener. The I/O function in the right panel shows the effects of adjusting the M.P.O in O.L.C hearing aids. It is obvious that adjustment of the M.P.O also affects the compression knee-point.

In analog hearing aids, this was because adjusting the M.P.O actually adjusted the voltage level that the compressor of the circuit needed that would tell it to begin compressing. As the M.P.O was turned to a maximum position, the compression knee-point was raised. As Figure 7 to 4 (right panel) shows, at any knee-point setting, the compression would not occur until input sounds that are higher than the intensity level specified by the knee-point are reached.

Increasing or raising the M.P.O increases the length of the 45° linear gain function, but it does not increase the gain. Recall that the length of any function or line in an I/O function has nothing to do with the amount of gain; only a right or left movement of the line shows a change in gain. Changes to the M.P.O here do not move the 45° lines to the right or to the left.

The effects of M.P.O adjustment can be heard in addition to being tested and witnessed on a hearing aid test box. To hear the effect of M.P.O adjustment, one would have to speak loudly into the hearing aid, because only then will the input sound plus the gain reach the M.P.O. A low-intensity input sound, such as a soft voice, plus the gain of the hearing aid may not result in an out-put that reaches the M.P.O. As the M.P.O is turned from a maximum position to a minimum position, the knee-point of compression and the M.P.O are reduced; when listening, one should notice that the amplified loud voice becomes softer. Again, this is because the O.L.C compression adjustment affects the M.P.O, not the gain.

An M.P.O type of adjustment on O.L.C hearing aids is especially useful for those clients who have a severe or profound hearing loss and a very limited dynamic range. The M.P.O can be set to a level that corresponds to the client's loudness tolerance levels. Some clinicians opt to set the M.P.O (measured in decibel S.P.L) to be about 15 decibel higher than the client's reported loudness tolerance levels (measured in decibel H.L). The rationale here is that the difference in decibel H.L versus decibel S.P.L over various intensity levels and across the speech frequencies is close to an average of about 15 decibel (with decibel S.P.L showing the greater decibel values).

Wide Dynamic Range Compression (W.D.R.C)

W.D.R.C hearing aids became extremely popular during the 1990s. As such, it was a newcomer to the world of compression: the "new kid on the block." Many clinicians who were used to O.L.C and its adjustments were initially quite confused by W.D.R.C and especially how it was adjusted. It is important to categorize where W.D.R.C properly fits in the world of compression, because then it can be appreciated for what it is and what it is not.

A typical I/O function for W.D.R.C is shown in Figure 7 to 6 (left). In contrast to O.L.C, W.D.R.C is associated with low threshold knee-points (below 60 decibel S.P.L) and low compression ratios (less than 4:1). In fact, W.D.R.C most commonly utilizes a 2:1 compression ratio. A look at Figure 7 to 6 (left) shows that due to its low knee-point the W.D.R.C hearing aid is in compression over a relatively wide range of inputs. As such, it is almost always in compression. Inputs from very soft speech to a yell will cause it

Wide Dynamic Range Compression and Gain Adjustment

to go into compression. It was called “wide dynamic range compression” because its low knee-point and low compression ratio serve to shrink a normally wide dynamic range into a smaller one that occurs with S.N.H.L.

Look at the slope of W.D.R.C as shown in Figure 7 to 6 (left panel) and compare that to the slope of O.L.C, as shown in the left panel of Figure 7 to 4. It is evident that W.D.R.C does not provide a great ratio or degree of compression. Instead, a W.D.R.C hearing aid provides a weak degree of compression over a wide range of inputs. The effect of W.D.R.C is thus very different from O.L.C—or the old linear hearing aids, for that matter. A listing of inputs, outputs, and consequent gain for an example of W.D.R.C is also shown in the right-most column of Figure 7 to 5. Unlike linear or O.L.C hearing aids that dramatically reduce the output (and hence the gain) for intense inputs once the input S.P.L exceeds a certain amount, W.D.R.C gradually reduces the output (and hence the gain) for a wide range of moderate to intense input S.P.L's. The I/O function for W.D.R.C (see Figure 7 to 6, left) actually shows that the hearing aid provides compression for more inputs than those that cause it to provide linear gain. W.D.R.C basically provides most gain for soft inputs and progressively less gain for progressively louder inputs. For very soft inputs linear (maximum) gain is provided; all other inputs are given compression (less gain). In contrast to O.L.C, W.D.R.C can be seen as having a focus on the "floor" of hearing sensitivity.

Figure 7-4 summary: This figure consists of two line charts illustrating the input-output function and the maximum power output adjustment for output limiting compression. The first chart displays a linear gain phase that transitions into a highly compressed phase after a specific knee-point, while the second chart shows how adjusting this knee-point affects the resulting output levels. The data indicates that the system provides maximum linear gain for soft and average input levels, but applies significant compression beyond the knee-point to strictly limit the ceiling of loudness tolerance. Consequently, lowering the knee-point results in a corresponding decrease in the maximum power output.

Figure 7-5 summary: The table compares different compression ratios and gain behaviors across linear gain, output limiting compression, and wide dynamic range compression. For linear gain and low-ratio limiting, the gain remains constant until a specific knee-point is reached, after which the output levels off and the gain decreases significantly. In contrast, wide dynamic range compression shows a more gradual reduction in gain as the input increases, resulting in a more controlled increase in output levels compared to the sharper limiting seen in the higher ratio settings.

Figure 7-6 summary: This figure consists of two line graphs illustrating Wide Dynamic Range Compression (WDRC) characteristics. The left graph displays an input-output function where a linear gain is applied to soft input levels until a specific knee-point is reached, after which a compression ratio is applied to limit the output. The right graph demonstrates the adjustment of the threshold knee-point, showing how varying the starting point of compression alters the gain applied to soft inputs relative to the maximum power output. It can be inferred that WDRC is designed to increase hearing sensitivity by providing higher linear gain for soft sounds. The data indicates that lowering the knee-point effectively increases the gain for soft input levels, while the low compression ratio ensures a gradual limitation of the output as input levels increase.

Output Limiting vs W.D.R.C: Displayed as Frequency Responses Both W.D.R.C and O.L.C can also be described in terms of their effects upon the frequency response of a hearing aid (Figure 7 to 7), much like we did with input versus output compression (see Figure 7 to 3). The left panel of Figure 7 to 7 shows that O.L.C provides its maximum gain for both soft and average input levels and then suddenly reduces its gain once the input level becomes more intense than its relatively high knee-point. The right panel of Figure 7 to 7 shows that W.D.R.C gradually reduces its gain over a wide range of increasing input sound levels above its low knee-point. In other words, in contrast to O.L.C, where the gain reduces abruptly as soon as an input sound is above its knee-point, the gain of a W.D.R.C hearing aid changes more gradually, over a wider range of input levels. Due to this behavior, W.D.R.C was sometimes called “input level-dependent compression.”

W.D.R.C is not without its problems. It has been suggested that the use of W.D.R.C can degrade some of the cues necessary for speech recognition. The reasoning is that since W.D.R.C amplifies soft inputs more than loud inputs, it will reduce the differences between the “peaks” (louder parts) and “valleys” (softer parts) of input speech sound waves. This will serve to “muddy” the crisp clarity of natural speech, thus making it more difficult to understand, especially in background noise.

Villchur (2008) does not agree with this stance, saying that the listener with S.N.H.L has recruitment and hence hears the loudness contrasts between the peaks and valleys of the speech waveform in an exaggerated manner. The exaggerated perception of loudness with recruitment therefore cancels out the reduction of intensity in the speech waveform. Therefore, while it is true that W.D.R.C decreases the peak-to-valley contrasts in the speech waveform envelope, the listener with recruitment will still benefit from W.D.R.C. Villchur also points out that although the use of high-frequency gain/output with W.D.R.C will reduce peak-to-valley contrasts, it will do so only between vowels and high-frequency unvoiced consonants; high-frequency emphasis will not affect the peak-to-valley contrasts between vowels and low-frequency voiced consonants.

We will return to this topic again in this chapter, in the section on “Syllabic Compression.”

Before leaving this topic, a word must be said here regarding linear gain versus the gain with W.D.R.C. Provided that outputs are not saturated—as long as the input plus the gain does not add up to equal the M.P.O where peak clipping—and conse- quent distortion takes place—linear gain hearing aids can sound very “clean” and clear. This is because there is nothing but pure amplification (no compression) taking place, which can audibly distort the sound quality. More will be described on the usage of linear gain in digital hearing aids in Chapter 8. Also, the impact of W.D.R.C versus linear gain upon the waveforms of speech will be discussed and described further in Chapter 10.

Figure 7-7 summary: This figure is a line chart. It illustrates the frequency response of a hearing aid by plotting gain against frequency for several different input intensities. The chart displays multiple curves representing different input levels, showing how the gain changes across the frequency spectrum for each level. The data indicates that for lower input intensities, the gain remains relatively consistent and linear. However, as input intensity increases beyond a certain threshold, the gain is significantly reduced. This demonstrates that the device employs compression to limit gain for louder sounds, providing varying levels of amplification based on the intensity of the input signal.

Adjustment of Gain in W.D.R.C: The “T.K” Control

In analog hearing aids with W.D.R.C, the T.K control was used to adjust the amount of linear gain given for very soft input sounds, that is, for inputs that are below those specified by the knee-point (see Figure 7 to 6, right panel). Like all compression hearing aids, W.D.R.C provides greater (linear) gain for inputs below the threshold knee-point of compression. The T.K adjustment serves to shift the 45° linear gain function to the right or left. Note how this is so very different from the adjustment of the M.P.O with O.L.C (see Figure 7 to 4, right panel).

Consistent with what we have noted earlier, the left-most 45 superscript circle linear gain line shows the greatest gain. This is the T.K set to the lowest knee-point position. The right-most gain line shows where there is the least amount of gain for soft inputs; here, the T.K is set to the highest knee-point position. In summary, as the compression knee-point with the T.K control is lowered, the gain for low-intensity input sounds is increased. As the compression knee-point is raised, the gain for low-intensity input sounds is decreased.

Recall from our previous discussion on input versus output compression that W.D.R.C had always been associated with input compression hearing aids. The basic circuit schematic shown on the bottom right of Figure 7 to 2 shows that for input compression, the V.C is located near the end of the circuit, between the amplifier and the receiver. As we mentioned earlier under the section "Input Compression on an I/O Function," there was no real clinical advantage for placing the V.C in this position.

Regarding the fact that W.D.R.C was always found in input compression hearing aids, however, there was indeed a design advantage. The T.K adjustment control could now be placed where the V.C was normally located for output compression circuits—namely, between the microphone and the amplifier. The T.K thus affected the amount of input signal that arrived at the compressor of the circuit, just like the V.C does for output compression hearing aids. This very design feature resulted in the fact that T.K adjustment looks much like the V.C adjustment did for output compression hearing aids! A look at Figure 7 to 2 (left panel) and also Figure 7 to 6 (right panel) will confirm that T.K adjustments in the input compression W.D.R.C hearing aid look much like the V.C adjustment on output compression hearing aids.

As W.D.R.C emerged after O.L.C, the T.K type of compression adjustment that came along with it also emerged later than the original M.P.O adjustment found on O.L.C hearing aids. As mentioned earlier, many clinicians during the 1990s were initially confused by the T.K control because it worked so very differently from the M.P.O adjustment to which they were accustomed. W.D.R.C utilizing a T.K type of gain adjustment for soft inputs was first commonly associated with hearing aids using the Kamp circuit.

On today's digital hearing aids, the T.K adjustment works the same way, but it may not always be called a "T.K" control. It may be seen on digital software simply as the left-most, or lower, knee-point on a rather complex-looking I/O function. More will be specifically described about compression in digital hearing aids per say in Chapter 8. For now, let's simply look at the T.K adjustment and how it works in general.

When listening to a hearing aid with a T.K control, in order to hear the effect of adjusting the T.K, it is important to let just the ambient noise of the room into the microphone. Since the knee-point range is so low (30 to 50 decibel S.P.L), the effect of adjusting the T.K will not be audible with any input greater than this. Adjusting the T.K will increase and decrease the loudness heard only for these very soft inputs. Contrast this to the very different adjustment of M.P.O found on O.L.C hearing aids, where loud input sound became louder to the listener as the knee-point was raised and the M.P.O was thus increased.

There is no real rule for adjusting the T.K control; after all, what rule would one use to determine whether to adjust the T.K control to inputs of 30 decibel S.P.L versus 50 decibel S.P.L? The main reason the T.K was adjustable in the first place was because a T.K setting for maximum gain in quiet (the lowest knee-point) environments could result in the client being able to hear the internal amplifier and microphone noise of the hearing aid itself. The audible hiss can be annoying, especially for the client who has excellent low-frequency hearing. This is not an issue for the client who presents with a flat, moderate S.N.H.L.

In today's digital hearing aids, “expansion” is commonly used along with W.D.R.C, in order to reduce the audibility of the “hissing” sounds in quiet. More will be described about expansion in Chapter 8.

Clinical Applications of Output Limiting Compression and W.D.R.C

When comparing clinical applications of O.L.C and W.D.R.C, it may be most useful and helpful to look closely at their names. The main clinical difference between the two is that O.L.C does its work above its high knee-point; its high compression ratio serves to reduce or limit the output for high input S.P.L's. On the other hand, W.D.R.C provides maximum (linear) gain below its relatively low knee-point; a weak or low compression ratio for average and high input S.P.L's means that maximum (linear) gain is provided only for soft inputs. As mentioned earlier, concerning the dynamic range of hearing sensitivity, O.L.C concentrates on the “ceiling,” while W.D.R.C concentrates on the “floor.”

W.D.R.C is normally used for mild-to-moderate S.N.H.L. As mentioned in previous chapters, the client who has O.H.C damage usually has “sensory” presbycusis, which presents with a mild-to-moderate S.N.H.L along with fair speech discrimination. Here, the “floor” of hearing sensitivity is elevated (usually resulting in thresholds between some 30 to 60 decibel H.L) although the “ceiling” of loudness tolerance is similar to normal (usually between 90 to 100 decibel H.L). The appropriate goal of amplification is to restore normal loudness growth, and to accomplish this goal, we need to amplify soft sounds by a lot and loud sounds by little or nothing at all. The reason for the term “W.D.R.C” is that its low knee-point and low compression ratio serve to reduce a normally large dynamic range into the smaller one associated with mild-to-moderate S.N.H.L. For example, a low compression ratio of 2:1 will compress a dynamic range of 100 decibel into one of 50 decibel.

The purpose behind W.D.R.C, was intellectually construed as an attempt to electroacoustically imitate the function of the O.H.C's of the cochlea. In Chapter 2, we discussed that the O.H.C's amplify soft sounds (approximately less than 40 to 50 decibel S.P.L) so that the I.H.C's can sense them. Use of W.D.R.C was seen then as most appropriate for mild-to-moderate S.N.H.L. Here is some food for clinical thought: It is no coincidence that oto-acoustic emissions and the knowledge of the O.H.C's, as well as the Kamp and W.D.R.C became clinically popular at around the same time, namely, the late 1980s and early 1990s.

The author once attended an American Academy of Audiology conference seminar on compression given by F. Kuk way back in 1996, who provided a very illustrative analogy regarding the very different workings of O.L.C versus W.D.R.C. O.L.C was compared to a teenager speeding down the road in a relative's car, who sees a stop sign at the end of the road, slams on the brakes, and screeches to a stop. W.D.R.C was compared to an elderly person who starts out at a normal speed but, on seeing the stop sign far ahead, ever so cautiously applies a foot gently to the brakes and slows to a stop over a long distance. They both got to the same place; they just got there in very different ways.

If restoring normal growth of loudness is the goal for mild-moderate S.N.H.L, then pertaining to the above analogy, the car to drive is W.D.R.C. Some problems would occur when trying to accomplish the same goal with O.L.C. Due to the high knee-points and high compression ratios found with O.L.C, relatively intense inputs (e.g., 70, 80, and more decibel S.P.L) will all be amplified to result in similar output intensities. These outputs would thus all sound similar and will all be perceived as being “loud.” This is not the restoration of normal loudness growth. Now look what happens with the lower knee-points and lower compression ratios provided by W.D.R.C. As shown in the right-most column of Figure 7 to 5, soft input sounds below the knee-point will be amplified by linear (maximum) gain so as to be audible. Progressively more intense inputs above the knee-point will still result in progressively more intense outputs; the outputs however, simply will not increase as fast as the inputs.

Does this mean that all those with mild-to-moderate S.N.H.L are completely satisfied with the restoration of loudness growth given by low knee-points and low compression ratios? Not really. For some clients, W.D.R.C might not be satisfactory because for aver- age speech levels of inputs, the outputs just might not sound “loud” enough. Although W.D.R.C will amplify soft input S.P.L's by a lot, it will not amplify average intensity input S.P.L's by the same amount.

When W.D.R.C first appeared, experienced clinicians were fond of pointing out the difficulties of fitting W.D.R.C hearing aids on clients who are accustomed to wearing linear hearing aids; these clients initially found that the W.D.R.C hearing aids were “not loud enough.” For soft input sounds, the W.D.R.C hearing aids were satisfactory because it is for these sounds that they provided the most gain; however, for average input sounds, W.D.R.C did not provide as much gain as their older linear hearing aids did. A common clinical report was, “I can hear people talking at tables further away better than I can her the person sitting right across from me!” A suggestion then was to raise the T.K adjustment so as to decrease the gain for soft sounds (of people sitting further away) and at the same time, increase the overall gain with the V.C. In today's digital hearing aids, accommodation for this complaint is made by additional linear gain for average speech-level inputs. More will be described about this in Chapter 8.

Now let's talk about the client with "neural" S.N.H.L. We have said before in the beginning of Chapter 2 that while most cases of mild-to-moderate S.N.H.L are caused by O.H.C damage, there are some cases where the mild-to-moderate degree of hearing loss can be caused by a combination or mixture of both O.H.C and I.H.C damage. A most common sign of this would be seen in relatively poorer speech discrimination. Usually, however, I.H.C damage tends to follow O.H.C damage, resulting in a more severe degree of S.N.H.L, and it is this population we are talking about now.

In these cases, O.L.C is a better choice than W.D.R.C. These clients have a very narrow dynamic range. Restoring normal loudness growth is just not one of the main goals here. Clients with severe S.N.H.L tend to prefer a strong, linear gain over a wide range of input S.P.L's, at least until the output S.P.L becomes close to their loudness tolerance or uncomfortable loudness levels.

High-power O.L.C gives lots of gain for soft sounds and the same “lots of gain” for average input sounds, making average conversational speech quite audible. This would be the goal and, hence, O.L.C may definitely be preferred by clients with severe-to-profound hearing loss.

Bill and Till: Two Types of Early W.D.R.C

Intuitively, it might be thought that a hearing aid with W.D.R.C would normally offer a fairly similar degree of compression across the frequencies, but this was certainly not always the case—not even in the analog era. Here we arrive at another fork in the historical development of compression; specifically, within W.D.R.C itself. Figure 7 to 7 (right panel) shows a frequency response of a fictitious W.D.R.C hearing aid, where the difference in gain for each of the various input levels is consistent across the frequency response of the hearing aid. Although most analog compression hearing aids actually had slightly different knee-points across the frequencies, for the sake of clarity here, presume that the knee-point in this W.D.R.C hearing aid is exactly the same across all the frequencies.

On these hearing aids, the low or high frequency gain could be adjusted with potentiometers or trimmers (either physical trimmers manually adjusted by a screwdriver on the hearing itself or programmable trimmers). These adjustments were often called "low-cut" or high-cut." For example, to fit someone with a sloping high-frequency hearing loss, the clinician would cut or reduce some of the low-frequency gain. Such adjustments would change the frequency response of the hearing aid, but they would not change the behavior of the W.D.R.C across the frequencies.

In other words, the difference in the amount of gain provided for different input levels would remain the same across the frequencies. For example, in Figure 7 to 7, a low-frequency gain reduction would simply be seen as an equal drop for all three lines or functions on the frequency response, across the low frequencies.

We have mentioned earlier that since W.D.R.C gives very different amounts of gain across a wide range of input intensity levels, it was sometimes called “input level dependent compression.” W.D.R.C, however, soon evolved to provide input level-dependent compression specifically at some frequencies more than at others. W.D.R.C can be applied more to bass frequencies than to high frequencies, or conversely, more to high frequencies than to low frequencies. There are various names that pertain to these categories, such as L.D.F.R (level-dependent frequency response), F.D.C (frequency-dependent compression), and A.S.P (automatic signal processing). Basically, these terms all boil down to at least one similar thing; namely, that the compression occurs more in some frequencies than in other frequencies.

Now it was (and still is) true that in any compression hearing aid, the compression knee-point is often set at somewhat different input S.P.L's for different frequencies. Hearing aid specifications in North America don't tend to show this; rather, they (and ansi measurements) tend to show the compression knee-point on I/O functions at 2000 Hz only. The reason why is because 2000 Hz is considered to be a very important frequency required for the recognition of audible speech.

While compression often occurs at somewhat different knee-points across the various frequencies, W.D.R.C was further developed whereby the compression knee-point were very different across different frequencies. Killion, Staab, and Preves (1990) gave the simplest classification of two types of frequency-dependent W.D.R.C—namely, bass increase at low levels (Bill) and treble increase at low levels (Till). Both Bill and Till rapidly emerged to become two types or subsets of W.D.R.C.

The advent of Bill took place in the mid-1980s; Till came along slightly later with the analog Kamp circuit, developed by Killion, around 1989. So Bill is a bit older than Till. Bill first appeared in a circuit known as the "Manhattan" circuit, created by a long-ago swallowed-up company called "Argosy." It was called the Manhattan circuit because in a long-ago naming contest at the American Speech-Language-Hearing Association, someone said its circuit board looked like the skyline of Manhattan. We will describe Bill first and then Till.

Bill hearing aids had a low knee-point for the low frequencies and a much higher knee-point (essentially linear gain) for the high frequencies. As with typical W.D.R.C, linear (maximum) gain was provided for very soft inputs below the knee-point, while a low (e.g., 2:1) compression ratio was applied for average and louder inputs. In the Bill hearing aid, however, this took place only for low-frequency inputs. Bill is basically W.D.R.C confined to the low frequencies.

Figure 7 to 8 shows a simple set of frequency responses for Bill (left) and Till (right). The Bill hearing aid had a very broad or flat frequency response with soft inputs (e.g., 40 decibel S.P.L). In other words, if input sound was produced so that it was at 40 decibel S.P.L all across the frequency range, then the gain and

Bill & Till Two Types of W.D.R.C

frequency response of the Bill hearing aid would look something like the top flat line on the frequency response (left panel). As the input across frequencies is increased in intensity to 60 decibel S.P.L, the frequency response would reveal a decrease in gain for the low frequencies. As the input intensity is increased to 80 decibel S.P.L, the gain for the low frequencies would drop even more.

The main idea behind the Bill circuit was to enable better listening for speech while in background noise. That is, the “hub-bub” of low-frequency background noise will be suppressed by compression, with the high-frequency sounds that render clarity for speech still receiving a full measure of gain at all input intensities.

Oticon took the concept of Bill and utilized it in the analog two-channel Multi-Focus hearing aid in 1995 (more on multi- channel hearing aids will be covered in the next section of this chapter). Later on, in 1997, Oticon again championed Bill in its first digital product, the DigiFocus. Oticon's stated purpose was to reduce the upward spread of masking, which has been discussed earlier in Chapters 1, 2, and 3. Recall that this phenomenon refers to the fact that low frequencies mask high frequencies better than highs mask lows. Bill was seen as a way to fight the upward spread of masking, so as to increase speech intelligibility in background noise.

The Till hearing aid (see Figure 7 to 8, right panel) was completely different. The original analog Till hearing aids were those that utilized the K Amp circuit. These hearing aids first appeared around 1989 and quickly became an extremely popular option offered by many of the hearing aid manufacturers of the day. Till hearing aids had an especially low knee-point for the high frequencies.

Since most mild-to-moderate S.N.H.L is worse for the high frequencies, Till hearing aids tended to provide very little gain for the low frequencies. Again, as with typical W.D.R.C, linear (maximum) gain was provided for very soft inputs below the knee-point, while a low (eg. 2:1) compression ratio was applied for average and louder inputs. In the Till hearing aid, however, this took place mainly for high-frequency inputs. Till is basically W.D.R.C confined to the high frequencies.

