I'm trying to decide if an amp's THD spec, by itself, is a strong indicator of amp quality (or at least "sweetness" ) or just one of many factors, and not something to get really hung up on.
Here's my specific situation:
--------------------------------------
For my live performance rig, I try to have the best sound I can get, within the financial limits of a normal gigging musician, but with my tops I'm not currently thrilled.
Years ago I used Tannoy T300's with Bryston class AB amps, and the sound was amazing, but I can no longer deal with all that weight.
So now I have EAW KF394's, which get raves from all who have used them, and they are powered by a Lab Gruppen C48:4.
Well, they excel at throwing rock vocals out to maybe 50 yards, and the clarity is amazing. However, they compared to my old rig, they sound a bit edgy to me, especially up-close. And that's with the greybox processor inline. (To be fair, I'm being rather picky here. OK, I'm being REALLY picky here. 🤐 )
Yesterday I decided to try them with my old Bryston 4B ST, and they definitely sound a little sweeter.
Of course, I'm comparing class AB to class TD, (incredibly GOOD class TD) but I want to try and understand what is the most significant technical difference.
As per my title, I suspect that THD is the main factor: The Bryston's spec (20-20K) is 0.007%, whereas the Lab Gruppen is 0.1%
I also tried my old Lab Gruppen fP6400, which has the same 0.1% spec as its newer cousin. While the C series sounds a little fuller, it is not any smoother or sweeter.
--------------------------------------
So:
I'm wondering if a different amp might help here. -Except it HAS to be class D or TD. Looking at the newest Lab Gruppen PLM +, the THD spec is 0.05%.
I have no way of auditioning this amp against my C series, so I'm wondering your thoughts on this. That number is still not even close to the Bryston, but it's HALF of the C series. (The amp design is also a few years newer, which might be a factor.)
Is it reasonable to assume that this difference in THD should likely translate to a less edgy sound, or is it in fact a very minor difference and not worth too much consideration?
(FWIW, I hate that the PLM's have built in processing. I don't need it, don't want to pay for it, and would rather not have the extra ADA, but it is what it is...)
All thoughts would be most welcome.
Here's my specific situation:
--------------------------------------
For my live performance rig, I try to have the best sound I can get, within the financial limits of a normal gigging musician, but with my tops I'm not currently thrilled.
Years ago I used Tannoy T300's with Bryston class AB amps, and the sound was amazing, but I can no longer deal with all that weight.
So now I have EAW KF394's, which get raves from all who have used them, and they are powered by a Lab Gruppen C48:4.
Well, they excel at throwing rock vocals out to maybe 50 yards, and the clarity is amazing. However, they compared to my old rig, they sound a bit edgy to me, especially up-close. And that's with the greybox processor inline. (To be fair, I'm being rather picky here. OK, I'm being REALLY picky here. 🤐 )
Yesterday I decided to try them with my old Bryston 4B ST, and they definitely sound a little sweeter.
Of course, I'm comparing class AB to class TD, (incredibly GOOD class TD) but I want to try and understand what is the most significant technical difference.
As per my title, I suspect that THD is the main factor: The Bryston's spec (20-20K) is 0.007%, whereas the Lab Gruppen is 0.1%
I also tried my old Lab Gruppen fP6400, which has the same 0.1% spec as its newer cousin. While the C series sounds a little fuller, it is not any smoother or sweeter.
--------------------------------------
So:
I'm wondering if a different amp might help here. -Except it HAS to be class D or TD. Looking at the newest Lab Gruppen PLM +, the THD spec is 0.05%.
I have no way of auditioning this amp against my C series, so I'm wondering your thoughts on this. That number is still not even close to the Bryston, but it's HALF of the C series. (The amp design is also a few years newer, which might be a factor.)
Is it reasonable to assume that this difference in THD should likely translate to a less edgy sound, or is it in fact a very minor difference and not worth too much consideration?
(FWIW, I hate that the PLM's have built in processing. I don't need it, don't want to pay for it, and would rather not have the extra ADA, but it is what it is...)
All thoughts would be most welcome.
A related question:
I notice that many high-end amp manufacturers no longer give specs for damping factor or slew rate.
This seems very odd to me, and a bit frustrating.
Has technology gotten so good that these no longer matter, or do they just not what us to know, for marketing reasons?
I notice that many high-end amp manufacturers no longer give specs for damping factor or slew rate.
This seems very odd to me, and a bit frustrating.
Has technology gotten so good that these no longer matter, or do they just not what us to know, for marketing reasons?
They only give the PMPO value that is what the vast majority of consumers like.
