Hi Scott,
What he did some years back was to post three differently distorted files for people to listen to on line. Sadly these where taken down later.
He did taw larger tests like that BTW covered in AES and other papers, one set as "proof of concept" of the GedLee Distortion metric another with Compression Driver distortion.
Some may.
I know it is not a new observation. I formulated "my challenge" precisely the way it did because it is entierly based on known issues.
A system with a few percent 2nd HD and a little bit of 3rd HD at full scale that declines with signal level (essentially the classic SE Triode Tube Amplifier Distortion) remains essentially inaudible in all tests I am aware of (and it has a very low GedLee Distortion Metric score).
A system with gross crossover distortion can still measure 0.01% THD at full scale, yet the sonic results of such a system are very objectionable (I know, I once when much younger build an amp from a Magazine article that had both severe crossover distortion and slew rate induced distortion, yet measured extremely low 1KHz THD).
So, absolutely, what I noted is nothing new nor was it intended to be new.
It was intended to show just how closed the mindset of some is, who will not even consider if such a statement is reasonable and backed in fact, but who will, simply because of the source and context of the information will reject it and claim the opposite (and include the odd personal digg or two for good measure).
Ciao T
What he did some years back was to post three differently distorted files for people to listen to on line. Sadly these where taken down later.
He did taw larger tests like that BTW covered in AES and other papers, one set as "proof of concept" of the GedLee Distortion metric another with Compression Driver distortion.
Some may.
I know it is not a new observation. I formulated "my challenge" precisely the way it did because it is entierly based on known issues.
A system with a few percent 2nd HD and a little bit of 3rd HD at full scale that declines with signal level (essentially the classic SE Triode Tube Amplifier Distortion) remains essentially inaudible in all tests I am aware of (and it has a very low GedLee Distortion Metric score).
A system with gross crossover distortion can still measure 0.01% THD at full scale, yet the sonic results of such a system are very objectionable (I know, I once when much younger build an amp from a Magazine article that had both severe crossover distortion and slew rate induced distortion, yet measured extremely low 1KHz THD).
So, absolutely, what I noted is nothing new nor was it intended to be new.
It was intended to show just how closed the mindset of some is, who will not even consider if such a statement is reasonable and backed in fact, but who will, simply because of the source and context of the information will reject it and claim the opposite (and include the odd personal digg or two for good measure).
Ciao T
Even if such testing setup would be infinitely fast and perfect, it would give, as a result, a sum of all possible distortions, and in this sum HD will dominate. How to distinguish contributions of various mechanisms?
A resume from such measurements, will be well known one - fight again the THD. But the same is recommended by using spectrum analysers.
Another camp people, try to design mesasurement setup, that would skip out presence of THD, but reveal the most agressively affecting sound, distortions of much smaller magnitude, than THD.
It will produce a "forest" of "unharmonic" IMD, not only 2nd HD and it will by clearly audible with complex signal, with result like this
http://www.diyaudio.com/forums/anal...ch-preamplifier-part-ii-1492.html#post2671674
Not only, but such measurement shows clearly, the here are not audible, but unmeasureble differencies in sound,distortions(= all deviations from orignal), because one "can hear" (and eventualy analyse, with spectral analyse) all distortion in isolation, unmasked.
Last edited:
Hi,
Now I know what I am doing wrong and why my traces look so furry that I need to use averaging and smoothing to make them meaningful.
I use TOO MANY DATAPOINTS.
Hey, hold it. If use many fewer datapoints, I am smoothing the resulting curve dramatically. Maybe I should only use 25 sweep points and then claim the resulting curves are not smoothed or averaged?
Hey, there is a great idea.
I disagree. It applies smoothing as a direct result of using very few datapoints (compared to other options).
In fact, it is fair to conclude your measurement system has 1/25th octave smoothing build in.
By doing so it hides most of the small differences I get when using 32K data points for the same bandwidth...
They are neither, they are a result of using dramatically less smoothing than your measurement system does.
Yes, this is what is commonly termed the baffle step.
However, it does only make sense if the speaker is placed in free space, which it is never when used to listen to music.
Actually, most domestic speakers operate already in an envoironment that has a real radiation angle of less than quarter space.
So if we compensate the so-called baffle step fully we invariably get a 6dB bass boost. This is in fact how mother nature works. So, your detailed explanation omits crucial factors so I rather feel I do know my A and E (actually alpha to omega) a bit better than someone who analyses speakers only for free-field.
Your explanation, once extended to cover the real acoustic circumstances that speakers operate in leads invariably to my "6dB Bass Boost" conclusion, so much is simple fact.
