What is the fundamental nature of distortion that is so
objectionable that we seek to cover it up or reduce it to 0?
Two things come to mind:
1) Out of tune
I was listening to a live high school choir concert on Thursday.
Some of the accompaning live orchestral components were out of
tune. They stuck out like a sore thumb.
Watching a TV show on the Fox channel called American Idol. For
those who don't get this, it is a group of kids in their twenties
who sing and compete against each other to get a recording
contract. Some of the singers were out of tune and it was very
apparent.
2) Congestion
Listening to recordings of large orchestral pieces. The louder
it gets, they more it deteriorates into a blob of sound.
Apparent on recordings but not as much on listening to live sound.
Other characteristics?
Thoughts:
Is 1) the reason that higher order distortion components
are so objectionable (because they are out of tune with their
nearest corresponding notes)?
Is 2) caused by large amounts of 2nd and 3rd order distortion
(which is otherwise in tune)? A parallel thought would
be to think about what would happen if an amplifier existed
that produced 2nd and/or 3rd harmonic distortion that was
as out of tune as higher order distortion. Would it be
objectionable? I think it would be very objectionable.
This is a thread that started from some posts in another thread.
I don't know if it will go anywhere, but I thought it would be
interesting to at ask the questions.
objectionable that we seek to cover it up or reduce it to 0?
Two things come to mind:
1) Out of tune
I was listening to a live high school choir concert on Thursday.
Some of the accompaning live orchestral components were out of
tune. They stuck out like a sore thumb.
Watching a TV show on the Fox channel called American Idol. For
those who don't get this, it is a group of kids in their twenties
who sing and compete against each other to get a recording
contract. Some of the singers were out of tune and it was very
apparent.
2) Congestion
Listening to recordings of large orchestral pieces. The louder
it gets, they more it deteriorates into a blob of sound.
Apparent on recordings but not as much on listening to live sound.
Other characteristics?
Thoughts:
Is 1) the reason that higher order distortion components
are so objectionable (because they are out of tune with their
nearest corresponding notes)?
Is 2) caused by large amounts of 2nd and 3rd order distortion
(which is otherwise in tune)? A parallel thought would
be to think about what would happen if an amplifier existed
that produced 2nd and/or 3rd harmonic distortion that was
as out of tune as higher order distortion. Would it be
objectionable? I think it would be very objectionable.
This is a thread that started from some posts in another thread.
I don't know if it will go anywhere, but I thought it would be
interesting to at ask the questions.
Hi mfc,
Out of tune.
This might relate to the quality of an original performance, but does it really relate to reproduction ?
Francis Vaughan gave us a remarkable insight to piano tuning in another thread, but we accept that source material can be out of tune.
Of equally outstanding notability in the live performance is an individual lack of timing accuracy, and this is where I feel that power amplifier-cable-loudspeaker system interaction can impact upon reproduction performance; ie. when the system generated components become fractionally voltage/time shifted and appear to alter the harmonic character of instruments and voices. Maybe this could be heard as an un-natural alteration of tuning.
Congestion.
Or compression ?
A low power amplifier can struggle more in this regard unless it is driving efficient loudspeakers, especially with orchestral recordings.
And then there is the loudspeaker - for these virtually become non-linear the moment you can see the cone moving. Hearing many small cones in series/parallel clearly reproducing life-like sound levels whilst apparently not moving can be quite a revelation.
Just my early morning ramblings.
Cheers ............. Graham.
Out of tune.
This might relate to the quality of an original performance, but does it really relate to reproduction ?
Francis Vaughan gave us a remarkable insight to piano tuning in another thread, but we accept that source material can be out of tune.
Of equally outstanding notability in the live performance is an individual lack of timing accuracy, and this is where I feel that power amplifier-cable-loudspeaker system interaction can impact upon reproduction performance; ie. when the system generated components become fractionally voltage/time shifted and appear to alter the harmonic character of instruments and voices. Maybe this could be heard as an un-natural alteration of tuning.
Congestion.
Or compression ?
A low power amplifier can struggle more in this regard unless it is driving efficient loudspeakers, especially with orchestral recordings.
And then there is the loudspeaker - for these virtually become non-linear the moment you can see the cone moving. Hearing many small cones in series/parallel clearly reproducing life-like sound levels whilst apparently not moving can be quite a revelation.
Just my early morning ramblings.
Cheers ............. Graham.
Graham Maynard said:
Hi,
I admit to walking on thin ice in this area and have only an
interest in these things. The reason I'll make a stronger
relationship between being out of tune in a live performance
and harmonic distortion in reproduction is that:
1) higher order harmonics (like the 7th) are out of
tune with the notes played by other instruments.
2) it is very easy to recognize a note as being out of tune or
playing in the wrong key (interpreted as the same thing by
our ear/brain?).