10000, 20000WPMPO into a 4" 10W speaker. Marvelous 🙄 😲😳
10000, 20000WPMPO into a 4" 10W speaker. Marvelous 🙄 😲😳
IMHO the THD value is only relatively useful, good to compare power/dist levels (std measurements); but it has little relation to perceived quality.
You need to consider the spectrum of the distortion.
Anyway, it is a complex subject where you need to weight the human perception.
Our sensitivity to distortion is fairly high at the frequencies of the voice's range; but very low at the extremes of the audio band.
There was a paper from James Moir in Wireless World very interesting, I think it was called "Just detectable distortion levels" or something like this, you may google it.
You need to consider the spectrum of the distortion.
Anyway, it is a complex subject where you need to weight the human perception.
Our sensitivity to distortion is fairly high at the frequencies of the voice's range; but very low at the extremes of the audio band.
There was a paper from James Moir in Wireless World very interesting, I think it was called "Just detectable distortion levels" or something like this, you may google it.
It is true things have change.
Say 50 years ago they listened to amplifiers with THD 1% or even more.
They did good listening.
Nowadays there are amplifiers with a fraction THD. Like 0.001% or even less.
They also give good listening.
Nelson Pass goes against this. His amplifier sure sometimes have more THD.
He is kinda old school.
Say 50 years ago they listened to amplifiers with THD 1% or even more.
They did good listening.
Nowadays there are amplifiers with a fraction THD. Like 0.001% or even less.
They also give good listening.
Nelson Pass goes against this. His amplifier sure sometimes have more THD.
He is kinda old school.
There is no proven direct correlation of sound quality and THD. In fact, the THD measurement
is the easiest possible test for an amplifier. Other tests such as multitone or IM are more difficult.
Use what you hear as the preferable amplifier, that's the whole point.
is the easiest possible test for an amplifier. Other tests such as multitone or IM are more difficult.
Use what you hear as the preferable amplifier, that's the whole point.
I think there may even be some evidence that there's a range of THD people prefer, but IMD and compression sure have an audible impact.
Take a look at Fletcher Munson curves. Notice how the 4kHz area is typically 10-15dB more sensitive than 1kHz. This is a problem for THD. For one thing, THD merely sums the amplitudes* of all the harmonics (up to say, 20kHz) of a test signal, without taking into account their different levels of audibility.
The classic, go-to method for reducing distortion in amplifiers (and sometimes on a per-component basis) is negative feedback, and basically there's an inverted error signal that is fed back to the input. What seems to happen is that the original error, for example the 2nd harmonic, has its spectrum spread all the way out to some limit where the phase is shifted 180 degrees which to neutralizes the negative feedback. And the spread is pretty much flat, with NO built-in avoidance of frequencies where hearing the most sensitive.
Some amplifier designs rely on being 'fast', so that the inherent distortion of the individual components is spread out to the MHz range, possibly resulting in ultra-low THD ratings. Others, pretty much eschew the negative feedback technique altogether, except for modest house-keeping, and then the first few harmonics dominate the THD ratings. Musically, the lowest harmonics are generally the most consonant, except for ones that match a pattern like octaves and so on. Which school sounds better? Listen and hear. Trust your ears!
*PS, in some THD ratings it's just the amplitudes. In others, amplitude and phase both count, and the difference could explain some effects that people sometimes describe, like: an amplifier being 'fatiguing', like it somehow tires the ears despite being 'clean'. I've been processing this in the background for a while and -- what happens when a large number of harmonics (m)align in phase to create a relatively large but narrow spike in amplitude?? The cochlea detects no problem, just lots of frequencies activated at a low level, but the eardrum is jolted. Food for thought!
The classic, go-to method for reducing distortion in amplifiers (and sometimes on a per-component basis) is negative feedback, and basically there's an inverted error signal that is fed back to the input. What seems to happen is that the original error, for example the 2nd harmonic, has its spectrum spread all the way out to some limit where the phase is shifted 180 degrees which to neutralizes the negative feedback. And the spread is pretty much flat, with NO built-in avoidance of frequencies where hearing the most sensitive.
Some amplifier designs rely on being 'fast', so that the inherent distortion of the individual components is spread out to the MHz range, possibly resulting in ultra-low THD ratings. Others, pretty much eschew the negative feedback technique altogether, except for modest house-keeping, and then the first few harmonics dominate the THD ratings. Musically, the lowest harmonics are generally the most consonant, except for ones that match a pattern like octaves and so on. Which school sounds better? Listen and hear. Trust your ears!
