Or... are THD numbers actually meaningful?
Given that the current technology produces amps that are way below 0.1% THD, what significance is there to this spec in the real world?
Can humans hear the difference between 0.05% and 0.1%? Does it matter if the distortion content is even or odd harmonics?
Given that the current technology produces amps that are way below 0.1% THD, what significance is there to this spec in the real world?
Can humans hear the difference between 0.05% and 0.1%? Does it matter if the distortion content is even or odd harmonics?
It seems to me that distortion figures and audible distortion are two entirely different creatures co-existing in a world of purists and realists. Then there are also those persons who don't care.
Interesting to me is the fact that almost every distortion figure I see is obtained during maximum power output and definitely doesn't clearly define the distortion over the complete audio bandwidth of the system.
Personally, I've listened to all kinds of amplifiers that sounded good and bad with absolutely no relation to their ratings.
Some real pieces of junk have sounded better than some more expensive stuff.
Interesting to me is the fact that almost every distortion figure I see is obtained during maximum power output and definitely doesn't clearly define the distortion over the complete audio bandwidth of the system.
Personally, I've listened to all kinds of amplifiers that sounded good and bad with absolutely no relation to their ratings.
Some real pieces of junk have sounded better than some more expensive stuff.
maylar said:
Human's ear is very sensetive, and can detect subtle changes.
Even harmonics are much more pleasant to the ears and sometimes even considered as "warmth" added to the sound.
odd harmonics are less natural and considered "harsh".
Tubes, for example, have much worst distortion figures
than transistors but many people prefer tubes.
Why is that? Because tube distortion is mostly even harmonics.
Just my opinion...
regards,
Udi.
I think there really is pretty good info on what constitutes CLEARLY AUDIBLE distortion - mp3 and other compression schemes are based on tossing out stuff that is perceptually masked by the music signal - a reasonable assumption is that since the presence or absence of signal in masked frequency components doesn’t matter then similar levels of distortion at the masked frequencies also won't matter – a distortion component is very likely to be audible if a good perceptual coder decides to use bits to encode it
the lesson from these perceptual models is that harmonics of a signal frequency component are likely to be masked - especially when most musical instruments produce harmonic series of their fundamentals anyway
so strictly speaking "harmonic" distortion is not very audible and THD # don't predict distortion audibility and serious audio engineers have known this from day one so much of the controversy is clouded by debates misusing a very uninformative #
engineers designing amplifiers for audio and radio have known since the beginnings of the field that intermodulation distortion components seriously degrade channel quality - intermodulation components of complex signals contain sum and difference frequency components of all of the possible orders
many harebrained theories have be proposed that sound reasonable in light of harmonic structure in music and masking but utterly fail to do the math for Intermodulation Products, it is impossible to have only “harmonic” distortion of a complex signal
there is a link between harmonic and intermod - a nth order harmonic of a single tone is evidence of a nth order distortion mechanism (some part of the system is producing a signal proportional to v**n ) - the intermod products for 2 tones acted on by a nth order nonlinearity include all possible frequencies formed by a*f1 +/- b*f2 where a,b assume all possible integer pairs that add up to n
the # of intermod frequencies increases rapidly with the order of the nonliearity causing the distortion and the # signal components, energy in the IMD products greatly exceeds the "harmonic" components at n*f1, n*f2 - for 70+ yrs engineers have explored "weighting factor" formulas to multiply the easily measured harmonic distortion components by rapidly increasing functions of n in attempts to get # that correlate somewhat with perceived quality; they largely failed to get much beyond the general conculusion that high order distortion is worse that lower oreder distortion
the current approach that appears to have some (still modest) success compares the undistorted and distorted signals’ properties after sending them through a complex model of the 1st layer of perceptual coding to account for masking of all of the distortion products, harmonic and intermod - nobody is offering absolute distortion audibility indexes, they just rank relative levels, also this is still quite poor in resolution compared to the judgments of subtleties you will find some audiophiles claiming to perceive (but just see them run when you mention abx/double bilnd testing)
at moderate levels of distortion THD is useless, presumably at Halcro’s sub ppm levels it could put a bound on the likely IMD
