THD Total Harmonic Delusion?

Once more i find my self tredding along the long path of audio engineering when anougher set of awkard questions are found buzzing about inside my head that demand attention from the masses.

First the question :-

Does vanishingly small amounts of THD in an amplifier GUARANTEE good sound and fine music reproduction?

Explanation :-

I've been developing a classA amplifier for a while and its been on the test bench for far too long now. The problem is i want it to be my best work to date but i'm convincing myself that it dosn't sound that good. A few weeks ago after the project stalled i thought its about time i did some serious research so i started reading loads of papers about the subject of distortion and its effects. I also bought a copy of Selfs 'Audio Power Amplifier Design Handbook'. After reading the book cover to cover i decided to apply some of Selfs methods for reducing THD across the entire amplifier design. After doing this on the sim my amp went from approx 0.065% / 0.016 THD 50/1 Watt to 0.002 / unmeasurable ( < 0.000 ). I was impressed by this figure ( Well done Self ) and so spent this easter mirroring the improvments in the real world on the test bench amp. Upon power up i started to make measurments. Everything seemed good, no measurable THD on the computer via RMMA ( crap soundcard but it said the previous attempt was not that good lol ), good squares and sines on the scope with no noise, 20v/us rise and fall etc. I am convinced there has been a dramatic improvment just as the sim said there would be. I then swapped test gear for audio gear and had a listen. The thing is, i expected the sound to be crystal clear, much improved on the last time i listened but i was dumbfounded to hear it had hardly changed ( if at all ). Hence the question above, if an amplifier has very low THD then why does it not sound as good as amplifiers i have thrown together in the past that use voltage regulators as current sources and have relativly high THD's? Am i imagining this? Are there other hidden distortion's / imperfections that can have a large impact on the perception of sound?? I'm confused but these questions need answers if i'm to carry on my journey into sound.

Thanks for reading
Leigh
 
Hi Leigh,


I think most people would agree that the correlation between THD and actual sound quality is very loose at best. Surely, if you have loads of THD, the sound reproduction will suffer, but the opposite is not necessarily true.

Some things that come to mind: If you say it doesn't sound good enough, compared with what? Are you swapping amps and feeling the quality goes down? If so, what amp is that?

What kind of speakers do you use? How's your amps damping factor? Have you tested the amp for oscillatory tendencies with complex (speaker-type) loads? What's the speaker min impedance, and what's the amps max undistorted output (volts, watts?)

Just some thoughts.

Jan Didden
 
Hi,

I'm pretty sure the amp is stable, i've done extensive testing with capacitors, very low impedencies and even blown a few car headlight bulbs with the amp ( i shouldnt i know ). No tendencies for any oscillation at all. I'm comparing the amp to other home brew amps i've done and conclude that some of my other rush jobs building with only junk parts sound better, even though they are quite primitive. This is the first time i've paid any real attention to quality of sound so i'm rather frustrated to find how hard it is to make any kind of large impact in this area. The amp in question slews at around 20v/us and is capable of 60V Pk-Pk before clipping the rails. Mode of operation is push pull class A. Intended power is approx 50WRMS into 8/ 100 into 4, Full classA up to at least 50. This is not about the amp tho... its about the theroy that less means more in THD terms.

Regards
Leigh
 
nitrate said:
[snip] This is not about the amp tho... its about the theroy that less means more in THD terms.

Regards
Leigh



OK, I thought this was about why your new amp sounds 'bad'...

Maybe you have just proven (again) that THD isn't the whole story. But there are so many other factors that really start to dominate how an amp sounds once the THD is reasonbly small. As I mentioned, a relative low power amp with inefficient/low impedance speakers may clip more than you think. If your other amp has a lot more power, it may sound better even if its THD is higher.

Then there's the whole field of power supply and power supply ground loops that can have a *large* impact.

Can you describe the 'badness' of the sound?
Jan Didden
 
I'll have a go... I cannot realy explaine why the sound is worse other than it seems like there is a veil over the sound obscuring the detail. I mean the detail is there and everything is clear but theres somthing thats causing fatigue when i listen. I feel forced to listen hard for all those subtle details. They aren't missing, they just dont leap out. Silly i know but thats as close as i can get to describing why i THINK the sound is wrong. Maybe i'm just convincing myself its bad when it isn't but like i say, i AM convinced theres somthing not quite clear. Seems to me the more feedback i use the worse this phenomenon gets. The comparrison amps in question are all lower power BTW

Leigh
 

pinkmouse

diyAudio Moderator Emeritus
2002-04-03 7:15 pm
Rotherham, England
Perceptions. ;)

When I first used the Audax HD3P tweeter, I thought it was dull and not very enthusiastic. As I listened to it more, and did some very basic measurements, I realised what I was used to was the relatively higher distortions in the metal dome tweeters I had been using. My ears had been fooling me.
 