And to anyone knowing a little bit of human hearing this bass boost makes sense, as it will improve the perceived tonal balance if listening at levels lower than "realistic", so it keeps being used, even though it is patently a (linear) distortion build into the speaker. However, this is conjecture, though reasonable and based in known observations and facts.
Ciao T
Now I know what I am doing wrong and why my traces look so furry that I need to use averaging and smoothing to make them meaningful.
I use TOO MANY DATAPOINTS.
Hey, hold it. If use many fewer datapoints, I am smoothing the resulting curve dramatically. Maybe I should only use 25 sweep points and then claim the resulting curves are not smoothed or averaged?
Hey, there is a great idea.
I disagree. It applies smoothing as a direct result of using very few datapoints (compared to other options).
In fact, it is fair to conclude your measurement system has 1/25th octave smoothing build in.
By doing so it hides most of the small differences I get when using 32K data points for the same bandwidth...
They are neither, they are a result of using dramatically less smoothing than your measurement system does.
Yes, this is what is commonly termed the baffle step.
However, it does only make sense if the speaker is placed in free space, which it is never when used to listen to music.
Actually, most domestic speakers operate already in an envoironment that has a real radiation angle of less than quarter space.
So if we compensate the so-called baffle step fully we invariably get a 6dB bass boost. This is in fact how mother nature works. So, your detailed explanation omits crucial factors so I rather feel I do know my A and E (actually alpha to omega) a bit better than someone who analyses speakers only for free-field.
Your explanation, once extended to cover the real acoustic circumstances that speakers operate in leads invariably to my "6dB Bass Boost" conclusion, so much is simple fact.
And to anyone knowing a little bit of human hearing this bass boost makes sense, as it will improve the perceived tonal balance if listening at levels lower than "realistic", so it keeps being used, even though it is patently a (linear) distortion build into the speaker. However, this is conjecture, though reasonable and based in known observations and facts.
Ciao T
Hi,
Yet it has been repeatedly shown as inaudible with music, in serious studies, large scale double blind tests etc.
So this leaves us with a dichotomy. Such distortion is not audible but measurable.
So again, the measurement is not useful in informing us of the audibility of distortion. You really should read the Geddes/Lee AES Paper, D.E.L. Shorter, Patterson and Olson on distortion and audibility.
The most recent paper on this is 8 Years old, the oldest nearly 60, so it is reasonable to expect cognisance, if not understanding of them.
Ciao T
Yet it has been repeatedly shown as inaudible with music, in serious studies, large scale double blind tests etc.
So this leaves us with a dichotomy. Such distortion is not audible but measurable.
So again, the measurement is not useful in informing us of the audibility of distortion. You really should read the Geddes/Lee AES Paper, D.E.L. Shorter, Patterson and Olson on distortion and audibility.
The most recent paper on this is 8 Years old, the oldest nearly 60, so it is reasonable to expect cognisance, if not understanding of them.
Ciao T
I know this paper, and I simply disagree. And similar view is e.g. here.
http://www.diyaudio.com/forums/anal...rch-preamplifier-part-ii-827.html#post2415329
http://www.diyaudio.com/forums/anal...rch-preamplifier-part-ii-827.html#post2415329
And using own ears to try capture isolated distortion is also useless? One can not hear something, what is clearly audible (1%IMD separated in substraction test), but he can hear something unaudible, bellow noise level in the same test? This is result? NO, thanks..
Last edited:
Hi,
You disagree with the conclusions drawn in the paper regarding the GedLee metric, or you disagree with the results of the experiments?
The conclusions are clearly open to debate, however I think you will find it hard going to invalidate the experimental results and the levels of correlation they show.
To quote SY's signature:
"In science, contrary evidence causes one to question a theory. In religion, contrary evidence causes one to question the evidence." Floyd Toole
Ciao T
You disagree with the conclusions drawn in the paper regarding the GedLee metric, or you disagree with the results of the experiments?
The conclusions are clearly open to debate, however I think you will find it hard going to invalidate the experimental results and the levels of correlation they show.
To quote SY's signature:
"In science, contrary evidence causes one to question a theory. In religion, contrary evidence causes one to question the evidence." Floyd Toole
Ciao T
Vladimir,
I am assuming here that (in most cases) it will be the power amps that produce the most unwanted artifacts, and not the op-amps. I think the idea of this test is to show any kind of difference.
About audibility of various kinds of distortions, how do I see it from a listener point of view.
I have a 1% THD at 10W SS amp, and
industrial hi-fi Pioneer A-66x SS amp, with 0,001% THD at 10W
When I listen the first one, I get an illusion of presence at a live event, acoustic instruments are distributed in front of me, and string's sound is so natural, that I want to touch the sring by hand.
After I replace the amp, by the industrial hi-fi, I hear a kind of a dull sound, nothing remains from the livliness.