Therefore the reason hi order harmonics are so objectionable is
because it is so easy for human beings to recognize a pitch
as being out of tune.
I also got a lot out of Francis' posting. The out of tune aspects
of a live orchestra are a fine balancing act. Makes me think of
the reproduction chain as another instrument added to the mix
of a performance.
After staring at the attached spreadsheet you see that many
hi order harmonics aren't in pitch or key, and that this maybe
why we find hi order harmonics objectionable. OTOH lower
order harmonics are in pitch or key, and this is why it takes
much more of them before the ear/brain overloads on them as
congestion.
I think that compression, and timing distortions are also equally
valid types of distortions. Compression of course being a
distortion of all harmonics, and timing, well thats a whole
'nother topic.
Thanks,
Mike
2) is pretty complex and I think you need to break out some of the individual elements. The first one to think about is dynamic range and compression/limiting. This involves evet stage of the chain from microphone to the loudspeakers in your living room. It is very difficult to make an orchestral recording with out some amount of dynamic range limitation -- potentially you have to cover SPLs generated by a single flautist playing quite softly to every instumemnt going full blast. I'm not sure that the numbers are, but even with DVD-A, SACD etc it is a challenge - one that not usually met.
That's just the technical limit. Then there is psychological and this is weird/ironic. Take a piano. Suppose you have one in you living room and a family member who can play it well. Measure the maximum SPL that can be produced in some of the more vigorous classiocal pieces. Now take a high quality CD (or DVD-A, whatever) and play it at the same max SPL. Even if you can do this with distressing your loudspeakers of bringing the electronics into clipping, nearly everyone including the pianist we just mentioned will find the result unpleasent.
The perception of unpleasentness will still be there even if the recording and reproduction chain is very low distortion. Appearently, we have been trained to accept much higher SPLs when they are produced by "real" instruments rather than recordings. In fact this effect setts in at SPLs much lower than the original. We just don't accept high (or even moderately) SPLs in our living room unless the source is "real" and even when the levels are not challenging equipment capabilities. I contend that even if the ultimate goal of absolutely perfect reproduction of ochestral music is achieved, we will still be subject to a psychological disconnect - our brain KNOWS there is no orchestra in our living room - it KNOWS that is an impossability and therefore cannot accept the associated SPLs.
OK so we play our CDs at lower than realistic SPLs no matter how good the recording chain. If the relative SPLs of the ochestra at max and that lone flautist are maintained, what happens? No flautists! Or at least one you can barely hear over your own breathing. Given this, I cannot fault in general the practice of compression/limiting when recording some type of performances - despite it's faults the alternative is even less listenable. It is my observation (which may be false) that among the more extremist audiophile types (some call me that) "favorite recordingd" tend to be of performances that involve by nature (small groups, vocals etc - whether classical, jazz or pop) have a restricted dynamic range where compression need rarely if ever be applied.
My point regarding comment 2). I know exactly the phenomonon you are talking about. It annoys me to no end. But I fear it won't go away even if "THD", "N", "IM", "TIM", "DIM" and all their cosins, real and imaginary, are eliminated. Only if someone finds a solution to the dynamic range problem including especially the purely psychological parts will the mush go away.
That's just the technical limit. Then there is psychological and this is weird/ironic. Take a piano. Suppose you have one in you living room and a family member who can play it well. Measure the maximum SPL that can be produced in some of the more vigorous classiocal pieces. Now take a high quality CD (or DVD-A, whatever) and play it at the same max SPL. Even if you can do this with distressing your loudspeakers of bringing the electronics into clipping, nearly everyone including the pianist we just mentioned will find the result unpleasent.
The perception of unpleasentness will still be there even if the recording and reproduction chain is very low distortion. Appearently, we have been trained to accept much higher SPLs when they are produced by "real" instruments rather than recordings. In fact this effect setts in at SPLs much lower than the original. We just don't accept high (or even moderately) SPLs in our living room unless the source is "real" and even when the levels are not challenging equipment capabilities. I contend that even if the ultimate goal of absolutely perfect reproduction of ochestral music is achieved, we will still be subject to a psychological disconnect - our brain KNOWS there is no orchestra in our living room - it KNOWS that is an impossability and therefore cannot accept the associated SPLs.
OK so we play our CDs at lower than realistic SPLs no matter how good the recording chain. If the relative SPLs of the ochestra at max and that lone flautist are maintained, what happens? No flautists! Or at least one you can barely hear over your own breathing. Given this, I cannot fault in general the practice of compression/limiting when recording some type of performances - despite it's faults the alternative is even less listenable. It is my observation (which may be false) that among the more extremist audiophile types (some call me that) "favorite recordingd" tend to be of performances that involve by nature (small groups, vocals etc - whether classical, jazz or pop) have a restricted dynamic range where compression need rarely if ever be applied.