Figure 7 to 8 (right panel) shows that for the Till hearing aid, low-intensity input S.P.L's of 40 decibel across the frequency range would result in a gain and frequency response that had more of a high-frequency emphasis. As the input was increased to 60 decibel S.P.L, the high-frequency emphasis would decrease relative to the gain for the low frequencies. With inputs of 80 decibel S.P.L, the gain for the high frequencies would decrease even more.

The main idea behind Till was to emphasize the high-frequency sounds of speech for the listener who most typically has high-frequency hearing loss. This client will have a reduced dynamic range for the high frequencies; that is, compared to that for normal hearing, the “floor” of hearing sensitivity for the high frequencies will be elevated, although the “ceiling” of loudness tolerance will not be. Killion (1996) posits that for Till hearing aids, the frequency response for inputs of 80 decibel S.P.L or more was intended to resemble the resonance of the open, unaided

Chapter 3: Applications of Compression

We have already determined that compression provides non-linear amplification. That is, the gain decreases as the input level increases. But, by how much should the input signal be compressed? At what input level?

And, how quickly? The answers to these questions depend upon the overall goal of the hearing aid fitting. Compression may be used to:

1. Limit the output of the hearing aid without distortion, 2. Minimize loudness discomfort, 3. Prevent further damage to the auditory system, 4. Optimize the use of the residual dynamic range, 5. Restore normal loudness perception, 6. Maintain listening comfort, 7. Maximize speech recognition ability, and 8. Reduce the adverse effects of noise.

Compression circuits are defined by their features – threshold kneepoint, compression ratio, and attack and release times. At the present time, research offers no compelling reasons for setting these parameters a certain way. Thus, each must be adjusted to achieve a desired goal. Bear in mind that, to improve usability, manufacturers may limit the adjustable parameters.

Figure 7-8 summary: This figure consists of two line charts illustrating frequency responses for two different conditions, labeled as BILL and TILL. Each chart plots gain against frequency for three different input intensities. In the BILL chart, the lines for different intensities converge at higher frequencies but are widely separated at lower frequencies. In the TILL chart, the lines originate from a single point at low frequencies and diverge as frequency increases. For the BILL condition, lower input intensities result in higher gain specifically at low frequencies, indicating that wide dynamic range compression is most active in the low-frequency range. Conversely, for the TILL condition, lower input intensities lead to higher gain at higher frequencies, demonstrating that wide dynamic range compression is concentrated in the high-frequency range.

Avoiding Distortion, Discomfort and Damage

Distortion, discomfort and damage all have the same foundation – intense sounds. Intense sounds force a hearing aid into saturation causing distortion. Intense sounds may be amplified beyond the individual's L.D.L's causing discomfort.

Finally, if left unchecked, intense sounds entering a hearing aid may cause amplification-induced hearing loss. While the latter two problems can be overcome simply by limiting the maximum output by other means, compression is the only way to prevent distortion.

This type of application is referred to as compression limiting. Figure 3 to 1 shows a sample I/O function of a hearing aid with compression limiting. The following are some desirable characteristics of a circuit designed for this purpose.

Figure 3-1 summary: This figure is a line chart. It illustrates the relationship between the input and output sound pressure levels for a hearing aid equipped with output compression limiting, featuring a specific threshold point labeled as TK. The chart shows a linear increase in output relative to the input up to a certain point, after which the rate of output increase slows down significantly. It can be inferred that the hearing aid provides a one-to-one gain for lower input levels, but once the input exceeds the threshold, a compression ratio is applied to limit the output, preventing excessively loud sounds from being delivered to the user.

• A.G.C-O compression is used so that volume control adjustments do not affect the maximum output of the hearing aid. For example, consider a situation where the individual increases the volume control of the hearing aid to listen to a child's relatively soft voice and a door is slammed shut. If an A.G.C-I circuit is used, the output of the device may exceed the L.D.L. In an A.G.C-O circuit, the maximum output could be set just below the L.D.L and, more importantly, volume control adjustments would have no impact on the maximum output of the hearing aid. Thus, this application of compression is often called output compression limiting (O.C.L).

• The T.K is set high. Because intense sounds are of primary concern, T.K's of 70 decibels S.P.L or greater are typically used. A high C.R, greater than 8:1, is used to prevent the amplified sound from exceeding the L.D.L.

• A fast A.T minimizes the overshoot associated with a rapid increase in input level. Thus, A.T's of 10 ms or less are generally used to limit the duration for which the output of the hearing aid exceeds the L.D.L.

- The R.T is a less critical element because it is associated with decreasing input and output levels. That said, however, consider the situation where a pot dropping on a tile floor during a conversation forces a hearing aid into compression limiting, resulting in a severe reduction in gain. If gain releases from compression limiting slowly, the speech that follows may be inaudible. Thus, an R.T of 100 ms or shorter is preferred; a circuit with an adaptive release time may also be used.

• Finally, single-or multi-channel compression is suitable for this application.

Figure 3 to 2 shows frequency response curves for devices with and without O.C.L (Hearing Aids A and B, respectively). Both hearing aids have identical responses for inputs of 50 decibels S.P.L. However, at 90 decibels S.P.L, the output of Hearing Aid B is greater than that of Hearing Aid A. This is an important consideration because the output of Hearing Aid B could exceed the individual's L.D.L and, if saturation occurs, the associated distortion will be high.

Figure 3-2 summary: This figure is a line chart. It displays the output levels of two different hearing aid configurations, one with an open vent and one without, across a range of frequencies for two different input sound pressure levels. The chart compares the output response for a louder input signal and a softer input signal for both devices. The data indicates that the hearing aid without the open vent generally produces a higher output across most frequencies compared to the one with the open vent. Additionally, the difference in output between the loud and soft input signals is smaller than the difference between the input levels themselves, demonstrating the effect of wide dynamic range compression. The output for both devices drops off significantly at very high frequencies.

Optimizing Use of the Residual Dynamic Range and Restoring Normal Loudness Perception

As indicated previously, hearing impairment results in a loss of sensitivity for weak sounds, with little or no loss of sensitivity for intense sounds. Thus, in order for the range of environmental sounds to fit within the residual dynamic range of the individual, more amplification is required for weak sounds than for intense sounds. The net result is that weak sounds are audible, moderate sounds comfortable, and intense sounds are perceived as loud without causing discomfort.

This type of application is referred to as wide dynamic range compression (W.D.R.C). Figure 3 to 3 shows a sample I/O function of a hearing aid with W.D.R.C. The following are some desirable characteristics of a circuit designed for this purpose.

Figure 3-3 summary: This figure is a line chart. It illustrates the input/output function of a hearing aid utilizing wide dynamic range compression, plotting the output sound pressure level against the input sound pressure level. The chart shows a linear relationship at lower input levels, which then transitions to a shallower slope after reaching a specific threshold. This indicates that for low-intensity sounds, the output increases proportionally with the input, while for higher-intensity sounds, the device applies compression to limit the increase in output, thereby maintaining the sound within a comfortable listening range for the user.

• A.G.C-I is used because the goal is to achieve a certain degree of audibility and/or loudness for incoming signals. If average, rather than individual, data are used to derive the amount of amplification necessary, the option may exist to increase or decrease the output of the hearing aid by adjusting the volume control.

- The T.K is as low as possible in order to make weak sounds audible. Thus, the T.K is typically set at or below 50 decibels S.P.L. Hearing aids with T.K's as low as 20 decibels S.P.L are said to use full dynamic range compression F.D.R.C because they aim to compress the entire gamut of environmental sounds in the residual dynamic range of the individual.

- Low C.R's of 4:1 or less can be used because compression acts over a wider range of inputs. Mueller (2002) gives the example that W.D.R.C is akin to gentle braking as one approaches a stop sign while driving. [By the same token, compression limiting is more like screeching to halt at the last minute!]

- The A.T and R.T may be fast or slow. However, as indicated earlier, several studies have shown that fast A.T's and R.T's significantly degrade sound quality. Thus, the use of A.T's slower than about 100 ms and R.T's slower than 2 s is advisable.

• Multichannel compression is used to accommodate audiometric configurations that deviate substantially from flat. Large variations in the degree of hearing loss across the frequency range also significantly change the perception of loudness.

As shown in Figure 3 to 4, the I/O function of a hearing aid with W.D.R.C (Hearing Aid A) is visibly different from that of a linear hearing aid with O.C.L (Hearing Aid B). In addition to the T.K and C.R being lower, Hearing Aid A also provides more gain than Hearing Aid B at and below the T.K. This additional gain results in improved audibility of weak sounds.

The effects of hearing impairment on loudness perception and compensation through the use of wide dynamic range compression will be discussed at some length in the next chapter in the context of fitting hearing aids with compression. For now, consider the loudness growth functions shown in Figure 3 to 5 (page 23). A loudness growth function is a graph showing the perceived loudness of a sound as a function of input level. As expected, the loudness rating increases with input level.

However, the shapes of the loudness growth functions are significantly different for persons with normal hearing and those with sensorineural hearing loss. In addition, note that the "loss of loudness" (as measured by the difference between the two curves) is greater for sounds rated as being soft, with little or no loss for sounds rated as loud. Linear amplification with O.C.L provides a fixed amount of gain regardless of the level of the incoming signal, making all sounds louder by the same amount. Thus, the shape of the loudness growth function remains the same as that for sensorineural hearing loss, but is shifted to the left.

The result is that weak sounds may still be inaudible, moderate sounds are comfortable, and intense sounds are too loud. In contrast, when W.D.R.C is used, the resulting loudness growth curve is very similar to that of a person with normal hearing. In other words, weak sounds are now audible, moderate sounds comfortable, and intense sounds are perceived as loud. Note that experienced users of linear amplification may object to the use of W.D.R.C because they have become accustomed to intense sounds being too loud and weak sounds being inaudible.

Figure 3-4 summary: This figure is a line chart. It displays the relationship between input and output sound pressure levels for two different hearing aid configurations, illustrating how each processes sound through various compression ratios. One line represents a linear hearing aid with output compression limiting, while the other represents a hearing aid with wide dynamic range compression. The chart identifies specific transition points where the compression ratios change. The wide dynamic range compression aid shows an initial linear phase followed by a period of moderate compression before reaching a higher compression ratio at peak levels. In contrast, the linear aid maintains a constant ratio until it reaches its limiting threshold. Consequently, the wide dynamic range compression aid provides a higher output for lower input levels compared to the linear aid, but both converge at the highest input levels.

Figure 3-5 summary: This figure is a line graph. It illustrates the relationship between input sound levels and perceived loudness ratings for individuals with normal hearing compared to those with sensorineural hearing loss, as well as the effects of two different amplification strategies. The graph tracks how loudness perceptions shift from very soft to uncomfortably loud across a range of decibel levels for four distinct conditions: normal hearing, sensorineural hearing loss, linear amplification with output clipping, and wide dynamic range compression. The data indicates that individuals with sensorineural hearing loss require significantly higher input levels to perceive sounds as audible compared to those with normal hearing. While linear amplification helps increase audibility, wide dynamic range compression more effectively aligns the loudness perception of impaired hearing with that of normal hearing across a broader range of input levels, particularly for softer sounds, while preventing sounds from becoming excessively loud.

Maintaining Listening Comfort

Even when the maximum output of a hearing aid is appropriately restricted to within the residual dynamic range, it may still be desirable to further reduce the output of the hearing aid to a level below the maximum output (or L.D.L) for much of the time. There are several ways in which this problem may be addressed that do not require the user to constantly adjust the volume control.

First, as discussed in the preceding section, W.D.R.C employs a low T.K and C.R. This results in intense sounds approaching the maximum output of the hearing aid more gradually than linear amplification coupled with O.C.L (Figure 3 to 5). The characteristics of a circuit designed for this purpose are much the same as those described in the previous section for optimizing use of the residual dynamic range and restoring loudness perception. That is, multichannel A.G.C-I compression should be used with low T.K and C.R, and long A.T's and R.T's.

The second method of maintaining listening comfort has no official name, but has been called mid-level or comfort-controlled compression. Figure 3 to 6 shows a sample I/O function of comfort-controlled compression. The following are some desirable characteristics of a circuit designed for this purpose.

• A.G.C-I compression is used because the goal is to achieve the desired loudness for moderate to intense inputs.

- The T.K is set at approximately 60 decibels S.P.L so that compression can operate on moderate and intense sounds.

- Low C.R's of 4:1 or less can be used because compression acts over a wider range of inputs than for O.C.L.

- A.T's and R.T's may be fast or slow, but should probably be slow to preserve sound quality. Thus, the use of A.T's slower than about 100 ms and R.T's slower than 2 s is advisable.

- Single-or multi-channel compression is suitable for this purpose.

Figure 3 to 7 compares mid-level compression (Hearing Aid A), W.D.R.C (Hearing Aid B) and linear amplification with O.C.L (Hearing Aid C). Note that below the kneepoint, gain for mid-level compression is similar to that for linear amplification with O.C.L; in contrast, gain is greater for the hearing aid with W.D.R.C. Although it does not increase speech audibility, this alleviates the primary disadvantage of W.D.R.C – increased likelihood of feedback resulting from the provision of more gain. Above the T.K, mid-level compression takes advantage of reduced gain for intense sounds when compared to linear amplification with O.C.L. As a result, like W.D.R.C, this assists in maintaining listening comfort.

Figure 3-6 summary: This figure is a line chart. It illustrates the input-output relationship for a hearing aid utilizing mid-level compression, plotting the output sound pressure level against the input sound pressure level. The graph shows a linear relationship at lower input levels, which then transitions to a shallower slope after reaching a specific threshold known as the TK point. It can be inferred that for sounds below the threshold, the device provides a constant gain, while for sounds exceeding the threshold, the device applies compression to reduce the output gain, thereby limiting the loudness of higher-intensity sounds.

Figure 3-7 summary: This figure is a line chart. It illustrates the relationship between input and output sound pressure levels for two different hearing aid compression strategies, identified as Hearing Aid A and Hearing Aid B. The chart displays how the output level changes relative to the input level, highlighting different compression ratios across various input ranges. The data shows that Hearing Aid B provides higher output at lower input levels compared to Hearing Aid A and exhibits a change in compression ratio at a lower input threshold. In contrast, Hearing Aid A maintains a linear relationship for a longer duration before transitioning to a compressed state. Ultimately, both devices converge toward a similar maximum output level as the input intensity increases, indicating that both employ compression to limit the maximum output.

Maximizing Speech Intelligibility

Arguably, the single most important criterion for maximizing speech intelligibility is increased audibility. That is, speech must be audible before one can be expected to understand it.

Treble-Increase-at-Low-Levels (Till) was introduced as a means to improve the intelligibility of speech in single-channel devices. As the name suggests, this circuit provides a high-frequency boost for weak sounds. When the overall input level is relatively low, high-frequency consonants are amplified more than the low-frequency vowels.

This additional high-frequency gain not only makes the consonants more audible, but also reduces the upward spread of masking by the more intense vowels. Although high-frequency emphasis is potentially useful at all input levels, it also increases the likelihood of loudness discomfort at high input levels. Thus, for intense sounds, the frequency response curve is relatively flat. Figure 3 to 8 shows frequency-gain curves for a Till circuit at inputs of 50, 70 and 90 decibels S.P.L. Note that, for an input of 50 decibels S.P.L, the high-frequency emphasis is in addition to the gain that would ordinarily be prescribed for the hearing loss.

Further, the shape of the curve changes with the input level – that is, it gets flatter as the input level increases. Improved customization of the frequency response through the use of multiple channels has resulted in the demise of single-channel Till processing.

Today, W.D.R.C is the primary means for maximizing speech intelligibility. The following are some desirable characteristics of a circuit designed for this purpose.

• A.G.C-I compression is used because amount of gain applied depends on the level of the incoming sound.

- The T.K is as low as possible, at or below 50 decibels S.P.L, in order to make the weaker components of speech audible.

- C.R's of 4:1 or less can be used because compression acts over a wider range of inputs.

- A.T's and R.T's should be faster than the duration of a typical syllable to provide more amplification for the weaker components than for the more intense components of speech.

• Multichannel compression is used so that weak consonant sounds can be amplified independently of the more intense vowel sounds. Increasing the consonant-to-vowel ratio C.V.R – that is, the intensity of a consonant relative to that of a vowel – has been shown to improve speech understanding.

The advantage of W.D.R.C over linear amplification with O.C.L, from the point of view of improving speech audibility, is shown in Figure 3 to 9. Consider two hearing aids that are set such that conversational speech is comfortable for the person with a sensorineural hearing loss. We have already seen that, for intense sounds, the output of the hearing aid approaches L.D.L more slowly with W.D.R.C than with linear amplification and O.C.L. This is indicated in the Figure by the region of increased comfort. More importantly, weak sounds are made audible by W.D.R.C (indicated by the region of improved audibility); the same cannot be said of the linear hearing aid with O.C.L. Keep in mind also that only average levels are shown in Figure 3 to 9. Speech has a dynamic range of 30 decibels. As a result, there may be components that are even weaker with linear amplification and O.C.L than shown in the figure.

Figure 3-8 summary: This figure is a line chart. It displays the gain across a range of frequencies for three different input levels, illustrating the phenomenon of treble-increase-at-low-levels. The chart shows that as the input level decreases, the gain increases across the frequency spectrum, with a pronounced peak occurring in the higher frequency range. From this data, it can be inferred that the system exhibits a non-linear frequency response where lower input levels result in significantly higher amplification, particularly in the treble frequencies, compared to higher input levels.

Figure 3-9 summary: This figure is a line chart. It displays the output levels across a range of frequencies for two different types of hearing aids, one utilizing wide dynamic range compression and another utilizing linear amplification with output compression limiting, each tested at several different input levels. The data indicates that the linear hearing aid produces a wider variation in output as input levels increase, whereas the hearing aid with wide dynamic range compression maintains a more consistent output across different input levels. Consequently, the wide dynamic range compression model provides a broader region of increased comfort for loud sounds and a broader region of improved audibility for soft sounds compared to the linear model.

Reducing the Adverse Effects of Noise

A common complaint regarding hearing aids relates to the effects of noise. Although noise management schemes are beyond the scope of this handbook, the effects of compression will be discussed. The problem with noise is twofold. First, amplification of intense sounds, such as a vacuum cleaner, may cause the listener distress even if the L.D.L is not exceeded. And, second, communication is challenging in noisy environments, such as at a party or restaurant.

Hearing aids with Automatic Signal Processing (A.S.P) were introduced in the 1990s to counter the deleterious effect of background noise. The principle is based on two assumptions: (1) the overall level of sound is relatively high in noisy environments, and (2) the hubbub of parties and/ or restaurants is dominated by energy in the low frequencies. Thus, A.S.P hearing aids reduce the amount of low-frequency amplification for high input levels to alleviate the loudness of the noise. Low frequency amplification is restored at low input levels. One can think of this as Bass-Increase-at-Low-Levels (Bill), which is another common name for A.S.P circuits. A further advantage of reduced low-frequency amplification at high input levels is the decreased likelihood of upward spread of masking. Figure 3 to 10 shows the frequency-gain curves of Bill circuits for speech presented at 50, 70 and 90 decibels S.P.L. Like Till circuits, Bill processing is seldom used nowadays because of the widespread use of multichannel hearing aids to achieve the same objective.

The compression of sound in multiple channels reduces the adverse effects of noise in several ways that provide an advantage over single-channel Bill processing. First, no assumptions are made regarding the frequency composition of the noise. Thus, gain is reduced in any frequency region where the input levels are high. Second, gain is reduced only in frequency regions where a great deal of noise is present; gain and audibility in the remaining channels are unaffected. Third, when the spectra of the signal and noise are different, multichannel compression may provide a slight improvement in the overall signal-to-noise ratio.

(S.N.R) when the outputs of channels with poor S.N.R are reduced relative to the outputs of those where the S.N.R is good. It is important to note that the S.N.R is not improved within any given channel.

Although it is convenient to compartmentalize the applications of compression, more than one goal can be achieved in a single fitting. For example, multichannel W.D.R.C may be used to optimize use of the residual dynamic range, normalize the perception of loudness, maintain listening comfort, maximize the intelligibility of speech, and reduce the adverse effects of noise. Figure 3 to 11 summarizes the compression characteristics associated with each application.

most adults with moderate sensorineural hearing loss. The quality of fitting — not channel count alone — is the primary determinant of outcome.

Figure 3-10 summary: This figure is a line chart. It displays the frequency response curves for different input levels, plotting gain against frequency. The chart illustrates how the gain varies across a range of frequencies for three distinct sound pressure levels. The data indicates that as the input level increases, the overall gain decreases across most of the frequency spectrum. Specifically, the lower input levels exhibit a more pronounced increase in gain at lower frequencies compared to higher input levels, demonstrating the bass-increase-at-low-levels phenomenon.

Figure 3-11 summary: This table outlines the optimal compression settings based on the intended application. To prevent distortion and damage, output control is used with high thresholds and compression ratios. Conversely, optimizing residual dynamic range and maximizing speech intelligibility both utilize input control with low thresholds and compression ratios. For maintaining listening comfort, input control is employed, with threshold and compression ratio levels varying depending on whether wide dynamic range compression or mid-level compression is used.

4. Bill — Bass Increase at Low Levels

Bass Increase at Low Levels

4.1 Definition and Concept

Bill describes a hearing aid processing strategy in which the amplification of low-frequency (bass) sounds is selectively enhanced when the overall input level is low. As input level rises, the low-frequency gain decreases relative to high-frequency gain. In other words, the frequency response of the hearing aid tilts — providing more bass at quiet input levels and rolling off the bass as the environment becomes louder.

Formal Definition: Bill: A frequency-response characteristic in which gain in the low-frequency region is inversely related to input level. Maximum low-frequency gain is applied to soft sounds; this gain decreases as input level increases, resulting in a flatter or high-frequency-tilted response at high input levels.

4.2 Physiological and Psychoacoustic Rationale

The Bill behaviour is grounded in several auditory phenomena:

1. Low-frequency upward spread of masking: In quiet environments, soft low-frequency room noise (fans, hvac systems, traffic rumble) can mask higher-frequency speech cues through upward spread of masking. By boosting low frequencies at low levels, Bill ensures that vowel energy and prosodic cues are audible when ambient noise is minimal.

2. Equal-loudness contours: At low intensities, the human auditory system is less sensitive to low-frequency sounds relative to mid-frequencies (as described by the Fletcher-Munson curves). Bill compensates for this by providing additional gain in the bass region at low listening levels.

3. Reverberation management: In echoic environments, low-frequency energy accumulates from reflections. A Bill response helps reduce the masking effect of reverberant energy on speech formants as the listening environment becomes louder.

4. Natural hearing: In normal-hearing listeners, the acoustic reflex reduces low-frequency sensitivity in loud environments. Bill mimics this natural protective and perceptual mechanism in hearing aid processing.

4.3 Technical Implementation

In digital hearing aids, Bill is typically implemented through one or more of the following mechanisms:

- Low-frequency channel compression: The compression threshold in the low-frequency channels (e.g., 250 to 1000 Hz) is set lower than in the high-frequency channels, causing gain reduction to engage at lower input levels in the bass region.

- Automatic low-frequency rolloff: An adaptive filter continuously monitors overall input level and applies a graduated high-pass filter characteristic as S.P.L increases, effectively rolling off bass gain.

• Input-controlled bass attenuation: Input-level detectors gate a frequency-shaping network that attenuates low frequencies proportionally to input level increase.

- Frequency-specific compression ratios: Higher compression ratios are programmed in the low-frequency channels relative to the high-frequency channels, producing the Bill effect automatically through standard W.D.R.C operation.

4.4 Clinical Applications of Bill

Bill processing is particularly beneficial for:

• Users with predominantly low-frequency hearing loss (rising audiogram configurations), who need low-frequency amplification for soft sounds but risk low-frequency masking of high-frequency cues at louder levels.

• Users in varied listening environments — moving between quiet (where Bill boosts bass for richness of sound) and noisy settings (where the hearing aid automatically reduces low-frequency amplification to reduce noise masking).

• Listeners who report that voices sound 'boomy' or 'muffled' in noise — a perceptual correlate of excessive low-frequency gain in louder environments.

