Hi Wimms...do you have a solution for this..?
http://www.diyaudio.com/forums/showthread.php?postid=496496#post496496
http://www.diyaudio.com/forums/showthread.php?postid=496496#post496496
Swedish Chef said:
Thanks for the excellent link Magnus!
I took some time to download and review the material there. Looks Mr. Richard C. Cabot has the credentials to cast some well founded oppinion.
For the benefit of those who did not review this article and other documentation, these are the facts in brief.
Several different performance tests are applied to a high performance device under test, where different forms of nonlinearity are introduced on purpose and in a controled way.
Tests include basic THD, SMPTE, CCIF, DIM and a new one dubbed FASTtest original from Audio Precision. Each test is designed with the purpose of uncovering peculiar distortion modes such as harmonic, IM, TIM, PIM and so forth, or to be as comprehensive as possible within practical limits.
Artificial nonlinearities are introduced by the way of diodes properly placed to mimick nonlinear transfers, or unbiased pushpull emitter followers to create crossover distortion, or heavy capacitive loading to limit slew rate.
Results for each test method and for each artificial nonlinearity are then plotted side by side and compared.
Overall, the most stringent (i.e. the one showing the worst distortion results) was the 20 KHz THD one, a result attributable to the lower NFB available at the upper end of the audio band, together with the effects of slew rate limits.
Quote Mr Cabot:
For detailed engineering measurement applications THD is the most flexible and usefull technique.
Let me stress anyway this does not dismiss IM / PIM / TIM specific tests, rather that THD (at a high enough frequency) is capable of bounding in level other distortion modes. In the root the cause both for harmonic or intermodulation either static or dynamic, lies in the device nonlinearities and this can be safely assesed by a (properly designed) THD test.
Rodolfo
You're welcome Ingrast although the credit should go to Mr Cabot.
I think that article is a very good starter. And it's the rare kind of work that actually pushes the audio industry forward rather than backwards. With more in-depth work and combined with physiological data and listening tests it would be top-notch.
/Magnus
I think that article is a very good starter. And it's the rare kind of work that actually pushes the audio industry forward rather than backwards. With more in-depth work and combined with physiological data and listening tests it would be top-notch.
/Magnus
ingrast said:
While I agree in general, there are exceptions, the nonlinearities
of a power Mosfet being a good example. The THD at higher
frequencies usually reflects Cdg nonlinearities, and at low
frequencies reflect other phenomena, Vgs nonlinearity in
particular. This shows up in circuits with "no feedback", but it
looks the same in a distortion vs. frequency plot as a bipolar
with declining feedback.
My apologies mikeks, I was not around here by then!
Definitely. And some fresh air I guess...
While this is a worthy distinction, I guess it does not deprive THD the character of an upper bound test. It should also be noted the focus here is in op amp performance. Probably you are quite right in pointing to the frequency dependent Cdg - Vgs nonlinearities in the context of power amplifier testing.
Rodolfo
Folks, the coke-pepsi taste test can ALSO be interpreted as: They were confused when the test was blind.
Once again, 30+ years ago, even before many of you existed, I watched a bar bet in a San Francisco bar, where the subject had to tell the difference, with eyes closed, between cola, 7-up, and ginger ale. Now trust me, these are VERY different tasting products, BUT after a few attempts, the subject could NOT tell which was which. Double blind testing does the same for me.
Now Coke and Coke Classic have a distinct taste difference, try it for yourself. I prefer Coke Classic. However, when Coca Cola TRIED to change the recipe of Coke, they were FORCED by public opinion to make Coke Classic. Then they ran a double blind test on someone who had put real money toward getting old Coke taste back, and he failed to tell the difference.
Please consider that the test, itself, is inherently flawed.
Once again, 30+ years ago, even before many of you existed, I watched a bar bet in a San Francisco bar, where the subject had to tell the difference, with eyes closed, between cola, 7-up, and ginger ale. Now trust me, these are VERY different tasting products, BUT after a few attempts, the subject could NOT tell which was which. Double blind testing does the same for me.
Now Coke and Coke Classic have a distinct taste difference, try it for yourself. I prefer Coke Classic. However, when Coca Cola TRIED to change the recipe of Coke, they were FORCED by public opinion to make Coke Classic. Then they ran a double blind test on someone who had put real money toward getting old Coke taste back, and he failed to tell the difference.
Please consider that the test, itself, is inherently flawed.
Folks, the coke-pepsi taste test can ALSO be interpreted as: They were confused when the test was blind.
Once again, 30+ years ago, before many of you existed, I watched a bar bet in a San Francisco bar, where the subject had to tell the difference, with eyes closed, of cola, 7-up, and ginger ale. Now trust me, these are VERY different tasting products, BUT after a few attempts, the subject could NOT tell which was which. Double blind testing does the same for me.
Now Coke and Coke Classic have a distinct taste difference, try it for yourself. I prefer Coke Classic. However, when Coca Cola TRIED to change the recipe of Coke, they were FORCED by public opinion to make Coke Classic. Then they ran a double blind test on someone who put real money toward getting old Coke taste back, and he failed to tell the difference. Please consider that the test, itself, is inherently flawed.
Once again, 30+ years ago, before many of you existed, I watched a bar bet in a San Francisco bar, where the subject had to tell the difference, with eyes closed, of cola, 7-up, and ginger ale. Now trust me, these are VERY different tasting products, BUT after a few attempts, the subject could NOT tell which was which. Double blind testing does the same for me.