My point regarding comment 2). I know exactly the phenomonon you are talking about. It annoys me to no end. But I fear it won't go away even if "THD", "N", "IM", "TIM", "DIM" and all their cosins, real and imaginary, are eliminated. Only if someone finds a solution to the dynamic range problem including especially the purely psychological parts will the mush go away.
Hi Chris,
I've been deafened by Bose PA speakers at major expeditions, and once bought a batch of their competitively priced 'Hi-Fi' speakers for Jukebox installation extensions, but for realistic reproduction - - - time for a Brandy.
Hi Sam,
Maybe we can enjoy the unpleasantness of sound from a piano or violin etc. when we are in proximity to the live instrument and thus cannot challenge its flawed tonality, but our brains cannot always be convinced that a system is reproducing faithfully even when it has been proved to be so via blind A-B live/recorded comparison.
Maybe some of the systems that are most easy to listen to suppress some of the higher harmonics that can distress us via realistic playback. I have two requirements in audio;- high accuracy and likeable easy listening; each being so different from the other for neither to be universal.
I do agree with the comments in this thread though.
Cheers ............. Graham.
I've been deafened by Bose PA speakers at major expeditions, and once bought a batch of their competitively priced 'Hi-Fi' speakers for Jukebox installation extensions, but for realistic reproduction - - - time for a Brandy.
Hi Sam,
Maybe we can enjoy the unpleasantness of sound from a piano or violin etc. when we are in proximity to the live instrument and thus cannot challenge its flawed tonality, but our brains cannot always be convinced that a system is reproducing faithfully even when it has been proved to be so via blind A-B live/recorded comparison.
Maybe some of the systems that are most easy to listen to suppress some of the higher harmonics that can distress us via realistic playback. I have two requirements in audio;- high accuracy and likeable easy listening; each being so different from the other for neither to be universal.
I do agree with the comments in this thread though.
Cheers ............. Graham.
Graham,
Subjectively: It is much easier to forgive or perhaps not even notice a flubbed note when present at a live performance. On the other hand one must politely tolerate the fat lady wearing too much perfume to your left and the old geezer coughing his lungs out on the back of your neck.
Objectively: In addition to reducing the dynamic range, as far as I know any sort of limiting or compression introduces harmonics as dioes "soft clipping" even though it may be acomplished with more subtlety than guitar amp overdrive circuits. Some decades ago I had a 30W (that was quite a bit at the time) NAD with soft clipping switch. While I left it in off position most of the time, sometimes "ON" was better than the alternative. The point: when ever alternatives exist, it's rare that one is perfect or ideal but rather with luck one is clearly less bad.
I don't think "live vs. recorded" is really a useful comparrisson. A century after Edison it's valid to consider recorded music a separate type of performance. Bach did not have a Bosendorfer to compose for and it is silly to asert that Bach should only be performed on th instruments available in his time. The question is how well it is done. A ninteenth century symphonic composition presented in a CD format owes as much to the talent and skills of the "technical" personnel as to the individual musicians. I suspect, but don't know, that many of the best CD products are ones where everyone involved realized and accepts that end objective is a CD (or LP or . . .), not that the CD is merely an archival record of a live event.
Live jazz probably has additional nuances . . . .
Subjectively: It is much easier to forgive or perhaps not even notice a flubbed note when present at a live performance. On the other hand one must politely tolerate the fat lady wearing too much perfume to your left and the old geezer coughing his lungs out on the back of your neck.
Objectively: In addition to reducing the dynamic range, as far as I know any sort of limiting or compression introduces harmonics as dioes "soft clipping" even though it may be acomplished with more subtlety than guitar amp overdrive circuits. Some decades ago I had a 30W (that was quite a bit at the time) NAD with soft clipping switch. While I left it in off position most of the time, sometimes "ON" was better than the alternative. The point: when ever alternatives exist, it's rare that one is perfect or ideal but rather with luck one is clearly less bad.
I don't think "live vs. recorded" is really a useful comparrisson. A century after Edison it's valid to consider recorded music a separate type of performance. Bach did not have a Bosendorfer to compose for and it is silly to asert that Bach should only be performed on th instruments available in his time. The question is how well it is done. A ninteenth century symphonic composition presented in a CD format owes as much to the talent and skills of the "technical" personnel as to the individual musicians. I suspect, but don't know, that many of the best CD products are ones where everyone involved realized and accepts that end objective is a CD (or LP or . . .), not that the CD is merely an archival record of a live event.
Live jazz probably has additional nuances . . . .
Multiple issues, multiple problems...