*PS, in some THD ratings it's just the amplitudes. In others, amplitude and phase both count, and the difference could explain some effects that people sometimes describe, like: an amplifier being 'fatiguing', like it somehow tires the ears despite being 'clean'. I've been processing this in the background for a while and -- what happens when a large number of harmonics (m)align in phase to create a relatively large but narrow spike in amplitude?? The cochlea detects no problem, just lots of frequencies activated at a low level, but the eardrum is jolted. Food for thought!
Lots of questions and topics there. I'll respond to your thread title question.
The spectrum of the harmonic distortion is more important than the THD. Low order harmonics, such as 2nd, 3rd, and fourth, are much less objectionable than higher order harmonics, such as seventh, eighth and ninth. In fact, many people like a bit of 2nd harmonic distortion.
Thanks, guys. I guess I have my answer.
But this discussion has raised a few other questions, even if only important for a better theoretical understanding:
-------------------------
Dave,
My strong understanding is that we tend to like even-order harmonics, and dislike odd-order. Tube amps tend to make even-order, hence why some folks love those old Mac 275’s. But AFAIK, all transistor circuits create mainly odd-order harmonics, so “less is always more.”
Is that not correct?
Regarding distortion frequency-distribution, that’s a great point and one I hadn’t thought about. But is it possible that this spectrum is relatively the same within each class of transistor amp? (In which case, that THD spec starts to have more significance.) Or are they all over the place?
As far as @abstract’s info, well I was ok until that last paragraph ! That sounds important, but it made my head explode. 😳 - Quite a mess all over the walls and ceiling ….
But this discussion has raised a few other questions, even if only important for a better theoretical understanding:
-------------------------
Dave,
My strong understanding is that we tend to like even-order harmonics, and dislike odd-order. Tube amps tend to make even-order, hence why some folks love those old Mac 275’s. But AFAIK, all transistor circuits create mainly odd-order harmonics, so “less is always more.”
Is that not correct?
Regarding distortion frequency-distribution, that’s a great point and one I hadn’t thought about. But is it possible that this spectrum is relatively the same within each class of transistor amp? (In which case, that THD spec starts to have more significance.) Or are they all over the place?
As far as @abstract’s info, well I was ok until that last paragraph ! That sounds important, but it made my head explode. 😳 - Quite a mess all over the walls and ceiling ….
Last edited by a moderator:
It has been shown fairly definitively that THD is a useless measure.
But the spectrum of harmonics that are the source for this single number is very useful. In the act of collapsing the grah of harmonics into a singlr number so much information is lost and mashed together that amplifiers with VERY different distortion profiles (and sonics) can end up with the same number from very different spectrum distribution.
So Harmonic Distrotion as expressed in a chart, is useful, but once collapsed becomes not vert good at all.
In the late ‘70s we learned to become very wary of receiver swith vanishingly low singlr number THD, since they usually got smoked by designs paying more attention to sonics.
dave
But the spectrum of harmonics that are the source for this single number is very useful. In the act of collapsing the grah of harmonics into a singlr number so much information is lost and mashed together that amplifiers with VERY different distortion profiles (and sonics) can end up with the same number from very different spectrum distribution.
So Harmonic Distrotion as expressed in a chart, is useful, but once collapsed becomes not vert good at all.
In the late ‘70s we learned to become very wary of receiver swith vanishingly low singlr number THD, since they usually got smoked by designs paying more attention to sonics.
dave
You are associating tube amplifiers with a characteristic that is really down to amplifier topology. Singlr Ended amplifiers tend to have more 2nd than 3rd, a Push-Pull amplifier by its nature cancels 2nd order. Harmonics higher than 3 are most often generated by feedback and tend to be undesirable even at very small amounts (since they are unnatural).
Your confusion likely lies with the fact that so far (at least commercially) the set of Single Ended ampliers is largely made up of tubes. You will find more than a few SS SE amplifiers (often designed by, or inspired by Mr. Pass) kicking around this forum.
I am pretty sure the Mac 275 is PP
The ACA is an intersting example. a single one is SE, a pair bridged PP, so your guess as to why people like it does not stand up.
dave
Last edited:
THD at low frequencies (say 1kHz or lower) is a good indicator of the basic linearity of an amplifier, ie. how well it respects the Vout/Vin relationship, which should be a straight line.
If the basic linearity is not good, you don't need to look further: under more difficult conditions of frequency or signal mix, the situation will worsen.
Of course, a single 1kHz figure doesn't tell all the story, and even a well-behaved amp under those conditions might create atrocious artefacts at higher frequencies, with mixed signals, etc. Think of early, simplistic class D amplifiers.
This observation lead some gurus to advocate harsher tests, to root out the "weaker" amplifiers: a typical example is the 20kHz THD, seen as a golden standard by some.
