Mikek, I was referring to serious designers. You know, the people who design real products which are successful in the marketplace.
I am being rather short with those who can't seem to study simple aspects of audio distortion for themselves, and continually ask the same questions that have been asked for decades. I realize that some of you are new at this, but try to understand that we addressed these questions decades ago. Yes, we had many of the same questions, but we have found that there remain differences in amplifier designs that still defy our complete understanding.
However, one factor I will just about bet on, and that is the presence of higher order 5th, 7th, 9th harmonic distortion at normal playing levels of almost ANY electronic amp, and you will have a disappointing design. This was stated in 1941 in the 'Radiotron Designers Handbook' as well, but we now allow even LESS distortion these days.
Virtually all Class B designs will generate higher order harmonic distortion. Some will have more, and some less.
Most amplifiers today, are usually high enough in slew rate to not suffer that problem in a severe way, BUT that was not always the case, AND our primary measuring tool in the 50-70's was the IM analyzer, not harmonic distortion measuring equipment. Good oscillators were not yet developed at a reasonable price until the mid '70's, and even then IM was the preferred measurement method. ONLY TIM measurement showed the weakness of IM measurement. I was there when Matti Otala gave our paper at the AES and we were questioned by Crown (politely, I might add) as to why IM did not measure TIM. An analysis of the SMPTE IM test can show that it is insensitive to rate-of-change distortion, but none of us realized it before this time. In fact, the presence of the 7KHz carrier signal seemed to imply that the amp's sensitivity to that frequency would be noted in the measurement, but it wasn't so. Oh well, enough of history.
The main thing is that no one, even today, has every answer to any area of audio design, and there is no one test that will reflect the overall quality of an audio component.
I am being rather short with those who can't seem to study simple aspects of audio distortion for themselves, and continually ask the same questions that have been asked for decades. I realize that some of you are new at this, but try to understand that we addressed these questions decades ago. Yes, we had many of the same questions, but we have found that there remain differences in amplifier designs that still defy our complete understanding.
However, one factor I will just about bet on, and that is the presence of higher order 5th, 7th, 9th harmonic distortion at normal playing levels of almost ANY electronic amp, and you will have a disappointing design. This was stated in 1941 in the 'Radiotron Designers Handbook' as well, but we now allow even LESS distortion these days.
Virtually all Class B designs will generate higher order harmonic distortion. Some will have more, and some less.
Most amplifiers today, are usually high enough in slew rate to not suffer that problem in a severe way, BUT that was not always the case, AND our primary measuring tool in the 50-70's was the IM analyzer, not harmonic distortion measuring equipment. Good oscillators were not yet developed at a reasonable price until the mid '70's, and even then IM was the preferred measurement method. ONLY TIM measurement showed the weakness of IM measurement. I was there when Matti Otala gave our paper at the AES and we were questioned by Crown (politely, I might add) as to why IM did not measure TIM. An analysis of the SMPTE IM test can show that it is insensitive to rate-of-change distortion, but none of us realized it before this time. In fact, the presence of the 7KHz carrier signal seemed to imply that the amp's sensitivity to that frequency would be noted in the measurement, but it wasn't so. Oh well, enough of history.
The main thing is that no one, even today, has every answer to any area of audio design, and there is no one test that will reflect the overall quality of an audio component.
"subjective sensitivity of the human ear to distortion artifacts"
How should there be any subjection to distortion, if the distortions
are below audible threshold ? The ear has a good resolution, but
still is limited. We are not dogs...
And THD... Again this weakness to optimize for sinus-signals...
Music is not a testtone, it's a complex summary of many sin's/cos's.
For example an amp having dynamic frequencydependent delay,
it does not show distortions for a pure sinewave coming from this
problem. But now feed a complex signal, the amp applys different
phasehifts to different frequencies, you can't speak of precise
reproduction. And the human ear is VERY sensitive to phaseshifts,
most of the ability to locate sounds 3dimensional comes from this
sensitivity. (Phaseshifts by the distance of the ears, the 2 "signals"
have a timedelay)
I heavily studied this sensitivity when programming a 3dsoundengine,
so these statements are not some believing...
