When optimizing, go for Lowest 3rd harmonics .. not THD!!

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Disabled Account
Joined 2006
While I feel it is better to have a favourable distortion spectrum where the H2 dominates the others, I agree with Nelson that one should be doing targeting this in view of having the lowest possible distortion. H3 is not so bad anyway, the higher odd and even are more damaging. Build Nelson Pass simple Jfet son of Boz which can be optimised to have only very low levels of H3 and youll realise that H3 is not bad at all. Amps with very high levels of H2 and very little H3 sound muffled, optimum that I find is a declining order of all the harmonics from H2 and preferable where after the third they are either absent or so low as to be insignificant. This is well known and thoroughly tested among tube lovers. Ever wondered why the symanym sounds as good as it does, look at the FFT spectrum.

Too much emphasis is being put on how amps sound, how does your source sound, and your speakers?? Junk into a amp, junk out of your amp, sometimes a less than optimum amp to compensate for bad sources is better, an amp should ideally be a wire with gain. Having some level of H2 in amps is good because cdplayers in particular sound cold and the H2 seems to disguise this well. Ever pondered why tubes have be
 
We are extremely fortunate in that we can build something close to what we are looking for in sound reproduction.

It is not often that the poor jerk in the street has the opportunity to specify how much and what order of harmonic content he would prefer.

This guy has to read though piles of magazines to find and audition that uses words that he can comprehend, only realizing after he made his megabuck purchase that the reviewer used the words in the wrong context.
 
Disabled Account
Joined 2006
Nico Ras said:



I was joking:idea:


I was not, I really find this with some, I also fall into this mind trap some times, listen to a original hiraga amp. This amp creates incredible sounstage which isnt there on your source player cdplayer. :)

Lineup has placed a very interesting subject here and I feel he has a valid point in trying to lower odd order harmonics in relation to even order.

Another good example I read here was with nelsonvandal, maybe unknowing the effect it has he exhanged the output trannies of his headphone buffer to less performing one, in reality he sacrificied ultimate low THD for a circuit which would have somewhat higher H2 and even harmonics compaired to the H3 and odd ones and he prefered it.

Nico very good point about the magazine articles, I was fooled many times exactly as you mention. The salesmen would run away from me because Id exchange an amp or other component a half dozen times before finding one that suited me and the existing components with which that component was to be coupled. Later they rather started to bring samples of various components to my home for auditioning before purchasing. :D

One man s gold is the other man s lead.

Sorry for the OT.
Lineup you placed a interesting transconductance Jfet amp circuit in the passlabs forum some time back, could you make a link here for it or email it to me please, Ive lost my bookmark. I would like to try it as a possible preamp or headphone amp sometime.
 
THD vs. Harmonics by 3 different Biias Currents

.
Here's a practical example from my simulation work.

One 4 transistor HeadPhone amplifier for 32 Ohm load.
IRF MOSFET output transistor in Class A current.
Global negative feedback, but low feedback factor.

Attachment shows 3 bias levels, alternatives.
- RED adjusted for lowest THD (0.007%)
- BLACK adjusted for lowest 3rd
- BLUE adjusted for lowest 5th


Many would actually without further contemplation go for 0.007% THD :rolleyes:

Myself, I would probably go for BLUE = 220 mA
Even if this means 0.037% THD.
Because the low 4th, 5th, 6th order dist looks alright to me.

Sometimes I have wondered why Nelson Pass and other members here
overbias their MOSFET Class A output stages.
Have seen this in Headphone Amplifiers and in other types.
Could it be that the higher biased MOS have a better sounding performance
due to better harmonics spectrum?

Lineup :) regards
 

Attachments

  • biascurrent_thd-vs-harmonics1.png
    biascurrent_thd-vs-harmonics1.png
    4 KB · Views: 512
roender said:
There is only one way which could bring down the 7th, and this is the hard way ... dumping as much as possible bias current into the output stage, many transistors correctly biased in class B as per Douglas Self definition.
I would probably go for BLUE = 220 mA
Both of these seem to be going in the same direction as the generally held view that FET output stages benefit from more bias which is quite different from ClassAB BJT that have a narrow range around optimum.
Borbely says >=100mA per device AND >=500mA total output bias.
 
homemodder said:
Compare to a good old record player or a reel to reel recorder, and come to speak to me again.


For HiFi use I would never dream of comparing a link to another. Analog gear is very colored and I want neutral gear, as close to transparent as possible.

When I test DAC's I compare the orignal signal to the one coming out of a AD-DA loop. By combining various AD and DA it's possible to nail down if the DUT is transparent or if it colors the signal audibly.

But sure, if you have tailored a stereo set up to a source that have audible HF roll off and inserts a linear CDP then of course it will sounds cold or bright. But that is not the same as saying the CDP sounds cold. Modern digital is so close to transparent that most people can not tell the difference between the original and the same material that has been looped to an DA and recorded again with an AD. That is telling if anything.


BTW the reel to reel the one Ive tested together with a family member and was used in studios and produced the well known album of yello, produced mostly H3 distortion.

Music production (producing a sound) should not be confused with reproduction.

Digital sources have become much better as time has passed especially the high end ones,

Budget stuff performs close to flawless and many times as good as the really expensive gear.

if one looks why youll see the use of tubes or discretes with the sound of tubes, again looking for H2 to disguise some shortcomings.

