I'm not along the lines of the "there-is-good-distortion" thinking. I was just argueing that IF there is unavoidable distortion of considerable amount, I'd prefer even order and quadratic weighting (quick decay with increasing order). In the silly THD numbers, to me 1% H2 is "better" than 0.05% H7, for example.
You guys sure like to argue a lot over nothing. You should get out more and do some reading.
Regarding this topic of "pleasing" harmonics, I seem to recall reading recently a piece by Keith Howard in Stereophile where he put this issue to rest for once and for all.
As I recall, he took musical excerpts and made a program that would add controlled amounts of his choice of harmonic distortion (2nd, 3rd, etc.). The results were that the 2nd harmonic *did not* mask the 3rd, nor did it sound good. He made his files downloadable so you could listen for yourself.
You can read the article for yourselves and see how good (or bad) my memory is. At any rate, you can stop arguing about something that has already been demonstrated.
http://stereophile.com/reference/406howard/index.html
Regarding this topic of "pleasing" harmonics, I seem to recall reading recently a piece by Keith Howard in Stereophile where he put this issue to rest for once and for all.
As I recall, he took musical excerpts and made a program that would add controlled amounts of his choice of harmonic distortion (2nd, 3rd, etc.). The results were that the 2nd harmonic *did not* mask the 3rd, nor did it sound good. He made his files downloadable so you could listen for yourself.
You can read the article for yourselves and see how good (or bad) my memory is. At any rate, you can stop arguing about something that has already been demonstrated.
http://stereophile.com/reference/406howard/index.html
Charles Hansen said:
While I admire his ability to create software that introduces controlled amounts of distortion to a digital signal, it's also clear that his listening tests were uncontrolled. He knew what he was listening to each step of the way, so whatever conclusions he reached are suspect.
"I did feel that I could detect differences..."
And from the followup, "First, my results were obtained informally, not under the blind listening conditions..."
It amazes me how many audiophiles can take this kind of undisciplined, uncontrolled approach and actually believe it means something.
Klippel gives an example automated subjective test methodology:
http://www.klippel-listeningtest.de/lt/background.html
http://www.klippel-listeningtest.de/lt/background.html
andy_c said:
The main point is that finally somebody did a controlled listening tests to see if 2nd harmonic was "euphonious", or if it could mask other harmonics.
As far as what it means, you can reach whatever conclusion you'd like. If you go to his website, you can not only download the musical clips he used, but also the software used. So you can create whatever files you'd like, using test tones or music or whatever you think is relevant and do your own listening tests in whatever fashion you so desire.
All I can say is that this experiment is far, far better than the conclusions reached previously by people when there were thousands of variables! They would listen to completely different amplifiers and try to generalize about the audibility of harmonics when *nothing* was controlled. Now *that* is an uncontrolled experiment.
Charles Hansen said:
Quite right, although that does not make me leap to trust the
digitally simulated tests either.
I did have an experience not too long ago of being able to
isolate a case down to essentially one variable - 2nd vs 3rd
order harmonic, using the same parts, supply, no feedback,
Class A, etc, and the results were interesting. While the
character of sound was different, I didn't develop a preference
for 2nd over 3rd at equivalent distortion levels, contrary to
prevailing street wisdom.
In real world circuits (examples: First Watt F1 vs F2 and Bride of
Zen vs SOBOZ) the "third harmonic" amplifier has the advantage
of lower actual distortion. In that case, most people prefer the
sound of the third harmonic product, even though they often have
a belief predisposition toward 2nd harmonic.
Yes, it would be nice if the components simply had no distortion.
The Stereophile article made a good point, which is that IM
distortion is really the elephant on the dance floor. If you are
listening to simple material you can put up with quite a bit of
2nd and 3rd, but as you switch from a duet to a symphony,
it starts to fall apart.
The Stereophile article made a good point, which is that IM
distortion is really the elephant on the dance floor. If you are
listening to simple material you can put up with quite a bit of
2nd and 3rd, but as you switch from a duet to a symphony,
it starts to fall apart.
Nelson Pass said:Yes, it would be nice if the components simply had no distortion.
The Stereophile article made a good point, which is that IM
distortion is really the elephant on the dance floor. If you are
listening to simple material you can put up with quite a bit of
2nd and 3rd, but as you switch from a duet to a symphony,
it starts to fall apart.
I agree, that IM is most likely the biggie in terms of what we hear, but let's not lose sight of the fact that THD and IM measurements are just two different ways of eliciting the symptoms of the same underlying nonlinearity. Each one may present the symptom in a different way, and each one may stimulate the nonlinearity in a different way, and they are both useful as measurement tools.
I do think, of course, that CCIF IM measurements with the spectrum displayed is a better tool than just THD+N or even THD with spectral analysis.
I also agree that we should not take too much comfort in the presence of second harmonic, although I must admit that I have, perhaps wrongly, always thought that even order and lower order distortions tended to be more benign in their sonic impact.
Cheers,
Bob
From a DIY perspective
If I were to build a complementary amp today (say 50 watts/channel) class A - what BJT's would you recommend using?
Same request for MOSFETS.
Thanks
BTW it you are looking for a nice IM analyzer project" see the article by Dick Crawford in the 3/2004 issue of AudioXpress. A PC board is available. Works very nicely.
Charles
If I were to build a complementary amp today (say 50 watts/channel) class A - what BJT's would you recommend using?
Same request for MOSFETS.
Thanks
BTW it you are looking for a nice IM analyzer project" see the article by Dick Crawford in the 3/2004 issue of AudioXpress. A PC board is available. Works very nicely.
Charles
Re: From a DIY perspective
I like to build balanced bridge designs. So for 50 watts you would need +/- 15 volt rails for the output devices. (The drivers and pre-drivers would need a bit more.) The idle current would need to be around 2.5 amps.
This means that each quadrant of the "H" bridge would be dissipating 37.5 watts. This is far too much for a single transistor. I would recommend running three large packages (TO-264) in parallel, which would reduce the dissipation to around 12 watts per device.
For BJT's I would use the On-Semi ThermalTrak NJL1302/3281 complements. For MOSFET's I would use the Exicon (Semelab) BUZ900P/905P complements.
stellavox said:
I like to build balanced bridge designs. So for 50 watts you would need +/- 15 volt rails for the output devices. (The drivers and pre-drivers would need a bit more.) The idle current would need to be around 2.5 amps.
This means that each quadrant of the "H" bridge would be dissipating 37.5 watts. This is far too much for a single transistor. I would recommend running three large packages (TO-264) in parallel, which would reduce the dissipation to around 12 watts per device.
For BJT's I would use the On-Semi ThermalTrak NJL1302/3281 complements. For MOSFET's I would use the Exicon (Semelab) BUZ900P/905P complements.