THD is meaningless, the harmonic data is what matters

[BigE]

I 1st got a footing on the reality of the usefullness of THD in the late 70s… and largerly because published numbers were way too small… 0.0000x kinda thing. These amps used excessive feedback to get vanishingly low numbers and sounded crap.

As soon as you collapse the various harmonics in an HD graph onto a single number -- or rven a single curve -- you loose all of the relevant information, the result is useless at telling how good or bad the amplifier is.

And as Ed says, it is the 1st watt that is important (to paraphrase Pass), and this is rarely shown.

Passes mix of 2nd & 3rd is only a but useful to designer but tells us consumers nothing and says nothing about how the amp sounds. The mix, the ratio of 2nd & 3rd us very important and how much higher level harmonic crap there is VERY important. A 1% THD amp with only a nice mix of 2nd & 3rd order distortion will often sound better than a 0.01% amp that has a significant amount of higher order harmonics in it.

dave

I remember the 70's too. Which is why I become interested with low figures.... ie. how were they acheived?

You may *prefer* huge amounts of the right distortion to the low distortion from high NFB amps, but in my book neither is acceptable.

 

[planet10]

You may *prefer* huge amounts of the right distortion to the low distortion from high NFB amps, but in my book neither is acceptable.

I like a low distortion amp… what i am saying is that THD tell us little about that.

dave

 

[xrk971]

When a THD+N value is unmeasurably small, it must mean something? Take "The WIRE LPUHP v2" it uses parallel ultra low noise head phone amps to get 0.002% THD+N at 10 watts and has a SNR in excess of 130dB.

http://www.diyaudio.com/wiki/The_Wire_-_All_Boards_and_Kits_Explained_Here?redirect=no

 

[planet10]

When a THD+N value is unmeasurably small, it must mean something?

The amp can reproduce a steady-state sin wave pretty well.

dave

 

[xrk971]

The amp can reproduce a steady-state sin wave pretty well.

dave

That amp is DC to 120kHz capable (-3dB, 1 watt into 8ohms). So probably does pretty well with non sinusoid signals.

 

[planet10]

That amp is DC to 120kHz capable (-3dB, 1 watt into 8ohms). So probably does pretty well with non sinusoid signals.

That spec tells you nothing more about how good it is.

dave

 

[xrk971]

That spec tells you nothing more about how good it is.

dave

This is the amp:

Do these plots tell more?

http://www.diyaudio.com/forums/vend...-here-bal-bal-se-se-lpuhp-15.html#post3700321

Here is the HD plot for 1kHz input (yes, that is a -160dB noise floor):

And THD ratio at all frequencies, and what we are saying is that when this value is so low everywhere - it means something because it means there is no perceptible HD at any harmonic anywhere:

I am really just trying to tell some folks here in Full Range about a potentially very cool amp designed by one of the members here on diyAudio (opc). Not to try to convince P10, because he will forever find a gripe or excuse to poopoo measurements and hard data.

For those of you who are looking for a clean amp to power high sensitivity speakers, check it out. Certainly if you like the ACA with only 5 watts, you might like this one with 10watts at 0.002% THD or 1% THD at 16watts - price is probably comparable or less to an ACA.

 

[planet10]

Yes, that says a lot more. The low noise floor is a big factor. But if one looks at the harmonics 3rd > 2nd. Not ideal, but almost unavodable in a PP amp. Lots of higher order harmonics which is something you don't like to see -- especially when they are as high as the 2nd, and lots of spurious peaks that one has to worry about where they come from. Anything higher than 3rd is not "natural" so the ear/brain is particularily sensitive to them.

And it is only a 1k spectrum.

Is this at full-power? if so it is unlikely the amp will ever be there in normal operation.

dave

 

[xrk971]

The 0.002% is at 10 watts and look at it for all freq, same 0.002%. So even with the higher orders, they are way down in the noise of unperceptible.

 

[planet10]

they are way down in the noise of unperceptible.

Kinda low power for those high order harmonics to show up… it doesn't take much of them to get in the way. The very low noise floor is a big help. You'd have to listen to it to see how much.

I would not take the graph at face value until all the spuroius peaks are explained.

dave

 

[xrk971]

The largest peak (3rd harmonic) is -110dB below the main. That is better than the dynamic range of a CD player or in other words way less than 1 significant bit in 16bit sound. So does it even matter?

 

[planet10]

The very low levels are good. That the higher harmonics are so high reletive to the 2nd & the 3rd are worrying. It is a good result and shows the amp has promise. You still have to listen to it.

