Why feedback control should be disabled for music programs – Part I

Marshall Chasin
May 8, 2012

When all is considered, one can truly say that the real advantage of digital hearing aids is feedback management allowing the hearing aid wearer to achieve more gain before feedback is encountered.  There is scant evidence about the benefit of any of the other “advanced features” in hearing aids.  And other than feedback management (and perhaps a greater bandwidth) hearing aids of the late 1980s were as good as those that we have today.

Acoustic feedback, as the name suggests, is caused by the leakage of amplified sound from the wearer’s ear canal back to the microphone.  This leakage is further amplified until it becomes audible. Robyn Cox and her colleagues had studied this in detail in the 1980s and had even cautioned against “sub-audible feedback” where some of the deleterious effects of feedback are observed even prior to being audible to anyone.  Dr. Cox was able to demonstrate that the peaks associated with feedback started to become more peaky prior to even hearing it, thereby causing what may have been a smooth etymotic response to look more like a spiked hair style- without the red and blue tint.

Simply stated, acoustic feedback results in constructive and destructive interference between the leaked sound returning to the hearing aid microphone and the initial sound.  Constructive interference will occur (and will generally result in audible feedback) when these two sounds are in phase- they simply add up.  Destructive interference will occur when these two sounds are significantly out of phase, resulting in a notch or loss of energy in the frequency response.

Engineering during the 1970s and 1980s understood this and even tried to use all pass filters in hearing aids.  The “all pass” characteristic of the filter would not affect the frequency response, but depending on the filter that was used, there was a change in the phase- these filters were “nonlinear with respect to phase”.  Changes in the bandwidth could be made with a trim pot that would reduce the higher frequency cut off from 12,000 to 10,000 Hz-  a difference that made no difference at all- but the phase was altered which could help the hearing aid fitter to get several decibels more of gain prior to feedback.

The digital era brought in three “easy-to-implement” feedback control approaches.  These were notch filtering, automatic reduction of gain, and phase cancellation.  Most hearing aids today use a form of phase cancellation where the feedback signal is detected and an identical signal is generated by the hearing aid except that it is 180 degrees out of phase.  The first two approaches actually worked quite well, especially the “automatic reduction of gain”.  Phonak was an early adopter of that approach where the entire frequency response was reduced in gain whenever feedback was encountered.  This approach maintained the frequency response so that the balance of music- the relative intensities of the fundamental and its harmonics were also maintained.  In chatting with some of the Phonak engineers, this is no longer used and they, along with most of the hearing aid industry has gone over to phase control.

Phase control is not without its problems as well.  In some environments the generated sound that is 180 degrees out of phase with the acoustic feedback is heard after the offending signal has ended.  This isn’t really a problem with speech because of the neurological and mechanical limitations of our vocal tracts- speech cannot begin and end as abruptly as is the case with many musical instruments.  A generated-180 degrees- out of phase signal may still be heard after a staccato note or one that stops suddenly.  Many musicians report that they hear a “chirp” in some parts of their music.

One can easily measure an acoustic feedback signal in the real ear during probe tube measures and also in a 2 cc coupler during the hearing aid testing according to ANSI 3.22.  It is observed as a narrow bandwidth peak that is out of line with the magnitude of the rest of the frequency response.  It is definitely “un-speech like” in the sense that the peaks of speech tend to have a shallower spectral slope than acoustic feedback (which can be almost indistinguishable from a pure tone).  However, it can be very similar to a fundamental or a harmonic of a musical instrument.

This was noted in the earlier first generation of digital hearing aids in the early 1990s.  The Oticon Adapto comes to mind where it simply turned itself off when certain musical instruments, such as the flute, were played.  Oticon responded quickly (as did Siemens) and in their next generation of hearing aids restricted their feedback control systems to only those frequencies above 1500 Hz.  The lower frequency sounds of many musical instruments were not deleteriously affected by the feedback management system.

However, do we really require feedback management systems when playing or listening to music? And that will be part II of this blog… stay tuned for next week!

Leave a Reply