New Electroacoustic Tests for Hearing Aids Having Non-Linear Signal Processing

by Larry Revit, hearing scientist

What appears here is a version of a poster that was displayed at hearing-aid conferences in 1991.


Traditional electroacoustic tests fall short of adequately describing the performance of hearing aids having non-linear signal processing. Several new tests, now available in clinical hearing-aid test equipment, more completely describe the performance of such hearing aids. These new tests include:

  • Families of gain- and output-versus-frequency response curves using wideband, “speech-like” test signals;
  • Input/output curves, input/gain curves, and attack/release-time tests, using variable test-signal frequencies, wideband, speech-like signals, and stimuli of varying duration;
  • Frequency response curves using a pulsed, wideband signal (simulating speech) in the presence of a continuous bias tone (simulating noise).


This poster will describe examples of these tests and their applications to evaluating “multi-band-compression,” “ASP,” “adaptive compression,” and normal “AGC” circuits. Other proposed tests, now available only in the laboratory, will be discussed, as well.





  • Give useful information about hearing aid performance, but fall short of adequately describing non-linear signal processing.
  • The curves above are of an instrument with an adaptive low-frequency filter (often called “ASP”).
    • The smoothness and deep low-frequency roll-off of the SSPL(OSPL)-90,
    • The linearity of the 2000-Hz input/output curve (through 75 dB SPL input),
    • The negligible attack and release times,




  • Shows that most of the action of this “ASP” adaptive filter occurs for inputs between 70 and 80 dB SPL RMS.
  • Curves for inputs of 80 and 90 dB SPL show the effects of intermodulation distortion: jaggedness of response, low-frequency region “filled in” with distortion components.
  • This hearing aid likely sounds garbled and harsh whenever inputs are high enough to activate the low-frequency filter (such as when the wearer is speaking).All these are misleading results. (Other tests in this poster give information that is likely more useful to the hearing-aid fitter.)




  • Shows how AGC, ASP and other time-varying non-linear circuits perform during their attack and release phases.
  • Above are two examples of defective release circuits.




  • Shows attack and release phases of an adaptive low-frequency (“ASP”) filter.
  • Traditional ANSI (2000-Hz) test signal (upper graph) is too high to test this low-frequency circuit.
  • Graph of output versus time shows smooth attack/release characteristics for a 500-Hz pure-tone signal (lower graph). But the relatively fast attack time (90 ms) means this circuit will probably react to speech, causing “pumping” or “breathing” sounds.
  • Above, an adaptive low-frequency filter “waits” about 600 milliseconds before beginning to react (total attack time: 1075 ms). This circuit, therefore, should respond to ongoing background noise, but should not respond to speech.
  • This circuit should exhibit far less “pumping” and “breathing” than similar circuits having shorter attack times (such as in the previous example).




  • A test for signal-processing circuits that work in a specific frequency range (such as “ASP” low-frequency filtering).
  • The above 300-Hz input/output curves (both of the same hearing aid) show gain reduction that would not be seen with a higher-frequency test.
  • Of special interest in the above tests, is that they are different only in the delay time from when the stimulus started to when the output was measured. A long delay (2.2 seconds, lower graph) was necessary for accurate results, because of the long attack time of this circuit (1.075 seconds).




  • Shows different release times for stimuli of different durations.
  • Above is a test of a hearing aid that has an adaptive release time (e.g., Phonak, Telex, K-Amp). For a long stimulus (2 seconds), the release time is long (593 milliseconds); for a short stimulus (100 milliseconds), the release time is short (95 milliseconds).
  • The purpose is that transient stimuli (such as dishes clanking) should not interrupt communication, whereas longer, but gapped stimuli (such as ongoing speech) should not cause raised levels of the background noise in the gaps.




  • The above test shows two of the seven possible frequency responses of a Starkey “Spectral Digital Scanner” circuit. The Spectral Digital Scanner responds differently to pulsating signals (such as speech) than it does to continuous signals (such as ongoing background noise).
  • In the above test, a “speech-in-noise” signal was a simulated by a pulsating composite tone (speech simulator) accompanied by a 500-Hz pure tone (noise simulator). A “speech-in-quiet” signal was simulated by a pulsating composite tone alone.
  • The presence of the noise-simulator caused the frequency response to change from broadband (curve with no symbols) to high-frequency-emphasis (curve with rectangle symbols).
  • A “ZETA Noise Blocker” circuit would show a similar result.




  • Predicts in the test box, what the real-ear benefit of a hearing aid should be for the average ear.
  • The above family of speech-weighted composite tests of a “K-Amp” hearing aid indicates a smooth, gently sloping insertion-gain response for low-level inputs. For a 90-dB-SPL input, the estimated insertion-gain response is flat and gives zero gain.
  • Compare the above estimated insertion-gain curves to the 2cc-coupler curves for the same device shown below (for inputs of 60, 70, 80, and 90 dB SPL).





  • Is a measure of the portion of the output that is determined by the input.
  • The maximum value for coherence is 1.
  • Noise and distortion lower the coherence.
  • Above, the upper graph is a frequency response made with random noise, whereas the lower graph is a coherence response made at the same time. An 80-dB-SPL input drove the device into saturation. Thus, the coherence was less than one.

The above coherence and frequency responses were done in the laboratory of David Preves, PhD., of Argosy Electronics




  • Although the new, wideband tests described above are the best for characterizing the general performance of hearing aids, the pure-tone SSPL/OSPL-90 test remains the best test for maximum output.
  • In the graph above, the pointy, solid trace is a spectral recording of the whistling sound of an office FAX machine announcing the arrival of a transmission. The dashed line is a saturation test using a wideband composite tone at 90-dB-SPL. The upper trace (rectangle symbols) is a pure-tone SSPL/OSPL-90 test.
  • The pure-tone SSPL/OSPL-90 test has accurately represented the maximum potential output of the hearing aid, whereas the composite-tone test has not.


With the exception of the coherence test, all the tests presented in this poster were done on a production-model FONIX 6500 hearing aid analyzer.

Back to Larry’s Corner Table of Contents

Leave a Reply

Your email address will not be published.