Pure tone/broad band signals and non-linear processing

[Phorize]

So I consulted my wife last night on the purchase of a new amp.

I’m heavily leaning towards a purifi, most likely the March audio. I was rambling about the impressive SINAD performance and power efficiency over another class a/b that I had been considering. As her background is in audiological science, she mentioned the test batteries involved in hearing aid design, and in particular the use of temporally varying signals as well as/instead of puretone and steady-state broadband noise input for frequency response testing.

I'm a technical layperson so can't really draw any conclusions on this or how it relates to audio amplifier testing in general, but though it was interesting nonetheless.

This article explains some of the standards/practices in hearing aid design:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4906306/

 

[Killingbeans]

Is this about Class D vs. Class AB?

What's bugging you about the amp you have now?

 

[March Audio]

So I consulted my wife last night on the purchase of a new amp.

I’m heavily leaning towards a purifi, most likely the March audio. I was rambling about the impressive SINAD performance and power efficiency over another class a/b that I had been considering.

Saw my name mentioned. Can you clarify what your questions are?

Alan

 

[Phorize]

Saw my name mentioned. Can you clarify what your questions are?

Alan

Hi Alan,

Perils of typing on a small phone-my post was incomplete. Now edited as above.

 

[Phorize]

Is this about Class D vs. Class AB?

What's bugging you about the amp you have now?

Thanks, I sent the post before finishing it. Now edited and hopefully clear.

 

[Phorize]

One question that occurs to me is, if hearing aid designers working on the non-linear processing of complex signals (such as the human voice) for hearing aid development don't rely on steady state broadband signals for measurement, do audio engineers have similar test standards-if not why not? As I said, I'm a scientific/engineering layperson-it seems intuitive that a very linear amplifier wouldn't require such as test, but with class d (and again I'm broadcasting my ignorance here) are we talking about an element of non linear processing that would require more complex signal to be presented during testing to get a clear measure of how the amplifier would behave "in the wild"?

 

[NTK]

One question that occurs to me is, if hearing aid designers working on the non-linear processing of complex signals (such as the human voice) for hearing aid development don't rely on steady state broadband signals for measurement, do audio engineers have similar test standards-if not why not? As I said, I'm a scientific/engineering layperson-it seems intuitive that a very linear amplifier wouldn't require such as test, but with class d (and again I'm broadcasting my ignorance here) are we talking about an element of non linear processing that would require more complex signal to be presented during testing to get a clear measure of how the amplifier would behave "in the wild"?

As you have stated in the first sentence of your post (post #6), hearing aids perform non-linear processing of the audio signal, and therefore their behavior is intentionally non-linear. The audio equipment we discuss in this forum are supposed to function linearly. (I'll leave those devices that add non-linear "euphoric distortions" to a separate discussion.)

When a device behave linearly, we can "extrapolate" (or interpolate) the measurement results to different operating conditions (see superposition principle). Think of linear as in a straight line. Once we have established the line with measurements inside a limited area, we can extrapolate outside of that area. Therefore, we can find out a lot about its performance with a small number of relatively simple test signals.

Of course, we must limit ourselves to only extrapolate within the expected area of linear operation (e.g. power limits, frequency limits, etc.). For these devices, we want to any nonlinearities to be insignificant, and we identify them with nonlinear measurements such as THD and IMD. And we measure output power limits and frequency bandwidths to find out the boundaries of the linear operation area.

If the device is nonlinear, we can't extrapolate without knowing exactly how these nonlinearities behave — which is rare. I.e. if we don't know how a (nonlinear) curve curves, we can't extrapolate by measuring just a small section of it. Therefore, when we want to characterize the performance of the device, we need to measure the device using measurement signals as close to the actual operating conditions as practical — therefore speech signals for the hearing aids, instead of steady state test tones.

 

[March Audio]

One question that occurs to me is, if hearing aid designers working on the non-linear processing of complex signals (such as the human voice) for hearing aid development don't rely on steady state broadband signals for measurement, do audio engineers have similar test standards-if not why not? As I said, I'm a scientific/engineering layperson-it seems intuitive that a very linear amplifier wouldn't require such as test, but with class d (and again I'm broadcasting my ignorance here) are we talking about an element of non linear processing that would require more complex signal to be presented during testing to get a clear measure of how the amplifier would behave "in the wild"?

Well I think we are talking about 2 different things. I'm no expert in this particular area but my thoughts.

Psychoacoustics where you are looking at how the sound is perceived by the subject. This will probably be more complex in people with hearing loss. The incoming sound will need to be specifically processed to achieve maximum intelligibility and benefit for those individuals. As such testvsignals may need to be specific and targeted.

With hifi electronics we are primarily interested in passing the incoming signals with as little modification as possible. Lowest noise, distortion etc. The use of various test signals often steady state will generally adequately characterise this performance. More complex signals are also used though if you want to test certain things or look at data in a different way.

Speakers are maybe a bit different and I will often look at the psychoacoustic weighting results for example.