• Hearing aid users with low-frequency residual hearing who benefit from access to low-frequency speech cues in quiet but experience upward masking in noise.

4.5 Bill and Background Noise Management

One of the most clinically significant functions of Bill is its role in automatic adaptation to background noise. When ambient noise levels increase (for example, entering a restaurant or cafeteria), the hearing aid's input-level detectors register a higher overall S.P.L. The Bill algorithm responds by reducing low-frequency gain, which:

5. Reduces the amplification of low-frequency noise (which is predominant in most background noise spectra).

6. Improves the signal-to-noise ratio (S.N.R) for speech intelligibility by relatively emphasising higher frequency consonant cues.

7. Reduces the perception of 'noise' and improves listening comfort in challenging acoustic environments.

It is important to note that Bill is not a noise reduction algorithm per say — it does not distinguish between speech and noise. Rather, it is a level-dependent frequency shaping strategy that achieves S.N.R improvement through spectral shaping.

5. Till — Treble Increase at Low Levels

Treble Increase at Low Levels

5.1 Definition and Concept

Till describes a hearing aid processing strategy in which the amplification of high-frequency (treble) sounds is selectively enhanced when the overall input level is low. Unlike Bill, where bass gain is boosted at soft levels, Till provides maximum treble gain for soft speech and reduces this treble emphasis as input level increases. The resulting frequency response presents more high-frequency emphasis at low input levels, transitioning toward a flatter response at higher input levels.

Formal Definition: Till: A frequency-response characteristic in which gain in the high-frequency region is inversely related to input level. Maximum high-frequency gain is applied to soft sounds; this gain decreases as input level increases, resulting in a flatter or low-frequency-tilted response at high input levels.

5.2 Physiological and Psychoacoustic Rationale

Till is rooted in the audiological reality of sensorineural hearing loss:

8. High-frequency audiogram configuration: The vast majority of sensorineural hearing loss presents with greater high-frequency loss. Soft speech — conversational speech at distance or quiet environments — contains high-frequency consonant cues that fall below audibility without amplification. Till specifically targets this deficit.

9. Consonant audibility: Voiceless fricatives (/s/, /f/, /sh/, /th/) and affricates (/ch/, /j/) carry the primary phonemic contrast information in English and many other languages. These sounds have very low intensity relative to vowels and are concentrated at 2000 to 8000 Hz. Till provides maximum gain for these critical but weak sounds.

10. Whittle's articulation index: The articulation index A.I demonstrates that high-frequency speech bands contribute disproportionately to speech intelligibility. The 1000 to 4000 Hz region accounts for approximately 80% of speech intelligibility, making treble amplification the most clinically impactful frequency region for most hearing aid users.

11. Loudness growth asymmetry: Sensorineural hearing loss often creates different loudness recruitment patterns at different frequencies. High frequencies may exhibit steeper loudness recruitment, meaning that loudness catches up with normal rapidly once sounds are audible. Till's level-dependent reduction of treble gain as input increases is consistent with this recruitment pattern, maintaining comfortable loudness.

5.3 Technical Implementation

Till processing is implemented in digital hearing aids through:

- High-frequency channel W.D.R.C: Compression thresholds in the high-frequency channels (e.g., 2000 to 8000 Hz) are set lower than in mid-frequency channels, ensuring that compression engages early in the high frequencies, producing gain reduction as level increases.

- Adaptive high-frequency emphasis filters: A frequency-shaping algorithm continuously adjusts a high-shelf filter based on the detected input level, boosting treble at low levels and attenuating it at high levels.

- Frequency-specific attack and release tuning: High-frequency channels may employ faster attack times to respond rapidly to loud transients, protecting the ear while preserving soft high-frequency cues.

- Soft-gain prescription targets: Fitting formulae such as N.A.L-N.L.2 and D.S.L v5 incorporate Till-like targets, prescribing significantly more gain for soft inputs (50 decibel S.P.L) in the high frequencies than for loud inputs (80 decibel S.P.L).

5.4 Clinical Applications of Till

Till is the most clinically prevalent of the three processing strategies because it directly addresses the most common complaint of hearing aid users: difficulty understanding speech, especially in quiet environments or at a distance. Till is indicated for:

• Users with high-frequency sensorineural hearing loss (the most prevalent audiometric configuration in presbycusis and noise-induced hearing loss).

• Individuals reporting difficulty understanding consonants — specifically complaining of 'hearing but not understanding' speech.

• Listeners who have difficulty with soft-spoken talkers or speech at a distance, where the already-weak high-frequency consonant energy is further attenuated by distance (following the inverse square law).

• Users who report that hearing aids make loud sounds uncomfortably sharp or tinny — a sign that treble gain at high input levels is excessive and needs to be reduced through Till-type processing.

- Children with sensorineural hearing loss fitted under D.S.L (Desired Sensation Level) targets, which apply particularly aggressive Till-type high-frequency soft-gain prescriptions.

5.5 Till and Speech Intelligibility Research

The evidence base supporting Till-type high-frequency amplification is extensive:

• Multiple studies demonstrate that audibility of high-frequency consonants is the single strongest predictor of speech intelligibility in adults with sensorineural hearing loss.

- The Articulation Index A.I and its successor, the Speech Intelligibility Index (S.I.I), quantify the contribution of each frequency band to overall speech intelligibility, consistently showing the primacy of the 2000 to 4000 Hz region.

• Research by Humes (1991) and subsequent investigators confirmed that high-frequency amplification significantly improves consonant recognition scores on standardized speech perception tests.

• The development of N.A.L-N.L.1 and N.A.L-N.L.2 (National Acoustic Laboratories) fitting targets is grounded in maximizing the S.I.I across input levels, producing implicitly Till-like gain prescriptions.

6. Pill — Programmable Increase at Low Levels

Programmable Increase at Low Levels

6.1 Definition and Concept

Pill represents the most sophisticated and flexible of the three processing strategies. While Bill prescribes low-frequency emphasis at low levels and Till prescribes high-frequency emphasis at low levels, Pill is a programmable strategy that allows the audiologist to configure the frequency region(s) that receive enhanced amplification at low input levels. Pill encompasses both Bill and Till as special cases within a generalized level-dependent frequency-shaping framework.

Formal Definition: Pill: A programmable, frequency-specific, level-dependent amplification strategy in which the hearing care professional can prescribe the frequency region, the degree, and the input-level range over which enhanced gain is applied, tailored to the individual user's audiometric profile and hearing needs.

6.2 Pill as a Generalized Framework

Pill can be understood as a superset of Bill and Till:

• Pill configured for low-frequency enhancement = Bill

• Pill configured for high-frequency enhancement = Till

• Pill configured for mid-frequency enhancement = a novel strategy (sometimes called M.I.L.L — Mid Increase at Low Levels — in academic literature)

• Pill configured for multi-region enhancement = custom audiogram-specific compensation

The programmable nature of Pill reflects the capabilities of modern digital signal processing platforms. Rather than hardwiring a specific tilt (bass-up or treble-up), the D.S.P platform exposes a parameter space that the fitting software can configure optimally for each user.

6.3 Technical Implementation

Pill processing in digital hearing aids is implemented through the following mechanisms:

- Multi-channel compression with independent per-channel programming: Each frequency channel (or group of channels) has individually programmable compression threshold, ratio, attack time, and release time. The audiologist selects the channels and parameters to achieve the desired Pill characteristic.

- Gain versus input level (I/O) function programming: Modern fitting software presents input-output functions for each channel. The audiologist can specify gain at 50 decibel S.P.L input (soft sounds), 65 decibel S.P.L (moderate), and 80 decibel S.P.L (loud), independently for each frequency band. This directly programs the Pill behavior.

• Prescription formula selection: Fitting formulae (N.A.L-N.L.2, D.S.L v5, Camfit, etcetera) prescribe different I/O targets for each frequency band and input level. Selecting a formula essentially chooses a Pill configuration optimized for that prescription philosophy.

- User adjustment programs: Hearing aids may offer multiple user-selectable programs, each with different Pill configurations — for example, a 'music' program with Bill-like bass enhancement and a 'speech in noise' program with Till-like treble emphasis.

6.4 The Programmability Advantage

The clinical power of Pill lies in individualisation. Consider the following audiometric profiles and how Pill adapts to each:

Table summary: The table outlines how PILL configurations are tailored to different audiogram profiles, matching specific gain distributions and compression strategies to the frequency of hearing loss to optimize speech audibility and manage loudness.

7. Comparative Analysis: Bill versus Till versus Pill

7.1 Side-by-Side Characteristics

Table summary: The table compares three hearing aid strategies, showing that BILL focuses on enhancing low-frequency sounds for bass-heavy needs, TILL focuses on high-frequency enhancement for speech clarity, and PILL provides a fully programmable and flexible approach that can be customized to any individual audiometric profile.

7.2 Interaction Between Strategies

In practice, modern hearing aids do not implement only one of these strategies in isolation. The multi-channel W.D.R.C architecture allows simultaneous Bill characteristics in the low-frequency channels and Till characteristics in the high-frequency channels. This composite behaviour — sometimes called Bill+Till — approximates the optimal frequency-specific amplification targets prescribed by evidence-based fitting formulae:

- Low-frequency channels: Lower compression threshold, higher compression ratio to Bill behaviour (bass boosted at low levels, reduced at high levels)

• High-frequency channels: Lower compression threshold, moderate compression ratio to Till behaviour (treble boosted at low levels, reduced at high levels)

- Mid-frequency channels: Intermediate compression parameters leads to Balanced gain across input levels

The Pill framework, by allowing independent channel programming, enables this composite Bill+Till response to be precisely calibrated to each individual's audiogram — achieving the audiological goals of audibility, comfort, and speech intelligibility simultaneously.

8. The Digital Signal Processing Chain

Table summary: The table outlines how various patient characteristics and environmental factors influence the selection and adaptation of hearing aid target strategies, highlighting the trade-offs between different gain approaches based on user age, experience, dexterity, listening settings, and prescription philosophies.

12.2 Counselling the User

Effective audiological management requires that users understand the rationale behind their hearing aid's processing strategy. Key counselling points:

- Explain that their hearing aid automatically adjusts how it amplifies different sounds based on loudness — it is not simply 'turned up' or 'turned down'.

- For Till users: explain that speech will sound crisper and clearer than they may be accustomed to — this brightness is intentional and represents correctly amplified consonants. Allow an acclimatisation period of 4 to 8 weeks.

- For Bill processing in noise: explain that the hearing aid reduces bass amplification in loud environments to help them understand speech better — this is why voices may sound slightly different in background noise.

- Emphasise that the first fitting is a starting point. Real-world experience and follow-up fine-tuning are essential to optimise the Pill configuration for the individual's lifestyle and preferences.

13. Summary and Key Principles

Bill — Key Principle: Bass amplification is maximised at low input levels and automatically reduced as overall sound level increases. This provides access to low-frequency speech cues in quiet while reducing low-frequency noise masking in loud environments.

Till — Key Principle: Treble amplification is maximised at low input levels and automatically reduced as overall sound level increases. This restores the audibility of soft high-frequency consonants — the primary determinant of speech intelligibility — while preventing treble discomfort at high levels.

Pill — Key Principle: A fully programmable, channel-specific, level-dependent amplification strategy that encompasses both Bill and Till as special cases. Pill allows the audiologist to tailor frequency-specific amplification to each individual's audiometric profile, prescription formula targets, and listening lifestyle.

Summary of Core Principles

22. All three strategies are fundamentally level-dependent: gain changes as a function of input S.P.L.

23. Bill and Till represent opposite ends of the frequency emphasis spectrum; Pill is the generalised framework that includes both.

24. Modern digital hearing aids implement Pill through multi-channel W.D.R.C with independently programmable compression parameters in each frequency channel.

25. Evidence-based fitting formulae (N.A.L-N.L.2, D.S.L v5) inherently encode Bill/Till/Pill targets derived from audiometric data.

26. Real-ear measurement at multiple input levels is essential to verify that prescribed Bill/Till/Pill targets are being achieved at the eardrum.

27. Advanced processing technologies — including directional microphones, digital noise reduction, machine learning classifiers, and frequency lowering — interact with and extend the Bill/Till/Pill framework.

28. Clinical decision-making must integrate audiometric data, patient complaints, listening lifestyle, and user factors to select and optimise the appropriate processing strategy.

The Bill/Till/Pill framework provides audiologists and hearing instrument specialists with a conceptually clear and clinically actionable vocabulary for understanding, prescribing, and explaining frequency-specific, level-dependent hearing aid amplification — the foundational technology underlying effective hearing rehabilitation. ear, thus providing an acoustic “transparency” for intense inputs. Of course, with the smaller high frequency dynamic range, for intense inputs, no amplification is needed at all.

Programmable and Multichannel Hearing Aids

Bill and Till were very popular in the mid-1990s, at the end of the analog hearing aid era. In those days, “high-end” analog hearing aids were “multichannel,” often having two separate channels. Bill would be found in the low-frequency channel, while Till would be found in the high-frequency channel.

With this advance in analog, two-channel circuitry, it became easy for clinicians to see that one hearing aid circuit could serve clients with many varying degrees of hearing loss, as well as many different hearing loss configurations. Instead of adhering to the differing philosophies behind Bill and Till, fitting flexibility became the “name of the game.” Due to the incredible number of channels in today's digital hearing aids, Bill and Till as separate processing schemes have largely disappeared from the scene.

In the late 1980s, prior to the emergence of multichannel technology, the first analog “programmable” hearing aids had become available. These were all single-channel hearing aids. It was later on, in the mid-1990s, that the first “multichannel” hearing aids became available.

At first these were not programmable; within about a year, however, they became programmable. The analog era ended with the emergence of “programmable, multichannel” hearing aids. In the following two sections, the concepts of programmable and multichannel will be described.

The pace of hearing aid technology development that quickened in the early 1990s continues at that same pace to this day. The major competition among hearing aid manufacturers also continues. Digital hearing aids that incorporate the use of many channels and Bluetooth programmability are being developed and released by the various hearing aid manufacturers at least as fast as this book is being written. We cannot, therefore, intelligently cover an up-to-date competitive comparison and contrast of features as found in specific models of available hearing aids today.

In the interest of following the historical development of compression however, Unitron's Sound F.X is highlighted here as an example of the high-end analog products that were found at the end of the analog era, just before digital hearing aids emerged on to the scene. The years would be 1995 to 1996. The Sound F.X was actually the first multichannel hearing aid. It utilized Bill in its low-frequency channel and Till in its high-frequency channel.

It and other hearing aids like it that immediately followed, soon acquired the added feature of being "digitally programmable." The description and discussion here on the features of program-mability and multichannel should set the scene for the ensuing discussion on digital hearing aid features in Chapter 8.

Programmable Hearing Aids

First, the terms “programmable” and “digital” should not be confused. Many high-end analog hearing aids of the late 1980s and mid 1990s were “digitally programmable,” but this did not mean that the circuit of the hearing itself truly provided digital signal processing. Hearing aids are truly digital if they process sound signals by the use of a complex series of binary mathematical sequences (i.e., a series of 0s and 1s). The concept of “digital,” along with the typical features of digital hearing aids, will be specifically covered in Chapter 8.

The term “digitally programmable” simply meant that instead of turning tiny screwdriver potentiometers to set the gain, M.P.O, compression adjustments, and so on, the clinician could adjust and set (program) the hearing aid settings with either a handheld programming unit or a computer using fitting software. Programmable hearing aids thus provide essentially the same compression characteristics as nonprogrammable hearing aids; the only difference is in the way that the controls were accessed. The term “digital screwdriver” became popular. In truth, the only thing truly digital about these analog “digitally programmable” hearing aids was the computer used to program them.

The first programmable hearing aids had one program, a stored frequency response that could be adjusted digitally instead of a manual turning of tiny screws or potentiometers.

Later on, programmable hearing aids offered some four to eight different programs, where each program was available for clinicians to set as a separate frequency response. Clients could access these stored programs or frequency responses with the flick of a switch, located on either the actual hearing aids or a handheld remote-control device. By doing so, clients could personally adjust the hearing aids for optimal listening in different listening environments (Figure 7 to 9). It soon became evident however, that four to eight channels were far too cumbersome for the average client to deal with, so manufacturers began to drop the numbers of available programs to two or three.

Three Programmable Frequency Responses

Figure 7 to 9 shows three different stored frequency responses, one for listening in quiet, another for listening in noise. The program for listening in quiet might be set to provide the necessary gain that comes closest to meeting the target(s) of a fitting method. The program for listening in noise tends to decrease the audibility of low-frequency background noise and at the same time boost the audibility of soft, high-frequency consonants. The amount of bass reduction and treble boost are often quite arbitrarily chosen, but are usually somewhere between five to 10 decibel.

Different programs were not limited to changes in frequency responses. They could also permit the client a choice between using directional versus omnidirectional microphone characteristics (these will be described later in Chapter 9). For example, in addition to being programmed to approximate the fitting method target, the listening in quiet program could also employ omni-directionality. The listening in noise program could utilize directionality in addition to providing less low-frequency gain and more high-frequency gain.

At times, a “listening to music” program can also be selected. Here, the frequency response will tend to show an increase in low-frequency gain as compared to the other programs. This is because compared to the high frequencies that render intelligibility to speech (such as the unvoiced consonants), music is generally lower in its frequency range. For example, low C is about 125 Hz, middle C is about 250 Hz, while high C is about 500 Hz.

Figure 7-9 summary: This figure is a line chart. It displays the frequency response, measured as gain in decibels, across a range of frequencies for three different programmable hearing aid settings. The chart compares how each program modifies sound amplification across the frequency spectrum. Based on the data, Program 1 provides a balanced gain profile intended for quiet environments. Program 2 exhibits significantly lower gain at low frequencies and higher gain at high frequencies, which is designed to emphasize consonants and reduce background noise for better speech perception in noisy settings. Program 3 provides the highest level of gain for low frequencies, making it suitable for listening to music. Overall, the figure demonstrates how programmable settings can be tailored to optimize hearing performance based on the specific acoustic environment.

Multichannel Hearing Aids

The left panel of Figure 7 to 10 shows a schematic for an early programmable hearing aid with two available programs. It has only one channel. Listeners could toggle between the two programs at will, depending on the listening environment.

The right panel shows a schematic for a two-channel hearing aid. Recall from our earlier discussion that analog W.D.R.C hearing aids used input compression. This can be determined from the position of the V.C, which is almost dead last in position in the two-channel circuit shown on the right panel of Figure 7 to 10. Just past the microphone, a band splitter separates the incoming input sound into two frequency bands, or channels. Each

Programmable versus Multi-Channel

frequency band or channel represents the low and high frequencies, respectively. Each separate channel has its own amplifier and compressor. On these analog two-channel hearing aids, both Bill and Till were commonly combined into one hearing aid, where Bill would be present in the low-frequency channel and Till would be present in the high-frequency channel.

Unitron's Sound F.X was built upon the popular DynamEQ2, a two-channel W.D.R.C circuit designed by Gennum in Canada in 1995. Only one analog multichannel hearing aid had three channels, and that was the Clock hearing aid produced by a hearing aid manufacturer called Argosy. Of course, with the advent of digital hearing aids, both of these products have long since disappeared, even in the proverbial rear view mirror.

The circuitry of the Sound F.X allowed for a fairly steep 24-decibel/octave slope between the two channels, as shown in Figure 7 to 11. When thinking about channels, the term “slope” refers to the steepness of the “sides” or “skirts” of the channels. The slope can be readily appreciated if the gain of one channel is turned down, while the gain of the other channel is turned way up. Channels turned down are shown by the dotted lines in the figure; the gain for either the low-or high-frequency channel could be turned down while leaving the other channel turned up. In addition, as shown in Figure 7 to 11, the frequency meeting point or crossover between the two channels could also be adjusted.

Compared to other analog hearing aids of the day, the steep slope between the two channels allowed for more independence between the channels. That, as well as the adjustable frequency crossover, served to enhance fitting flexibility, and this became readily apparent to clinicians. The gain/output of the hearing aid could be “sculpted” to more closely fit difficult-to-fit audiometric configurations, for example, precipitous or reverse hearing losses. One could thus accommodate the “corner” in an audiogram by providing very different amounts of gain for the low versus the high frequencies.

Consider, for example, various precipitous hearing losses, where the drop-off in thresholds occurs at 500, 1000, or 2000 Hz. The frequency crossover adjustment could accommodate the meeting frequency of the two channels—namely, at 500, 1000, or 2000 Hz. The low-frequency channel could then be turned down, while the high-frequency channel could be turned up. Although the initial intent might have been to provide both Bill and Till in separate channels, the fitting flexibility offered by the two-channel W.D.R.C hearing aids cast a shadow on the features and benefits of these two twins. From here on, the concepts of Bill and Till simply “exited stage left.”

There is more to the story here. An especially interesting feature of analog multichannel hearing aid circuits like the DynamEQ2 was that the gain controls for the low-and high-frequency channels actually adjusted the compression ratio for each channel (Figure 7 to 12, left panel). This was indeed unusual, because up until this point in time, compression could only be adjusted by changing the knee-point. In the case of O.L.C, this was an M.P.O adjustment; in the case of W.D.R.C, this was a T.K adjustment. Recall that the knee-point is the “when” of compression. The additional capability of adjusting the compression ratio enabled clinicians to adjust the amount or the “how much” of compression.

The way in which the gain was adjusted by compression ratio was also unique. Figure 7 to 12 (left panel) shows that an increased compression ratio in the low-or high-frequency channels constituted an increased gain in these channels. In everything

Figure 7-10 summary: This figure consists of two schematic diagrams. The diagrams illustrate the architectural differences between single-channel and multichannel programmable analog circuits used in hearing aids. The left diagram depicts a single-channel system where a microphone signal passes through an amplifier and a volume control to a receiver, with the amplifier being influenced by a compression component and a programmable memory system. The right diagram shows a multichannel system where the microphone signal is first divided via band splitting into separate paths, each with its own amplifier and compression control, before being recombined and sent to the receiver through a volume control. The comparison indicates that multichannel circuits offer more granular control over different frequency bands compared to single-channel circuits, and that both configurations can incorporate programmable settings to allow for adjustments via external software or hardware.

Figure 7-11 summary: This figure is a line graph illustrating the gain response across a frequency spectrum. The graph depicts the performance of a two-channel wide dynamic range compression hearing aid, featuring a low-frequency channel and a high-frequency channel separated by a frequency crossover point. Multiple lines are shown to represent different input levels, with solid lines indicating maximum gain and dotted lines showing reduced gain. The data demonstrates that as input sound levels decrease, the gain for both the low-frequency and high-frequency channels increases. The figure concludes that the ability to independently adjust the gain for each channel, combined with a controllable frequency crossover, significantly enhances the flexibility of the device fitting process.

Loudness Growth & Compression Ratios

we have covered so far in this chapter, we have understood that compared to linear gain, compression provided less gain. Accordingly, an input/output ratio of 2:1 normally provides less gain than an input/output ratio of 1:1. This understanding of compression is indeed true, provided that one assumes that on an input/output function, the compression hinges or pivots from the left or lower knee-point. The point here is that the compression ratio can also hinge from the right! If this is the case, then linear (1:1) gain is less gain than the gain with a 2:1 or a 4:1 compression ratio.

The input/output function on the left side of Figure 7 to 12 could represent either of the two channels of the analog, two-channel Sound F.X. It shows that for each channel, there are actually two knee-points, a left or lower one and a right or higher one. This is a compression feature to digest here, because it pointed to the way compression is presently utilized in the digital hearing aids of today.

Look at the left-most or lower knee-point, which is situated above the input S.P.L where compression is usually assumed to “begin.” As with all W.D.R.C, the T.K control adjusts this lower or left-most knee-point. As discussed earlier, increasing the knee-point (moving it to the right) decreases the gain for soft input sounds. Decreasing the same knee-point (moving it to the left) increases the gain for soft input sounds, because linear gain provided below the knee-point occurs now for even softer input levels.

The T.K control, at one time the only control available to adjust the compression of W.D.R.C hearing aids, now came to be accompanied with adjustable compression ratios. The unique thing here is that, unlike the other hearing aids of their day, the compression ratios of Unitron's Sound F/X™ hearing aid "hinged" from the right-most, bigger knee-point, not from the first lower knee-point. In this situation, an increased compression ratio implies increased gain. Again, this was quite contrary to most other types of compression, in which an increased compression ratio is associated with decreased gain.