The "congestion" issue is simply a combination of amplifier deficits and speaker deficits combining... it's not all that difficult to put together both speakers and amps that *don't do that*. But, it isn't inexpensive nor does it correlate exactly to "price point" either. To get this right requires some significant insight into the system and space you're involved with.
The short and simple version for amps:
- power supplies that don't duck on peaks
- output current and voltage capability that is large compared to the load (generally speaking)
- circuitry that is not riddled with sources of non-linearity, or sources of higher order (albiet low level) distortions (especially those that rise significantly near clipping)
For speakers:
- low distortion
- ample sensitivity & power handling to remain within linearity for the space
For speakers, my experience is that they often don't meet even the most basic requirements in these two regards...
As far as I am concerned, the issue of "being in tune" is not related to the reproduction of sound via a hi-fi system. As far as American Idol is concerned, to my ears the band itself is not tuned properly to a standard (so-called) "440" in some cases - OR the tune has been transposed to another key so that the contestants can sing it, and it doesn't "sound right" in the other key ( a common effect ).
The higher order harmonics don't make a given bit of reproduced music or voice sound "flat" or "sharp."
Two things to remember here:
- that humans are "programmed" to expect to hear sound in a rather primordial or primitive way. This is why we can hear things in "stereo" and make out a three dimensional sound field! We're reassembling the sound according to a pre-programmed understanding and expectation of how sound works "naturally." This takes *brain cycles and effort* to pull off.
The *less* sound that does not correspond to the brain's expectations is presented the easier it is to "hear". The more that does not correspond to the natural "expectation" (in an already artifical situation, btw) the more difficult it is to reassemble the information into meaningful sound.
- The second thing to remember is that these higher order harmonics may or may not conform to or interfere with a "natural" sounding presentation... when they do interfere you'd probably hear things like "grain" "strain" or "harshness" to some degree...
Many musical instruments embody higher order harmonics - violins and trumpets - so merely *having* higher order harmonics present in and of itself doesn't tell you all that much on the surface... although the difference between a really great violin and a not so great violin doesn't seem to be all that much in terms of differences in the harmonics, yet it's these differences that let us human ears define which are great and which are not.
_-_-bear
The "congestion" issue is simply a combination of amplifier deficits and speaker deficits combining... it's not all that difficult to put together both speakers and amps that *don't do that*. But, it isn't inexpensive nor does it correlate exactly to "price point" either. To get this right requires some significant insight into the system and space you're involved with.
The short and simple version for amps:
- power supplies that don't duck on peaks
- output current and voltage capability that is large compared to the load (generally speaking)
- circuitry that is not riddled with sources of non-linearity, or sources of higher order (albiet low level) distortions (especially those that rise significantly near clipping)
For speakers:
- low distortion
- ample sensitivity & power handling to remain within linearity for the space
For speakers, my experience is that they often don't meet even the most basic requirements in these two regards...
As far as I am concerned, the issue of "being in tune" is not related to the reproduction of sound via a hi-fi system. As far as American Idol is concerned, to my ears the band itself is not tuned properly to a standard (so-called) "440" in some cases - OR the tune has been transposed to another key so that the contestants can sing it, and it doesn't "sound right" in the other key ( a common effect ).
The higher order harmonics don't make a given bit of reproduced music or voice sound "flat" or "sharp."
Two things to remember here:
- that humans are "programmed" to expect to hear sound in a rather primordial or primitive way. This is why we can hear things in "stereo" and make out a three dimensional sound field! We're reassembling the sound according to a pre-programmed understanding and expectation of how sound works "naturally." This takes *brain cycles and effort* to pull off.
The *less* sound that does not correspond to the brain's expectations is presented the easier it is to "hear". The more that does not correspond to the natural "expectation" (in an already artifical situation, btw) the more difficult it is to reassemble the information into meaningful sound.
- The second thing to remember is that these higher order harmonics may or may not conform to or interfere with a "natural" sounding presentation... when they do interfere you'd probably hear things like "grain" "strain" or "harshness" to some degree...
Many musical instruments embody higher order harmonics - violins and trumpets - so merely *having* higher order harmonics present in and of itself doesn't tell you all that much on the surface... although the difference between a really great violin and a not so great violin doesn't seem to be all that much in terms of differences in the harmonics, yet it's these differences that let us human ears define which are great and which are not.
_-_-bear
Hi,
Yes, otherwise we'd be listening purely to electronic tones.
This has been found to be dull and boring and its the original
harmonics that make it interesting.
The point is that hi order harmonics are more out of tune then lo
order harmonics when compared with their corresponding notes
in the equal tempered scale. It is easy for the ear/brain to hear
things that are out of tune. So hi order harmonics should be used
carefully...both by instrument makers as well as avoided, masked,
or greatly reduced in reproduction equipment.
