jkeny

Measurements without psychoacoustic relevance - how useful?

Recommended Posts

Measurements used as guides to accessing whether engineering goals have been achieved are essential but when it comes to evaluating the performance of a device when playing a music signal, they tend to be far less useful. Why?

Two factors spring to mind:

  • the use of simplistic test signals which bear little relation to music signals & exercise the device in different ways than when it is handling music signals.
  • the interpretation of measurements without taking into consideration the hearing mechanism & psychoacoustics of auditory perception.

  Although I believe it is only a part of the picture, Harmon/Sean Olive have done some work in showing certain aspects of speaker/headphone measurements & their correlation to listener preference & hence to psychoacoustics 

Edited by jkeny

Share this post


Link to post
Share on other sites

They're pretty good at telling you when a thing is broken. And when it's mended again. Is that "an engineering goal" or is it "the performance of a device when playing a music signal" ? I think you make some valid points in the text of your post. But the thread title could come across as provocative in this particular forum.

VB

  • Upvote 1

Share this post


Link to post
Share on other sites
Super Wammer

If the purpose of an amplifier or signal source is to render an undistorted output from an input, then whether there is a listener there or not is irrelevant. The measurements tell you everything you need to know about how the item works. 

Loudspeakers are somewhat different in that the interaction with the room is also important, but if the overall in-room response is measured, then that provides a decent enough idea of how well the combination works. 

What no measurement will do is to determine whether YOU will like the results, but I suggest that's totally irrelevant to the engineering of a system.

S

Share this post


Link to post
Share on other sites
2 minutes ago, SergeAuckland said:

If the purpose of an amplifier or signal source is to render an undistorted output from an input, then whether there is a listener there or not is irrelevant. The measurements tell you everything you need to know about how the item works. 

But where this line of logic falls down is that ALL amplifiers distort - the resultant distortion is judged to be irrelevant if it is judged that we can't hear it - so the listener is very much relevant to this evaluation.

Quote

Loudspeakers are somewhat different in that the interaction with the room is also important, but if the overall in-room response is measured, then that provides a decent enough idea of how well the combination works. 

What no measurement will do is to determine whether YOU will like the results, but I suggest that's totally irrelevant to the engineering of a system.

S

Yes, what I was referring to in Harmon's measurements was the big take away that an off-axis smooth frequency response was the engineering target  which correlates to listening preference & is very much tied up with reflections & room response. So this is an example of engineering goals being very much relevant to listening preference i.e whether you like the result. Same applies to their target curves for headphones  

Share this post


Link to post
Share on other sites
Super Wammer
1 minute ago, jkeny said:

But where this line of logic falls down is that ALL amplifiers distort - the resultant distortion is judged to be irrelevant if it is judged that we can't hear it - so the listener is very much relevant to this evaluation.

Yes, what I was referring to in Harmon's measurements was the big take away that an off-axis smooth frequency response was the engineering target  which correlates to listening preference & is very much tied up with reflections & room response. So this is an example of engineering goals being very much relevant to listening preference i.e whether you like the result. Same applies to their target curves for headphones  

Yes but amplifier distortion is so low, that it can be taken as zero for all intents and purposes. Any amplifier that has audible distortion is no longer an amplifier but an effects box.

As to loudspeakers, it depends on whether the purpose is to make loudspeakers that are accurate, or that impress on demonstration. The two are not necessarily the same.

S

Share this post


Link to post
Share on other sites
24 minutes ago, SergeAuckland said:

Yes but amplifier distortion is so low, that it can be taken as zero for all intents and purposes. Any amplifier that has audible distortion is no longer an amplifier but an effects box.

But by what criteria do you evaluate "distortion is so low"? Does this evaluation refer to listener's ability to hear such distortion or are you using some empirical engineering measure of "so low"?

I would refer you to the initial post which specifies that the test signals used are not exercising the device in the same way as when it is handling music signals so how do we know that the distortion with these test signals is actually the distortion produced with music signals.