The right-most or higher knee-point is where compression “ends” (Figure 7 to 12, left panel). Beyond this input S.P.L, the hearing aid reached a point of “unity gain.” Here, the compression ratio once again is linear, but there is no gain whatsoever. For example, a 100-decibel S.P.L input results in a 100-decibel S.P.L output, a 101-decibel S.P.L input results in a 101-decibel S.P.L output, and so on. Here, the hearing aid was said to be truly acoustically “transparent.”

The feature of adjusting compression ratios for the right-most or higher knee-point was found to be very useful as a model for restoring normal loudness growth (Figure 7 to 12, right panel). In other words, the loudness growth model matches well with the compression ratios that hinge from the right-most knee-point (left panel). The right panel of Figure 7 to 12 shows the reduced dynamic range and abnormal loudness growth that occurs with S.N.H.L. It also shows the required gain to restore normal loudness growth. Matching the two panels of Figure 7 to 12 together, one can see that to restore normal loudness growth, the smaller dynamic range will require a greater compression ratio (greater amount of gain), especially for soft inputs.

As a high-end analog hearing aid, the Sound F.X of 1996—along with its unique compression characteristics—was a precursor to the generation of digital hearing aids that immediately followed in 1997. Remember its two knee-points and adjustable compression ratios. We will encounter these again in Chapter 8.

Figure 7-12 summary: This figure consists of two side-by-side line graphs. The left panel illustrates an input-output function for a circuit, showing how different compression ratios and knee-points affect gain, while the right panel depicts the relationship between input levels and perceived loudness growth. The left panel shows multiple lines representing different compression ratios that hinge from a common right-most knee-point, with linear gain occurring below the initial knee-point. The right panel compares normal loudness growth against impaired loudness growth, using arrows to indicate the dynamic range of hearing loss and the gain necessary to restore normal perception. It can be inferred that adjusting the compression ratio and the right-most knee-point allows for the manipulation of gain to compensate for hearing impairment. Specifically, by selecting appropriate compression settings, the system can model and restore normal loudness growth for a user, effectively bridging the gap between impaired and normal auditory perception.

Common Clinical Combinations of Compression

It is reinforced here once again that an appreciation of these historic types of compression, as they first appeared in analog hearing aids, is essential in order to understand compression as utilized in today's digital hearing aid technology. Let's summarize where we have come so far. In the beginning, the signal processing in hearing aids involved simple linear gain along with peak clipping. Next came O.L.C, followed in time by W.D.R.C. Figure 7 to 13 summarizes all three of these types of signal processing.

The left-most panel shows linear gain, where equal amounts of gain are applied to all input. Peak clipping is employed when the output would be excessive, although this caused distortion. The middle panel shows O.L.C, which is a very similar type of signal processing to linear gain except that instead of using peak clipping to limit the M.P.O, compression is utilized. The right panel shows W.D.R.C, where progressively less and less gain is applied to increasingly more intense inputs. In this manner, a wide dynamic range is neatly shrunk into a smaller one.

Regarding compression and its evolution in analog hearing aids, three of its “faces” have been woven together in this chapter: (1) input versus output compression, (2) O.L.C versus W.D.R.C, and (3) methods of adjusting each of these types of compression. While input versus output compression are now buried beneath the surface and no longer a matter of choice for the clinician, both O.L.C and W.D.R.C—as well as their respective methods of

Summary A Clinical “Spectrum” of Compression

adjustment—are still commonly seen in today's digital hearing aids. Figure 7 to 14 summarizes the clinical applications of each of these three faces or dimensions of compression in yesterday's analog hearing aids as they have been applied to sensory (mild-to-moderate) and neural (severe or more) S.N.H.L.

Figure 7 to 14. Mild-to-moderate“sensory”hearing loss and severe“neural”hearing loss are shown on the audiogram of this summary figure. A small transition area is shown in between. W.D.R.C, which was a type of input compression that utilized a“T.K”adjustment, was deemed to be well suited for fitting mild-to-moderate S.N.H.L. O.L.C, which was always used with output compression and which utilizes M.P.O adjustment, was seen as suited for fitting severe hearing losses.

Figure 7-13 summary: This figure is a conceptual diagram comparing three different hearing aid amplification strategies. The diagram illustrates the relationship between input levels, gain, and output levels relative to a client's loudness tolerance threshold for linear, output limiting compression, and wide dynamic range compression systems. In the linear model, gain remains constant across all input levels, leading to peak clipping when the output exceeds the tolerance threshold. The output limiting compression model maintains constant gain for lower inputs but dramatically reduces gain once a specific threshold is reached to prevent clipping with less distortion. The wide dynamic range compression model reduces gain more gradually starting from a lower threshold, effectively shrinking a broad range of input intensities into a smaller output range. The comparison indicates that while linear aids are prone to distortion via clipping, output limiting compression mitigates this distortion, and wide dynamic range compression provides a more natural loudness growth by gradually adjusting gain.

Figure 7-14 summary: This figure is a conceptual diagram consisting of a grid and associated text labels. The grid represents a frequency-intensity map with frequency values increasing across the horizontal axis and intensity levels increasing down the vertical axis. The diagram divides the intensity range into two distinct zones: an upper region associated with input compression, wide dynamic range compression, and TK adjustment, and a lower region associated with output compression, output limiting compression, and MPO adjustment. It can be inferred that the system applies different processing strategies based on the input level, where lower intensity signals are managed by input-related compression techniques and higher intensity signals are governed by output-related limiting and compression mechanisms to ensure safety and signal integrity.

A Compression Combination for Mild-to-Moderate S.N.H.L

Input compression, W.D.R.C, along with the use of a T.K adjustment, all worked well together in the same analog hearing aid circuit (see Figure 7 to 14). This combination had several features that addressed the needs of clients with mild-to-moderate S.N.H.L and a relatively larger dynamic range of at least 40 decibel to 60 decibel. With input compression, the V.C raised and lowered both the gain and the M.P.O together. Although there is really no clinical advantage here, the simultaneous effect of the V.C upon both the gain and M.P.O was of no great consequence for this clinical population. The physical placement of the V.C in the circuit, however, allowed for the physical addition of the T.K compression adjustment control. W.D.R.C per say, has a low knee-point and a low compression ratio. With these two compression characteristics in mind, a T.K adjustment of the knee-point of compression specifically affects the gain for soft incoming sounds. In this way W.D.R.C “imitates” the role of the O.H.C's.

The provision of linear gain for only very soft input levels, along with a weak degree of compression for all greater input intensity levels, enabled clients to get most amplification for soft input sounds and progressively less and less amplification for sounds that progressively increased in intensity. This serves to shrink a larger dynamic range into a smaller one, thereby restoring normal growth of loudness for those with mild-to-moderate S.N.H.L.

A visual categorization of input compression is shown in Figure 7 to 15. As the philosopher John Quine once said, “Categorization is the essence of intelligence.” First, look at the outside ring of the circle shown in the figure. This represents the historical development of input compression that, for the sake of brevity and clarity of explanation in this chapter, has not been mentioned until now. We would be remiss in historical coverage of compression in analog hearing aids, however, if we did not mention it. Preceding W.D.R.C, there did exist an input compression circuit that was not W.D.R.C. These were often high-power, analog hearing aids that utilized O.L.C and had an M.P.O type of adjustment. In the author's opinion, this was truly an “oddball” combination of compression characteristics. In colloquial terms, these hearing aids comprised a “department of redundancy department,” because they offered two ways of adjusting the M.P.O: by means of the V.C adjustment as well as the M.P.O adjustment itself!

W.D.R.C then entered the scene as the “new kid on the block,” and it was typically used along with input compression. At the time, then, W.D.R.C was thus a subset type of input compression, but not all input compression was W.D.R.C. Similarly, Bill and Till were two subset types of W.D.R.C, but not all W.D.R.C was specifically Bill or Till. Recall that Bill hearing aids offered most compression for the low frequencies and Till offered most compression for the high frequencies. A straight W.D.R.C hearing aid that is neither Bill nor Till offered a more similar degree of compression across the frequencies.

Categorizing Input Compression, W.D.R.C, Bill & Till When degree of compression is more in low frequencies, it means the gain is more in low input levels in low frequencies than at high input levels in low frequencies (Bill). Whereas if the degree of compression is more in high frequencies, it means gain is greater at low input level high frequencies compared to high input (Till).

Figure 7-15 summary: This figure is a Venn diagram illustrating the hierarchical relationship between different types of input compression technologies used in hearing aids. The diagram consists of nested circles where the outermost circle represents input compression, which contains a smaller circle for Wide Dynamic Range Compression (WDRC), which in turn contains two smaller, adjacent circles labeled BILL and TILL. The layout indicates that while all WDRC is a form of input compression, not all input compression is WDRC. Similarly, BILL and TILL are specialized subsets of WDRC, with BILL corresponding to low-frequency applications and TILL corresponding to high-frequency applications. The figure concludes that BILL and TILL are specific implementations of WDRC, but they do not encompass all possible WDRC applications.

A Compression Combination for Severe Hearing Loss

Output compression, O.L.C, along with the use of an M.P.O adjustment also all worked well together in the same analog hearing aid circuit (see Figure 7 to 14). This combination had several features that addressed the needs of clients with more severe S.N.H.L and a relatively small dynamic range that came with that degree of hearing loss. With output compression, the V.C adjusted the gain but not the M.P.O. With its high knee-point and high compression ratio, O.L.C is well poised for use with high power, with lots of linear gain for both soft and average input levels. The method of adjustment for O.L.C is raising and lowering the M.P.O. This was of course done by the clinician, based upon the client's loudness discomfort level. Output compression, O.L.C, along with the use of an M.P.O adjustment thus also worked well together in the same analog circuit.

This combination has several features for clients with severe hearing loss. With high-power hearing aids, protection of residual hearing is critical for these clients. Using output compression, the client can be assured that although the V.C of these hearing aids changes the knee-point of compression, it affects only the gain and not the M.P.O. The client received lots of linear gain for soft through average input levels. Once the output came close to the individual's loudness tolerance levels, however, the hearing aid suddenly provided a high degree of compression, so as to limit the M.P.O. Although normal loudness growth was not achieved for these clients, they did get a strong degree of amplification along with a good protection against excessive output.

Another hearing loss that might be mentioned as a candidate for this clinical combination of compression is conductive hearing loss. Note that unlike severe S.N.H.L, the dynamic range of conductive hearing loss is not normally diminished by much. Recall that conductive hearing loss is much like a plug in the ear. Since the thresholds as well as loudness tolerance levels for this clinical population would both be elevated from normal, this sector would also benefit from linear gain—along with a high M.P.O. W.D.R.C per say would not be the best fit for conductive hearing loss.

Dynamic Aspects of Compression

Until now, compression has been discussed in terms of threshold knee-points and compression ratios. These are sometimes known as the “static” aspects of compression, because they involve the input S.P.L when compression begins (knee-point) and the degree of compression (ratio) once it occurs. Sound in the environment, however, is constantly changing in intensity over time, and a compression hearing aid has to respond to these changes in intensity over time. The “dynamic” aspects of compression concerning reaction times of compression are known as the “attack” and the “release” times.

The attack and release times are the lengths of time it takes for a compression circuit to respond to changes in the intensity of an input S.P.L (Figure 7 to 16). When the input S.P.L exceeds the knee-point of compression, the hearing aid “attacks” the sound by going into compression and reducing the gain. Once the input sound falls below the knee-point of compression, the hearing aid “releases” from compression and restores the original gain. The

Dynamic Compression Characteristics

attack time is the length of time it takes for a hearing aid to go into compression; the release time is the length of time it takes for it to come out of compression.

Hearing aids are not the only electrical devices that use compression, nor are they the first to have attack/release times. Audiovisual equipment has used input and output compression, W.D.R.C, and O.L.C for many years. It has also utilized various schemes of attack/release times.

We have all heard the effects too. Recall, for example, the television broadcasts in which the sports announcer is talking and the background noise is changing in intensity over time. When a score is made and the audience suddenly increases their cheers, the background noise increases in intensity. It may take a short time for the compression of the audiovisual equipment to attack and reduce the gain of the noise.

In turn, there is a slight bit of time required to reduce the gain for the announcer's voice. When the cheering stops, it may again take some time for the system to release from compression, and the announcer's voice will accordingly take some time to return back to a normal, audible level.

Attack and release times cannot be instantaneous. For example, if a compression circuit is to respond to a sudden input S.P.L increase, it must wait for at least one cycle of the sound wave to “know” if the increased S.P.L will remain. A change in gain that occurs faster than the longest cycle or period of incoming sound will necessarily change the fine details of the sound waves, and distortion will result.

Most attack and release times have been set to achieve a best compromise between two undesirable extremes. Times that are too fast will cause the gain to fluctuate rapidly, and this may cause a jarring acoustical perception by the listener. Times that are too slow may make the compression act too slowly and cause a real lagging perception on the part of the listener.

Quick attack times (i.e., 10 ms or less) will likely prevent sudden, tran-zee-unt sounds from becoming too loud for the listener. When these are used in combination with similarly quick release times, however, this can cause the hearing aid to track the amplitude of individual cycles of sound waves, causing distorted sound quality. Consequently, release times need to be a bit longer than attack times. Poor management of attack/release times can cause a “fluttering” perception on the part of the listener.

This is because while the hearing aid may be in compression with speech inputs, the background noise between sentences is increased. Such a rapid modulation of sound can cause a “breathing” or “pumping” perception on the part of the listener.

Just as we have seen with the static aspects of compression, there are lots of “buzz” words that float around when the dynamic aspects of compression are discussed. Different attack and release time combinations are sometimes used to categorize different types of dynamic compression. Some of the various types of dynamic compression methods as they evolved in analog hearing aids are briefly discussed here.

Figure 7-16 summary: This figure consists of two waveform diagrams showing sound intensity over time. The top diagram illustrates an input sound signal that remains at a low intensity, suddenly increases to a higher intensity for a duration, and then abruptly returns to a low intensity. The bottom diagram depicts the corresponding output response of a compression hearing aid, highlighting the transition periods labeled as attack time and release time. It can be inferred that the hearing aid provides gain to soft sounds to increase their audibility. When a sudden increase in sound intensity occurs, the compression circuitry does not react instantaneously, resulting in an attack time before the signal is fully compressed. Similarly, when the sound intensity drops, there is a release time during which the circuitry gradually stops compressing the signal.

Peak Detection

Most analog compression hearing aids initially used a technique called peak detection to “track” the peak amplitude of incoming sound waves. If the peak input intensity was greater than the compression threshold knee-point, the circuit attacked and began to compress the signal, which in turn reduced the gain. Once the peak input intensity fell below the knee-point, the compression released and the gain increased again (back to linear gain).

Peak detection allowed for a wide variety of times that could separately be specified and assigned as attack and release times; however, these times would be constant and fixed for any incoming sound intensity patterns. Most peak detection systems in hearing aids were adjusted to provide quick attack times and longer, slower release times. Peak detection commonly used attack times of about 50 ms and release times of about 150 ms. This was seen as the best compromise between objective effectiveness of compression and subjective listening comfort.

An advantage of peak detection was that it although reacted quickly to increases in environmental sound levels, it also reacted inappropriately to very short intense sounds, because the longer release times kept the gain down after the short intense sound has stopped. This may unnecessarily reduce the gain of sounds the listener might want to hear in the immediate aftermath.

With fixed attack/release times, the hearing aid could not respond differently to different patterns of sound input intensities when needed. As discussed previously in Chapter 6, the intensity of speech changes constantly and rapidly over time. The dynamic acoustic sounds of speech within the world of ever-changing background noise (e.g., the sudden slamming of a door or the constant roar of traffic) can pose real problems for the peak detection method and the listener.

“Automatic volume control” (A.V.C) and “Syllabic compression” are two specific sets of alternative attack/release times that were encountered on analog hearing aids. These were also utilized on the earliest digital hearing aids. The first high-profile usage of A.V.C occurred with the advent of the Widex Senso digital hearing aid in 1997. The first such usage of syllabic compression occurred with the advent of the Oticon DigiFocus in 1997.

Automatic Volume Control

This attack/release paradigm has often been used during broadcasts with audiovisual equipment. In a comparison to peak detection, A.V.C has a relatively long attack and long release times; its release times are usually more than 150 ms and can be as long as several seconds. Because of this, it does not respond to rapid fluctuations of sound input. On the contrary, it responds mainly to overall changes in sound intensity, which in turn reduces the need for the listener to adjust the volume control manually.

The long attack/release times with A.V.C were actually intended to imitate the length of time it takes for a listener to react to sudden noise increases by physically raising a hand to manually adjusting the V.C on a hearing aid; hence, its name! Widex promoted the use of A.V.C on its Senso digital hearing aid; the reason was because field trial subjects who first tried the Senso liked it best.

Syllabic Compression

“Syllabic compression” refers to the exact opposite of A.V.C—namely, relatively short attack and release times; its release times vary from less than 50 ms up to 150 ms. The attack/release times are specifically intended to be shorter than the duration of the typical speech syllable, which is about 200 to 300 ms. Short attack/release times allow the hearing aid to compress or reduce the gain for the peaks of more intense speech (usually the vowel sounds), and this provides more uniformity in the intensity of ongoing speech syllables. In other words, syllabic compression reduces the differences between the normally more intense vowels and the softer unvoiced consonants such as /s/. The main premise of syllabic compression is to allow a hearing aid to make the softer sounds of speech more audible with simultaneously keeping the normally louder parts of speech from becoming too loud.

Syllabic compression was somewhat controversial, and not everyone agreed with its use. Recall from earlier discussion in this chapter (section on W.D.R.C) that since W.D.R.C amplifies soft sounds more than it amplifies loud sounds, it reduces the “peak-to-valley” contrasts of input speech sound waves, thus compromising natural clarity of ongoing speech. In addition, Kuk (1999) suggested that the use of fast attack/release times (less than 10 ms) and short release times (less than 100 ms) would also compromise the intensity differences between the various phonetic elements of speech. Adding the fast attack/release times of syllabic compression that reduces the peaks of the speech waveform, then, would seem to even further deteriorate the speech waveform that has been degraded already with W.D.R.C. The resultant modified and more uniform speech waveform envelope would allow noise to more easily fill in the small gaps that remain. In noisy situations, then, a hearing aid might amplify the noise situated between the reduced peaks of the speech waveform.

, however, reminds us that in amplifying soft sounds more than loud sounds, it is only the peak-to-valley contrast between low-frequency vowels and high-frequency unvoiced consonants that is compromised by W.D.R.C. In addition, posits that although this—along with fast-acting (syllabic) compression—indeed does reduce the peak-to-valley contrasts in the ongoing speech waveform envelope, this can actually be of benefit for the listener with recruitment who perceives an abnormally large loudness contrast between the more intense versus the softer elements of speech. In a way, then, the peak-to-valley reduction of the speech waveform with fast-acting W.D.R.C would offset the abnormal, exaggerated loudness perception experienced by those with recruitment.

As mentioned earlier, in the mid to late 1990s, Oticon promoted the use of syllabic compression (along with Bill), in its analog Multi-Focus hearing aid, and in its first digital DigiFocus hearing aid that immediately followed. Bill was thought to reduce the upward spread of masking and thereby improve aided speech recognition. Thus with Syllabic compression, not only was W.D.R.C present especially for the low frequencies, but the W.D.R.C was made to act quickly. If the normally loud, low-frequency background noise can be thought of as the bull and relatively soft, high-frequency consonants as fragile china teacups, then syllabic compression used in conjunction with Bill was seen as a way to “control the raging bull in the china shop.”

Adaptive Compression

This type of compression had fixed, quick attack times but had release times that varied with the length of time it took for a loud sound to become quiet again. For short (tran-zee-unt) intense sound inputs like a door slam, the attack and release times were short. For sound inputs that took longer to become quiet again, the attack time remained quick but the release times were longer.

The desired result was a reduction of compression “pumping” heard by the listener. Adaptive compression was originally patented by a now long-gone company called Telex; this company was also associated with F.M systems of the day. Later on, adaptive compression became most commonly associated with the Kamp circuit, which came to be utilized by many of the hearing aid manufacturers.

Average Detection

Average detection was first associated with the previously mentioned analog DynamEQ2 circuit by Gennum. Recall that this circuit was one of the original analog two-channel W.D.R.C hearing aid circuits that emerged during the mid-1990s, when the low-frequency channel utilized Bill and the high-frequency channel utilized Till.

Unlike the peak detection method that tracked the peak amplitude of incoming sound waves, the average detection method looked at the average amplitude of incoming signal over a given length of time. When the average S.P.L exceeded the knee-point of compression, then the gain was reduced. For the purpose of explaining “average detection” in its historical context, its implementation in the analog, two-channel DynamEQ2 will be explained here.

The DynamEQ2 had “twin” average compression detectors; one was a fast detector and the other was a slow detector (Figure 7 to 17). The slow average detector averaged sound inputs over a 220-ms time interval (i.e., about 1/5 of a second) and was in control of the compression system most of the time. When the slow average of incoming sounds exceeded the threshold knee-point of compression, the gain was slowly reduced and the reduction was hardly noticeable. With the slow detector alone, however, a short spike of intense tran-zee-unt sound could be averaged into the overall body of sound taking place over 220 ms. This slow average would not be enough to “tell” the hearing aid to go into compression and reduce the gain.

This is where the fast average detector came into the picture. The fast average detector averaged sound inputs over much shorter time intervals of about 10 ms (i.e., 1/100 of a second) and it acted when intense transients were not “caught” by the slow detector. When the “fast” average was 6 decibel greater than the “slow” average, the fast average detector took over and reduced the gain for the spike of intense sound.

The main advantage with average detection is that both the attack and release times varied with the length of the incoming intense sounds. This was in direct contrast to the peak detection systems that gave constant, fixed quick attack and slow release times for all incoming stimuli. Reacting to a tran-zee-unt door slam, the peak detection system would provide its usual quick attack and slow release times.

Because of the reduction of gain and the long recovery of the peak detection circuit, soft speech spoken right after the door slam would be temporarily inaudible to the listener. On the other hand, the average detection circuit enabled a quick recovery of gain after the door slam because its

Twin Average Compression Detectors for the DynamEQ trademark Circuit

release time was quick for short sounds. With sudden tran-zee-unt loud sounds, the average detection system provided quick attack and quick release times. For sounds that took longer to become louder and softer, average detection provided longer attack and release times.

A big benefit to the listener was that with average detection, there was even less “pumping” perception than with adaptive compression. The average detection system was touted as a compromise between fast compression that reacts to every short intense sound and slow compression that may react too slowly for some sounds that should be compressed. The idea was that audible by-products of compression should not become a nuisance to listeners. In other words, dynamic aspects should be considered when trying to make hearing aids acoustically “transparent.”

It is interesting to note that the opposites of A.V.C and syllabic compression were used in the very first two digital hearing aids of the day (Widex Senso and Oticon DigiFocus). There you have it: opposite attack/release time schemes for the first two digital hearing aids out in the marketplace. The jury was obviously out regarding the best attack/release times. A firm conclusion on the very best paradigm for attack/release times has never really been reached.

Some of the various dynamic compression characteristics described above are actually available on most of today's digital hearing aids on the fitting software. For the most part, on most manufacturers' fitting software, the choice is largely between syllabic compression and average detection. Most often, syllabic compression is the default for the low-frequency channels of the digital hearing aid, and average detection is the default for the higher frequency channels. Before leaving this topic, it should be noted that an additional focus over the past 10 or so years has almost become the instant compression of sudden, very loud tran-zee-unt sounds.

Figure 7-17 summary: This figure is a line chart displaying signal waveforms over time. The content illustrates the operation of average detection by comparing an input sound level waveform with two different averaging detectors: a slow detector and a fast detector. The slow detector produces a relatively flat line, while the fast detector produces a more volatile line that tracks the input signal more closely. From this, it can be inferred that the slow detector provides a stable baseline for general compression decisions, whereas the fast detector is designed to respond rapidly to sudden peaks. The system is designed so that when the fast average significantly exceeds the slow average, the fast detector takes control, allowing the system to apply quick attack and release times for transient sounds while maintaining longer times for gradual changes in sound intensity.