The only conclusion one can make from this, that distortions other than the harmonic ones are important for the hearing mechanism. Most probably they relate to dynamic properties, transients, signal front edges, etc., i.e. to distortions of the most important for hearing parts of the signal curve.
Empirically, I can state, that shunt PS improves the pulse responce, and this is important for more live sound. If we will measure difference in signal front, with shunt and serial PS, this difference is hardly detectable. Nevertheless, during designing and testing of a prototype, I would prefer to have equipment, that allows to fix this transient behaviour differences, rather than differences in THD.
I have a 1% THD at 10W SS amp, and
industrial hi-fi Pioneer A-66x SS amp, with 0,001% THD at 10W
When I listen the first one, I get an illusion of presence at a live event, acoustic instruments are distributed in front of me, and string's sound is so natural, that I want to touch the sring by hand.
After I replace the amp, by the industrial hi-fi, I hear a kind of a dull sound, nothing remains from the livliness.
The only conclusion one can make from this, that distortions other than the harmonic ones are important for the hearing mechanism. Most probably they relate to dynamic properties, transients, signal front edges, etc., i.e. to distortions of the most important for hearing parts of the signal curve.
Empirically, I can state, that shunt PS improves the pulse responce, and this is important for more live sound. If we will measure difference in signal front, with shunt and serial PS, this difference is hardly detectable. Nevertheless, during designing and testing of a prototype, I would prefer to have equipment, that allows to fix this transient behaviour differences, rather than differences in THD.
Hi,
IMD expressed as single number and without reference to SPL and frequency is equally useless as single number THD. They are in fact two different ways of measuring the effects of the same underlying nonlinearity, in a competently designed Amplifier or Speaker Driver.
So the whole argument of HD vs. IMD you are presenting is a non-argument and does not have any context to audibility. If the Harmonic Distortion content is inaudible with complex signals then so is the resulting Intermodulation Distortion...
Ciao T
IMD expressed as single number and without reference to SPL and frequency is equally useless as single number THD. They are in fact two different ways of measuring the effects of the same underlying nonlinearity, in a competently designed Amplifier or Speaker Driver.
So the whole argument of HD vs. IMD you are presenting is a non-argument and does not have any context to audibility. If the Harmonic Distortion content is inaudible with complex signals then so is the resulting Intermodulation Distortion...
Ciao T
Hi Vladimir,
I find a high quality DSO (Digital Storage Oscilloscope) set to single shot mode with a short timebase can for audio (and higher if it is fast enough) give quite some insight.
BTW, how is Tashkent these days, I visited in 1986 IIRC. In those days it was very Soviet in architecture. I heard it has changed a lot?
Ciao T
I find a high quality DSO (Digital Storage Oscilloscope) set to single shot mode with a short timebase can for audio (and higher if it is fast enough) give quite some insight.
BTW, how is Tashkent these days, I visited in 1986 IIRC. In those days it was very Soviet in architecture. I heard it has changed a lot?
Ciao T
Did you make a listening of your system in the four different configurations? Did you notice any difference in its sound?
A raw test on the importance, if any, of the transient distortion can be made by connecting the output of one channel of your preamp (supposed to be the best, at the moment) to the input of the other channel (two identical audio stage in cascade, with gain setted to more or less unity) and by making a comparison, in mono listening, of only one channel and, then, two channel in cascade. If the preamp is limiting in some way the performances of the system, some difference had to be noted.
Last edited:
Just reading several posts here, I'm not sure we all mean the same thing when we talk about 'transient distortion'.
I've always understood it to mean a distortion that occurs as a result of a signal transient, like IM distortion that occurs during a high-level transient, or the more drastic absence of a low level signal when the amplifier is slew rate limited due to a high level transient. This seems to be the accepted engineering meaning.
I have not understood it to mean the distortion of the transient or puls itself which would be called limited rise/fall times or slew rate limiting or ringing.
Is that also how you guys understand it? Seems important to know what we mean for a meaningfull discussion.
jan didden
I've always understood it to mean a distortion that occurs as a result of a signal transient, like IM distortion that occurs during a high-level transient, or the more drastic absence of a low level signal when the amplifier is slew rate limited due to a high level transient. This seems to be the accepted engineering meaning.
I have not understood it to mean the distortion of the transient or puls itself which would be called limited rise/fall times or slew rate limiting or ringing.
Is that also how you guys understand it? Seems important to know what we mean for a meaningfull discussion.
jan didden
Jan,
you are right: many times we use same words with different meanings.
In my case, I intend what can be read here attack transient (music) -- Britannica Online Encyclopedia.
I found the response to pulse input a good way to unveil this type of "distortion". It's only my thought: I could be wrong!!