Maybe getting distortion in reproduction equipment down
below .002% across the audio band is all that is neccessary to
preserve the original harmonic recipe? This seems to be the
solution that most designers have chosen.
bear said:
Yes, otherwise we'd be listening purely to electronic tones.
This has been found to be dull and boring and its the original
harmonics that make it interesting.
The point is that hi order harmonics are more out of tune then lo
order harmonics when compared with their corresponding notes
in the equal tempered scale. It is easy for the ear/brain to hear
things that are out of tune. So hi order harmonics should be used
carefully...both by instrument makers as well as avoided, masked,
or greatly reduced in reproduction equipment.
Maybe getting distortion in reproduction equipment down
below .002% across the audio band is all that is neccessary to
preserve the original harmonic recipe? This seems to be the
solution that most designers have chosen.
Probably not.
Unfortunately a figure like 0.002% does not correspond to much of anything in terms of perception... there have been some recent studies that indicate that vanishingly low percentages of higher order harmonics in the *right situation* can be objectionable.
The same studies seem to show that the harmonic spectrum spewed out by much higher distortion circuits (for example triodes) do not exhibit the same objectionable perception.
I'm painting a very broad brush above, so just focus on the idea that merely having apparently low distortion figures does not correlate well to perception.
Oh, and other than speakers, achieving the quoted figure (on the surface) is rather trivial as almost any Japanese branded receiver will do that trick, as will almost any CD player...
_-_-bear
I have listened to some .WAV files that demonstrated this. Very low levels of distotion producing incredibly horrible sounds. My problem is that even granting that that files were exactly as claimed, I have never, ever heard anything like those sounds on even the cheapest boom box, PC speaker, 1960's era long distance telephone - you name it.. . . . So while this demonstates the power of very slight high order distortion to sound bad, I suspect a "strawman", that is to say I'm doubtful that demonstations correspond to real world audio behavior.
This of course, fits nicely with the position of some audio reviwers that amplifiers and CD players have just become "apliances", i.e., pretty much interchangeable except for cosmetics and convenience features. Personally, I don't worry about this much because even if true, I still like "rolling my own". Amplifiers that you build yourself (no matter how humble) alway sound better than anything you could buy, IMO.
Hi,
I'm looking at the slide presentation at www.gedlee.com
about the need to account for masking in the perception of
distortion.
It would be nice to be able to get a better handle on what a
distortion spectrum would look like that fits his metric the best.
One thing I've been able to infer from his slide #15 and #16
is that a distortion spectrum that is monotonically decreasing
will provide better masking than a distortion spectrum that is flat.
This should help to eliminate objectionable distortion.
So I think I can safely say that having a distortion spectrum that
is as low as possible to begin with and that is monotonically
decreasing with increased harmonic order would be a start.
How much each harmonic needs to decrease by is something I'm
not sure of yet.
Thoughts?
bear said:
I'm looking at the slide presentation at www.gedlee.com
about the need to account for masking in the perception of
distortion.
It would be nice to be able to get a better handle on what a
distortion spectrum would look like that fits his metric the best.
One thing I've been able to infer from his slide #15 and #16
is that a distortion spectrum that is monotonically decreasing
will provide better masking than a distortion spectrum that is flat.
This should help to eliminate objectionable distortion.
So I think I can safely say that having a distortion spectrum that
is as low as possible to begin with and that is monotonically
decreasing with increased harmonic order would be a start.
How much each harmonic needs to decrease by is something I'm
not sure of yet.
Thoughts?
Hi Mike,
I quite agree with the monotonically decreasing harmonic requirement, and on the surface a 0.002% total thd should satisfy, but I believe that it is how the 0.002% figure is arrived at that leaves listeners still wanting.
In many amplifiers the propagation delay can interfere with nfb loop correction of voltage sensed output error when the resultant loudspeaker current causing the voltage error has become dynamically phase shifted wrt music input. A figure of 0.002% measured in time isolation at an amplifier's output terminal using a resistor load that maintains ideal current-voltage coherence cannot reveal any amplification topology time shift/nfb induced errors, only amplifier amplitude linearity, which I believe is a lesser contributor to observable distortion.
The best measuring amplifiers still do not sound best for this very reason, and their distortion remains observable via loudspeakers which themselves produce far greater levels of distortion, because the amplifier's nfb loop induced reactivity has generated new high frequency currents which cannot perfectly correct the on-going voltage waveform.
No nfb - no propagation delay induced interaction to generate new fractional time shifted components, eg. Susan's 'Zero Feedback Impedance Amplier'.
Extremely low propagation delay - can shorten the loudspeaker induced error voltage correction period to minimise and thereby lift nfb induced interaction beyond audibilty, eg. as MikeB's 'Ultrafast NoCap class-AB'.
Cheers ............. Graham.