I made a post in the other thread about multi-tone test signals as a step in the right direction towards more relevancy to music signal structure. i also made the point that the workings of the hearing mechanism & the psychoacoustics of auditory processing handles single tones differently to mixed tone signals.

Quote

As to loudspeakers, it depends on whether the purpose is to make loudspeakers that are accurate, or that impress on demonstration. The two are not necessarily the same.

I'm not sure what your point is here as the same could be said about any audio device?

Edited by jkeny

Share this post


Link to post
Share on other sites
Super Wammer
29 minutes ago, jkeny said:

But by what criteria do you evaluate "distortion is so low"? Does this evaluation refer to listener's ability to hear such distortion or are you using some empirical engineering measure of "so low"?

I would refer you to the initial post which specifies that the test signals used are not exercising the device in the same way as when it is handling music signals so how do we know that the distortion with these test signals is actually the distortion produced with music signals.

Music signals are just a random selection of single tones, as are ALL audio signals, so existing tests for harmonic and intermodulation distortion adequately characterise the performance under music conditions. The Belcher test is one such way.

S

Share this post


Link to post
Share on other sites

I'm trying to have a sequential discussion with you but you keep jumping around which makes it difficult

You stated that listeners were irrelevant to measurements, yet the evaluation of all measurements involve listeners, including the Belcher test whose introduction very much makes subjective listening & thus listeners at the heart of the reason for the test. I'm not really following your logic here? 

Quote

A new method of measuring non-linear distortion of sound signals is described. This method uses a test-noise signal, based on a pseudo-random sequence and occupying most of the audio band, in conjunction with comb-filter techniques to measure distortion over a wide frequency range; it gives better correlation with subjective test results than the conventional total harmonic distortion method.

 You stated that "amplifier distortion is so low" & I asked by what criteria you judge "low"

The Belcher test you mention is an example of multi-tone tests that I said were a step in the right direction but how often do we see measurement results from such tests? 

Without the results from such multitone tests does your statement hold up that "existing tests for harmonic and intermodulation distortion adequately characterise the performance under music conditions"?

You stated that "Music signals are just a random selection of single tones, as are ALL audio signals" but does that mean that using single tone test signals will test the device in the same way as complex music signals?  

Is our auditory perception of single tones the same as our perception of complex tones? 

Edited by jkeny

Share this post


Link to post
Share on other sites
Super Wammer
8 hours ago, jkeny said:

I'm trying to have a sequential discussion with you but you keep jumping around which makes it difficult

You stated that listeners were irrelevant to measurements, yet the evaluation of all measurements involve listeners, including the Belcher test whose introduction very much makes subjective listening & thus listeners at the heart of the reason for the test. I'm not really following your logic here? 

 You stated that "amplifier distortion is so low" & I asked by what criteria you judge "low"

The Belcher test you mention is an example of multi-tone tests that I said were a step in the right direction but how often do we see measurement results from such tests? 

Without the results from such multitone tests does your statement hold up that "existing tests for harmonic and intermodulation distortion adequately characterise the performance under music conditions"?

You stated that "Music signals are just a random selection of single tones, as are ALL audio signals" but does that mean that using single tone test signals will test the device in the same way as complex music signals?  

Is our auditory perception of single tones the same as our perception of complex tones? 

I really don't care what our auditory perception is, as those investigations were done many many years ago, and our hearing hasn't evolved in the past 50+ years, so I am taking them as still valid.

If distortions of all sorts, at all frequencies and loads within specification are below 0.1%, then they are inaudible. What the actual threshold is depends on the harmonic structure and frequency, but I've not seen any credible suggestion ever that distortions of that level are audible. 

As any amplifier of any competence has distortion levels well below that, I don't concern myself with amplifier distortion.

S  

Share this post


Link to post
Share on other sites
Moderator

Interesting question John.

