AndrewT
Good point. I'd like to witness how a sound card can "simply" avoid injecting noise or error at those voltage levels with the method Mr. Wurcer describes.
scott wurcer
Most class B amps are biased into A for usual listening levels so if I'm not mistaken, the premise is that crossover distortion is ugliest (most audible) just as it leaves this bias. I can understand this but circuits with minimal bias (<10mA) and speakers with <90dB sensitivity are usually well into the amp's error correcting range at normal listening levels and for those who like to listen at lower levels, a higher bias keeps them without crossover distortion. Also, at those low levels of output your sources are more likely to be your problem. What I'm saying is that crossover distortion seems to be a problem that has solutions that don't possibly introduce other problems (ie: greater parity of 2nd and 3rd order or elevated higher level harmonics). I get your point about Dunlap's non-switching output stage but without permanently 'marrying' an amp to a speaker system, his approach seemed the most sensible one. How would you have done it without introducing other compromises? This is not a challenge, I'm just looking to learn.
Reoder Felgen
For the record, I consider <10ppm THD+N at any frequency with rated power approaching 1ppm. I concede, I should have said max rated operational power
Good point. I'd like to witness how a sound card can "simply" avoid injecting noise or error at those voltage levels with the method Mr. Wurcer describes.
scott wurcer
Most class B amps are biased into A for usual listening levels so if I'm not mistaken, the premise is that crossover distortion is ugliest (most audible) just as it leaves this bias. I can understand this but circuits with minimal bias (<10mA) and speakers with <90dB sensitivity are usually well into the amp's error correcting range at normal listening levels and for those who like to listen at lower levels, a higher bias keeps them without crossover distortion. Also, at those low levels of output your sources are more likely to be your problem. What I'm saying is that crossover distortion seems to be a problem that has solutions that don't possibly introduce other problems (ie: greater parity of 2nd and 3rd order or elevated higher level harmonics). I get your point about Dunlap's non-switching output stage but without permanently 'marrying' an amp to a speaker system, his approach seemed the most sensible one. How would you have done it without introducing other compromises? This is not a challenge, I'm just looking to learn.
Reoder Felgen
For the record, I consider <10ppm THD+N at any frequency with rated power approaching 1ppm. I concede, I should have said max rated operational power
Last edited:
I'm not sure, was that post addressed to me.
If so, what was the problem?
Do you have have another definition of max output power of an amplifier?
Yours
Reodor
Yes, maybe a sloppy use of the word. There are 400W low ppm amplifiers, does it really matter exactly where they clip. A manufacturer might state rated power where the THD specs are fully met as well as max power by some other definition. I think the only point was that very high power low ppm amplifiers exist.
A typical 26db or so PA has 100's of nV/rt-Hz noise at the output far worse than a good sound card. To put some numbers on this we are actually talking about 10's of mV at the output of the PA where the sound cards preamp gain can be used (rather than attenuation). There are a lot of decent cards with <10nV of noise referred to input in this case. For larger levels a simple first order null (even a passive twin tee) removes the fundamental and again you gain up the residual to take the sound card's noise out of the equation.
I don't mean to take a stand on any of this being important but I found it interesting that a Halcro (one of the original ppm guys) has literally >1% THD at very low levels. It's possible this only matters to the >100dB sensitivity horn folks I don't know. EDIT - I did not confirm this measurement but saw nothing obviously wrong, Earl Geddes has also conducted these tests on SS amps.
As for output stages, there are several active error correction feedback schemes out there. One dating from as far back as 1972.
Last edited:
Do I have an example for "noise", whatever means?
Built in thick film resistors. Super. Clean, fine ... but did chink and clink very well.-) Nothing between resistor and radiator - the material! Insulator between, and the sound was clean, without "distortions"-!
My experience:
We do listen material, a mix of materials;-!!! Much much more than THD, TMD, crossover distortions.
We do hear the difference of complementary-parts! as pnp and npn transes much much more than crossover distortions;-!
We do hear the difference of separate psus;-!!!
And so on...-)
LG
Built in thick film resistors. Super. Clean, fine ... but did chink and clink very well.-) Nothing between resistor and radiator - the material! Insulator between, and the sound was clean, without "distortions"-!
My experience:
We do listen material, a mix of materials;-!!! Much much more than THD, TMD, crossover distortions.
We do hear the difference of complementary-parts! as pnp and npn transes much much more than crossover distortions;-!
We do hear the difference of separate psus;-!!!
And so on...-)
LG
I think Scott addressed the noise question. The other issue you appear to be asking about might be what Scott referred to as having to do with "gain up." My understanding would be that after nulling out the test signal from the DUT output with a twin tee, a carefully designed low level amplifier can be used, if necessary, to bring up the filtered DUT output signal up to a level suitable for use with a sound card. One could also measure any distortion attributable to the "gain up" amplifier and, if significant, try to subtract that from the spectrum with the DUT amplifier in circuit, but remember the "gain up" amplifier never sees the high level unfiltered test signal so its distortion products should be very low compared to what is coming out of the DUT amplifier.
Maybe Scott can correct me if misstated anything in the above.