What i wanted to show, is the fact that an amp might have 0.0% thd
with a pure sinewave, but still "distort" with complex signals !
That might be a reason, why thd seems so useless to describe
the quality of an amp.
Mike
How should there be any subjection to distortion, if the distortions
are below audible threshold ? The ear has a good resolution, but
still is limited. We are not dogs...
And THD... Again this weakness to optimize for sinus-signals...
Music is not a testtone, it's a complex summary of many sin's/cos's.
For example an amp having dynamic frequencydependent delay,
it does not show distortions for a pure sinewave coming from this
problem. But now feed a complex signal, the amp applys different
phasehifts to different frequencies, you can't speak of precise
reproduction. And the human ear is VERY sensitive to phaseshifts,
most of the ability to locate sounds 3dimensional comes from this
sensitivity. (Phaseshifts by the distance of the ears, the 2 "signals"
have a timedelay)
I heavily studied this sensitivity when programming a 3dsoundengine,
so these statements are not some believing...
What i wanted to show, is the fact that an amp might have 0.0% thd
with a pure sinewave, but still "distort" with complex signals !
That might be a reason, why thd seems so useless to describe
the quality of an amp.
Mike
Well designed, conservatively operated conventional speakers do produce distortion but it's mostly low order nonlinearities resulting from mechanical operation and it only rises with level. These are readily masked, especially with increasing SPL.
But, paradoxically, the masking nature of human hearing is such that nonlinearities are more detectable at low and medium SPLs than at high levels and so, quite remarkably small SPL higher order nonlinear artifacts originating from the audio chain behind the speaker are detectable by listeners.
I think figures 10 and 11 here
http://www.phon.ucl.ac.uk/home/shl9/rosen/rosen1.htm
illustrate what I'm getting at.
http://www.phon.ucl.ac.uk/home/shl9/rosen/fig10.gif
http://www.phon.ucl.ac.uk/home/shl9/rosen/fig11.gif
(There's more but that's the money part for our purposes.)
The shape of hearing filters show masking is more extensive on the high side of the masker and that our hearing's gain falls off with increasing SPL.
And we might infer, (at least I would since I'm reckless), that higher order nonlinearities of low frequency origin and very small SPLs may well fall outside the masking area and thus may be audible at normal listening levels.
The last time I put this sort of idea out here someone started posting those nice articles by Mr Toole - which it was a pleasure to read again - but the issue here right now is something like this: What levels of distortion are audibly meaningful and what produces them?
If someone can hear higher order nonlinearities being played through a good speaker sensibly operated, it's very likely the speaker is not their origin.
MikeB said:
There is an exactly defined relation between amplitude and phase characteristics. And - once you have impulse response (derivate of step response), you can calculate response to any combination of sinusoidal signals. The amps phase shifts are nothing compared to speaker's and to phase shifts of sound waves at listeners sitting place. And I am a bit suspicious when I am reading about phase shifts without any closer specification.
I was talking about dynamic delays. The delay from speaker to listener
is very high compared to the amp, but is constant. (Reflections ignored,
assuming good roomacoustics)
My point is, that you can't use a single step response and apply it
to all freqs.
And yes, i am talking of delays below 1uS.
For example, you have a mixture of 1Khz and 5khz. A bad amp now
for example has a delay of 200nS at 1Khz and 250nS at 5khz.
The signal at output is not identical to the one at input. It's not distorted,
but definitely different.
And yes, speakers have their own delay. But a proper designed
speaker has constant groupdelay at critical freqs.
I think i must write a program applying these delays via fft to a wave,
then we know if this effect is responsible for amps having bad separation/details.
is very high compared to the amp, but is constant. (Reflections ignored,
assuming good roomacoustics)
My point is, that you can't use a single step response and apply it
to all freqs.
And yes, i am talking of delays below 1uS.
For example, you have a mixture of 1Khz and 5khz. A bad amp now
for example has a delay of 200nS at 1Khz and 250nS at 5khz.