The people at MQA have said that there has been a lot of papers on pyscho-acoustics that look at our current understanding of what is relevant to how audio sounds. Their fundamental question was is distortion of timing information more important than the current focus on frequency distortion. The current idea that if you have great high frequency response then all the timing information is preserved may not be true. I have only seen one of the papers showing peoples sensitivity to timing information was far greater than believed. I think it was Williamson many, many years ago introduced the idea that THD, frequency response and noise were the key parameters. They may be for engineers making amplifiers but they do not correlate with the sound quality. If that was true then every solid state amp should sound the same as most SS amps (and a few valve) amplifiers measure perfectly using these parameters.

So do we  ignore the enormous data from people who say that they hear differences or take a more open minded approach and research what parameters are important for the sound quality of amplifiers? We can shut the door and tell a lot of people they are deaf or deluded or conned or determine whether we are measuring the right things. If frequency distortions are important but do not now help with determining sound quality of current audio then what measurements will help? Certainly a more realistic test signal will help. A better understanding of what humans listen for and how that can be characterised by measurement would be useful. Although there are some research papers being produced on human perception of audio I am not aware of audio companies doing or supporting similar work to allow a closer correlation between sound quality and measurements.

I would be interested in any of the audio perception papers but do not have access to them. One paper that did interest was the finding that the brain is 'connected' to the ear with twice as many nerves sending signals to the brain than those that receive data from the ear. The brain seems to be sending information to the ear to adjust how we hear/interpret audio. Can our hearing of audio become more acute by experience and is that able to overcome any frequency response losses? Many older conductors who have lost high frequencies are able to perceive, identify and ask for corrections to very subtle changes to instruments or their playing.     

Share this post


Link to post
Share on other sites
Admin

I'd like to listen to it on my own . No demo . I heard it at Birmingham and i'd say subtle re MQA

Share this post


Link to post
Share on other sites
Super Wammer

agreed , i heard it too at brum show 

Share this post


Link to post
Share on other sites
Super Wammer
Measurements used as guides to accessing whether engineering goals have been achieved are essential but when it comes to evaluating the performance of a device when playing a music signal, they tend to be far less useful. Why?
Two factors spring to mind:
  • the use of simplistic test signals which bear little relation to music signals & exercise the device in different ways than when it is handling music signals.
  • the interpretation of measurements without taking into consideration the hearing mechanism & psychoacoustics of auditory perception.
  Although I believe it is only a part of the picture, Harmon/Sean Olive have done some work in showing certain aspects of speaker/headphone measurements & their correlation to listener preference & hence to psychoacoustics 

There are oodles (that's a unit of measure) of threads in this forum which address the limitations or not of measurement. They can easily become binary and usually end up with people politely agreeing to differ. Sometimes light is shed along the way!

I'm intrigued as to what has prompted you to post this now as opposed to any other time in the last few years. Have you heard something different and which measurements cannot explain? Or something which sounds the same but measures differently? Have you just come across the Harmon paper? The context of your post may or may not help me and others to understand where you're coming from. Thanks in anticipation!

Share this post


Link to post
Share on other sites
1 hour ago, SergeAuckland said:

I really don't care what our auditory perception is, as those investigations were done many many years ago, and our hearing hasn't evolved in the past 50+ years, so I am taking them as still valid.

OK, so you don't want facts to interfere with your story - fact is that the understanding of auditory perception has improved & continues to improve to this day. Burying your head in the sand is not a very objective approach to measurements - do some research into Auditory Scene Analysis

2 hours ago, SergeAuckland said:

If distortions of all sorts, at all frequencies and loads within specification are below 0.1%, then they are inaudible. What the actual threshold is depends on the harmonic structure and frequency, but I've not seen any credible suggestion ever that distortions of that level are audible. 

Please define what "distortions of all sorts" means & detail what test signals were used to measure these distortions.         

2 hours ago, SergeAuckland said:

As any amplifier of any competence has distortion levels well below that, I don't concern myself with amplifier distortion.

Indeed!

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


  • Recently Browsing   0 members

    No registered users viewing this page.