Interaction Between Static and Dynamic Aspects of Compression

Compression consists of static aspects in one dimension and dynamic aspects in a separate dimension. With incoming sounds, the attack/release times of a hearing aid can and do interact with the ratio of compression. The I/O functions on hearing aid specifications show compression ratios that are obtained with constant pure tones, however, and not with the stops and starts of sounds like speech. Static compression ratios on hearing aid specifications therefore give an idea but do not always accurately represent the actual compression ratios experienced in real life by clients who wear the hearing aids. Interactions between attack/release times and compression ratios affect the sound quality for the listener.

Fast attack/release times have the effect of temporarily reducing the ratio or amount of compression for any given sound stimulus. In general, a combination of short attack/release times (e.g., 10 ms) and high compression ratios (e.g., 10:1) causes the most distortion. If the same short attack/release times are used with low compression ratios (e.g., 2:1), then the sound quality is not as distorted. On the other hand, long attack/release times can be combined with either high or low compression ratios.

Dynamic aspects and static aspects of compression began to be found in predictable combinations. Syllabic compression, with its relatively short attack and release times, came to be associated most often with W.D.R.C hearing aids that had a low-compression knee-point and low-compression ratios. It was less common with O.L.C hearing aids with their high knee-points and high-compression ratios.

Summary

■ The 1990s were the “golden” age of compression, because compression types flourished and differentiated, and yet, all hearing aids were still analog. This meant that each hearing aid was confined to providing one type of compression or another. In order to choose the appropriate hearing aid products for their clients, clinicians thus needed to understand each type of compression well and had to categorize these types among all available types of compression.

■ In this chapter, we looked at the historical development of the many faces of compression. With I/O functions, compression was explored along the dimensions of input

versus output compression, O.L.C versus W.D.R.C, and their respective M.P.O versus T.K adjustments.

■ Clinically, the main difference between input and output compression was the effect of the V.C. With output compression, the V.C affected the gain but not the M.P.O. With input compression, the V.C affected both the gain and the M.P.O.

■ Output limiting compression was compared to W.D.R.C regarding compression knee-points and compression ratios. O.L.C has a high knee-point, which means that compression is activated only for relatively intense sound input levels; it also has a high compression ratio. W.D.R.C has a low knee-point and a low compression ratio. These differences separate their respective clinical purposes. O.L.C acts mostly above its knee-point to limit the output. W.D.R.C acts mostly below its knee-point to provide most (linear) gain for soft input sounds.

■ The effects of M.P.O adjustment on O.L.C hearing aids and the T.K control on W.D.R.C hearing aids were compared. For M.P.O adjustment, as the knee-point is lowered, so is the M.P.O. For T.K adjustment, as the knee-point is reduced, the gain for soft inputs is increased. The effect of M.P.O adjustment is audible only when speaking loudly into the hearing aid; the effect of the T.K control is audible only when letting ambient room noise into the hearing aid microphone.

■ O.L.C was most often associated with output compression and utilized an M.P.O type of adjustment. This combination of features was found to be appropriate for severe hearing loss that usually exhibits a narrow dynamic range. W.D.R.C was associated with input compression and utilized a T.K adjustment. This combination was found to be appropriate for mild-to-moderate S.N.H.L, which presents with a relatively wider dynamic range.

■ Two types of W.D.R.C were Bill and Till.

■ Dynamic aspects of compression were discussed separately from the static compression aspects of compression

threshold knee-point and compression ratio. Different types of attack/release time parameters were discussed. The usual compromise with the historically early peak detection method was to provide fast attack times with longer release times.

- ☑ A.V.C and syllabic compression were described as being opposites. A.V.C has long attack/release times, and syllabic compression offers relatively short attack/release times. A.V.C was chosen for client comfort. Syllabic detection was chosen to reduce the upward spread of masking. Adaptive compression offered fixed attack times and variable release times, while average detection offered variable attack and release times. Both adaptive and average detection were designed to reduce the adverse perception of audible hearing aid amplifier “pumping.”

Review Questions

Answers are given in Appendix B. Sketching I/O functions for Questions 4 to 10 may be very helpful. Three figures in Appendix B show I/O functions that are helpful in interpreting the answers.

1. On an I/O function, longer 45 degree lines represent more linear gain. True False

2. On an I/O function, moving a 45 superscript circle line to the right represents more linear gain. T F

3. With W.D.R.C, lowering the knee-point increases linear gain for soft inputs. T F

4. A compression hearing aid provides 90 decibel S.P.L output with 40 decibel S.P.L input; what's the gain here?

5. Same hearing aid: knee-point at 50 decibel S.P.L, compression ratio of 2:1; the output for a 60-decibel S.P.L input is:

6. Same hearing aid: knee-point at 50 decibel S.P.L, compression ratio of 2:1; the gain for a 60-decibel S.P.L input is:

Chapter 5: Expansion

Issues Resulting from W.D.R.C As we have seen throughout this handbook, W.D.R.C provides additional gain for weak inputs, as compared to linear amplification. The advantage of this approach is increased speech intelligibility, especially for soft speech, and normalized loudness perception. The bad news is that it also amplifies other soft sounds, including the drone of a refrigerator, the creaking of floor boards and the hum of an air conditioner. How can this be a disadvantage when a person with normal hearing is able to hear these everyday sounds?

Presbycusis, the most common cause of hearing loss in adults, typically has a gradual onset. As a result, by the time the hearing loss becomes noticeable to the individual, many of the weaker sounds in the environment may not have been heard for some time. W.D.R.C enables amplification of weak sounds to within the audible range.

This can be overwhelming to an auditory system that has been deprived for a long period of time. On a different note, individuals who have good residual hearing (i.e., thresholds that are within or close to the normal range) at some frequencies may be able to hear noise that is inherent to the hearing aid, such as circuit noise that is amplified by the hearing aid. Either of these scenarios may result in an unpleasant experience and, in turn, hearing aids that spend most of their time in a drawer. To avoid this rejection, the problem must be addressed.

The first approach to tackling the "problem" of increased audibility for weak sounds is counseling. Potential hearing aid users often do not realize that, in addition to hearing the soft voice of a grandchild, they will also hear other sounds in the environment that they may not have heard in a long time. It is important that they understand that these sounds are also heard by persons with normal hearing and that, over time, they will become accustomed to the abundance of environmental sounds. Some individuals are especially sensitive to these sounds and, as such, are eager for an immediate solution. Moreover, counseling will not alleviate the problem when it is caused by audible circuit noise.

Gain reduction has traditionally been the second line of attack in this situation. Specifically, the gain for weak sounds is decreased in an attempt to reduce the audibility of environmental and/or circuit noise. The drawback of this solution is that it defeats the purpose of using W.D.R.C in the first place – weak sounds are now no more audible than they would be with linear amplification or, worse, with no amplification at all. Further, the audibility of important speech sounds is adversely affected.

Thus, alleviating one problem has given rise to a second, and arguably more serious, issue.

A more viable solution to the predicament, and one that is commonly available in modern digital hearing aids, is the use of expansion. Expansion is designed to make a hearing aid sound silent in quiet environments. One can think of it as a noise reduction strategy for quiet environments. The use of expansion can lead to greater listener satisfaction by reducing the intensity of weak environmental sounds and circuit noise, without sacrificing speech audibility and intelligibility.

What is Expansion?

Consider the following example. Figure 5-1A compares the residual dynamic range of a person with sensorineural hearing loss to the range of sounds in the environment. As expected, weak sounds are inaudible, while intense sounds are perceived as loud. Figure 5-1B illustrates the effect of compression amplification – weak sounds are amplified to a greater extent than moderate and intense sounds. This allows the range of environmental sounds to fit within the residual dynamic range of the individual.

While this improves the intelligibility of speech, it also makes unwanted sounds audible, such as the quiet drone of a refrigerator. Figure 5-1C shows the effect of expansion in combination with compression. The band representing weak sounds is wider than those for moderate or intense sounds. Further, the weakest of the weak sounds are below the dynamic range, and only the more intense components of the weak sounds are audible to the individual.

This is the principle behind expansion – the softer sounds are amplified less than louder sounds. In hearing aids, expansion is only applied to the weakest sounds in the environment. A comparison of Figures 5-1B and 5-1C shows that the bands representing moderate and intense are identical. In other words, compression is applied to moderate and intense sounds, while expansion is used only for weak sounds.

Figure 5 to 2 shows an I/O function for a linear hearing aid with O.C.L. A fixed amount of gain is applied to all input levels, as is characteristic of linear amplification. Specifically, for every 10 decibels change in input level, the output also changes by 10 decibels. When expansion is applied at low input levels (i.e., at or below the T.K of 50 decibels S.P.L), the output changes by 20 decibels for a 10 decibels change in input level (orange curve in Figure 5 to 2). Note that the I/O function above the T.K is unaffected by whether or not expansion is applied.

Figure 5-2 summary: This figure is a line chart. It illustrates the relationship between input and output sound pressure levels for two different processing functions: linear amplification with output compression limiting and expansion. The chart identifies a threshold kneepoint where the behavior of the expansion function converges with the linear function. Based on the data, the linear function maintains a consistent ratio between input and output until it reaches a saturation point where the output levels off. In contrast, the expansion function shows a steeper increase in output relative to input below the threshold kneepoint, effectively reducing the output for lower input levels compared to the linear amplification model.

Like compression, expansion delivers different amounts of gain depending on the input level. However, in expansion, the gain increases as the input level increases. This is in contrast to compression where the gain decreases as the input level increases.

The change in gain as a function of input level is most easily illustrated by an I/G function. Figure 5 to 3 (page 38) shows the I/G function of a linear hearing aid, and the effect of expansion. In a linear hearing aid with no expansion, the gain remains constant below kneepoint. In a hearing aid with expansion, however, little or no gain is applied to input levels of 20 decibels S.P.L or lower. As the input level increases, so does the gain, until it reaches a maximum of 30 decibel at the T.K (i.e., 50 decibels S.P.L). Note once again that the I/G function above the T.K is unaffected by whether or not expansion is applied.

Figure 5-3 summary: This figure is a line chart. It illustrates the relationship between input levels and gain for two different signal processing configurations: linear amplification with output compression limiting and expansion. The chart displays how gain remains constant up to a specific threshold kneepoint before decreasing, while the expansion configuration shows gain increasing from a lower starting point until it reaches that same threshold. The data indicates that while both configurations eventually converge to the same gain level, the expansion setting significantly reduces gain for lower input signals. Furthermore, both configurations exhibit a reduction in gain at high input levels due to the output compression limiting effect.

Figure 5-1 summary: This figure consists of three comparative diagrams. The diagrams illustrate the relationship between the range of environmental sounds and the dynamic range of hearing for individuals with sensorineural hearing loss across three different scenarios: without amplification, with compression amplification, and with compression amplification combined with expansion. In the first scenario, a significant portion of environmental sounds falls outside the dynamic range of hearing, being either too soft to be perceived or too loud. The second scenario shows how compression amplification shifts the perception of sounds, bringing more of the environmental range into the audible dynamic range. The third scenario demonstrates that combining compression amplification with expansion further optimizes the alignment between environmental sounds and the dynamic range of hearing. It can be inferred that the use of compression and expansion technologies effectively increases the proportion of environmental sounds that are audible and comfortable for individuals with sensorineural hearing loss compared to no amplification.

Characterizing Expansion

Expansion is characterized in much the same way as compression, via a threshold kneepoint, expansion ratio, and attack and release times. There are, however, a few key differences which will be highlighted here. [Note that, at the time of this writing, there are no standards for defining the characteristics of expansion.]

Expansion Threshold

Expansion threshold (X.T) is the input level below which expansion operates. For the hearing aid shown in Figure 5 to 4, the X.T is 50 decibels S.P.L. Because it looks like a bent knee in an I/O function, the X.T is also referred to as a T.K.

Figure 5-4 summary: This figure is a line chart. It illustrates the relationship between the input and output sound pressure levels for an expansion function, specifically highlighting the threshold kneepoint and the expansion ratio. The graph shows that as the input level increases, the output level also rises, with a distinct change in the slope at the threshold kneepoint. From the content, it can be inferred that the system operates with a specific expansion ratio where the output increases more slowly than the input below the kneepoint, and the threshold kneepoint serves as the critical transition point for the expansion process.

The potential for confusion arises when expansion is coupled with compression, both of which are characterized by a T.K. The T.K's for both expansion and compression frequently occur at the same input level. Thus, as shown in Figure 5-5A, expansion operates below the T.K, while compression takes over above it. A less commonly implemented design is one where the T.K for compression T.Kcomp is higher than the T.K for expansion T.Kexp (Figure 5-5B. Note that amplification in the portion of the I/O function in between T.Kexp and T.Kcomp is linear (i.e., C.R=1:1).

A low T.Kexp (input levels of 50 decibels S.P.L or lower) permits the use of more gain to ensure maximum speech audibility. In this scenario, the hearing aid may be in expansion for only a small portion of time and only in very, very quiet environments. As a consequence, the problem of a "noisy hearing aid" may not be alleviated at all. On the other hand, a high T.Kexp (input levels greater than 40 decibels S.P.L) will result in a quiet hearing aid, but at the expense of speech audibility.

Expansion Ratio

Like its counterpart in compression, the expansion ratio (X.R) is an indicator of the extent to which the input signal is expanded. Specifically, X.R is calculated using the formula:

Math summary: This calculation determines the expansion ratio. It is computed by dividing the change in input values by the change in output values.

For the hearing aid shown in Figure 5 to 4, increasing the input from 40 to 50 decibels S.P.L ( Delta Input = 10 decibels) increased the output from 60 to 80 decibels S.P.L ( Delta Output = 20 decibels). Using the above formula, the X.R of the hearing aid is:

Math summary: This computation calculates the compression ratio by dividing the change in input values by the change in output values. The process takes an input increase of ten decibels and divides it by an output increase of twenty decibels to produce a final scaling factor of zero point five.

C.R's are always greater than 1 when compression is applied. Further, the larger the number, the greater is the degree of compression applied. On the other hand, X.R's between 0 and 1 indicate the application of expansion. Figure 5 to 6 shows I/O functions for three hearing aids with the same T.K but different X.R's. Hearing Aid A, with an X.R of 1, is linear below the T.K. The I/O function for Hearing Aid C (X.R =0.4) has a steeper slope than that for Hearing Aid B (X.R =0.7). A steeper expansion slope indicates a greater degree expansion. Thus, in general, the smaller the X.R (i.e., closer to 0), the steeper is the slope of the I/O function and the greater is the degree of expansion.

The higher the X.R (i.e., closer to 1), the more likely it is that weak sounds will be amplified to within the residual dynamic range. In other words, the likelihood that a person will be able to hear the weakest of sounds is greatest at high X.R's. On the other hand, low X.R's are known to adversely affect the intelligibility of speech at low input levels. Either problem may be exacerbated by the exact location of the T.K. Specifically, applying a lot of expansion (low X.R) at high T.K's may drastically reduce speech audibility, while applying only a little expansion (high X.R) at low T.K's may result in the hearing aid sounding noisy.

Figure 5-5 summary: This figure consists of two line charts showing the relationship between input and output sound pressure levels for hearing aids. Panel A illustrates a scenario where a single threshold kneepoint serves as the transition for both compression and expansion. Panel B displays a configuration where there are distinct threshold kneepoints, with the compression threshold occurring at a higher input level than the expansion threshold, creating an intermediate linear region. From these charts, it can be inferred that the hearing aid in Panel A switches directly from an expansion phase to a compression phase at a specific input level. In contrast, the hearing aid in Panel B provides a range of input levels where the output increases proportionally with the input before compression begins, allowing for a more gradual transition between expansion and compression.

Figure 5-6 summary: This figure is a line chart. It displays the relationship between input and output levels for three different hearing aid configurations, all sharing a common threshold kneepoint. The chart illustrates how different expansion ratios affect the output signal relative to the input signal before the kneepoint is reached. The data indicates that as the expansion ratio decreases, the slope of the output curve becomes steeper. Consequently, hearing aids with a lower expansion ratio provide a higher output for a given input level compared to those with a higher expansion ratio, until they all converge at the same threshold kneepoint.

Attack and Release Times

Because expansion only occurs below the T.K, attack and release are defined differently for expansion than for compression. Attack into expansion occurs when the signal level drops below the T.K, while release from expansion occurs when the signal level exceeds the T.K.

A fast A.T limits the amplification of ambient sounds in the environment, and implies that the hearing aid may go into expansion during pauses in speech. Slow R.T's are known to adversely affect speech intelligibility. This occurs because gain is considerably reduced when the hearing aid is in expansion, and is slow to recover when the input level increases; the small amount of gain may render speech inaudible. The effects of both of these problems are magnified if the T.K is high and the X.R is low. Thus, in theory, it appears that a combination of a slow A.T and a fast R.T offers the best balance, making weak environmental sounds inaudible with little effect on the audibility of speech.

Measuring Expansion

Expansion operates on low-level inputs. Ambient noise levels in most environments are often too high (i.e., above the T.K) for expansion to engage. As a result, it can be difficult to hear and/or measure the effects of expansion.

While a sound treated room or anechoic chamber are not necessary, an important prerequisite for listening to and/or measuring expansion is that the environment be quiet. Objections to the hum of the refrigerator or circuit noise occur primarily when the hearing aid user is in a quiet environment, such as reading a book in the living room. Thus, a quiet office, away from the sounds of copy machines and people talking, will suffice for evaluating expansion.

Informal Listening

With expansion turned off, listen for quiet sounds (e.g., whirr of a computer, hum of the air conditioner, etcetera). Turn on expansion and listen for those same sounds. The hearing aid should sound much quieter with expansion turned on. The difference between the two conditions will be more prominent at high T.K's and low X.R's.

Figure 5-7 summary: This figure is a line chart. It illustrates the relationship between frequency and gain for a device measured in a coupler using a broadband signal, comparing the effects of having the expansion feature turned off versus turned on with varying wait durations before measurement. The data shows that as the wait time increases, the gain levels rise across the frequency spectrum, moving closer to the levels observed when expansion is completely disabled. It can be inferred that the expansion feature reduces gain, and this reduction is most pronounced immediately after the signal begins, gradually recovering toward the baseline gain as the wait duration lengthens.

Coupler Measurement

The effect of expansion can be measured using a standard 2-cc coupler, a test box and a broadband input signal such as white, pink, speech-shaped or composite noise. Figure 5 to 7 shows the gain of a hearing aid for a composite noise with expansion turned off and an overall input level of about 70 decibels S.P.L (or above the T.Kexp of the hearing aid). To complete the measurement, turn expansion on and turn the stimulus in the test box off for longer than the A.T to allow expansion to engage. Once this has occurred, turn the stimulus on again and make another measurement immediately ("Expansion On (no wait)" in Figure 5 to 7). You should find that the gain of the hearing aid is lower with expansion on than with expansion off. The longer the interval of time between turning on the stimulus and making a measurement, the

Compression and Other Features in Digital Hearing Aids

Introduction

So far, the reader no doubt wonders why it has taken up until this point to finally and specifically address digital hearing aids. True, we have consistently referred to digital hearing aids as something to be discussed "later on." Compression, however, did not begin with digital hearing aids, and compression in digital hearing aids cannot be truly appreciated unless the characteristics of their analog forebears are understood.

As mentioned earlier in Chapter 7, the various types of compression in analog hearing aids had to be learned by clinicians, because those hearing aids typically provided either one type or another type of compression. Successful client fittings thus relied on such knowledge. This is why decade of the 1990s was truly the golden age of compression. Furthermore, programmability and multichannel amplification (as discussed in Chapter 7) also originated in analog hearing aids.

Further advances in amplification, such as digital noise reduction (D.N.R), feedback reduction—and more—are now commonplace in today's digital hearing aids. Digital hearing aids are all normally adjusted by the manufacturer's fitting software. Digital instructions (algorithms) for the gain, M.P.O, and compression, along with many nuanced acoustic behaviors for various different listening environments, are implemented by means of software onto the hardware of the digital chip or the digital signal processing (D.S.P) core. All of these features are then separately combined in each channel of the digital hearing aid, at the flick of the software's "quick-fit" option.

As mentioned in Chapter 7, a drawback of this major advance is that so much seems buried under the software surface. Clinicians who do not take the time to understand the rationales behind the software's settings are in peril of becoming mere technicians who just push buttons and who don't understand "why." This is why it is important for clinicians to actually listen to these hearing aids in various environments.

Compare the sound quality of different digital hearing aids when they are set up to meet the fitting requirements of some specific hearing loss. Clinicians are urged not to completely rely on the marketing claims made by the manufacturers; rather, take the time to check them out with your own ears. The cochlea is an excellent acoustic analyzer.

In addition to the digital implementation of compression, this chapter also covers other features that were not normally found on analog hearing aids. Digital hearing aids offer the following list of advantages over their analog counterparts, and these are each discussed below, in the following order in this chapter:

☑ In situ audiometric testing

■ Channels and bands

☑ Automatic feedback reduction

☑ Digital combinations of compression

☐ Expansion

☑ Digital noise reduction (D.N.R)

The main advantage of D.S.P over analog circuitry is that various compression schemes or combinations—as well as any of the above features—can be incorporated numerically, without many of the physical constraints of analog circuits. First, however, it is important to contrast digital from analog signal processing per say.

"Digital" Versus "Analog"

All hearing aids that do not have a digital circuit, or a D.S.P core, are analog. For hearing aids, the term “analog” means that the patterns of electrical current in the circuit are analogous (similar) to those of the acoustic (sound) input. The acoustic waveforms emitted from the speaker (receiver) are in turn, analogous to the patterns of electrical current. It is also important to remember that just like the incoming and outgoing sound waves, the electrical patterns produced by the analog circuit are continuous (not discrete pieces of numerical information).

When thinking of analog, picture the process of music being played from record albums on a stereo turntable. The needle wiggles in the grooves on the record. These wiggle patterns are converted into analogous patterns of electricity, which are then amplified. These amplified patterns of electricity are then turned back into analogous sound waves by the speakers (which are like microphones in reverse). Figure 8 to 1 (top) shows a very simple schematic of an analog hearing aid circuit compared to that of a digital hearing aid circuit (bottom).

Note that the analog circuit includes two different kinds of energy: acoustic (sound) and electrical (voltage and current). Both analog and digital hearing aids share the fact that they all have microphones and receivers, known as transducers. Transducers simply change energy from one form into another. Microphones change sound into electricity and receivers (speakers) changes electricity back into sound.

At the amplifier stage, gain is added to the input. The sum total electrical current or voltage is sent on to the receiver, where it is converted back into sound. On either analog or digital hearing aids, these transducers are actually analog components.

As Figure 8 to 1 (bottom) shows, digital hearing aids have an additional transduction process; after the sound is transduced into electricity by the microphone, an analog-to-digital (A/D) converter changes the electrical current into binary sequences of numbers (or digits). These can be manipulated in almost any way possible to provide the gain or other digital processing instructions that are needed for someone's hearing requirements. Once these D.S.P algorithms have been executed (once the binary digits have been manipulated), the numbers are then changed back

Analog and Digital Hearing Aid Circuits

into electrical current by a digital-to-analog (D/A) converter. This current is then transduced back into sound by the receiver.

Whenever reading about digital hearing aids, one often encounters the term “algorithm,” which is simply a series of mathematical instructions. Numbers lend themselves easily to complex manipulation, and it this ability that sets digital hearing aids apart from their analog forebears. Digital circuitry allows the frequency response of the hearing aid to be even more flexible than that of the analog, multi-channel wide dynamic range compression (W.D.R.C) hearing aids described in Chapter 7. With D.S.P, the fre frequency response across multiple channels can be exquisitely shaped to meet the fitting method targets as closely as possible.

Additional digital algorithms provide other unparalleled flexibility and adaptability for difficult listening environments that cannot be accomplished with analog circuitry. For example, some binaurally worn digital hearing aids can communicate with each other to provide as to enhance “natural” listening behaviors. Some digital hearing aid can actually “learn” about the environments most encountered by the client, and can then adapt accordingly. These specific and very advanced features will not be covered here as they lie far beyond the scope of this chapter.

The basic thing to remember about digital hearing aids is that sound is represented and manipulated by separate or discrete (not continuous) numbers or digits. When reading about digital hearing aids, various terms are encountered: two of these are “sampling rate” and “quantization.” These refer to the way a digital circuit converts a continuous analog signal into a sequence of discrete numbers. Sampling rate and quantization, respectively, basically refer to how the frequency and the intensity of sound become represented by numbers.