Now Coke and Coke Classic have a distinct taste difference, try it for yourself. I prefer Coke Classic. However, when Coca Cola TRIED to change the recipe of Coke, they were FORCED by public opinion to make Coke Classic. Then they ran a double blind test on someone who put real money toward getting old Coke taste back, and he failed to tell the difference. Please consider that the test, itself, is inherently flawed.
The main problem is that simple, why manufacturer don't make a high OLbandwith opamp design intended for low gain purposes? Audio rarely uses gain factor more than 100x Closed Loop for 1 single stage. It is true that opamps usually have magnificent gain factor, surely ends up in lowOL bandwith. Is there such thing as High OL bandwith opamp with limited gain factor (OL=500x, for example)?
Maybe this is why PMA likes AD844 (current feedback opamp). Current feedback usually have high OL bandwith.
john curl said:
Did you wear some flowers in your hair..?
john curl said:
Obscure was the word, how on earth can we (at least some..) say that some particular amplifier is drawing with som.. may it be "PIM"...if it's obscured by an op-amp in the CD player, are there any CD-player qualified, or should we modify every one before.., and when is it enough good, and finally are we really sure because we seem not be able to meaure everything as You have pointed out many times, so then theres only our ears... which is subjective property of every each human on this earth...?
john curl said:
The Million dollar Q izz... can we feel the taste of the unwashed glass, whetere it was previously used for german weat beer or dark Beamish stout from Ireland... maybe I will perform the test this evening, it's Friday anyhow!
john curl said:
Oh, so it's not going to be obscured by the ompamp (used to be used )in CD-players...?
john curl said:
That's not a good situation for our CD-players, right?
I'm not aware how the output stage use to be on top-end CD-players, but then I assume it must ONLY be discrete components handling the signal directly after the D-A.... so it wont "obscure" our audiophile experience when tasting to a nice unwashed glass of wine!
What I try to say, I think you just stiring in the pot with nothing concrete!
Cheers!
Swedish Chef said:
Hi Magnus !
I choosed transfercurves that creates for single sine wave specific
harmonics only. Exactly these transfercurves now have a known
thd for sinewave. So used these on a musicsignal, checking what
distortions results.
About THD for tweeter above 6khz, i understand now that is even
for higher freqs important, as i see that imd creates harmonics
at much lower freqs. Let's say i have a signal with 10+12khz, now
i get a harmonic at 2 khz if 2nd harmonics-thd is high enough.
What did i get wrong ?
Of course i can't measure 3rd harmonics above 8khz, as nyquist
is at 24khz with fs=48khz. This issue is good visible in the plot.
With a 96khz-soundcard it would be possible, but i doubt that
the mic and the amp would be precise enough.
I believe that the exploding 2nd harmonic in the plot is from the amp (Yamaha...)
The measurements with my speakers were done at unpleasant
SPL-levels, but i have no exact numbers. If i would have used
higher levels, the excursion of the woofer would have been too
large for 30hz. The measurement was mainly for bandwidthlinearity,
but the mic was not calibrated. At least it showed that the X-over
works "perfect".
The tweeter is a Visaton-G20SC, their cheapest highend-tweeter.
http://www.visaton.de/deutsch/artikel/art_266_2_8.html
Mike
Hi MikeB!
You are right but still missed something! The issue is that THD is not valid for bandwidth-limited systems.
Say that you play a 15 kHz tone to the tweeter. Thus h2 is at 30 kHz and h3 at 45 kHz. But the response of the tweeter drops like a rock above 20 kHz! So h2 and h3 will be attenuated and you will get very good THD specs! But clearly they are not very representative for the actual tweeter distortion. Redo the measurement with say a CCIF (IMD test with two tones at 14 kHz and 15 kHz) and you will get quite different numbers!
I don't say that this tweeter is crap because it isn't. The FR response is very smooth and it is probably great for low-level listening. But a dome tweeter with a sensitivity of 88 dB just is simply not going to produce high SPL:s with low distortion. Pretty much only compression drivers on horns are up to this.
Stuff like JBL2426, B&C DE25, Beymas and so on.
I wish that everyone could witness the capabilities of a truly awesome speaker system such as the one Paradigm (Roland Mollbrandt) on this site has put together. 2*JBL2215, 2*JBL2225, 2*JBL2110, 2*JBL2445 on tractrix-style horns and 2*JBL2402. That would mean the end to people worrying about their interconnect cables and resistor brands. Period.
/Magnus
You are right but still missed something! The issue is that THD is not valid for bandwidth-limited systems.
Say that you play a 15 kHz tone to the tweeter. Thus h2 is at 30 kHz and h3 at 45 kHz. But the response of the tweeter drops like a rock above 20 kHz! So h2 and h3 will be attenuated and you will get very good THD specs! But clearly they are not very representative for the actual tweeter distortion. Redo the measurement with say a CCIF (IMD test with two tones at 14 kHz and 15 kHz) and you will get quite different numbers!
I don't say that this tweeter is crap because it isn't. The FR response is very smooth and it is probably great for low-level listening. But a dome tweeter with a sensitivity of 88 dB just is simply not going to produce high SPL:s with low distortion. Pretty much only compression drivers on horns are up to this.
Stuff like JBL2426, B&C DE25, Beymas and so on.
I wish that everyone could witness the capabilities of a truly awesome speaker system such as the one Paradigm (Roland Mollbrandt) on this site has put together. 2*JBL2215, 2*JBL2225, 2*JBL2110, 2*JBL2445 on tractrix-style horns and 2*JBL2402. That would mean the end to people worrying about their interconnect cables and resistor brands. Period.
/Magnus
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.