The signal at output is not identical to the one at input. It's not distorted,
but definitely different.
And yes, speakers have their own delay. But a proper designed
speaker has constant groupdelay at critical freqs.
I think i must write a program applying these delays via fft to a wave,
then we know if this effect is responsible for amps having bad separation/details.
john curl said:
More indefatigable hand waving from the omnipotent John Curl...
john curl said:
Questions, dear John, and answers from whomsoever can provide them, are the raison d'etre of a forum such as this.....
john curl said:
john curl said:
Did you now!!!...
Come now Mr Curl!....
There is no shame in admitting you do not have all the answers...... john curl said:
Speak for yourself John...
I think Jan summerized your attitude rather accurately here:
http://www.diyaudio.com/forums/showthread.php?postid=481878#post481878
Moreover:
http://www.diyaudio.com/forums/showthread.php?postid=481861#post481861
http://www.diyaudio.com/forums/showthread.php?postid=481869#post481869
http://www.diyaudio.com/forums/showthread.php?postid=481877#post481877
Now John, do you have an answer to the question i posed at the bottom of this thread....?
http://www.diyaudio.com/forums/showthread.php?postid=491291#post491291
If you do not know, that believe you me, is perfectly alright...
The world as we know it will not come to a grinding halt merely because the Omniscient John Curl knoweth not the answer to some obscure facet of amplifier design...
So..fear ye not old chap...
Just steer well clear of the juvenile limb-waving.....it does you no favours at all.....
FrankWW said:
I disagree.....Swedish chef asked a legitimate question here....
http://www.diyaudio.com/forums/showthread.php?postid=491144#post491144
...i posted my considerd view to this, agreeing with jcx's analysis here:
http://www.diyaudio.com/forums/showthread.php?postid=489927#post489927
...and here:
http://www.diyaudio.com/forums/showthread.php?postid=491090#post491090
This was John's response:
http://www.diyaudio.com/forums/showthread.php?postid=491207#post491207
http://www.diyaudio.com/forums/showthread.php?postid=491396#post491396
FrankWW said:
How so??
You see Frank, i do not generally kow tow to posters merely because they invoke 'vast experiance' as sufficient grounds for us lower life forms to accept their assertions as bible truth without justification....
If this sort of 'yes my lord John Curl' approach is acceptable to you, that's your choice...
By the way, do you have an answer to my humble query....Your Lordship?
Folks, let me give some clues here as to what has happened in audio over the years.
First, about 1/3 of a century ago, my associates and I personally went to visit the late Richard Heyser about the question: "Why is it that loudspeakers, phono cartridges, and analog tape have so much harmonic distortion (lower order of course) that we still hear differences in our electronic amplifier designs, when they can be so low in measured distortion?" Richard Heyser, who was then developing TDS measurement, told us that it appeared that GLOBAL NEGATIVE FEEDBACK was the problem. Could he prove it at the time? No! But experience with negative feedback of his designs and those of others, showed this to be true. Actual harmonic distortion measurements, and even SMPTE IM measurements could be fairly lousy, YET an amp could still sound OK, in many cases. BUT many amplifiers that apparently measured well, could sound lousy. Obviously, we had much to learn about distortion measurement.
I went the direction of Otala, in making linear, high open loop, discrete designs.
A few years later, working with Matti Otala and personally making hundreds of measurements, myself, we found that SMPTE IM was most useful for crossover distortion (higher order harmonics at listening level) but useless for TIM (SID) distortion measurement. We were looking for a SINGLE measurement to denote amplifier quality. We tried noise loading, harmonic, SMPTE IM, CCIR IM, sine-square TIM (which we developed), and anything else that we could find.
Ultimately, we could find NO one test would give us complete understanding of an electronic component. This was in 1976, a long time ago.
Walt Jung, independently researched SID (TIM) with extended harmonic distortion measurements. The problem was that existing THD measurement equipment had fairly lousy extended distortion measuring capability, because the IC's used in the test equipment could not remain low distortion at extended frequency to 100KHz, which was necessary. Both Walt and I found that upgrading our ST THD harmonic analyzers with better IC's, that we could make acceptable measurements in order to see TIM (SID, which is the same thing, but it took us years to reconcile this).