Figure 8 to 2 shows the basic concepts behind sampling rate and quantization. The sampling rate is how often the digital circuit samples the amplitude of the analog signal, per some unit of time. In other words, the sampling rate is the frequency of sampling (seen as the horizontal axis of Figure 8 to 2). If the sampling rate is fast, these times between the samples taken are very small. A digital circuit with a fast sampling rate thus “samples” the sound more often as the sound changes over time than does a digital circuit with a slower sampling rate.

The higher the sampling rate, the greater the ability for the digital circuit to accurately represent high frequencies of sound with numbers. High frequencies have very short periods or cycle times for each wavelength to occur. To represent these short wavelengths accurately with numbers requires a fast sampling rate that can represent these sound waves over very small units of time. This used to be a challenge in the early days of digital technology but is now one that today's digital hearing aids have long since overcome.

Quantization is the ability of the D.S.P circuit to accurately represent the sound intensity (Figure 8 to 2, vertical axis). Quantization is the assignment of numbers to the samples of sound,

Figure 8-1 summary: This figure is a schematic diagram illustrating two different types of hearing aid signal processing architectures. The diagram compares an analog signal processing path with a digital signal processing path, tracing the flow from input sound through a microphone to an output receiver. In the analog path, the acoustic signal is converted to an electrical signal, amplified, and then converted back to sound. In the digital path, the electrical signal is converted into binary numbers via an analog-to-digital converter, processed by a digital signal processor, and then converted back to an electrical signal via a digital-to-analog converter before reaching the receiver. The figure demonstrates that while both systems share the same input and output components, they differ fundamentally in how the signal is manipulated, with the digital system introducing a numerical processing stage that allows for algorithmic modifications compared to the direct amplification used in the analog system.

Sampling & Quantization

where the numbers represent voltages or current levels. Quantization thus generates a stream or series of numbers that represent the sampled signal intensity level. As with the sampling rate, more quantization permits more accuracy of intensity representation by numbers. A sound wave that can be assigned thousands of possible intensity values is far more accurately represented than the same sound wave when assigned, say, 100 possible intensity values.

In the former case, each of a vast number of different specific intensities can be assigned a specific numerical value. In the latter case, intensities that are located between any one of the 100 possible values will have to be rounded up or down to the closest value.

A high amount of quantization thus enables low distortion and a high dynamic range; in other words, very soft sounds as well as loud sounds can be accurately captured or represented digitally. A high sampling rate together with many possible quantized values is like a fine-toothed comb: It permits a higher resolution or a more accurate numeric representation of the sound. Obviously, a high degree of quantization and a high sampling rate are preferred, but these come with a cost—namely, high-power consumption. A complete discussion of digital hearing aids, however, is beyond the scope of this book (and also the author's knowledge).

Figure 8-2 summary: This figure consists of two side-by-side line plots. The left panel illustrates a continuous sound wave with vertical lines indicating the points where the signal is sampled over time. The right panel displays the resulting digital representation of that sound wave, appearing as a stepped waveform. The comparison shows that the continuous analog signal is converted into a discrete digital format through sampling and quantization, where the sampling rate determines the frequency of data points and quantization determines the precision of the amplitude values. Consequently, the digital signal approximates the original smooth wave but consists of discrete intervals and levels.

In Situ Audiometric Testing

One thing not commonly mentioned about digital hearing aids is the fact that they can sometimes produce as well as receive sounds! For example, in situ behavioral thresholds can be measured; this means a behavioral audiogram can be produced with the hearing aid in place in the client's ear canal. The hearing aid gain and output can then be adjusted to best meet target as specified by some fitting method. An example of in situ audiometry is given in Chapter 10, where we will discuss the Adaptive Dynamic Range Optimization hearing aid.

In situ audiometric testing overcomes the necessity to manually transfer data back and forth from the audiometer in decibel H.L to the hearing aid in decibel S.P.L. Direct in situ measures reduce the necessity for transforms from behavioral thresholds obtained under headphones, to 2-cc coupler data, to probe tube real ear measures. In other words, in situ thresholds can remove the seams from the hearing aid fitting process. Digital hearing aids that can produce sounds are also capable of self-diagnostic testing to detect processing problems in the hearing aid itself.

In situ audiometry is not intended to replace audiometric thresholds obtained for diagnostic purposes. For example, bone conduction thresholds are not obtained with in situ audiometry, and so the type of hearing loss is not being ascertained. Instead, the focus of in situ audiometry is to arrive more precisely at the required gain/output of the hearing aid while it is being worn. Kuk (2012) reminds us that the traditional audiogram and the in situ audiogram will likely look quite similar if both were to be measured in the same coupler. Otherwise, they might not. The reasons why not might include the actual volume of air between the hearing aid and the client's eardrum, the venting effects, hearing aid tubing effects, and so on. All of these may well serve to affect the actual thresholds of the client reason(s) with the hearing aid in place (in situ) and thus make them different from thresholds obtained with insert headphones calibrated with a 2-cc coupler or under circumaural headphones that are calibrated with a 6-cc coupler.

In situ audiometry can also be used with the fitting software of the hearing aid in question for the purpose of fitting verification. In the event that real ear measures are not available, in situ audiometry can be very helpful in arriving at a conclusion that the fitting has accomplished what it originally set out to do.

Channels and Bands

Whereas high-end analog hearing aids are always restricted to two or three frequency bands or channels, digital hearing aids are known to have in excess of 20 bands or channels! Here is a good point to distinguish between “bands” and “channels.”

In analog hearing aids, the term “frequency band” was used interchangeably with the term “channel.” These bands or channels were often separated by filtering from other frequency regions. In Chapter 7, and in Figure 7 to 10, this filtering was referred to as “band splitting.” In each band, only the gain—and consequently, the output—was adjusted. It has become industry convention to posit that if only the gain is adjusted, each frequency region is called a “band.” This is just convention; there is no official definition of “bands” and “channels.” So, if the whole point is shaping the frequency response, then one might call such a hearing aid a “multi-band” system.

In today's digital hearing aids, if more than just the gain is adjusted within a band, then the band is referred to as a "channel." Additional adjustable features might be the D.N.R, feedback reduction, and so on. Adjacent frequency bands are often grouped together to share a common set of digital algorithms that adjust these additional features. Each of these groups of bands then would be called a “channel.” For example, a group of low-frequency bands may be programmed to provide mild-gain W.D.R.C and D.N.R (both to be discussed later in this chapter), while the remaining high-frequency bands might be programmed to provide moderate-gain W.D.R.C with no D.N.R. In the final analysis, it is possible to have more bands than channels, but not the other way around. From now on, we will simply use the term “channel.”

The decibel/octave slope of the individual channels in digital hearing aids can also be much steeper than in analog hearing aids, and this further increases fitting flexibility. The steepest analog slope found in hearing aids was 24 decibel/octave (as found with Gennum's DynamEQ circuit, mentioned in Chapter 7). Although this permits a relative degree of independence between adjacent channels, it does not allow complete independence. With D.S.P, however, the slope between adjacent channels can be almost infinitely steep in comparison. A very steep slope between channels permits an even higher degree of fitting flexibility to accommodate all kinds of shapes or configurations of hearing loss. By adjusting the gain and output of each channel, much like the buttons on a stereo equalizer, the frequency response can be literally “sculpted” to meet the fitting method target (or targets) as closely as possible (Figure 8 to 3).

Then again, a very different approach can be taken. The slopes can be deliberately made to be much more shallow and asymmetrical, especially for the low-frequency channels. Furthermore, the bandwidth of the channels can be constructed to be wider and wider with increasing frequency. This has sometimes been done in order to imitate the naturally overlapping bandpass filters that exist along the basilar membrane of the cochlea itself, as described by Yost (2006). These bandpass filters of the cochlea are often highlighted in textbooks to illustrate its tonotopic arrangement. These adjacent cochlear filters are also asymmetrical in shape, much like the psycho-physical tuning curves as described in Chapter 3. Interestingly, no specific advantage in resultant speech recognition has been shown to be had with narrow, steeply sloped versus wide overlapping channels.

How many channels is best? Recall our previous discussion about “diminishing returns” regarding compression ratios in

Many Frequency Bands / Channels

Change gain in each band... until desired frequency response is achieved Chapter 7: There we described how there is larger gain reduction when going from a 1:1 to a 2:1 ratio than when going from a 10:1 to a 20:1 ratio. A similar analogy can be drawn when considering the optimal number of channels in a digital hearing aid.

First, it should be noted that subjective differences in sound quality may be found even among hearing aids with the same number of channels. This may simply be because the channels are implemented specifically for the purpose of shaping the frequency response by one manufacturer versus for the purpose of maximizing client listening comfort by another (Clark, 2006). Objective measures of speech recognition, however, are another matter. According to Clark (2006), there is a statistically significant improvement in speech recognition when going from one to two channels and from two to three channels, and so on. Once the number exceeds five or six channels, however, no such improvement seems to continue to increase with increased numbers of channels.

It is also well known that increased numbers of channels require additional digital processing complexity, which in turn will cause an actual time delay in processing the signal. Any frequency response processing necessarily requires processing time, even in analog circuits. Digital systems, however, always take more time to accomplish the same processing than do analog systems.

Time delay in processing a signal is thus something that occurs in all D.S.P hearing aids, and it is measured in the order of milliseconds. Such time delays are sometimes referred to as “group delays.” Time delays of a few milliseconds are generally acceptable. In the author's experience with an early D.S.P prototype, however, time delays over 10 ms, for example, resulted in an aided “clap” being heard immediately after the actual clap occurred!

The optimal number of channels might also depend on the audiometric configuration of the hearing loss. Gain adjustments whereby to match fitting method targets, while at the same time managing feedback, are improved when increasing the number of channels from four to at least seven. In particular, an increased number of channels will likely improve frequency-specific audibility for those with steeply sloping audiograms, while this may not be the case for those with a flat hearing loss configuration. For precipitous losses, they pin the number of eight to the minimum number of channels generally required for optimal speech recognition. Furthermore, fast-acting W.D.R.C with eight channels would better serve to improve audibility of high-frequency speech sounds and hence better speech recognition in noise. They suggest that increased numbers of channels might also enable D.N.R to reduce noise in narrow frequency regions where speech is not predominantly present, although they do state that this has yet to be proven by research. More on compression, feedback reduction, and D.N.R will be discussed in later sections of this chapter.

At this point, it is important to mention a “channel-free” digital hearing aid that has been available for over 10 years by now—namely, the Symbio produced by Bernafon. Many clinicians have asked exactly what this means; specifically, what is the difference between having one channel and being channel free? Technical explanations for the difference here abound, but few clinicians truly understand them (including the author). At the risk of stumbling into a quagmire of jargon, an attempt will be made here to sketch out a palatable explanation of the concept of “channel free.” The specific purpose of this little discussion is to highlight the various types of D.S.P architecture that can be found in today's digital hearing aids.

In the beginning of the digital hearing aid era (1996 to 1997), digital hearing aids were construed and developed as digital versions of their analog counterparts (as described in Chapter 7). In other words, their architecture was constructed so as to digitally imitate the analog properties that came before them. We have already said that in analog multichannel hearing aids, the microphone sent the input to a band splitter (filter), which separated the input into frequency-specific bands. Thus, the typical analog multichannel hearing aid had a low-and a high-frequency band, and the gain in each band was separately adjusted by means of compression and so forth. After being separately processed, the final results in each channel were then recombined together.

This end product was sent out to the receiver, which transduced the electrical product into an acoustic product for “consumption” by the human ear. Concepts like this were readily understood by clinicians who are accustomed to viewing adjacent frequencies along the audiogram. So, when digital technology began in the hearing aid manufacturing sector, it was far easier to re-create their analog “ancestors” in digital form and then simply add to them as technology advanced.

In hindsight, however, an electrical engineer purist would not have taken things along this route. Most hearing aid manufacturers, however, did. They did so because of natural hearing aid evolution. “It was the custom, and therefore it was correct.” This, in and of itself, is testimony yet again to what was said at the outset of Chapter 7—namely, it is very important for aspiring clinicians to understand the advances made in analog hearing aids before digital technology can truly be appreciated!

Most digital hearing aids construct their channels to operate in the “frequency domain,” employing a fast Fourier transform (F.F.T). The F.F.T acts as a digital filter to split the incoming sound.

It analyzes its frequency content and splits the incoming signal into separate parallel frequency bands. Once the gain, compression, and so forth in each of these bands have been separately manipulated according to the digital algorithms of the hearing aid, the separate contents of each parallel frequency band are then reunited and sent out to the receiver to be transduced back into acoustic form. Adjustment of separate individual channels in these F.F.T systems is similar to that of the typical equalizer, with buttons that represent frequency, as shown in Figure 8 to 3.

In contrast, the “channel-free” technology is said to operate in the “time domain” and does not use an F.F.T to separate incoming sound into separate frequency channels. Instead, the wideband input is taken as it is, and adjustments in gain are made extremely rapidly over time. The desired output frequency response—based on the hearing loss, the fitting method and any other selected option provided by the fitting software—is programmed into the channel-free hearing aid. A quantized value is assigned to each new input sample over time, in accordance with specific output demands that are placed upon it, so as to achieve the desired output frequency response. Each new sample thus quantized is added to all the other samples that have been previously quantized, in order to constantly update the entire output frequency response over time.

One can think of the channel-free hearing aid as an equalizer operating over time (like that shown in Figure 8 to 3), where the buttons adjust sound over three dimensions: amplitude, frequency, and sharpness. It enables frequency response shaping by updating the frequency response very rapidly over tiny, serial units of time. This technology would thus enable a “holistic” sculpting of the output frequency response without an apparent channel division.

An advantage for digital hearing aids operating in the time domain is that they present with comparatively very little processing time delay; thus, the gain added to the input produces a minimum of distortion to the output temporal waveform envelope. Furthermore, with channel-free processing, there is less spectral distortion that can occur between adjacent channels.

Both channel-free and the more typical multichannel fast-acting W.D.R.C (see Chapter 7) digital hearing aids have been compared for subjective preference and also for objective speech recognition. For subjects with no previous hearing aid experience, Plyler et al. (2013) found no significant differences in subject performance with channel-free versus a seven-channel W.D.R.C hearing aid, on the Hearing in Noise Test hint and the Abbreviated Profile of Hearing Aid Benefit aphab. Interestingly, individual subjects did have definite preferences for either the channel-free or for the seven-channel W.D.R.C hearing aid. Plyler, Hedrick, Rinehart, and Tripp (2015) compared performance among experienced hearing aid wearers with channel-free versus the same seven-channel W.D.R.C hearing aid. No statistical difference in consonant recognition in quiet and in noisy listening conditions was found between the two methods of D.S.P. The investigators also found no significant difference in subjective sound quality preference between the two D.S.P schemes. A third finding of their study was that previous experience wearing hearing aids did not seem to play a part in the objective performance or subjective preference findings.

In summary, there are always trade-offs to be considered when engineering new digital hearing aids. Again, it behooves clinicians to listen to digital products with their own ears and compare sound quality among digital hearing aids before automatically adhering to the claims of the manufacturers who build them. The high-end digital hearing aid from any one specific manufacturer may not necessarily offer the best sound quality.

Sometimes, the simpler products, with fewer frequency bands and fewer bells and whistles, in fact can sound quite good! Usually, good old straight linear gain (when not distorted by peak clipping) can also sound quite clean and clear.

Automatic Feedback Reduction

Feedback is one of the major complaints of hearing aid wearers. The classic problem has been that the client turns down the volume of the hearing aid to reduce the feedback; of course, the result is that he or she can then no longer hear effectively. Feedback is caused when aided sound outputs from the receiver. leak out of the ear canal, either through a slit leak because the hearing aid is loosely fit, or through a vent. These amplified outputs can then be picked up by the microphone and passed through the hearing aid again. Think of moving close to a speaker while speaking into a microphone. At a certain point, amplified sound from the speaker (as well as your voice) will enter the microphone, only to be amplified again. The amplified output is simply being re-amplified, and the continuous loop of this event is feedback. This sound loop then becomes repetitive until the cycle is somehow stopped, either by reducing the gain or the input.

Bentler, Mueller, and Ricketts (2016) point out that hearing aid feedback is always occurring to some extent or another, but it becomes audible only when the output that leaks out to the microphone is greater in intensity than the ambient room noise. They also note that feedback is likely to occur most in quiet listening situations, as this is when W.D.R.C will be providing the most gain!

Feedback can become especially audible when a hand or telephone is placed close to the aided ear. Any sound leaking from the receiver back to the microphone is increased in this way, because it is actually guided to the microphone by bouncing off the hand or the telephone receiver. Feedback is associated with large, sharp peaks in the mid-to high-frequency response of the hearing aid output (Figure 8 to 4). Besides being annoying, feedback can also drive an amplification system into its M.P.O and thereby reduce battery life.

Automatic feedback reduction (A.F.R) has always been a big advantage of digital hearing aids ever since they first emerged in the late 1990s. Feedback in analog hearing aids had to be handled by means of nonelectronic acoustic modifications, such as filters. These consisted of tiny screens situated inside the ear hook of B.T.E's or just in front of the receiver in I.T.E's. In terms of acoustic impedance to the passage of sound, filters offer resistance, which is not frequency specific. As such, filters reduce the peaks in a frequency response wherever the peaks may be. As tiny screens, however, filters could easily clog up, due to wax, perspiration, and so on. All too often, the presence of a clogged filter meant a functionless hearing aid or else a very compromised output.

Automatic Feedback Reduction

The only other alternative was the draconian solution of simply reducing high-frequency gain/output. Of course, this worked against the end goal, which is the provision of exactly those high frequencies. Digital hearing aids, however, routinely provide A.F.R algorithms, which go a long way to reduce feedback with less compromise to high-frequency gain/output.

A.F.R plays an even larger role now, due to the fact that most hearing aids are presently those of the ever-popular, behind- the-ear, open-fit or receiver-in-canal R.I.C styles. Since these are often fit with stock open domes inside the ear canal, venting can be as large as 4 millimeters in diameter. While large vents like these might deal well for the occlusion effect, high-frequency output would also certainly find its way to escape out of the ear canal and cause feedback. This is largely why A.F.R is a major feature in digital hearing aids. It has served to extend the usage of open-fit and R.I.C-style hearing aids past the range of mild-to-moderate hearing loss.

A.F.R is generally accomplished by first sensing and then reducing the narrow, sharp peaks that appear in the hearing aid's output frequency response. Hayes (2003) and Parsa (2006) have written concise articles reviewing some techniques and challenges in reducing feedback in digital hearing aids today. The initial digital approach to feedback reduction was the use of notch filters.

These simply reduce the gain in a narrow frequency range and thus can reduce the sharp feedback peaks in the output frequency response. If the feedback peaks always occur at the same narrow band of frequencies, then a static notch filter set to reduce the gain at the same narrow band of frequencies would work very well. This approach would create very little increase in battery consumption and digital processing demands.

The trouble is that most feedback peaks are not always so stable. Feedback is a dynamic issue; changes in the listening environment create changes in the locations of feedback peaks.

“Roving” notch filters can address this problem to some degree, but they also consume more battery and digital processing power. In their heyday, roving notch filters tended to be limited to reducing a maximum of three feedback peaks, because the usage of more such notch filters would compromise the frequency response. In addition, the roving notch filters set to operate on some three peaks would require some time (about 200 ms) to lock onto the offending feedback peaks.

These days, phase canceling has become the norm. With phase cancellation, the feedback peak (or peaks) are detected and then reverted into opposite phase. This opposite phase signal is then used to cancel out the feedback.

The phase canceling itself is done at the microphone stage of the feedback. Phase canceling is a potentially powerful tool whereby to suppress feedback peaks. Like the roving notch filter approach, however, phase canceling often can require increased battery and digital signal processing power. Phase canceling also faces the challenges posed by the necessity of tracking feedback peaks that are constantly changing. Again, feedback is a dynamic phenomenon and it can take some time for digital phase-cancellation algorithms to respond quickly to changes in feedback frequency.

Ricketts, Johnson, and Federman (2008) point out that there is a wide variation among hearing aid manufacturers in the effectiveness of their various implementations of A.F.R. Specifically, the A.F.R among six different manufacturers resulted in some 7 to 16 decibel of added high-frequency gain before feedback occurred. The variation in A.F.R results extends along another dimension as well—namely, among individuals. Thus, the most effective average A.F.R may not act in the same way on all clients.

Regarding A.F.R, it is important for clinicians not to “throw the baby out with the bathwater.” The A.F.R offered on fitting software can at times be overly aggressive, resulting in too much high-frequency gain reduction. Bentler et al. (2016) suggest doing real ear measures in order to determine the effectiveness of the A.F.R taking place.

Figure 8-4 summary: This figure consists of two line graphs showing frequency response and gain. The top panel illustrates an output frequency response where a specific peak in the mid-to-high frequency range is identified as the cause of feedback. The bottom panel displays the gain across various frequency bands, showing how a specific band corresponding to the feedback peak can be targeted for gain reduction, as indicated by a dotted line. It can be inferred that reducing the gain in the specific frequency band associated with a peak eliminates feedback. Furthermore, the figure suggests that hearing aids with a higher number of narrow frequency bands can more precisely reduce feedback while preserving overall high-frequency gain compared to devices with fewer, wider bands.

Digital Combinations of Compression

As said in the previous chapter (Chapter 7), digital hearing aids simply combine all sorts of compression types that were found separately on yesterday's analog hearing aids. Always remember those building blocks of compression discussed earlier in Chapter 7; they are fundamental to understanding compression in today's digital hearing aids. Readers of Chapter 7 may recall the left-hand input/output (I/O) function shown in Figure 8 to 5; it illustrates one way that compression is often adjusted, and it is the way that many clinicians understand compression.

The compression ratios hinge to the right from a single knee-point. Higher compression ratios here result in decreased gain.

Readers will also recall the I/O function shown in the right panel of Figure 8 to 5. It depicts a new way in which compression began to be adjusted, with the advent of those high-end two-channel analog hearing aids of the late 1980s and early 1990s.

Different Ways of Adjusting Compression

As we have seen, outer hair cell (O.H.C) damage, along with its resultant loudness growth requirements, and the emergence of W.D.R.C all really played a big part of hearing aid development some 20 to 25 years ago. Notice the compression ratios hinge from a right-most knee-point. Here, a higher compression ratio results in increased gain. The point here was to restore normal loudness growth, as seen in Figure 7 to 12, in the previous chapter.

Digital hearing aids often combine both of these types of knee-points, so that the I/O function for any channel now routinely has at least two knee-points (Figure 8 to 6). As with all previous I/O functions, the greatest amount of gain is seen below—or to the left of—the left-most knee-point. Here, the gain is linear. Some

Two Kneepoints, with Linear, W.D.R.C, Output Limiting

times even greater than linear gain can also occur; this is called “expansion,” and it is an alternative to the linear gain shown here. Expansion will be covered in the next section of this chapter. W.D.R.C, with its low compression ratio, occurs between the two knee-points.

Above—or to the right of—the right-most knee-point, output limiting compression (O.L.C) occurs with its high compression ratio. The purpose here is to limit the M.P.O for high-input sound levels.

All manufacturers provide software for the fitting of their digital hearing aids, but the frequency response (not I/O functions) is usually the main focus. This is largely because frequency response is the most readily understood display by most clinicians, Adjustments in hearing aid gain/output are typically done by way of altering some portion of the frequency response. Read ers should know that when doing so, underneath the scene on the computer monitor, the I/O function for each channel is also being altered. I/O functions sometimes comprise a section of the fitting software, where the results of knee-point and compression ratio adjustments can actually be seen. These, however, are usually buried beneath the more immediately displayed frequency responses. The savvy clinician who digs them up, can also adjust gain/output by separately and independently adjusting the knee-points, which in turn adjusts the compression ratios.

Raising the left-most knee-point vertically increases the compression ratio for soft-to-moderate level input sounds, which increases the gain for these sounds (Figure 8 to 7). This accom- plishes the same thing as increasing the compression ratio as shown in the right panel of Figure 8 to 5. Moving the left-most T.K to the left does nothing to the compression ratio itself, but it does have the effect of increasing the gain for very soft input sound levels. This is an alternative to lowering the knee-point with the T.K control, as discussed in Chapter 7.