distortion shouldn’t be oversimplified and designers pursuing “low distortion” should really be talking about IMD with complex signals, but THD and minutiae of the “harmonic distortion structure” are likely to continue to fuel debates for some time
the lesson from these perceptual models is that harmonics of a signal frequency component are likely to be masked - especially when most musical instruments produce harmonic series of their fundamentals anyway
so strictly speaking "harmonic" distortion is not very audible and THD # don't predict distortion audibility and serious audio engineers have known this from day one so much of the controversy is clouded by debates misusing a very uninformative #
engineers designing amplifiers for audio and radio have known since the beginnings of the field that intermodulation distortion components seriously degrade channel quality - intermodulation components of complex signals contain sum and difference frequency components of all of the possible orders
many harebrained theories have be proposed that sound reasonable in light of harmonic structure in music and masking but utterly fail to do the math for Intermodulation Products, it is impossible to have only “harmonic” distortion of a complex signal
there is a link between harmonic and intermod - a nth order harmonic of a single tone is evidence of a nth order distortion mechanism (some part of the system is producing a signal proportional to v**n ) - the intermod products for 2 tones acted on by a nth order nonlinearity include all possible frequencies formed by a*f1 +/- b*f2 where a,b assume all possible integer pairs that add up to n
the # of intermod frequencies increases rapidly with the order of the nonliearity causing the distortion and the # signal components, energy in the IMD products greatly exceeds the "harmonic" components at n*f1, n*f2 - for 70+ yrs engineers have explored "weighting factor" formulas to multiply the easily measured harmonic distortion components by rapidly increasing functions of n in attempts to get # that correlate somewhat with perceived quality; they largely failed to get much beyond the general conculusion that high order distortion is worse that lower oreder distortion
the current approach that appears to have some (still modest) success compares the undistorted and distorted signals’ properties after sending them through a complex model of the 1st layer of perceptual coding to account for masking of all of the distortion products, harmonic and intermod - nobody is offering absolute distortion audibility indexes, they just rank relative levels, also this is still quite poor in resolution compared to the judgments of subtleties you will find some audiophiles claiming to perceive (but just see them run when you mention abx/double bilnd testing)
at moderate levels of distortion THD is useless, presumably at Halcro’s sub ppm levels it could put a bound on the likely IMD
distortion shouldn’t be oversimplified and designers pursuing “low distortion” should really be talking about IMD with complex signals, but THD and minutiae of the “harmonic distortion structure” are likely to continue to fuel debates for some time
Well presented jcx
I wholeheartedly agree with jcx. Depending on how you look at it, THD may mean lots or next to nothing.
Extremely low THD is the hallmark of a fairly linear system, barring slew rate hard limits exposed with higher frequency test tones. In this condition it may well be expected IM to be low.
THD on the other hand is a bundled quantity resulting from the sum of spurious powers distributed within the audio band. Completely different spectral structures may lead to the same number though the actual listening results will be very different.
The bottom end in my view is THD can be considered a valid quality parameter as long as it is a very low figure (probably much better than 0.01%), up to 20 KHz.
Rodolfo
I wholeheartedly agree with jcx. Depending on how you look at it, THD may mean lots or next to nothing.
Extremely low THD is the hallmark of a fairly linear system, barring slew rate hard limits exposed with higher frequency test tones. In this condition it may well be expected IM to be low.
THD on the other hand is a bundled quantity resulting from the sum of spurious powers distributed within the audio band. Completely different spectral structures may lead to the same number though the actual listening results will be very different.
The bottom end in my view is THD can be considered a valid quality parameter as long as it is a very low figure (probably much better than 0.01%), up to 20 KHz.
Rodolfo
I guess it's all subjective..
Thanks for sharing your thoughts on this. The reason I ask is that it's common to see on other forums a comment like, "You can't hear less than 0.1% distortion anyway, so who cares if the spec is .02 or .08?". I'm just wondering if there's any truth to that.
Still others claim that speakers have 10 times the distortion of amplifiers anyway, so amp distortion is insignificant.
Good point about IMD, jcx. It's one spec that doesn't get much attention and probably should.
No doubt you're correct.
Thanks for sharing your thoughts on this. The reason I ask is that it's common to see on other forums a comment like, "You can't hear less than 0.1% distortion anyway, so who cares if the spec is .02 or .08?". I'm just wondering if there's any truth to that.