jcx

Member
2003-02-17 7:38 pm
..
its hard to convince people of the importance of level matching in listening tests

SPL has to be very, very closely matched or you Will hear differences - and at <1 dB level matching the differences will not be perceived as loudness differences, most commonly the actually louder output will sound "better" - even if you are listening to the same circuit between the 2 tests

http://www.provide.net/~djcarlst/abx_crit.htm

if you are serious about gauging your amplifiers by subjective listening tests it could be a good idea to understand what generations of audio researchers have determined about how to turn subjective impressions into usable data - level matched, blinded listening, controls, with comparisons made within our short term memory time horizon
 

kevinkr

Administrator
Paid Member
Tell us a bit more about the differences in topologies and complexity of the "blameless" amplifier you are working on and the ones which you are using for comparison.

I went through this phase early in my hifi design odyssey and discovered to my chagrin that the more complex and better measuring the topology the less likely it was to sound good to my ear. Simple direct circuit paths are often the best.

Take a look at what Nelson Pass or John Curl do, their stuff is transparent, and good sounding even to this tub-o-phile.

Perceptions may also play quite a large role here, something that sounds radically different from what you are accustomed to may not at first be recognized as an improvement.

Incidentally I found my nirvana by stepping back into the past. Simple vacuum tube circuits did it for me.

Edit: I meant to comment on the previous poster's comments about level matching. I agree totally. Critically important..
 
Leigh,

If you look at the spectral display of the amp's output in RMAA with a sine wave input, do you see any frequency components separated from the main one by 100 Hz? These won't show up in a THD measurement but are distortion nonetheless.

If you do see that, and it's not a grounding quirk of your measurement setup, it can indicate the need for more aggressive filtering of the power supply to the amp's front end.
 
So it appears that the general view here is that THD does not directly correlate with the quality of sound. So an amp with say 100 times the THD of anougher can actually be perceived as being the better quality amplifier in terms of its audio output??

I think i'm imagining half of this but my ears tell me somthings amiss.

Leigh
 
Pan said:
Kevin,

Take a look at what Nelson Pass or John Curl do, their stuff is transparent, and good sounding even to this tub-o-phile.
How did you come to the conclusion that Pass and Curl amps (and which of them) are transparent?


/Peter


I would have to do some digging to determine just what I have listened to, and none of it that recently. My first exposure to something designed by John Curl was with the loaner of a pair of JC-80 pre-amps in the early 1990's, compared to the Yamaha C-2, a Phase Linear and some of my own tube designs it was quite a bit more transparent, revealing detail on lps that I had not heard before in my own system. Careful level matching was performed and several other people listened. It motivated me to redesign my tube pre, which up to that I point I had thought was pretty resolving. I've built several of the complementary fet buffers for use with mc cartridges and thought they sounded better than the cheap transformers I could get my hands on at the time, and far better than the MC inputs on that Yamaha C-2.

Several acquaintances recently have built Nelson Pass power fet based designs from AudioXpress which to my surprise sounded good in my Onken/horn based system.

One of my close friends owns an X5 and AAD 2000 series speaker systems which is the most recent exposure I have had. (Bench Mark Dac 1, Lambda Drive, home brew media server, Pass pre - model escapes me at the moment. )

FWIW That's just my personal opinion.. And I don't spend much time listening to stuff designed by other people. I just thought it might be instructive to take a look at the work of several well respected audio professionals, hopefully that is not an issue...

Clearly you have a (somewhat?) different opinion, and a different experience so why don't you share those thoughts, and how you came to your conclusions. Honestly I don't like much of the solid state gear I hear, but the original poster's comments struck a chord as his experience eerily echoes mine of some 25yrs ago. This is the only reason I posted in fact.
 

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
Erm, very interesting this, been there done that and got the T shirt :D . What Nitrate is saying is the conclusion I have come to over the years. If specifications as we understand them are the whole story then a Pioneer amp I used to own would have been supreme. 150 Watts R.M.S. , distortion so low ( think it was around 0.0015% THD/TIM) that it was irrelevant. Sadly while it was good HiFi it was not good musically, and I to started down the road of listening and seeing what configurations actually sound good, what as I call it, is "compelling listening". I followed Dougs brilliantly written articles at the time, and constructed his "blameless class B amp" on the official PCB,s with a passive preamp. Again good Hifi but musically lacking, and the long and short is that I developed my own design that I am very pleased with. I have compared directly with a commercial class A amp that always gets the vote for best sounding amp, I won't mention any names, but I know which I prefer. Most circuits seem to have a "family" sound, perhaps not that suprising as the topology is very similar between so many designs. Four years with this amp have shown it to have remarkable musical qualities, and just a couple of months ago I swapped back to the "class B" and could'nt quite believe how musically uninvolving it was. Boring to listen to. Nothing wrong with class B in itself however, I use class B in my design. Anyone interested can see details on this forum under "My MOSFET amplifier designed for music"
Regards Karl
 
nitrate said:

I've been developing a classA amplifier for a while and its been on the test bench for far too long now. The problem is i want it to be my best work to date but i'm convincing myself that it dosn't sound that good. A few weeks ago after the project stalled i thought its about time i did some serious research so i started reading loads of papers about the subject of distortion and its effects. I also bought a copy of Selfs 'Audio Power Amplifier Design Handbook'. After reading the book cover to cover i decided to apply some of Selfs methods for reducing THD across the entire amplifier design. After doing this on the sim my amp went from approx 0.065% / 0.016 THD 50/1 Watt to 0.002 / unmeasurable ( < 0.000 ). I was impressed by this figure ( Well done Self ) and so spent this easter mirroring the improvments in the real world on the test bench amp. Upon power up i started to make measurments. Everything seemed good, no measurable THD on the computer via RMMA ( crap soundcard but it said the previous attempt was not that good lol ), good squares and sines on the scope with no noise, 20v/us rise and fall etc. I am convinced there has been a dramatic improvment just as the sim said there would be. I then swapped test gear for audio gear and had a listen. The thing is, i expected the sound to be crystal clear, much improved on the last time i listened but i was dumbfounded to hear it had hardly changed ( if at all ). Hence the question above, if an amplifier has very low THD then why does it not sound as good as amplifiers i have thrown together in the past that use voltage regulators as current sources and have relativly high THD's? Am i imagining this? Are there other hidden distortion's / imperfections that can have a large impact on the perception of sound?? I'm confused but these questions need answers if i'm to carry on my journey into sound.
Did you measure it with loudspeaker you listen or dummy load? :)
 

russel

Member
2005-10-07 9:38 pm
Total harmonic delusion

Hi,

My two pence worth is that like anything else an amplifier is only as good as its weakest section. A low THD into an eight ohm resistor that goes unstable in the real world will not sound that great unless it feeds a loudspeaker that approximates an eight ohm resistor. With my own audio meanderings I noticed early on that one of the most critical parts in designing an amplifier is layout, get that wrong and it will sound, dull flat, undynamic, feel free to add your favourite description of a bad amp. Self has a good section in his book on that, since you say you have got THD and stability right I would look in that area. Its often overlooked or downright neglected even by some well respected manufacturers and once you know what it sounds like its easy to spot. From what you have said I doubt whether you are that far off getting right, keep experimenting.

BR
 
Have you done any IMD tests? I haven't designed much SS gear, so I'm curious the relationship between THD and IMD for a super low THD amp.

Other things I'm curious about the amp:

1) when you did whatever you did to put THD from 0.06% to unmeasurable, what happened to the damping factor of the amp? Someone mentioned this before but I don't recall an answer to it. Sometimes lower DF sounds better with some speakers, so if it was more feedback that was added to reduce THD, the DF probly went up too, which might have more of an effect on sound than a 0.06% to 0.002% THD change.

2) Have you checked to see if the musical peaks become near to amp clipping? (using an o-scope) I'm not sure if i recall correctly but I believe that amplifiers with less feedback respond 'better' to overload situations. I think it has something to do with the front end circuitry not being 'asked' to enter saturation so much. Again if the change you did added more feedback, this may cause a difference in amplifier sound (somebody back me up or correct me on this one if it ain't right)
 
nitrate said:
Once more i find my self tredding along the long path of audio engineering when anougher set of awkard questions are found buzzing about inside my head that demand attention from the masses.

First the question :-

Does vanishingly small amounts of THD in an amplifier GUARANTEE good sound and fine music reproduction?

Explanation :-

I've been developing a classA amplifier for a while and its been on the test bench for far too long now. The problem is i want it to be my best work to date but i'm convincing myself that it dosn't sound that good. A few weeks ago after the project stalled i thought its about time i did some serious research so i started reading loads of papers about the subject of distortion and its effects. I also bought a copy of Selfs 'Audio Power Amplifier Design Handbook'. After reading the book cover to cover i decided to apply some of Selfs methods for reducing THD across the entire amplifier design. After doing this on the sim my amp went from approx 0.065% / 0.016 THD 50/1 Watt to 0.002 / unmeasurable ( < 0.000 ). I was impressed by this figure ( Well done Self ) and so spent this easter mirroring the improvments in the real world on the test bench amp. Upon power up i started to make measurments. Everything seemed good, no measurable THD on the computer via RMMA ( crap soundcard but it said the previous attempt was not that good lol ), good squares and sines on the scope with no noise, 20v/us rise and fall etc. I am convinced there has been a dramatic improvment just as the sim said there would be. I then swapped test gear for audio gear and had a listen. The thing is, i expected the sound to be crystal clear, much improved on the last time i listened but i was dumbfounded to hear it had hardly changed ( if at all ). Hence the question above, if an amplifier has very low THD then why does it not sound as good as amplifiers i have thrown together in the past that use voltage regulators as current sources and have relativly high THD's? Am i imagining this? Are there other hidden distortion's / imperfections that can have a large impact on the perception of sound?? I'm confused but these questions need answers if i'm to carry on my journey into sound.