I quite agree with the monotonically decreasing harmonic requirement, and on the surface a 0.002% total thd should satisfy, but I believe that it is how the 0.002% figure is arrived at that leaves listeners still wanting.
In many amplifiers the propagation delay can interfere with nfb loop correction of voltage sensed output error when the resultant loudspeaker current causing the voltage error has become dynamically phase shifted wrt music input. A figure of 0.002% measured in time isolation at an amplifier's output terminal using a resistor load that maintains ideal current-voltage coherence cannot reveal any amplification topology time shift/nfb induced errors, only amplifier amplitude linearity, which I believe is a lesser contributor to observable distortion.
The best measuring amplifiers still do not sound best for this very reason, and their distortion remains observable via loudspeakers which themselves produce far greater levels of distortion, because the amplifier's nfb loop induced reactivity has generated new high frequency currents which cannot perfectly correct the on-going voltage waveform.
No nfb - no propagation delay induced interaction to generate new fractional time shifted components, eg. Susan's 'Zero Feedback Impedance Amplier'.
Extremely low propagation delay - can shorten the loudspeaker induced error voltage correction period to minimise and thereby lift nfb induced interaction beyond audibilty, eg. as MikeB's 'Ultrafast NoCap class-AB'.
Cheers ............. Graham.
Direct audibility of nonlinear distortion itself is overrated. Thats evident when one tries to hear -60db signal with system volume settings at normal listening levels.
But, there are no clean signals in music, and all the nonlinear byproducts add up in complex ways. The sum of these byproducts must not cause speakers to generate audible sounds above human masking levels. The THD very low levels of acceptability (like <0.002%) are really a vague way to say that "if you keep THD/IMD below that, then with today's usual music content the sums of nonlinear distortions would not rise above human masking levels".
Sure, you can attach the high order harmonics with the above and see that its easier to achieve with devices producing only low order harmonics, but its no panacea.
1. Out of tune. I disagree with that idea. Human ear discerns main tones from harmonics very well, and all higher harmonics are perceived as timbre, unless their levels are too huge. And even then, matter of tuning is far from trivial. Piano is imo not a very good example to focus on, because it is intentionally in very complex tuning (*it* is basically out of tune). Some say that realistic playback of piano alone is the ultimate test of reproduction chain. Because pianos perceptible timbre is so much dependent on precise reproduction of complex higher harmonics that every change up there sticks out as change in timbre.
I can't see anything in amp that could possibly generate anything out of tune. Its something else.
2. Congestion. I can't help but think of revereberation here. Imo thats the main source of congestion. Yes, if you think of above, all products of THD adding up in their complex ways, the repetition frequency of exceeding masking levels of human ear increase with complexity of music and levels. Nonlinearities of speakers bite here, but also, as resemblence with reverberation, energy storage of speakers. Anyway, I think that room and speaker system has far more impact on congestion than amp nonlinearities.
Also I think that dynamic linear distortion is much more of issue than decently low nonlinear distortion. Subharmonics in IMD test are clear evidence of such effects happening in amps (AM). Also in speakers.
I agree with Graham that amp-speaker interaction is of huge influence on all that becomes audible. It is sufficient to compare FR/PR plots of same speaker when driving it with voltage source vs. current source to realize how much influence the amp output behaviour has. Dynamic changes in amp output behaviour cause imo dynamic changes to linear response, even with no THD present. That is imho responsible for some part of perceived sound muffling, or amp character, leaving amps with similar THD specs sound different.
But, there are no clean signals in music, and all the nonlinear byproducts add up in complex ways. The sum of these byproducts must not cause speakers to generate audible sounds above human masking levels. The THD very low levels of acceptability (like <0.002%) are really a vague way to say that "if you keep THD/IMD below that, then with today's usual music content the sums of nonlinear distortions would not rise above human masking levels".
Sure, you can attach the high order harmonics with the above and see that its easier to achieve with devices producing only low order harmonics, but its no panacea.
1. Out of tune. I disagree with that idea. Human ear discerns main tones from harmonics very well, and all higher harmonics are perceived as timbre, unless their levels are too huge. And even then, matter of tuning is far from trivial. Piano is imo not a very good example to focus on, because it is intentionally in very complex tuning (*it* is basically out of tune). Some say that realistic playback of piano alone is the ultimate test of reproduction chain. Because pianos perceptible timbre is so much dependent on precise reproduction of complex higher harmonics that every change up there sticks out as change in timbre.
I can't see anything in amp that could possibly generate anything out of tune. Its something else.
2. Congestion. I can't help but think of revereberation here. Imo thats the main source of congestion. Yes, if you think of above, all products of THD adding up in their complex ways, the repetition frequency of exceeding masking levels of human ear increase with complexity of music and levels. Nonlinearities of speakers bite here, but also, as resemblence with reverberation, energy storage of speakers. Anyway, I think that room and speaker system has far more impact on congestion than amp nonlinearities.