Absolutely not. Modern IC's is so good these days that they are transparent, no audible coloration. No need for discretes or tubes if you want clean sound. If you're after some subjectively pleasing coloration, well that's another story.

We now have a more serious problem affecting cd, record producers are damaging the sound with their quest for sound levels.

Now we're talking. That's a huge problem and also of course there have been CDP's of poor design but it's not "fair" to blame the format because some manufacturers are incompetent and ignorant.

Yep, this thats the right way to evaluate a amp, I get my sister to play the violin and sax, while my brother in law is a pretty handy with drums and electric guitar.

Why would that be the right way? Doing that you compare the whole playback chain, the room and the shortcomings of the stereophonic system as such against a real instrument. If you change the amp to another in such a test and get an audible difference you still have no clue about the performance of the amps since you may be compensating for errors somewhere else in the chain.

Why not choose some good material and compare the input signal to the amp with the output? That way you can learn something about the amp and what it actually do to the signal.



/Peter
 
thank you for the interesting thread.

for me it has raised the following question: what does changing one harmonic over another do to the original signal? In other words if you compare a scaled version of the music waveform on your CD/record/tape to the output of the amp - what are the differences.

i know this was partially addressed when it was suggested that odd harmonics speed the rise time while even harmonics slow the rise time. nevertheless it think this might be a useful experiment.

i would pick a track where I felt that i heard the biggest difference between two different amplifier conformations.

when doing something similar myself i have been convinced that amplitude distortions, the type of distortion that FFT's measure, is essentially irrelevant at the levels produced by most modern day amps. BUT i have not done the exact experiment mentioned above yet so i am prepared to be convinced otherwise. However, as of right now my feeling is that you can't ask an amp to create something that is not in the original signal and that the amplitude distortions produced by today's amps are essentially inaudible.

i think that the space for biggest improvement in audio might be for amplifiers to actively offset the large distortions produced by loudspeakers.
 
Most distortion phenomena have a compressive effect so that odd order harmonics would increase risetime I see as incorrect.

On a simple sine wave odd order harmonics simply slightly compress the positive and negative sides of the sine slightly and symmetrically. It would not bee seen on a scope unless we are talking about really high distortion levels.

Asymmetricall nonlinearities results in dominantly even order harmonics which means the positive and negative part of the sine wave will look different (if magnified).

High levels of even order is not "musicall"as many seems to believe. Nonlinearities always leads to intermodulation distortin, not only harmonic distortion.

Transparent audio gear is my goal since I find all the beauty in the recordings.


/Peter
 
Pan said:


On a simple sine wave odd order harmonics simply slightly compress the positive and negative sides of the sine slightly and symmetrically. It would not bee seen on a scope unless we are talking about really high distortion levels.

/Peter

Hi peter.

Look at the waveform and see that you only have half the picture. It demonstrates that the slope (rise time is faster) as well as being compressed. This is a sine wave with some H3 to demonstarte that this.

Nico
 

Attachments

  • untitled1.jpg
    untitled1.jpg
    24.5 KB · Views: 410
Yes, it was a simplification but strictly speaking isn't rise time the time from 10%-90% of a step function?

Is rise time really applicable on a sine and "sine compound waveforms" including several partials? Guess it is! :)

edit: But depending on which part of the derivate we are looking at we may come to the opposite conclusion.


/Peter
 
By andrew T. - I wonder if there is a cheap way to actually perform the adjustment of minimising the third and the other odd harmonics?

look around guys,... mike B's symasym "gross" compensation
(68K/330p RC) at VAS output takes care of H3,5,7 and
quasi's ???? 'secret RC" (47R/330p)between the differentials
affects the phase of all the above. Comp. is the hardest
to accurately simulate as layout and other factors
complicate things..
OS
 
ostripper said:


look around guys,... mike B's symasym "gross" compensation
(68K/330p RC) at VAS output takes care of H3,5,7 and
quasi's ???? 'secret RC" (47R/330p)between the differentials
affects the phase of all the above. Comp. is the hardest
to accurately simulate as layout and other factors
complicate things..
OS

I do not believe in all cases, this is not a universal magic fix. I guess these components are best selected with a live application.

Sim shows increased in THD one order of magnitude on the amp I am working on now.

Nico
 
It was one guy in forum that talked about 'natural distortion'.

In nature, of course, there are harmonic distortions, too.
and most any kind of distortions.
Not only in sound coming from loudspeakers.
We can guess there are forms of distortion that are more un-natural
than other forms of distortions.

For example in a big cave we have a certain echo from the walls
and this gives the special sound, like in a church.

Are there such things as more natural distortions?
Some things that our hearing have easier to accept,
because it is in our Human DNA.

The DNA for hearing that was programmed not long ago, when there was no loudspeakers and no electricity.
 

Attachments

  • biascurrent_thd-vs-harmonics1.png
    biascurrent_thd-vs-harmonics1.png
    4 KB · Views: 350
You guys are aware that push-pull - or any symmetry like balanced etc - generally cancels 2nd harmonic?

-> 3rd harmonic dominant.

Well it's single-ended for you guys then ;)

Have fun, Hannes

PS: 2nd, 3rd are masked in audio and therefore less perceptible as higher order harmonics (7th etc).

EDIT: I always find such discussions entertaining - Nelson Pass writes 2nd harmonic is said to be more pleasant, but in listening sessions more people prefer 3rd harmonics :D

As Douglas Self wrote so nicely: "In audio where agreement of any sort is rare..."
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.