Raising the right-most T.K vertically has the effect of decreasing the compression ratio (Figure 8 to 8). This increases the gain, but only for the more intense sound inputs; the most obvious effect is to increase the M.P.O. Moving the right-most knee-point to the right again has no effect on the compression ratio, but it does increase the gain for the most intense input sounds. These adjustments of the right-most knee-point (raising it vertically or moving it horizontally) together accomplish the same thing as adjusting the M.P.O with the output limiting compression control, as discussed earlier in Chapter 7.

The digital difference (from analog) here is that the vertical and horizontal adjustments can be done separately and independently. What's more is that these adjustments can be done for any particular knee-point displayed in the I/O function. Always remember that I/O functions do not show frequency per say. As mentioned in Chapter 7 (section on Bill and Till), unless otherwise stated, ansi electroacoustical measures typically associate the I/O functions with 2000 Hz. In some hearing aid software, however, separate I/O functions can be shown for each channel.

Some digital hearing aids utilize multiple (more than two) knee-point I/O functions, and the stated purposes proposed by the manufacturers of these hearing aids are quite interesting. Figure 8 to 9 shows an example. Expansion (to be described in the next section), for example, is called “soft squelch,” and it is also shown in Figure 8 to 6. Expansion normally appears below the first or left-most knee-point, which is shown at around 25 decibel S.P.L. The main purpose of expansion is to reduce annoying audible internal hearing aid noise that is below the intensity of typical speech.

W.D.R.C appears next, between the first and second knee-points (25 to 65 decibel S.P.L). Its purpose is to specifically increase audibility of soft speech and also to hear softer sounds that are a further distance away. From 65 to about 80 decibel S.P.L, however, the gain becomes once again linear! Past 80-decibel S.P.L inputs, the compression ratio is then dramatically increased to provide O.L.C in order to limit the M.P.O.

Let's look more closely at that "second" position of the linear gain in the multiple knee-point I/O function. This has been utilized by various hearing aid manufacturers over the past decade, as a means whereby to provide "extra" gain for average to slightly greater than average inputs, such as speech in a somewhat noisy environment. The reasoning here is at these levels, speech and noise are commonly mixed together. Most people generally prefer increased gain for these levels, so as to hear speech better in these more difficult listening situations. It is one solution to address

Figure 8-5 summary: This figure consists of two line graphs illustrating different methods of compression adjustment for hearing aids. The left panel depicts compression hinging from a left-most knee-point, where the input and output relationship starts linearly and then diverges into different compression ratios. The right panel depicts compression hinging from a right-most knee-point, where multiple lines with different compression ratios converge at a single high-input point. In the left panel, a linear ratio provides the most gain, while higher compression ratios result in progressively less gain. Conversely, in the right panel, a linear ratio provides the least amount of gain, and higher compression ratios lead to increased gain to better approximate normal loudness growth.

Figure 8-6 summary: This figure is a line graph illustrating the relationship between input and output levels in digital hearing aid fitting software. The graph depicts a multi-stage input-output function characterized by several distinct operational regions separated by knee-points. The initial segment represents linear gain, while a separate dashed line indicates an expansion region where gain is lower than linear. Between the first and second knee-points, the function operates via wide dynamic range compression. Beyond the final knee-point, the function enters an output limiting stage. The graph demonstrates that the system can be customized by adjusting the position of the knee-points to tailor the gain and output limits. It can be inferred that the hearing aid employs different processing strategies based on the input level to optimize sound amplification and prevent excessive output.

Figure 8-7 summary: This figure is a line graph illustrating the relationship between input and output levels for a Wide Dynamic Range Compression system. The graph displays multiple output curves that originate from a common right-most knee-point and diverge based on the vertical position of the left-most knee-point. As the left-most knee-point is raised vertically, the slope between the two knee-points becomes shallower, representing an increase in the compression ratio. Consequently, raising this point results in higher output gain for soft and mid-level input intensities. The data demonstrates that the compression ratio is determined by the relative positions of these two knee-points, and adjusting the left point vertically directly modifies the gain applied to lower input levels.

Figure 8-8 summary: This figure is a line graph. It illustrates the relationship between input and output levels for a compressor, showing how adjusting the right-most knee-point affects the maximum peak output. Multiple lines originate from a single left knee-point and diverge as they move toward the right knee-point, with different slopes representing varying compression ratios and different heights representing different output levels. The graph demonstrates that raising the right knee-point vertically increases the maximum peak output and increases the gain for mid to intense input levels without altering the compression ratio itself. Consequently, shifting this point upward results in a higher output level for intense sounds, while moving it horizontally to the right similarly increases gain for intense levels.

A Multi-Knee Point Input/Output Function

the complaint given by many clients who wear W.D.R.C hearing aids, such as, “I can hear people at other tables better than the person sitting right across from me!” Chapter 10, the final chapter in this book, will go on much further to describe the alternative usage of linear gain for various input intensity levels.

Digital hearing aids also uniquely implement various types of dynamic compression characteristics. As we discussed in Chapter 7, there are many possible variations of attack/release times, fixed or adaptive. Readers will recall that in the very beginning of the digital era (1997), the Widex Senso utilized A.V.C with its long attack/release times; on the other hand, Oticon's DigiFocus utilized syllabic detection, which is the complete opposite (short attack/release times).

Today, most digital hearing aid fitting software tends to offer two dynamic compression types: syllabic compression and average detection. Recall from Chapter 7 that syllabic compression is most successfully used along with low-frequency W.D.R.C (or Bill). Average detection provides adaptive attack and adaptive release times and has no specified frequency of usage. Many quick-fit options of digital hearing aids default to the use of syllabic compression for the low-frequency channels and average detection for the high-frequency channels.

Most clinicians do not tend to adjust or make changes to the default dynamic compression characteristics. In some instances, there is no adjustment possible, while in other instances, some adjustment is possible. Regarding dynamic compression characteristics, the “jury is out”; that is, there does not seem to be a round consensus as to the “best” set of attack/release times. The complexity of the matter is complicated further when considering what set of attack/release times might be best for various different types and degrees of hearing loss. Considering the interaction between dynamic and static aspects of compression (as discussed in Chapter 7) and also the incredible complexity of each manufacturer's digital algorithms, clinicians are often strongly ill-advised by the manufacturer to tamper with the default dynamic compression characteristics provided.

Figure 8-9 summary: This figure is a line chart illustrating the input-output function of digital hearing aid software. The graph depicts the relationship between input levels and output levels, highlighting several distinct operational regions separated by knee-points. These regions include an initial phase where either linear gain or expansion is applied, followed by a wide dynamic range compression section, a subsequent return to linear gain, and a final stage of output limiting compression. The chart demonstrates that the software allows for multiple adjustable segments to tailor the amplification. It can be inferred that this multi-segmented approach allows for precise control over different sound intensities, specifically enabling increased gain for average to slightly louder speech inputs while preventing excessive output at the highest levels.

Expansion

Expansion is the opposite of compression. On the basis of everything discussed so far—especially when considering compression, and the reduced dynamic range that results from S.N.H.L—one might wonder when this would ever be of use. Basically, expansion is a technique whereby to reduce internal microphone and amplifier noise that sometimes becomes audible to the listener in quiet. This is especially noticeable by those who have good low-frequency hearing.

Expansion actually serves to provide next to no gain for very soft input sounds, for example, 0 to 10 decibel S.P.L. It then rapidly increases the gain as inputs increase, up until the first knee-point of compression. Expansion was actually offered on an analog circuit produced by Gennum, the DynamEQ3, but the advent of digital technology precluded its use in hearing aids. Today, most digital hearing aids offer expansion.

Here's how and why it works: Figure 8 to 10 illustrates expansion superimposed on an I/O function showing typical W.D.R.C, as offered by some fictitious hearing aid. The vertical output axis is extended downward on this figure (further than usual) in order to show where the function of expansion would terminate. One can see here that the output for a 0-decibel S.P.L input would be 0 decibel S.P.L, and so the gain then would also be 0 decibel. The gain dramatically increases, however, as the inputs increase, up until the knee-point shown. Look at how expansion (the opposite of compression) provides greater than 1:1 linear gain; in the case here it has a 1:2 input/output ratio. One can see from Figure 8 to 10 that when used along with W.D.R.C, expansion thus provides maximum gain at (and only at) the kneepoint.

In an I/O function with multiple knee-points, this would be the left-most knee-point. The idea is to have this left-most knee-point set at an input level typical to very soft conversational speech, because then this soft speech is provided with the greatest amount of gain for the listener. Expansion in digital hearing aids commonly offers expansion ratios of 1:5, 1:75, or 1:2. With a 1:2 compression ratio as shown in Figure 8 to 10, for each added decibel of input, there are two decibels of added output!

As mentioned earlier, expansion is mainly used to reduce the gain for very soft, internal microphone and amplifier noise. The left panel of Figure 8 to 11 shows an I/O function for a typical W.D.R.C hearing aid. This is actually the same as the fictitious W.D.R.C hearing aid as that shown in Figure 8 to 10; however, the vertical output line is no longer extended below the horizontal input line.

Here it can be seen that straight W.D.R.C without the use of expansion would provide 40 decibel of linear gain below the knee-point. The W.D.R.C shown here provides a 2:1 compression ratio. Once again, the solid line extending downward to the left of the knee-point shows expansion when used with this W.D.R.C hearing aid.

As in Figure 8 to 10, the expansion in this example (Figure 8 to 11) gives a 1:2 input/output ratio. It is important to note that expansion is always offered only below the left-most knee-point. It's greater than 1:1 linear gain gives less and less gain for progressively softer inputs below the knee-point.

The right panel on Figure 8 to 11 shows the very same hearing aid, but this time the vertical axis shows gain and not output. Plotted this way, it can be readily seen that without the use of expansion, the W.D.R.C hearing aid here is providing a steady amount (40 decibel) of gain for all inputs below the knee-point. When the same hearing aid uses expansion, however, there is progressively less and less gain for the really soft input sounds from 40 decibel S.P.L down to of 0 decibel S.P.L. Again, with expansion, the maximum gain is seen at, and only at, the knee-point of compression. The gain increases as input sound levels increase up to the knee-point; as compression kicks in, the gain is once

Figure 8-10 summary: This figure consists of a line graph and a corresponding table of input-output values. The graph plots output levels against input levels, comparing linear gain, wide dynamic range compression, and expansion. The table provides specific examples of how input levels translate to output levels and the resulting gain for an expansion system. The data indicates that for expansion, gain increases as input levels rise toward a specific knee-point, after which the gain begins to decrease. It can be inferred that expansion acts as the opposite of compression by providing greater than linear gain for soft inputs. Furthermore, when expansion is combined with wide dynamic range compression, the maximum gain is achieved exclusively at the knee-point.

Expansion

Figure 8-11 summary: This figure consists of two line charts comparing wide dynamic range compression and expansion. The left panel displays an input-output function where the relationship between input and output is plotted. It shows that while wide dynamic range compression provides linear gain for soft inputs, expansion results in a steeper slope below the knee-point, indicating a more aggressive increase in output relative to input. The right panel displays an input-gain function, illustrating how gain changes based on the input level. In this view, wide dynamic range compression maintains a constant gain for all inputs below the knee-point before the gain decreases. Conversely, expansion shows gain that increases as the input level rises until it reaches the knee-point. The primary conclusion is that expansion provides increasing gain for softer inputs up to a maximum point, whereas wide dynamic range compression provides a steady level of gain for those same soft inputs.

Gain is reduced for inputs below 40 decibel

again reduced. In this way, W.D.R.C hearing aids that use expansion and have a knee-point of around 40 decibel provide maximum gain for soft sounds of speech but less gain for both softer and for louder input levels. Figure 8 to 11 and the very same concepts that it illustrates can also be seen in an article by Staab (2011).

The reasoning behind expansion is quite simple. If maximum gain is provided for all soft inputs from 0 to 40 decibel S.P.L, then the hearing aid is working its hardest, supplying most gain for all soft inputs equally. Maximum amplification, however, applies to both internal and external sound inputs. Straight use of W.D.R.C without expansion implies, then, that in all soft listening situations, any and all internal microphone and amplifier noise is also given maximum amplification; this results in an unwanted audibility of internal hearing aid noise.

Clients who complain that their W.D.R.C hearing aids make a “hissing” sound in quiet listening situations are apt to appreciate the benefits of expansion. These clients will also most likely have good low-frequency hearing. Recall from Chapter 7 that the focus of W.D.R.C hearing aids is to “lift the floor” of hearing sensitivity, to imitate the O.H.C's by amplifying soft sounds by a lot and loud sounds by little or nothing at all. Clinicians can thus construe W.D.R.C as an electric implementation of O.H.C action. By providing the same maximum gain for all soft input levels below the knee-point, however, clients with good low-frequency hearing will hear unwanted internal microphone and amplifier noise. By analogy, one can think of this as a case where the W.D.R.C by itself is “trying too hard” to imitate O.H.C's. Used along with W.D.R.C, expansion thus acts like an internal noise squelch feature.

In fact, some of the digital hearing aid literature from the various hearing aid manufacturers calls expansion a “soft squelch” feature. It is useful mostly for those who have mild-to-moderate S.N.H.L and would otherwise benefit from W.D.R.C.

Not everyone agrees with this summarization of expansion. Using a single-channel digital hearing aid, Plyler, Hill, and Trine (2005) found that while expansion may reduce the gain for very soft inputs below the knee-point, it also “throws the baby out with the bathwater” in that it reduces audibility of very soft high-frequency speech cues. For subjects with sloping hearing loss and for those with a more flat hearing loss configuration, expansion was found to compromise speech intelligibility for soft inputs below the knee-point, in both quiet and in noise environments. Subjective preferences for the same subjects, however, were found to be improved with the use of expansion, especially for those with the more sloping hearing loss configuration. It should be noted here, however, that the expansion used in this study was applied with a knee-point set at 50 decibel S.P.L and not at a lower input level such as 40 decibel S.P.L. Expansion applied to all soft inputs below 50 decibel S.P.L would certainly have adverse effects upon soft speech intelligibility.

Later on, Lowrey and Plyler (2007) experimented with expansion in a four-channel hearing aid, evaluating speech intelligibility performance and subjective preference when applying expansion in three conditions: to frequencies below 750 Hz, to frequencies below 1750 Hz, and no expansion at all. Subjects were found to perform better on speech intelligibility tasks when expansion was applied only to frequencies below 750 Hz—and also when no expansion was utilized—than when expansion was applied to all frequencies below 1750 Hz. Subjective satisfaction for the three conditions depended on the listening environment, but overall preference was found for limiting expansion to the low frequencies only. The authors conclude that overall, expansion should be limited to frequencies below 1000 Hz; food for thought...

Types of Digital Noise Reduction (D.N.R)

As mentioned earlier in Chapters 2 and 7, hearing aids have a twofold task. First, they provide gain for the hearing loss, and in doing so, they use compression to accommodate the client's reduced dynamic range. Second, they must also address the speech-in-noise problem, otherwise known as improving the signal-to-noise ratio (S.N.R). In so doing, almost all digital hearing aids employ the use of some form of D.N.R.

We all (even those with normal hearing) have increased difficulty listening to speech in background noise. As discussed earlier in Chapters 2 and 3, the problem is even worse for those with S.N.H.L. This is why the promise of D.N.R was met with so much hope, when it first appeared in digital hearing aids, with the Widex Senso in 1997. We have described how D.S.P is based on numerical manipulations that give rise to algorithms, or complex series of commands. As such, digital hearing aids are well suited to deliver D.N.R. The promise of D.N.R in hearing aids, however, did not necessarily turn out as planned. This is not because present digital algorithms are so poor; rather, it is because speech and noise sounds are so inextricably mixed together. The complete removal of background noise from the mixed-up speech and noise is easier said than done. The purpose of this section is to explain in general terms how D.N.R basically works in digital hearing aids. The clinical benefits of D.N.R are discussed in Chapter 9.

Hearing aids are not the first devices to employ D.N.R; it has been utilized in the military, the telecommunications industry, and so on (Bentler & Chiou, 2006; Schum, 2003a). There are various methods whereby D.N.R has been attempted. One of these is spectral subtraction. Here, the combined spectrum of the speech and the background noise mixed together is analyzed. The spectrum of the noise itself is then estimated as closely as possible. According to Levitt (2001), the spectrum of noise alone is measured during pauses in conversation. This noise spectrum is then subtracted from the spectrum of the speech mixed with the noise. The desired end result is to remove as much of the noise as possible, without at the same time removing too much of the speech.

The main limitation with this approach, however, is the broad-band width of most noise spectra and how much this width intersects with the broad or wide spectrum of typical speech. If the noise spectrum is very narrow or has several narrow bandwidths, then there is not much of a problem. Subtracting a narrow spectrum of some noise from the total broad-band speech and noise spectrum will not remove many of the speech frequencies. Most annoying or bothersome noise, however, is quite broad in spectral bandwidth. As a result, there will more than likely be quite a bit of intersection between the spectrum of the noise and the spectrum of speech. Subtracting this broad noise spectrum from the equally broad spectrum of speech will then obviously remove many speech frequencies as well. As the main idea is to remove as few of the speech frequencies as possible, this goal is not ably achieved with the spectral subtraction method.

Phase cancellation is another approach for D.N.R. Here, the exact waveform of the interfering noise—not its frequency spectrum—must be measured. Once this is done, then the phase of the noise waveform is flipped 180° in phase, so that it can be added to the original noise waveform. This would cancel out the noise. We have already seen that phase cancellation is used in some digital algorithms for feedback reduction.

Phase cancellation is the type of D.N.R used in high-end noise-reduction headphones. Here, the speech signal of interest is coming straight from the headphone transducer into the ear canal. The interfering noise, on the other hand, is arriving from the sound field outside of the headphone. Phase cancellation can be readily accomplished with these headphones because nearly the exact waveform of the noise can be captured by a small microphone situated on the outside of the headphone.

When the noise waveform this microphone is digitally reversed to be opposite in phase, it can then be added from the headphone to the existing speech and the noise inside the ear canal. The result is an effective reduction of the noise in the ear canal.

As Schum (2003a) points out, however, digital hearing aids do not have this luxury, because the sound picked up by the hearing aid microphone already consists of both the speech and the noise! If this complex waveform were to be reversed in phase and combined with the original waveform, the result would be total silence! In the situation of hearing aids then, the noise waveform cannot be separately measured from the waveform of speech. As a result, then, it cannot be reversed in phase so as to cleanly delete it from the speech and noise waveform.

Other approaches to D.N.R have also been made. Instead of trying to delete or remove the noise from speech, the speech itself can be enhanced. Some research has been done with “spectral enhancement,” although they did it with analog technology available to them at the time. It did not meet with much success.

In their experiment with spectral enhancement, Stone and Moore (1992) used an analog filter bank composed of 16 channels to produce the enhanced speech stimulus. Various channels among the 16 were used to deliberately increase the intensity of the ongoing, changing input speech spectrum, at specific frequencies where the amplitude was the greatest. In this way, they hoped to emphasize the cues used for speech intelligibility.

A total of 10 subjects with mild-to-moderate S.N.H.L were tested for their speech reception ability in the presence of continuous background noise. In one experiment, subjects wore their own hearing aids, most of which were linear. The background noise was presented at two different levels (44 decibel and 64 decibel S.P.L), and the speech was presented at a level that was 3 decibel more intense than these background noise levels.

In another experiment, the subjects did not wear their hearing aids; instead, the stimuli were given high-frequency emphasis so as to imitate the function of their own hearing aids. In the second experiment, the subjects adjusted the level of the background noise to match levels they normally found to be comfortable in their everyday lives. Speech intelligibility did not improve for Stone and Moore's subjects, and in some cases, it actually became worse! The second experiment however, showed that the subjects did have a subjective impression that speech stood out better against background noise.

Spectral enhancement is still attempted in today's digital hearing aids. Here, a D.S.P algorithm attempts to recognize cues that are specific and unique to the speech spectrum and then deliberately enhance these cues. The desired result is to improve the recognition of speech in noise, although it continues to be easier said than done.

Here's why. Spectral analysis of typical speech will almost always show greatest amplitude for the low frequencies, which are the vowels. Within this frequency region, there are cues that are specific to speech, such as the fundamental frequency, and other harmonics. There are also frequencies of emphasis that are specific to the various vowels themselves.

Spectral enhancement tries to emphasize these amplitude contrasts or differences that are specific to the various vowels sounds, so as to make them more distinct. It is precisely these low-frequency speech sounds, however, that are commonly also the easiest for those with S.N.H.L to hear!

The most important cues for discriminating speech are actually found in the softer and higher frequency consonants. A spectrum of speech in broadband background noise shows that these higher frequencies are the ones that are most easily obliterated, covered up, or masked by the noise. Those with S.N.H.L have an especially difficult time discerning the high frequencies of speech; therefore, they are especially affected by background noise when trying to listen to speech.

The trouble with the spectral enhancement approach is that it is difficult to emphasize or deliberately enhance spectral cues for these all-important high frequencies, especially as they tend to become buried in the background noise. It is much easier for digital spectral enhancement algorithms to enhance the low-frequency content of the vowels that are intense enough to usually withstand the masking of the background noise. The irony of the spectral enhancement approach, then, is that it assists the audibility of the louder low-frequency sounds that are generally easier to hear for those with high-frequency S.N.H.L. In short, it “gives to the rich,” who don't really need it in the first place.

Another attempt at enhancing speech relative to background noise is speech synthesis. In this approach, the D.S.P algorithm tries to identify specific speech sounds within the noise. When these are detected, the algorithm would then reproduce—or synthesize—those same sounds but without the noise. In other words, synthesized speech sounds are added to the detected speech sounds in order to enhance their recognition. This would require a virtually instantaneous detection of specific speech sounds within the noise, combined with an instantaneous substitution from a stored bank of the same speech sound without the noise.

The synthesized speech method of speech enhancement obviously comes with its problems. If the background noise were to be competing speech, this could wreak some unintended havoc. Then again, imagine the required complexity of a digital algorithm that could accurately accomplish the goals of the successful speech synthesis approach; namely, specific speech sound recognition (in noise), followed by an instantaneous augmentation and addition of that same speech sound (without the noise). It would be hard enough to do this for the normally louder vowels, let alone the soft, tran-zee-unt unvoiced speech sounds that are normally buried in the noise!

Noise Reduction with Amplitude Modulation

Let's come down to earth and look at the implementation of D.N.R and/or speech enhancement approaches that are actually used in today's digital hearing aids. Most D.N.R algorithms attempt to characterize the unique acoustic properties of speech versus those of background noise. This consists mainly of looking at how these acoustic properties change over time. Changes in intensity and frequency over time are compared and differentiated for speech as opposed to background noise.

Changes in intensity and frequency over time are known as “modulation”—specifically, amplitude modulation and fre- frequency modulation. The techniques utilized by most hearing aid manufacturers consist of the use of amplitude modulation detection and, to a lesser degree, frequency modulation detection. Since amplitude modulation is the biggest player in the D.N.R scene, we will look at this more closely.

Compared to most ongoing noise, speech—especially speech in quiet—tends to have fewer modulations per second; however, the modulations of ongoing speech in quiet also have a much greater degree or depth of amplitude modulation than do those of noise. The modulations for speech occur at a frequency of about 3 to 10 times per second (3 to 10 Hz), while those for a jet engine occur at a frequency of over 30 Hz. According to Mueller and Ricketts (2005), syllables are about 75 to 150 ms in length, these, together with the pauses in speech, result in about four to six sizable modulations per second.

We have already noted in Chapter 5 that the spectrum of average, ongoing conversational speech spoken in quiet has a dynamic range (greatest to least amplitude) of about 30 decibel. In terms of its waveform over time, speech in quiet has “peak-to-valley” dips or changes in intensity that are about 15 decibel “deep”. These constitute the amplitude modulations of speech. Compared to the intensity of most background noise, such as a fan, an air conditioner—and even the hubbub and babble of background speech noise for that matter—the intensity of average ongoing conversational speech in quiet thus changes quite dramatically in intensity over time (Figure 8 to 12). It is these choppy, “stop-start” acoustic properties, unique to speech, that are what most D.N.R algorithms seize upon in order to identify speech versus noise in each channel of the digital hearing aid.