Today, the best single measurement approach would probably be a really low distortion harmonic analyzer with discrete distortion harmonic resolution or an FFT spectrum analysis, after the initial THD measurement.
Still, Barrie Gilbert's article shows a unique FM distortion that would NOT be measured by the AM sensitive THD measurement. This FM distortion is NOT fixed by negative feedback, although the AM distortion component generated by the same source has been reduced by negative feedback. This is what makes Barrie Gilbert's analysis significant and useful. It shows a potential distortion that is NOT removed by global negative feedback.
First, about 1/3 of a century ago, my associates and I personally went to visit the late Richard Heyser about the question: "Why is it that loudspeakers, phono cartridges, and analog tape have so much harmonic distortion (lower order of course) that we still hear differences in our electronic amplifier designs, when they can be so low in measured distortion?" Richard Heyser, who was then developing TDS measurement, told us that it appeared that GLOBAL NEGATIVE FEEDBACK was the problem. Could he prove it at the time? No! But experience with negative feedback of his designs and those of others, showed this to be true. Actual harmonic distortion measurements, and even SMPTE IM measurements could be fairly lousy, YET an amp could still sound OK, in many cases. BUT many amplifiers that apparently measured well, could sound lousy. Obviously, we had much to learn about distortion measurement.
I went the direction of Otala, in making linear, high open loop, discrete designs.
A few years later, working with Matti Otala and personally making hundreds of measurements, myself, we found that SMPTE IM was most useful for crossover distortion (higher order harmonics at listening level) but useless for TIM (SID) distortion measurement. We were looking for a SINGLE measurement to denote amplifier quality. We tried noise loading, harmonic, SMPTE IM, CCIR IM, sine-square TIM (which we developed), and anything else that we could find.
Ultimately, we could find NO one test would give us complete understanding of an electronic component. This was in 1976, a long time ago.
Walt Jung, independently researched SID (TIM) with extended harmonic distortion measurements. The problem was that existing THD measurement equipment had fairly lousy extended distortion measuring capability, because the IC's used in the test equipment could not remain low distortion at extended frequency to 100KHz, which was necessary. Both Walt and I found that upgrading our ST THD harmonic analyzers with better IC's, that we could make acceptable measurements in order to see TIM (SID, which is the same thing, but it took us years to reconcile this).
Today, the best single measurement approach would probably be a really low distortion harmonic analyzer with discrete distortion harmonic resolution or an FFT spectrum analysis, after the initial THD measurement.
Still, Barrie Gilbert's article shows a unique FM distortion that would NOT be measured by the AM sensitive THD measurement. This FM distortion is NOT fixed by negative feedback, although the AM distortion component generated by the same source has been reduced by negative feedback. This is what makes Barrie Gilbert's analysis significant and useful. It shows a potential distortion that is NOT removed by global negative feedback.
THD vs other sources
Gentlemen:
As probably many followers of this thread are familiar with, linear systems may be looked from a time domain or frequency domain perspective for a complete characterization.
This should clue that harmonic distortion measurements at a single frequency or a discrete set is only part of the picture.
Taking the alternate time domain viewpoint, as much as the impulse response is in itself the complete description of a system's transfer function or kernel, then its dynamic behavior should not be dismissed either.
And this holds for linear systems where harmonic distortion is absent by definition.
The bottom line is, being audio amplifiers (as real vs ideal systems) nonlinear, the approach to their study cannot possibly be made without going either through synthesis of an exhaustive list of different cause - effect contributions (non constant gm, slew rate limits etc.) or an input - output analysis and breakdown of measured deviations.
To this - and in the light of what matters us - the physiological constraints of auditory perception must be factored in to be able to determine what is acceptable and what is not.
Rodolfo
Gentlemen:
As probably many followers of this thread are familiar with, linear systems may be looked from a time domain or frequency domain perspective for a complete characterization.