Still others claim that speakers have 10 times the distortion of amplifiers anyway, so amp distortion is insignificant.
Good point about IMD, jcx. It's one spec that doesn't get much attention and probably should.
No doubt you're correct.
THD figures are useful if used appropriately. Also they are useful in different ways at different times.
One simple and valid use is in testing, adjustment and diagnosis of an amplifier. For example when trying to adjust an amplifier for optimal bias, if you have equipment THD+N is probably the most straight forward approach. Similarly it may be easy way to evaluate the effect of modifying component values such wether or not a baypass cap is "big enough". THD and noise are cummulative, or so I've been told, for a system so if they are easily reduced in one area without compomising anything else it certainly is sensible to do so as it leaves you with more room for trade-off and options elsewhere.
The issue of measuring THD at rated power is interesting. I suspect it stems from the "power ways" of the late 70's. Hirsch used to report and amps power at some fairly high distortion figure (1%?) as a way to keep the power rating on a comparable basis. Even if you can't bear listening to 1% THD, at least the method was an attempt to provide a ranking figure with which different products could be comparred. Such quoted THD numbers seem to be qualified with the words "at 1kHz" or "< x% from 20 - 20k Hz", or "worst case" -- all of which seem pretty useless to me. If I were dictator, a frequency vs. THD @ 2/3 power would be a requirement plus another which was power vs. THD @ 1k and 10k. Or at least something along those lines that relates a little better to listening experience.
The spectral components of THD have been demonstrated to have significance. However, I skeptical of some of demonstrations based on artificially constructed .wav files. One I heard clearly showed a situation in which .1% THD was incredibly horrible but whoever provided it didn't say what the components were. The sound was worse than the worst $10 transistor radio I have ever heard, thus leading be to the conclusion that while technically honest, the demo was a straw man -- i.e., no real world amp no matter how awful ready does that! Looking at a number of spectra from various amps, I think it is likely that they could be grouped in to a small number of common patterns each typical of particular classification of amplifier (tube vs ss, SS symetric topology vs SS original Lin, etc). Within each pattern group, I imaging a simple THD figure consistently arrived at would be a pretty effective relative evaluator.
Regarding audibility. I suspect it is less than many think, but even when not inaudible are often not noticed without some serious non-casual listening.
One simple and valid use is in testing, adjustment and diagnosis of an amplifier. For example when trying to adjust an amplifier for optimal bias, if you have equipment THD+N is probably the most straight forward approach. Similarly it may be easy way to evaluate the effect of modifying component values such wether or not a baypass cap is "big enough". THD and noise are cummulative, or so I've been told, for a system so if they are easily reduced in one area without compomising anything else it certainly is sensible to do so as it leaves you with more room for trade-off and options elsewhere.
The issue of measuring THD at rated power is interesting. I suspect it stems from the "power ways" of the late 70's. Hirsch used to report and amps power at some fairly high distortion figure (1%?) as a way to keep the power rating on a comparable basis. Even if you can't bear listening to 1% THD, at least the method was an attempt to provide a ranking figure with which different products could be comparred. Such quoted THD numbers seem to be qualified with the words "at 1kHz" or "< x% from 20 - 20k Hz", or "worst case" -- all of which seem pretty useless to me. If I were dictator, a frequency vs. THD @ 2/3 power would be a requirement plus another which was power vs. THD @ 1k and 10k. Or at least something along those lines that relates a little better to listening experience.
The spectral components of THD have been demonstrated to have significance. However, I skeptical of some of demonstrations based on artificially constructed .wav files. One I heard clearly showed a situation in which .1% THD was incredibly horrible but whoever provided it didn't say what the components were. The sound was worse than the worst $10 transistor radio I have ever heard, thus leading be to the conclusion that while technically honest, the demo was a straw man -- i.e., no real world amp no matter how awful ready does that! Looking at a number of spectra from various amps, I think it is likely that they could be grouped in to a small number of common patterns each typical of particular classification of amplifier (tube vs ss, SS symetric topology vs SS original Lin, etc). Within each pattern group, I imaging a simple THD figure consistently arrived at would be a pretty effective relative evaluator.