But this is untrue, sorry.. Each harmonic distortion (caused by nonlinearity), visible with single tone excitation spectra as multiples of fundamental tone , will cause (the same nonlinearity) with complex signal IMD distortion, innevitable.
Such experiments with adding in digital domain harmonics to single fundamental and examinig audibility are useless, because with complex signal they are always accompanied with IMD.. Or You are hearing only single tone music?
Hi,
These are Slewrate Induced Distortion (SID) and "Transient InterModulation" (TIM).
Transient Distortion, as it used here would seem to refer to changes in the shape of an impulse passing through a DUT, including, but not limited to limited rise and fall time, ringing, nonlinear effects etc.
There is more than just limited rise/fall times.
For example, Transistors and especially vertical Fets shows strong increases in parasitic capacitance as the voltages across them change and especially if reverse biased. This can produce interesting effects when tested with transients, though the relevance to music reproduction is debated.
Ciao T
These are Slewrate Induced Distortion (SID) and "Transient InterModulation" (TIM).
Transient Distortion, as it used here would seem to refer to changes in the shape of an impulse passing through a DUT, including, but not limited to limited rise and fall time, ringing, nonlinear effects etc.
There is more than just limited rise/fall times.
For example, Transistors and especially vertical Fets shows strong increases in parasitic capacitance as the voltages across them change and especially if reverse biased. This can produce interesting effects when tested with transients, though the relevance to music reproduction is debated.
Ciao T
Hello, Thorsten
Since that time, of course, everything has changed, the architectural look of the city is much more modern and mainly european-like. However, electronics industry has almost died, only some fancy old parts are still passed from one DIYer to another, remaining about the times when the moon-rover control equipment was developed and manufactured in Tashkent. At the same time, almost any imported active parts are available locally in Tashkent, prices correspond to those in South-East Asia.
As for the things being discussed here, I appeal more to empirical findings, and one thing has become clear to me,
using of RF transistors at low-frequency audio applications, leads to substantial benefits, put the things close to the best of tube-based achievements.
Using of ordinary non-RF parts, that cause a need for insuring of a proper interstage interaction, is plausible only in theory, and is not appreciated for achieving best sound.
Last edited:
Well that link defines the expression 'transient attack' but what do you mean by 'transient distortion'?
jan didden
Hi,
Let me rephrase then.
If a complex test signal (music) that has been subjected to a distortion producing mechanism that produces a given set of harmonic distortion shows not audible distortion (empirically) then both the HD and IMD resulting from the distortion mechanism are inaudible, as the production of IMD is essentially the same process as that produces HD, however more (than one) individual tones are involved.
Incidentally, the human hearing is not only subject to gross HD creation in the process, but because there are underlying nonlinearities it is subject to gross IMD that are produced by the same nonlinearities and of course the same masking effect applies to this IMD. I would have expected so much to be patently obvious and self evident, but I spell it anyway, just in case.
So, back to square one, if a given form of distortion producing mechanism is inaudible with a complex signal than both the HD AND the IMD the distortion producing mechanism causes are inaudible, or more precisely, ANY distortion produced by this mechanism, harmonic or otherwise is inaudible. Period.
And the exact nature of this audibility varies with SPL and Frequency (and to a degree with the individual).
It is however possible to have additional effects in amplifiers that produce IMD that is not directly to the non linearities revealed by HD testing. I normally would call that "bad design" and it should not be encountered commonly these days. In such cases we always measure greater levels of IMD than we would expect from the HD measurement results.
Ciao T
Let me rephrase then.
If a complex test signal (music) that has been subjected to a distortion producing mechanism that produces a given set of harmonic distortion shows not audible distortion (empirically) then both the HD and IMD resulting from the distortion mechanism are inaudible, as the production of IMD is essentially the same process as that produces HD, however more (than one) individual tones are involved.
Incidentally, the human hearing is not only subject to gross HD creation in the process, but because there are underlying nonlinearities it is subject to gross IMD that are produced by the same nonlinearities and of course the same masking effect applies to this IMD. I would have expected so much to be patently obvious and self evident, but I spell it anyway, just in case.
So, back to square one, if a given form of distortion producing mechanism is inaudible with a complex signal than both the HD AND the IMD the distortion producing mechanism causes are inaudible, or more precisely, ANY distortion produced by this mechanism, harmonic or otherwise is inaudible. Period.
And the exact nature of this audibility varies with SPL and Frequency (and to a degree with the individual).
It is however possible to have additional effects in amplifiers that produce IMD that is not directly to the non linearities revealed by HD testing. I normally would call that "bad design" and it should not be encountered commonly these days. In such cases we always measure greater levels of IMD than we would expect from the HD measurement results.
Ciao T
- Status
- Not open for further replies.