This should clue that harmonic distortion measurements at a single frequency or a discrete set is only part of the picture.
Taking the alternate time domain viewpoint, as much as the impulse response is in itself the complete description of a system's transfer function or kernel, then its dynamic behavior should not be dismissed either.
And this holds for linear systems where harmonic distortion is absent by definition.
The bottom line is, being audio amplifiers (as real vs ideal systems) nonlinear, the approach to their study cannot possibly be made without going either through synthesis of an exhaustive list of different cause - effect contributions (non constant gm, slew rate limits etc.) or an input - output analysis and breakdown of measured deviations.
To this - and in the light of what matters us - the physiological constraints of auditory perception must be factored in to be able to determine what is acceptable and what is not.
Rodolfo
john curl said:
Hi
If any distortion is at the output (FM... or pick a name for it)... it means that something at the output is different from the input.
Something ,that a "null test" can easily tell ...
Pity ,that we have lost David Hafler...his input will be very interesting in this thread.
Folks, the idea of the NULL TEST is not a new idea. Walt Jung and I used it comparing capacitors with a AD534 instrumentation IC that has very high common mode rejection.
What is wrong with this test? Well, it doesn't separate LINEAR from NON-LINEAR distortion. What does this mean? Well, ANY time delay, phase shift, dielectric absorption, etc will OVERWHELM the test itself.
Yes, you can attempt to compensate for some of the linear distortion, BUT not all.
Is this possible with digital subtraction? Yes, John Meyer and I first used it 30 years ago to measure loudspeakers. What is the problem with this test? Well, you need a lot of bits of resolution in order to make it useful. Remember, you are looking at a DIFFERENCE, and the residual must be error free to be useful. Perhaps a 24 bit system, with really good supporting electronics, MIGHT make it possible to measure audio differences, but I have not yet tried it. Perhaps someone else has.
Critics and cynics, get out of your armchairs! Get to work making the perfect single measurement system. Until then, please live with the fact that no ONE measurement shows EVERY deviation from ideal, in audio electronics.
With all due respect to David Hafler. His amp may have passed his test, but that amp is not considered 'perfect' by any standard, and is usually 'upgraded' by amateurs like yourself, with noted 'improvement'.
What is wrong with this test? Well, it doesn't separate LINEAR from NON-LINEAR distortion. What does this mean? Well, ANY time delay, phase shift, dielectric absorption, etc will OVERWHELM the test itself.
Yes, you can attempt to compensate for some of the linear distortion, BUT not all.
Is this possible with digital subtraction? Yes, John Meyer and I first used it 30 years ago to measure loudspeakers. What is the problem with this test? Well, you need a lot of bits of resolution in order to make it useful. Remember, you are looking at a DIFFERENCE, and the residual must be error free to be useful. Perhaps a 24 bit system, with really good supporting electronics, MIGHT make it possible to measure audio differences, but I have not yet tried it. Perhaps someone else has.
Critics and cynics, get out of your armchairs! Get to work making the perfect single measurement system. Until then, please live with the fact that no ONE measurement shows EVERY deviation from ideal, in audio electronics.
With all due respect to David Hafler. His amp may have passed his test, but that amp is not considered 'perfect' by any standard, and is usually 'upgraded' by amateurs like yourself, with noted 'improvement'.
Incoherent is an understatement.......
Perhaps:
"More pompus, condescending, grandstanding from someone who seems to wish that we would rather refer to him as Your Lordship" is more appropriate.
Hey Professor......when you have all the answers, we will "kow tow" to you. But you don't, so stop acting like you do.
Jocko
Perhaps:
"More pompus, condescending, grandstanding from someone who seems to wish that we would rather refer to him as Your Lordship" is more appropriate.
Hey Professor......when you have all the answers, we will "kow tow" to you. But you don't, so stop acting like you do.
Jocko
I agree with John Curl. There is never a problem in finding a difference between input and output, the problem is understanding which differences matter and which don't. Also, listening to the difference signal tell you very little of use. There will always be a difference. After all, if there weren't your NFB amp wouldn't work.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.