The depth of amplitude modulation is greatest for speech in quiet, but is reduced for speech in background noise. In this all-too-common situation, it becomes more difficult for the D.N.R algorithm to determine what constitutes speech and what constitutes noise. In any case, some decision rule has to be adopted for use by the D.N.R algorithm in question, in order to determine the ratio of modulation that “constitutes” speech. The upshot of this approach is that the same gain is applied to a channel regardless of speech alone is present in that channel or if mostly speech and some noise are present. Although not a perfect solution, it does err on the side of preventing an excessive

Amplitude Modulation

Background Noise: Less Modulation loss in audibility. Basically then, the goal of most D.N.R algorithms is to identify the presence of speech versus noise in each channel (group of bands) of the hearing aid. If a channel is found to have an undue amount of noise, the gain in that channel is reduced.

Let's look again at Figure 8 to 12. Two sounds are shown here; the top noise sound is steady in intensity over time, while the bottom speech sound fluctuates much more in intensity over time (time is shown horizontally, and intensity is shown vertically). Noise is generally assumed by the D.N.R algorithm to be fairly steady in intensity over time, especially when compared to the fluctuations of speech, shown in the bottom panel.

One may ask, “What about background speech that you don't want to hear? That, too, may be considered as noise.” Well then, now is the time to think about the English alliterations “babble” or “hubbub”; these nouns would suggest that the intensity of background speech (as noise) is also relatively steady in intensity over time. Assume that the hubbub or babble is the relatively steady intensity of background cocktail party speech (top panel). This will be considered as “noise” by the D.N.R algorithm, compared to the fluctuating intensity of speech spoken by a person right in front (bottom panel).

Most of us do not abstractly think about the acoustics of speech as separate from the meaning of speech, but if we do so, the truly unique acoustics of speech become exposed. As mentioned early in Chapter 1, the actual acoustic pops, fizzes, stops, and sputters of close-up, one-on-one speech are difficult to appreciate unless one goes out and listens to a language he or she cannot understand. In that situation, all meaning becomes stripped and the rapid starts and stops of speech are truly present in the utter abstract to hear (dogs and cats must be laughing their heads off at us).

Regarding D.N.R, there are other matters to consider, such as the speed with which the D.N.R reacts to reduce the gain for channels in which noise is detected. This might be considered the dynamic aspects its attack and release times. D.N.R algorithms take some time to maximally reduce the gain in frequency channels where noise is sensed. According to Mueller and Ricketts (2005) and Bentler and Chiou (2006), this time can vary anywhere from 2 seconds to 20 or 30 seconds, depending on the specific digital hearing aid model and manufacturer. The time it takes to return to the original gain can be quick (5 ms) to as long as a few seconds.

Then again there is the amount by which the gain is reduced once noise is detected in any particular channel, and this can be anywhere between 5 to 20 decibel. The depth of modulation itself can be used to determine the amount of gain reduction. Many digital hearing aids produce progressively less and less gain in a channel as the sound becomes less and less modulated (more and more steady state). Most manufacturers utilize some amount of frequency modulation along with amplitude modulation. Any one particular method of D.N.R employed by a specific hearing aid manufacturer will consist of clusters or combinations of these possible parameters. There is no one specific D.N.R method that is common to all manufacturers; each one utilizes its own unique and specific D.N.R methods. Furthermore, no research has shown any one method is best for improving speech recognition in noise.

Statistical Distribution of Speech Versus Noise Intensity

How do digital hearing aids apply their amplitude modulation method of D.N.R? They use mathematics, and this involves statistics. We have seen in Figure 8 to 12 that the intensity of ongoing noise over time is relatively stable (top panel). The waveform of speech (bottom panel) is irregular in pattern; that is, the wave “envelope” shows abrupt increases and decreases in intensity over time.

These differences can also be illustrated by means of statistical distributions of intensity over time, as shown in Figure 8 to 13. The statistical distribution of typical ongoing background noise will show a relatively normal “bell curve” shape. The mean or average intensity will be in the center of the bell curve. In statistical terms, the mean will be quite similar to the median, which in turn will be quite similar to the mode.

In plain English, the mean (average) is the sum total of values that is then divided by the number of values; the median is the middle value in the list of values, and the mode is the value that occurs most often. Figure 8 to 13 shows that the mean, median, and modal intensity for most ongoing background noise are all centered in the middle of the range. This is what is referred to as a “normal distribution.”

For ongoing conversational speech, however, the statistical distribution looks very different, quite irregular in shape. This is what is known as a “nonparametric” distribution. Here, the mean is not similar to the median, which in turn is not similar to the mode. The reason for this abnormal-looking distribution of intensity over time is that speech has lots of starts and stops. In musical terms, the intensity of ongoing speech could be said to be quite staccato.

This brings us to the discussion of the long-term average speech spectrum (L.T.A.S.S) that we initiated back in Chapter 5.

Distributions of Intensity Speech versus Background Noise

There, we noted briefly that the mean or average intensity of ongoing speech is not situated in the centered or middle of its dynamic range of intensity. This is also shown in Figure 8 to 14. Due to its staccato sputters of stops and starts, and the consequent nonparametric distribution of intensity over time, the mean intensity is not situated in the middle of its general 30-decibel range of intensity. Unlike as would be the case for most background noise, the mean intensity of L.T.A.S.S is situated 12 decibel below the top of the intensity range and 18 decibel above the bottom of the intensity range. This is why figures in previous chapters,

Figure 8-13 summary: This figure is a line graph featuring two distinct distribution curves. The graph plots the percentage of time intensity remains at specific decibel levels against the sound pressure level. One curve represents steady-state noise, while the other represents conversational speech from a single talker. The noise distribution follows a symmetric bell curve where the mean, median, and mode coincide at a central peak. In contrast, the speech distribution is irregular and non-parametric, characterized by multiple peaks and a lack of symmetry. This indicates that while noise maintains a relatively stable intensity over time, speech intensity fluctuates rapidly and unpredictably, making its average intensity more difficult to determine from the distribution.

Long-Term Average Speech Spectrum

such as Figure 5 to 10 and Figures 6 to 4 and 6 to 5, do not show the mean of long-term speech intensity to be located right in the very center of the range of speech intensity.

It is precisely these very different statistical differences shown in Figures 8 to 13 and 8 to 14 that the artificial intelligence of most D.N.R algorithms utilizes to determine if the input sounds are speech or noise. If in any channel the input is so categorized as speech, the gain is left alone and not reduced. If the input fits into the categorization of noise, then the gain in that channel will be reduced. Again, the actual amount of gain reduction depends on the manufacturer and the choices the clinician has entered on the fitting software.

Come to think of it, however, with a voice kept at a constant intensity, like holding a note while humming a tone, the D.N.R algorithm just might consider that to satisfy the criteria for noise, and it will reduce the gain accordingly. This is exactly what some digital hearing software has taken into account, when fitting clients who might want a program for listening to music. In this situation, the D.N.R algorithm is usually shut off.

For optimal speech intelligibility in the midst of background noise, D.N.R should provide for the very least reduction in speech information. Here, two things would have to be the case: (1) the D.N.R algorithm would operate in each of many narrow frequency bands in a digital hearing aid, so that a reduction in gain in any channel would reduce the audibility of only a narrow band of frequencies, and (2) only a very narrow band of noise would enter the microphone of the hearing aid. This optimal situation would result again in a gain reduction over a very narrow frequency band of speech. As most can appreciate, however, this is most often not the case.

Indeed, most noise entering the hearing aid has a fairly broad frequency spectrum. Bottom line: It ain't easy.

Figure 8-14 summary: This figure is a line graph. It displays the long-term average speech spectrum, showing the relationship between frequency and sound pressure level, with a central line representing the mean average and outer dotted lines indicating the range of intensity. The graph illustrates a general trend where sound intensity is higher at lower frequencies and gradually decreases as the frequency increases. The distance between the mean average and the boundaries of the range varies across the frequency spectrum. It can be inferred that vowels, which typically occur at lower frequencies, are more intense than consonants, particularly unvoiced ones. The fact that the mean average is not centered within the intensity range suggests that the statistical distribution of speech is abnormal, likely due to the rapid and unpredictable fluctuations in speech intensity over time.

Speech Enhancement

Digital hearing aids have also utilized algorithms for detecting speech in noise that fall more along the lines of speech enhancement. These algorithms rely on the identification of acoustic properties—other than modulations—that are unique to speech. In general, speech enhancement algorithms do their job (provide more gain) when speech is present, while amplitude modulation algorithms used by most D.N.R do theirs (provide less gain) when noise is present.

One such type of speech enhancement algorithm is called “comodulation” or synchrony detection. As mentioned earlier, this involves searching the acoustic environment for cues that are specific to the speech spectrum. An example here would be the harmonic frequencies of the lowest (fundamental) frequencies of speech (in males about 125 Hz, and in females about 250 Hz). Harmonics of speech are far smaller in amplitude than the fundamental frequencies. In fact, the decibel/octave roll-off for speech leaving the mouth is about -6 decibel per octave. If these acoustic properties are found (either in quiet or in and among background noise), then more gain with less compression is provided. This added gain is given to both the speech—and of course the background noise—contained within any particular channel. If no such harmonics or speech-like properties are detected, then less gain with more compression is provided.

This speech enhancement algorithm is based on an observation that those wearing hearing aids who want to hear the speech that is spoken in noise, tend to prefer the same or even more gain in noise than they do when listening in quiet. In general, more gain is thus given when speech is present. Schum (2003b) is quick to add that this algorithm is no better at separating noise from speech than D.N.R algorithms that use amplitude modulation, but does tout its virtue in providing more listening comfort in background noise.

Two Examples of Early Digital Hearing Aids

We have looked at common digital features found in today's digital hearing aids. For historical interest, let's look way back to 1996/1997 to see the salient features of the two earliest digital hearing aids. This way, we can see how they each were very different pioneers in the utilization of many of the concepts we have looked at in this chapter, and also in the preceding chapter on compression. The Widex Senso was the very first digital hearing aid to hit the market. The author recalls well the scene at the American Academy of Audiology convention in 1997. Most clinicians saw that its biggest advance was the presence of amplitude modulation D.N.R. High hopes were placed for finally addressing the classic "speech-in-noise" problem faced by most people with S.N.H.L.

The Senso had three channels with adjustable crossover frequency controls, which could adjust the frequency where adjacent channels “meet.” These crossover controls enabled the channels to be widened or narrowed. The gain could be adjusted separately for each channel, depending on the shape of the hearing loss (where it rises or falls). Each channel had W.D.R.C geared to provide normal loudness growth for the listener.

An interesting aspect of the Senso was that the threshold knee-point of compression could be set to as low as 15 decibel to 20 decibel S.P.L (see also Figures 8 to 7 and 8 to 9). In the Senso, this low knee-point is associated with a larger-than-usual amount of gain for low-intensity input sounds—which was expansion! Widex explained that with analog hearing aids, this low knee-point could not be utilized without expansion, because excessive feedback would be generated by the large amount of gain applied to the low-intensity input sounds.

The attack/release times of the Senso were long, some several hundreds of milliseconds in length; these relatively long dynamic characteristics are the same as those provided by A.V.C (see Chapter 7). Long attack/release times were employed by the Senso, because shorter attack/release times combined with the low knee-point of compression were not well received by many initial users in their field trials. The dynamic aspects of compression remained relatively slow in a stationary noise environment, but they could speed up if a sudden, intense tran-zee-unt sound occurs.

During the initial fitting of the Senso, in situ audiometry determined hearing thresholds by emitting complex tones from the hearing aid while in the ear of the listener. These were used to determine targets for gain. Widex explained that the advantage of this method is that the results include the effects of the ear mold or the shell of the hearing aid in situ (in place) in the ear.

The Senso was the first to use D.N.R that was based on amplitude modulation. Widex explained it as an ongoing statistical method separately utilized in each of its three channels, whereby every few seconds, the speech and background noise are sampled. In any particular channel, if the overall sound waveform did not fluctuate (modulate) much over time, then it would be deemed to be noise, and the gain would be reduced.

The assumption was also made that in loud listening situations, speech is spoken louder. When a channel sensed that background noise was present, the gain for both the speech and the noise would be reduced in that channel. Because speech in noise is normally spoken at a more intense level, however, the speech would still be audible even though the gain for both the speech and the noise had been reduced. This is good, historically interesting stuff upon which to look back. Perhaps the reason the product was called the Senso was because it was always “sensing” if the input was speech or noise.

Oticon's DigiFocus came out immediately on the heels of the Senso. It had seven frequency bands. Unlike those of the Widex Senso, the crossover frequencies of the DigiFocus could not be adjusted; that is, bands adjacent to each other could not be widened or narrowed. Despite being fixed in place, the seven bands each represented a relatively narrow range of frequencies, which permitted a high degree of fitting flexibility for people with difficult-to-fit hearing loss configurations. The DigiFocus was the first to combine frequency bands into low-and high-frequency channels (the low-frequency channel composed of three low-frequency bands and the high-frequency channel composed of four high-frequency bands).

Central to the Oticon DigiFocus philosophy was the concept of “Adaptive Speech Alignment,” which was not a type of D.N.R. The goal of this stated feature was to provide aided speech that would be as intelligible as possible. This would be accomplished with accurate and specific frequency shaping with the seven frequency bands, by reducing the upward spread of masking, and by providing very different attack/release times for the low-frequency and high-frequency channels.

The low-frequency channel basically provided Bill, along with the fast attack/release times of syllabic compression. This combination was designed to reduce the upward spread of masking by the more intense, low-frequency vowels, which can obliterate softer high-frequency consonantal speech. Recall our discussion of Bill in Chapter 7 and its conceptual purpose. Most background noise is relatively low in frequency. Consider the background noise and vowels of speech as the bull and the unvoiced consonants as delicate pieces of china. The whole purpose of Bill, along with syllabic compression, is to “control the bull in the china shop.” The DigiFocus was an initial digital implementation of this concept.

The high-frequency channel provided O.L.C, along with slower acting attack/release times (what Oticon called “adaptive gain”). Consistent with O.L.C (described in Chapter 7), the gain for the high-frequency channel was essentially linear, with a high knee-point and high compression ratio, so as to limit the output from exceeding the listener's loudness discomfort levels. The attack times for the high-frequency channel were about 20 ms, with release times that could vary from about 230 to 320 ms. The purpose of this brief look back at these two pioneering digital hearing aids is to highlight just how very different philosophies of the day were incorporated into digital form. Over the passage of time, neither one of these was proven to be correct or better than the other. Consider, for example, the fact that Widex addressed audibility and speech reception with W.D.R.C and D.N.R while Oticon addressed the same with a very different approach—Bill with syllabic detection.

Consider also the A.V.C (long attack/release times) of the Senso versus the opposite—syllabic compression (short attack/release times) of the DigiFocus. Each was put forth as the best possible option by Widex and Oticon. No one went to jail on account of either one; the jury was (and still is) largely out on this one. Today, some 20 years later, many manufacturers still include various elements from both of these initial strategies on their fitting software.

Digital Hearing Aids: State of the Art and the Future

Regarding the many digital features that have been described here, it is not always a matter of whether the stated objectives are right or wrong; rather, clinicians should be aware that the digital hearing aid manufacturing sector is highly competitive. So-called white papers are routinely put out by the manufacturers so as to explain their newest features in an intellectual manner. These are often written to convey the research that took place to create the new features. In all truth, however, the research is done almost exclusively by those working at the particular manufacturer!

Manufacturers are highly secretive about the specific meth- ods they use. This is understandable, considering the highly competitive nature of the digital hearing aid market. To release details of an actual proprietary D.S.P circuit core to the general arena of hearing aid manufacturing would be giving away what likely took a lot of time and money to develop. Each manufacturer continues to develop its own type of circuit that uses very specific algorithms. As mentioned earlier, only the microphone and receivers in digital hearing aids have some remote similarity to those of analog hearing aids.

The actual D.S.P circuit core is quite unique to any one specific digital hearing aid, although manufacturers were known to sell these to other manufacturers. This happened more when digital hearing aids first appeared than it does now, because in the beginning of the digital hearing aid era, some manufacturers did not have their own digital research and development off the ground but, they still wanted to “enter the game.”

To further tickle the reader's cynical elbow, there are even times when a feature can be created out of a realization of a flaw in the design! The author recalls one manufacturer noticing a rather rounded-looking knee-point in the I/O function of one of its analog products. This product then began to be advertised and promoted as providing "curvilinear" compression. Whether it did or not was one thing (and it did); the point is that the concept of curvilinear compression did not even exist at the time when the product was created!

There is yet another patch of thorns in today's digital hearing aids: Many terms relating specifically to individual digital products are not readily understandable, because they are used nowhere else in the industry. Sometimes, when one actually finds out what some touted feature in fact does, the name it has been given has very little to do with the function the feature does. Furthermore, the different manufacturers often give similar features different names!

This is very confusing to clinicians, because they simply want to understand the concepts behind the features. An early example that comes to mind is “Adaptive Speech Alignment,” coined by Oticon and described earlier with regard to their first digital product—the DigiFocus. Ostensibly, one would never know that this basically referred to the use of Bill along with syllabic detection. A much more recent example of this is the “channel-free” Symbio by Bernafon; it has been notoriously difficult for hearing health care professionals. als to digest the distinction between “channel-free” and “single channel.” As discussed earlier, that term has been a “bone to be chewed” by clinicians for years by now. The good thing here, though, is that in getting answers, we are forced to educate ourselves further. Like any good consumer, clinicians should make it their collective business to become good consumers of the products manufacturers provide.

On the receiving end, some clinicians have been frustrated over the complexity of the fitting software when fitting digital hearing aids. The products are in some ways like closed boxes; the manufacturer may know what is going on inside the hearing aid, and clinicians are left to trust that the fitting will be satisfactory for clients. Complexity in fitting software has replaced elegant simplicity.

The automatic “quick-fit” option is the easiest route; no wonder most clinicians take it! In the author's opinion, a return to simplicity in software would be a truly welcome thing. Far too many clinicians do not routinely use real ear measures to verify that the predicted gain or output from the software is indeed taking place with the hearing aid situated in the client's ear. Without real ear verification, we stand in danger of forgetting how to fit hearing aids.

Regarding digital fitting software over the past several years, it has been interesting to note the departure from a required knowledge of compression types and other technological features on the part of the clinician and the trend toward addressing psychosocial issues of the client. Many software adjustments to compression (and other) settings for most digital products today depend on specific answers to specific psychosocial situations and other various listening conditions. The complexity of some digital hearing aid products, when coupled with this client-based focus, can produce software adjustment queries that lean toward the extreme in specificity. To highlight this point at conference presentations, the author has given this tongue-in-cheek example (also given in Chapter 5): “Do you have trouble hearing the preacher every second Sunday, when sitting at a 45 superscript circle angle to the right, in the third pew from the front? If so, push this button.”

In all actuality, many manipulations are possible with digital hearing aids. Clinicians can take comfort from fears of getting lost with too much freedom. The same fitting software that offers sometimes too many features for adjustment also does provide lots of guidance and fitting solutions on their fitting software. As clinicians, however, we cannot expect fitting software to “do our fittings.” That's our job to complete. As said at the end of Chapter 5, fitting software gets us “into the ballpark.” Verifying the fit with real ear measurement is required to complete the job.

As was mentioned at the outset of this chapter, clinicians must take the time to listen to the specific digital hearing aids they tend to recommend the most. It is one thing to be persuaded about the remarkable advances touted on glossy marketing brochures; it is quite another to have that "wow!" effect when listening to a hearing aid. The importance of this cannot be overstated. I recall distinctly meeting a student of mine with a flat, moderately severe S.N.H.L who had just acquired a new pair of high-end digital B.T.E's. Just for fun, I attached an old pair of high-power analog B.T.E's featuring O.L.C to the student's ear molds, and manually set the trimmers at mid-positions (so that I could be, at most, half wrong).

I then asked the student how those sounded. The student's answer was actually quite humorous: "Wow, where did you get these digital hearing aids? They sound so clean!" As I drove home that day, I took refuge in Occam's razor (attributed to the medieval philosopher, William of Occam), which posits that "the simplest explanation is the best one."

Aside from this rather cynical look at the present state of affairs, clients are often quite satisfied with the look and sound of their digital hearing aids, noting that they are an improvement from the beige B.T.E bananas of yore that whistled a lot. In the long run, digital advances we have mentioned do go a long way to provide increased comfort and audibility for the end user. Digital hearing aids can also reduce the internal amplifier noise commonly associated with some analog circuitry. It's not that we have everything solved here; there is always room for improvement.

Audibility has largely been addressed with compression utilized in yesterday's analog programmable, multi-channel hearing aids and in today's digital hearing aids. These developments have allowed us to make advances toward our second challenge—namely, the speech-in-noise problem faced by those with S.N.H.L. For these clients, we try to objectively increase the signal-to-noise ratio (S.N.R) and subjectively enhance listening comfort in noise. In the next chapter, we look at two specific ways of dealing with these endeavors.

Summary

Analog hearing aids transduce sound into electricity (by way of the microphone), amplify the electrical current, and then transduce this back into sound (by way of the receiver). For the most part, the microphone and receiver portions of digital hearing aids are still analog. Digital hearing aids differ from their analog counterparts in that they have an A/D converter, a central D.S.P core, and a D/A converter. Digital hearing aids thus transduce sound into electricity, electricity into digits, digits back into electricity, and finally, electricity into sound.

■ Six features typical to digital hearing aids are in situ testing, the possibility of including many more than two frequency bands or channels, automatic feedback reduction, combinations of compression types, expansion, and D.N.R.

■ In situ testing enables the testing of audiometric thresholds and subsequent fitting according to various fitting methods, all while the hearing aid is situated in the client's ear. This overcomes the necessity for transforms from 2-cc coupler data, decibel S.P.L to decibel H.L, and so on.

■ Digital hearing aids today commonly have many frequency bands; low-end digital products commonly have fewer bands while high-end products will have more. Channels were defined as combinations of frequency bands that share common digital algorithms.

■ Automatic feedback reduction digitally reduces high-frequency peaks in the output frequency response while the hearing aid is being worn. This feature has reduced the need for the physical presence of filters, which can easily become clogged up with wax.

■ Combinations of compression types discussed in Chapter 7 are commonly found in today's digital hearing aids. For example, linear gain is often provided for soft inputs, W.D.R.C for medium-intensity inputs, and O.L.C for high-intensity inputs. In order to address the observation that

W.D.R.C does not give sufficient gain, linear gain is provided also for average to slightly more than average level inputs. Input compression is often provided for soft-to medium-intensity inputs, while output compression is provided for more intense inputs.

■ Expansion is the opposite of compression and is provided for very soft inputs below the left-most (lowest) knee-point. Greatest gain is thereby provided at (and only at) the first knee-point of compression, which represents the input levels of soft speech. Expansion serves to reduce the gain for extremely soft inputs, thereby reducing audibility of internal microphone and amplifier noise. Those who benefit from expansion are clients with mild-to-moderate sloping S.N.H.L. Expansion is often included in medium-power digital hearing aids where linear gain and W.D.R.C are utilized.

■ D.N.R comes in several forms. The most common type looks at the amplitude modulation of incoming sounds and determines whether the intensity is steady in intensity over time or fluctuates rapidly in intensity over time. Steady-state intensity is determined to be noise; the gain is reduced in any channel where noise is sensed. Another type of D.N.R discussed was that of speech enhancement, which determines whether the incoming sounds have acoustic properties similar to speech. In any channel, if harmonics similar to those found in speech are detected, the input is determined to be speech; the gain is then accordingly increased and the compression is reduced. Neither of these methods works “better” than the other; they are simply different approaches.

■ Two older, first-generation digital products were reviewed, for the purpose of seeing how they each incorporated the features previously discussed. These first two digital hearing aids to appear were the Senso from Widex and DigiFocus from Oticon.

Hearing aid manufacturers are intensely competitive. Different names are often given to similar digital features. Their fitting software for digital products has

become increasingly complex. Psychosocial situations and other various listening conditions are examined, and the answers to these queries largely determine the settings for the digital products. This approach tends to cut clinicians off from their understanding of the “hows” and “whys” behind the adjustments and reinforces the impression that the 1990s were indeed the “golden” age of compression.

You have reached the end of the document.