Unfortunately, such a test has no actual relevance for audio: we cannot hear the harmonics of 20kHz, and any transmission chain or media would fail miserably or succeed exceedingly well for the wrong reasons: think of FM broadcast or digital encoding schemes.
With modern, complex audio encoding systems, the listening test under control conditions is the only possible criterion, but if you assess a simple link like an amplifier, start with the THD (IMD will be closely linked, because the distortion mechanisms are the same unless the device has some pathological behaviour).
If the THD level is high, the spectral profile will have an influence, but if the THD is <0.01% for all levels and frequencies below 10kHz, the profile is not going to matter, at least in double blind tests: people claim to have the ability to discern much lower levels, but objective tests do not confirm this, at least for "simple" devices like amplifiers.
As soon as complex signal processing is involved, this can change because various unnatural effects are introduced, and these can be detected at much lower levels
If the basic linearity is not good, you don't need to look further: under more difficult conditions of frequency or signal mix, the situation will worsen.
Of course, a single 1kHz figure doesn't tell all the story, and even a well-behaved amp under those conditions might create atrocious artefacts at higher frequencies, with mixed signals, etc. Think of early, simplistic class D amplifiers.
This observation lead some gurus to advocate harsher tests, to root out the "weaker" amplifiers: a typical example is the 20kHz THD, seen as a golden standard by some.
Unfortunately, such a test has no actual relevance for audio: we cannot hear the harmonics of 20kHz, and any transmission chain or media would fail miserably or succeed exceedingly well for the wrong reasons: think of FM broadcast or digital encoding schemes.
With modern, complex audio encoding systems, the listening test under control conditions is the only possible criterion, but if you assess a simple link like an amplifier, start with the THD (IMD will be closely linked, because the distortion mechanisms are the same unless the device has some pathological behaviour).
If the THD level is high, the spectral profile will have an influence, but if the THD is <0.01% for all levels and frequencies below 10kHz, the profile is not going to matter, at least in double blind tests: people claim to have the ability to discern much lower levels, but objective tests do not confirm this, at least for "simple" devices like amplifiers.
As soon as complex signal processing is involved, this can change because various unnatural effects are introduced, and these can be detected at much lower levels
Noise is being missed in this discussion.
THD + N is more useful as a figure of merit. A little Harmonic Distortion is often unnoticed.
THD + N may be an indicator of IMD or crossover distortion and such things that are much more objectionable.
ASR is starting to use 100dBs SINAD as a target for amplifiers.
THD + N is more useful as a figure of merit. A little Harmonic Distortion is often unnoticed.
THD + N may be an indicator of IMD or crossover distortion and such things that are much more objectionable.
ASR is starting to use 100dBs SINAD as a target for amplifiers.
Last edited by a moderator:
Ah yes those THD vs power graphs. I cringe when I see a steady downward slope from 0.1mW all the way to 50W or so, where just above that it maxes out and starts clipping.
For just about any audio application I can think of, best performance should be at low power. I can imagine the reviews: terrible, gritty vocals, but damn those gunshots sounded so smooth!
For just about any audio application I can think of, best performance should be at low power. I can imagine the reviews: terrible, gritty vocals, but damn those gunshots sounded so smooth!
Hey @planet10.
I will start setting up Single Ended amplifiers, practice what I learnt by studying Nelson Pass stuff.
But only in SPICE 😵
Preferably without global feedback
With noticable thd, the sound is always altered.
Some naturally pure tones like vocals or certain piano notes in the tonal midrange that are low in overtones, can not be reproduced with the original purity and the contrast to „richness“ gets lost, details are getting masked.
Crossover distortion at low levels in AB amplifiers and multibit DACs is worst.
Some naturally pure tones like vocals or certain piano notes in the tonal midrange that are low in overtones, can not be reproduced with the original purity and the contrast to „richness“ gets lost, details are getting masked.
Crossover distortion at low levels in AB amplifiers and multibit DACs is worst.
THD is a 3-dimensional function: one variable is signal level, the other variable is signal frequency.
Most natural sounding is an amplifier that has monotonously increasing THD vs. level, that is vanishing THD at low level and increasing with level up to a reasonable (low) amount. As others pointed to it, distortion spectrum is also important, it also varies with level. It is best if 2nd dominates over the full level and frequency range. Multitone distortion (IMD) is another important parameter. And there might be others, too.
Most natural sounding is an amplifier that has monotonously increasing THD vs. level, that is vanishing THD at low level and increasing with level up to a reasonable (low) amount. As others pointed to it, distortion spectrum is also important, it also varies with level. It is best if 2nd dominates over the full level and frequency range. Multitone distortion (IMD) is another important parameter. And there might be others, too.