Also I think that dynamic linear distortion is much more of issue than decently low nonlinear distortion. Subharmonics in IMD test are clear evidence of such effects happening in amps (AM). Also in speakers.
I agree with Graham that amp-speaker interaction is of huge influence on all that becomes audible. It is sufficient to compare FR/PR plots of same speaker when driving it with voltage source vs. current source to realize how much influence the amp output behaviour has. Dynamic changes in amp output behaviour cause imo dynamic changes to linear response, even with no THD present. That is imho responsible for some part of perceived sound muffling, or amp character, leaving amps with similar THD specs sound different.
The trick is that it is (sadly) the other way around. The attitude has been that harmonics in an amplifier of low order are OK, because they mimic the natural nature of harmonics. But the problem in nature, and music is that they don't always follow a harmonic series properly. It is not that the amplifier has generated something that is out of tune, rather that the music's harmonic structure is already unable to be consonant with even a low order harmonic.
The example of the piano is extreme - but that was where we came in. I am one of those that do suspect that the piano's legendary difficulty as a reproduced instrument may indeed come from the instruments very difficult, and curious harmonic structure.
But the problem exists in just about every musical instrument, and indeed in almost every composition, before it is even played. Harmonic theory is nasty and complex, mostly because it is actually internally inconsistent. The great unifier of the equal tempered scale is, if anything just as much a problem as any other scale.
One way of building the scale, discovered by Pythagorus - and named after him, is built upon the circle of fifths. The idea is that you can take the root note, multiply its frequency by 3/2 and get a new note. If the new note is greater than double the root, you halve its frequency. The ratio of 3/2 is the ratio of the root to the fifth. Now the idea is that if you do this 12 times you end up with the root note again (or an octave above it) and have visited every note in the diatonic scale. This is a very appealing fiction. From the point of view of music reproduction it suggests that the second harmonic (i.e. the octave) and the third harmonic (which is the fifth an octave higher than the root) are consonant, and should be relatively innocuous. And indeed this seems to be mostly true. But only mostly. In fact you don't ever reach an octave, after 12 applications you reach a note that is 23.46 cents flat. You can keep going round, and after 43 tries you get closer, but you can never reach consonance. The gap between the 12th time round with the circle of fifths and the octave is called the Pythagorean Comma - and is quite clearly heard. You won't mask this level of dissonance.
The trouble with the Pythagorean scale is that it fundamentally cannot actually map the notes in the diatonic scale. Most descriptions of the diatonic scale usually build it using a tritone. That is the ratios 2:3:5 (sometimes also described with 4:5:6 since that is the major triad). For instance the ratio 4:5 is the major third, and 9:5 (i.e. 3*3/5) is the seventh.
The unique prime factorisation theorem tells you that you cannot build a number for which 5 is a factor with only 2 and 3 as factors. It gets worse. Most people have been taught that, for instance, Bb and A# are the same note. But the Bb (in the scale of C) is the ratio 16:9, whilst the A# is 225:128. Although a piano has only one set of black keys, there are instruments for which the keys are actually split to allow the worst of the mismatches to be avoided. Equal temperament, where the scale is built by repeated application of the 12 root of two is the modern ideal. But whilst it allows composers the freedom to move from key to key, with no issues, it also makes a number of the intervals sound noticeably off.
This is only a problem for fretted instruments. As I mentioned earlier, modified tunings are actually vastly more common than most people know. But also so are instruments for which this isn't an issue at all. Fret-less instruments (most bowed instruments) when played by a competent musician can always reach the correct interval for the harmonic and melodic structure. The consonance both melodically and harmonically being in the ear of the performer.
So what has this got to do with distortion? Simply that if you add, even low order, harmonics to some music you will create products that are enough out of scale to be a noticeable dissonance.
If you do so with a recording of an ensemble of fret-less instruments you run a very real risk of creating products of one instrument that really are significantly different to the played note of another instrument, even though a simple application of a Pythagorean harmonic analysis would suggest otherwise.
But other instruments may be largely immune. Solo instruments, where there is little harmonic structure, or small ensembles where the harmonic structure is sparse, may perhaps be less upset. However it is possible that music that is harmonically rich is much more prone to audible (bad) effects than hoped for.
The ratio 80:81 is a rather useful one to remember. This is the syntonic comma. As a rough guide, it is the ratio that when applied to a note's frequency (either multiplied or divided) will take it to the same note in a different ratio-metric derivation of the scale. Thus it gives a feel for the spread of harmonic frequency mismatches that might be caused by additional harmonic distortion products.