Regarding audibility. I suspect it is less than many think, but even when not inaudible are often not noticed without some serious non-casual listening.
hardly on topic, but:
tommorow i am taking the exam of "Computer-based metrological systems" or something like this, .....(dont know the English name for this)
and I'm studying hard about best measurement systems...
can anybody tell me how they generate the 'pure' sine and measure the voltage on the output with such good resolution....
I bet it is hard to get THD below 0.01% with generator directly to measurement instrument unless you spend thownsends of $.
what do you think?
how about measurements of THD of power amp reliability?
tommorow i am taking the exam of "Computer-based metrological systems" or something like this, .....(dont know the English name for this)
and I'm studying hard about best measurement systems...
can anybody tell me how they generate the 'pure' sine and measure the voltage on the output with such good resolution....
I bet it is hard to get THD below 0.01% with generator directly to measurement instrument unless you spend thownsends of $.
what do you think?
how about measurements of THD of power amp reliability?
on topic now:
a test
please answer yourself these questions:
1. Is your amp always driven to full power and are there no quiet parts of what you listen to?
2. To input you connect a signal source, right... and to the output do you connect a 8-ohm RESISTOR as a elcectroacoustic transducer?
3. Your amp never clips, right?
4. May your favourite tunes and songs be nicely approximated by a 1kHz max. peak-to-peak level SINE wave? (in time, spectrum, wavelet..., any domain)?
If at leastone of your answers is 'no' or ' i'm not sure' then you really have to re-consider your belief in THD.
regards
a test
please answer yourself these questions:
1. Is your amp always driven to full power and are there no quiet parts of what you listen to?
2. To input you connect a signal source, right... and to the output do you connect a 8-ohm RESISTOR as a elcectroacoustic transducer?
3. Your amp never clips, right?
4. May your favourite tunes and songs be nicely approximated by a 1kHz max. peak-to-peak level SINE wave? (in time, spectrum, wavelet..., any domain)?
If at leastone of your answers is 'no' or ' i'm not sure' then you really have to re-consider your belief in THD.
regards
darkfenriz said:
Surprisingly you may have already a system capable of this level of measurement. This was as you say a multi thousands U$ as little as 15 years back.
A moderate quality PC audio board will both generate and accept signals with an overall THD of about 0.01% or better.
A better device still below U$ 200 like the Sound Blaster Audigy II easily receives signals at the 0.001% THD level, while you can improve the generator quality with a simple active filter based on a low noise low distortion device like the OPA3143. I say all this because I am currently using such configuration.
The other required component is a suitable signal generation and analysis software. SpectraLab is an excellent example but costly, yet you can do something for free with some hassle. You can generate with LTSpice the signal you want and save as a wave file, play it, record the return and save to other .wav. Then import to LTSpice and you are done. LTSpice is free.
Rodolfo
I saw some soundcard reviews a few months ago. In the under $100 range Turtle Beech Santa Cruz came out with the best figure. I used a RightMark loop back test on mine and got .006% -- not quite as good as the review but still not too bad. Some people claim better results with one of the M-Audio cards. If money is not much of an object, I'll bet a top of the line Lynx Labs sound card (L22?) with SpectraLab software wiill put you in nearly the same lead as an AP test set.
It is a waste of time to talk about THD, UNLESS you also know the order of the harmonic(s) that are present. .01% 7th harmonic may be more annoying and even more audible than 0.5% of 2'nd or 3'rd harmonic. This has been known for about 70 years at least. It is just conveniently forgotten about by mid-fi manufacturers. ;-)
john curl said:
That's what I remarked
here in this thread, spectral structure is the qualifier.
Unless of course the bundled THD value is really low (0.001% or thereabouts).
Rodolfo
Pavel,
How about viewing it this way: you should be grateful for these people, they make you look good.
And you are good, too, no doubt about it; but perhaps you know this already?
All this nonsense, after all, is just the product of folklore and emotive marketing, and it should be a pleasant challenge for the cognoscenti to sift through all this dross and make the right choices, no??
Cheers,
Hugh
How about viewing it this way: you should be grateful for these people, they make you look good.
And you are good, too, no doubt about it; but perhaps you know this already?
All this nonsense, after all, is just the product of folklore and emotive marketing, and it should be a pleasant challenge for the cognoscenti to sift through all this dross and make the right choices, no??
Cheers,
Hugh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.