Thanks for reading
Leigh


Unfortunately, low THD is neither necessary nor sufficient in the genreal case for an amplifier to sound good. There are many reasons for this.

First, if all you are measuring is THD-1 and a single number, that is nearly useless, and is one of the things that has given THD a bad name.

THD-20 at all powers and including spectral analysis of the residual gets closer to being quite useful, but largely guarantees that certain things are not going bad. THD-20 as measured with only an 80 kHz HF bandwidth tends also to obscure much of the interesting higher-order harmonic information in the spectral analysis.

For that reason you also want to do something like 19 + 20 kHz CCIF IM with spectral analysis, so you can see the in-band IM components out to fairly high orders.

But there are many, many other things that can go wrong in an amplifier that will not be detected by THD. BTW, however, burst parasitic HF oscillations often do cause an increase in THD, so this can sometimes be useful.


Cheers,
Bob
 
I've been developing a classA amplifier for a while and its been on the test bench for far too long now. The problem is i want it to be my best work to date but i'm convincing myself that it dosn't sound that good. A few weeks ago after the project stalled i thought its about time i did some serious research so i started reading loads of papers about the subject of distortion and its effects. I also bought a copy of Selfs 'Audio Power Amplifier Design Handbook'. After reading the book cover to cover i decided to apply some of Selfs methods for reducing THD across the entire amplifier design. After doing this on the sim my amp went from approx 0.065% / 0.016 THD 50/1 Watt to 0.002 / unmeasurable ( < 0.000 ). I was impressed by this figure ( Well done Self ) and so spent this easter mirroring the improvments in the real world on the test bench amp. Upon power up i started to make measurments. Everything seemed good, no measurable THD on the computer via RMMA ( crap soundcard but it said the previous attempt was not that good lol ), good squares and sines on the scope with no noise, 20v/us rise and fall etc. I am convinced there has been a dramatic improvment just as the sim said there would be. I then swapped test gear for audio gear and had a listen. The thing is, i expected the sound to be crystal clear, much improved on the last time i listened but i was dumbfounded to hear it had hardly changed ( if at all ). Hence the question above, if an amplifier has very low THD then why does it not sound as good as amplifiers i have thrown together in the past that use voltage regulators as current sources and have relativly high THD's? Am i imagining this? Are there other hidden distortion's / imperfections that can have a large impact on the perception of sound?? I'm confused but these questions need answers if i'm to carry on my journey into sound.

My quasi-complementary (2N3773 as OPs!), bootstrapped (!) design that uses cheap components [BC212, BD139, 1/2W carbon-film resistors, cheap electrolytics in signal path (!), etc] and has ~0.05..0.1% THD (max. power, 20kHz) sounded BETTER then my "Self/Slone-type" state-of-art design that uses expensive parts (2SA1360 as VAS, 2SC4793+comp. drivers, 2SC5200/SA1943 OP's, 1%-1/2W metal-fim resistors, high quality capacitors) and really sophisticated circuits and has ~0.001...0.005% THD and nice square-wave response etc
 

Pan

Member
2002-09-19 3:52 pm
kevinkr said:
Several acquaintances recently have built Nelson Pass power fet based designs from AudioXpress which to my surprise sounded good in my Onken/horn based system.

Good sound is nice. :)

Clearly you have a (somewhat?) different opinion, and a different experience so why don't you share those thoughts, and how you came to your conclusions. Honestly I don't like much of the solid state gear I hear, but the original poster's comments struck a chord as his experience eerily echoes mine of some 25yrs ago. This is the only reason I posted in fact.

Not much of experience and opinions on those amps. I have never heard Johns amps only one of he Pass amps.. Aleph 5. It did sound very nice but I do not think transparent would be the right word.

I was simply curious on how you came to the conclusion that something is transparent. Putting something in the rig and calling it transparent just because it sounds good doesn't make sense IMO. There may very well be some colorations in the other components that the "transparent" amp matches well with it's own colorations.

IMO the only way to find out if an amp is transparent is to investigate if it can pass a signal (with a realistic load) without altering the signal audible.


/Peter