To close, I'm very much in agreement about some of the other important distortion issues (and especially interested in time varying ones), but since, in a way, I came in with this discussion of harmonic analysis, I'm concentrating on that.
Its not the Harmonics!
Harmonic Theory is OK for music but it really isn’t that important for explaining audible distortion – Intermodulation Products Totally Dominate All “Harmonic” Distortion
Some of my thoughts on the subject:
http://www.diyaudio.com/forums/showthread.php?postid=559330#post559330
in
http://www.diyaudio.com/forums/showthread.php?postid=424592#post424592
you can see the 1KHz intermod difference product sticking out like a sore thumb, clearly audible and not “harmonically” related to the 2 high frequency tones that caused it
A high order “harmonic” distortion is evidence of a high order nonliearity that will generate vastly hugher Intermodulation distortion components having non-harmonic frequencies, IMD differences create new distortion frequency components below the fundamentals of the music that will be particularly audible because they are not masked
further down in the 2nd thread I give the best reference on distortion I know of:
"Multitone Testing of Sound System Components - Some Results and Conclusions, Part 1: History and Theory" JAES V 49#11 nov 2001 by Czerwinski et al at Cerwin Vega – with 119 references it is a truly thorough and relatively recent review article of the history of distortion measurement back to the dawn of audio reproduction
Harmonic Theory is OK for music but it really isn’t that important for explaining audible distortion – Intermodulation Products Totally Dominate All “Harmonic” Distortion
Some of my thoughts on the subject:
http://www.diyaudio.com/forums/showthread.php?postid=559330#post559330
in
http://www.diyaudio.com/forums/showthread.php?postid=424592#post424592
you can see the 1KHz intermod difference product sticking out like a sore thumb, clearly audible and not “harmonically” related to the 2 high frequency tones that caused it
A high order “harmonic” distortion is evidence of a high order nonliearity that will generate vastly hugher Intermodulation distortion components having non-harmonic frequencies, IMD differences create new distortion frequency components below the fundamentals of the music that will be particularly audible because they are not masked
further down in the 2nd thread I give the best reference on distortion I know of:
"Multitone Testing of Sound System Components - Some Results and Conclusions, Part 1: History and Theory" JAES V 49#11 nov 2001 by Czerwinski et al at Cerwin Vega – with 119 references it is a truly thorough and relatively recent review article of the history of distortion measurement back to the dawn of audio reproduction
I have a suspicion you are right. Would be very nice to find that reference - but I let my AES membership lapse a decade ago. Maybe I should reconsider.
Harmonic analysis still has a significant part to play, since on the whole that is how we hear music. It can explain the different perceptions of sound quality and explain the relative importance of a distortion product. But to a first approximation, the first thing it tells us is that intermodulation products are very likely to always sound bad. Even at very low levels.
I still suspect that some of the blather I wrote above can be important. Especially in looking at the perceived sound of current designs. I have noted how Hugh Dean is a great exponent of allowing second harmonics to remain, and claims aural benefits, whilst John Curl is more in favour of a little third harmonic. Yet both design well respected amplifiers. Both with perceived strengths and weaknesses.
If the harmonics in real music instrument is not predictable mathematical way, or cannot be duplicated naturally in audio amplifier cct with its own natural electronic harmonic degeneration, is it wrong for an amp to let any harmonic distortion behind?
Power amp with a little low order harmonic left behind usually sounds better than power amp with minimal THD (to average people, PMA would disagree on this
). Better here means better for the audience to listen to, not to compare to real musical insturment.
But from this discussion, the structure of harmonics in real musical instrument (musical instrument harmonics) is different than structure in electronic power amp with intended left low order harmonics (power amp harmonics)?
Power amp with a little low order harmonic left behind usually sounds better than power amp with minimal THD (to average people, PMA would disagree on this
). Better here means better for the audience to listen to, not to compare to real musical insturment. But from this discussion, the structure of harmonics in real musical instrument (musical instrument harmonics) is different than structure in electronic power amp with intended left low order harmonics (power amp harmonics)?
A highest THD of 0.002% at 20KHz for an amplifier GUARANTEES that the designer has taken care of non-linearities in the topology, particularly in a Class AB type with nasty half circulating currents on the supplies including lead dress of harnesses board geography according to need not prettiness, and the isolation of sensitive high impedance stages from radiated hash. It guarantees an input stage with excellent common mode and differential behaviour - all in a simple measurement. To achieve this in a real life reproducible product takes experience time and patience and a good measurement regime. Is it any wonder many would like to suggest it's unimportant to support their Black Art inferior offerings.
Anything less is altering the performance. Who would want to listen to 0.5% of power supply hash or intermodulation products.
Anything less is altering the performance. Who would want to listen to 0.5% of power supply hash or intermodulation products.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.