Last edited:
There may be something to what you say. We aren't so good at measuring some things, particularly non-stationary distortions that may not be well excited by a fixed frequency test signal. But, we are really good at measuring with a fixed frequency, so we measure what we can and tend to believe that's most of what there is.
If you are referring to the "Krill" by Steve Dunlap I think that the design was taken apart proving that it was not working as stated.
Yours
Reodor
To comment on this, the bias spreader was a very eccentric connection of devices that generated a "glitch" for lack of a better word counter to the crossover discontinuity. There was a trim to tweak it for cancellation at a particular load, so at an 8 Ohm resistive load at the output level tested there was a benefit. Give the load any phase or changing magnitude the cancellation was compromised. There was also a non-switching side discussion that was a total red herring, there are many class AB outputs that do not totally switch off, in fact poorly modeled effects like quasi-saturation, charge storage and ft vs current dramatically affect this since an AB stage with ideal transistors never switches off in either direction maintaining a geometric mean of current.
No, that's essentially correct and you can find information on all this in the ultra-low distortion oscillator thread (you will have to sort through a lot though).
Markw4
Thanks for the elaboration. That may be a trivial procedure for some but not for me
Reoder Felgen
Your pH is quite low. I think a little irreverence is generally a healthy thing but you unjustly disrespect a truly important influence in the audio space when you deride Mr. Dunlap's achievement. I know Mr. Wurcer states that error corrections of crossover distortion have been proposed since 1972 but I doubt he would assert that they were as elegant as what was achieved in the krill. That real world 'glitch' that Mr. Dunlap exploited had the effect of damping/smoothing the crossover so as to make it inconsequential. Designers like vzaichenko and others have been inspired by his work to achieve similar success. Tom Christiansen gives a nod to Mr.Wurcer on his webpage for inspiring his remarkable achievement using a composite topology. Respect where respect is due. Who have you inspired with your work?
Thanks for the elaboration. That may be a trivial procedure for some but not for me
Reoder Felgen
Your pH is quite low. I think a little irreverence is generally a healthy thing but you unjustly disrespect a truly important influence in the audio space when you deride Mr. Dunlap's achievement. I know Mr. Wurcer states that error corrections of crossover distortion have been proposed since 1972 but I doubt he would assert that they were as elegant as what was achieved in the krill. That real world 'glitch' that Mr. Dunlap exploited had the effect of damping/smoothing the crossover so as to make it inconsequential. Designers like vzaichenko and others have been inspired by his work to achieve similar success. Tom Christiansen gives a nod to Mr.Wurcer on his webpage for inspiring his remarkable achievement using a composite topology. Respect where respect is due. Who have you inspired with your work?
Last edited:
Chemist? I would rather let the Krill episode rest in peace but from what I have seen of Vzaichenko's work there is reasoning behind each step of the design. This was not the case with the Krill, unless simply no one was able to put it into reasonable engineering terms but several devices were connected in a virtually non-functioning state and the results were presented as purely empirical.
scott wurcer
Lots of people built Dunlap's circuit in real life and were happy with the result. As far as I know, not a single one of them has blown up and taken speakers down with them. There are lots of submissions here that can't make the same claim. The fact that he couldn't, or more likely wouldn't, describe it to the satisfaction of those who demanded explanations from him doesn't change that reality. I am aware of the history between you two and it saddens me that you're still taking pot shots at him. You're both bigger than that. We both made our points, the readers can look him up and decide for themselves where he belongs in the lexicon of audio.
Lots of people built Dunlap's circuit in real life and were happy with the result. As far as I know, not a single one of them has blown up and taken speakers down with them. There are lots of submissions here that can't make the same claim. The fact that he couldn't, or more likely wouldn't, describe it to the satisfaction of those who demanded explanations from him doesn't change that reality. I am aware of the history between you two and it saddens me that you're still taking pot shots at him. You're both bigger than that. We both made our points, the readers can look him up and decide for themselves where he belongs in the lexicon of audio.
OK this ends now, I have been here before asking hard technical questions is a personal attack to some so be it.
Link to a schematic?
By, the way I didn't see any potshot. Only some technical comments about transistors.
Last edited:
nanina,
If I posted our guidelines for a design review from 1974 you might understand, it was a full frontal assault and a few newbie designers were brought to tears (really). Nothing passed without a good story. It was never personal and we all moved on to good things.
A novel circuit teaches, tells a story, I looked at Vzaichenko's recent posts and everything there makes perfect engineering sense. But in a larger sense none of these circuits like NP's stasis, Hawksford, and other error correction schemes have much of a presence in current DIY. Many folks are perfectly happy with a Self blameless with modern devices.
If I posted our guidelines for a design review from 1974 you might understand, it was a full frontal assault and a few newbie designers were brought to tears (really). Nothing passed without a good story. It was never personal and we all moved on to good things.
A novel circuit teaches, tells a story, I looked at Vzaichenko's recent posts and everything there makes perfect engineering sense. But in a larger sense none of these circuits like NP's stasis, Hawksford, and other error correction schemes have much of a presence in current DIY. Many folks are perfectly happy with a Self blameless with modern devices.