If you are looking for a measure that will tell you how an amp sounds we are all still looking. They can tell you wehn something is so bad it is unlikey to sound good, but once it gets to a certain point there is somuch more to consider. A lot of information is hidden by using sin as a stimulus… but it does make the "math" much easier.

dave

 

[wintermute]

sorry folks I didn't realize when I split this that it inherited the temporarily closed status. I've reopened the thread!

 

[Sangram]

I have spent a long time on the objective side of the fence and my overall learnign was pretty consistent with what Dave says. There is nothing in the measurements that will tell you how an amp will sound, whether a given person or kind of listener will like it.

Owen's amp must be superb. There is no doubt that it adds no colouration at all to the signal, and you could say the same for many designs on this site - Roender's FC-100 and Tom's Modulus spring to mind.

However, a lot of people do not necessarily like uncoloured sound. Added to that is the very strong suggestion that amps with a simple distortion profile - such as some of the ultra-simple Pass Labs designs, and some SET tube amps - sound better to the ear. Such amps would typically exhibit highish 2nd and 3rd HD, but the higher orders would fall off in orders of magnitude.

I don't think that anyone doubts the solidity of the low-distortion goal, given that it's all we have at the moment. And no doubt about which sort of an amp you would choose for measurement purposes, or listening, quite literally, to 'What it says on the tin'. However for example Class D amps tend to exhibit roughly 3 to 4 times their distortion at higher frequencies than at the totally misleading 1khz number.

THD is a single measurement at a single frequency, and it is completely useless to compare amps without looking at IMD, clipping behaviour, and THD vs frequency at various power outputs. It will tell you a little more than THD, and some sense can be made out of it.

When testing my own builds ,I ignore THD numbers completely and look at the FFT with increasing frequency and power. That is much more illuminating. Or at least, I think so 🙂

 

[SY]

There is nothing in the measurements that will tell you how an amp will sound, whether a given person or kind of listener will like it.

In order, nonsense and true.

 

[DF96]

However, a lot of people do not necessarily like uncoloured sound.

How true.

 

[jcx]

even the "harmonic structure", "2nd lower than 3rd is bad" is not very credible given actual practice, numbers from the heyday of analog recording

analog magnetic tape, the studio masters that are lusted after by audiophiles have predominantly odd order distortion

"0 dB" on premium mag tape is set at 3% 3rd harmonic from saturation - and analog "0 dB" has always just been "advisory" often exceeded

in fact modern digital recording production may bounce to tape just for the coloration of the tape' "natural compression" - which is predominantly 3rd, odd order, very little 2nd


and "superior" balanced/bridged electronics can give substantial even harmonic cancellation too


many "harmonic structure" arguments should be circular filed alongside single point THD

 

[Pano]

Because it depends. I used to think that harmonic structure sounded the same across the audio band. To me, at least, it does not. 3rd order (and a little 5th) can sound rather nice in the low end. Higher orders sound harsh when the fundamental is >1KHz.

I don't have any kind of definitive list, just have noticed that the same harmonics don't have the same flavor in different part of the audio band.

 

[Xaborus]

Because it depends. I used to think that harmonic structure sounded the same across the audio band. To me, at least, it does not. 3rd order (and a little 5th) can sound rather nice in the low end. Higher orders sound harsh when the fundamental is >1KHz.

I don't have any kind of definitive list, just have noticed that the same harmonics don't have the same flavor in different part of the audio band.

Conical reed instruments (saxophone) produce odd and even order harmonics. Cylindrical reed instruments produce odd order harmonics. Different instruments use different parts of the audio spectrum. This is useful determining what type of harmonic distortion you prefer at different frequencies.

I think its pretty obvious that guitars have become the most popular/loved instruments. This is partially, I believe, do to the fundamental frequency range falling within the same spectrum as humans (85hz to 1k). But also, part of this is the odd and even harmonics created with string instruments, which to my knowledge the human voice also creates. That is my own bias and opinion though. There is nothing wrong if you prefer odd-only harmonic overtones.

This chart may be helpful exploring the frequency range of differing instruments.

 

[jcx]

but the distortion harmonic structure for audio electronics discussions often seem to miss IMD sum and difference products being the bigger audible problem

and the most worthless "harmonic (distortion) structure" comments are warnings that "the worst" are the high odd order when it is hard to make isolated high odd order distortion components
they really only show up accompanied by much higher lower order - so the "harmonic structure" is good?