"THE BAXANDALL PAPERS" sounds imposing .. 
Almost an understatement !! 2 grossly underbiased BJT amps (20ma) ... no test instruments. One with and one without (TMC) , children and "homeboys" can hear the difference.. Of course , one look at the output with a scope tells the tale.
You lose at least some loop gain with TMC implemented in any way. Since adding the TMC option to all my blameless based creations , I have optimized for maximum loop gain vs. phase margin.
Thanks , Doug..
OS
Almost an understatement !! 2 grossly underbiased BJT amps (20ma) ... no test instruments. One with and one without (TMC) , children and "homeboys" can hear the difference.. Of course , one look at the output with a scope tells the tale.
You lose at least some loop gain with TMC implemented in any way. Since adding the TMC option to all my blameless based creations , I have optimized for maximum loop gain vs. phase margin.
Thanks , Doug..
OS
Last edited:
Okay, you didn't exactly say that. So what? Please no semantic blah blah. On several occasions, in your article in WW, in your books, in your latest article in Jan's 'bookzine, you said that: "Bob Widlar stated in 1988 that output stage behaviour must be well-controlled up to 100MHz for the technique to be reliable; this would be flat-out impossible for discrete power stages......."
IOW, it doesn't work!
This only proves that those people were incompetent. NDFL does work if implemented in the right way. See the PGP amp. Besides, Edward Cherry is one of best EEs ever.
Admittedly. If I were you, I wouldn't dare to do that.
No! I'm referring to personal emails about TMC addressed to you.
That's really amazing, as you already know it was a lousy implementation.
No, it did not, as you have thrown out half of the baby with the bathwater.
Your non-optimal example is far from exciting. Please read again this comment: http://www.diyaudio.com/forums/soli...erview-negative-feedback-303.html#post2355163 and you will understand why I'm not impressed by your 'exciting' implementation. Most likely, this is also the reason why you told me (by email, about ten years ago) that the results of TMC were not exciting. Then I told you what the reasons might possibly be. But you didn't bother to answer me. Clearly you weren't interested in this topic, until I put TMC in the picture on this forum.
Hmm...if you don't mind, I prefer to consider this as a rather lame excuse. In all examples of TMC that you could (and should) have seen, even TMC's nephew (i.e. TPC) , as published by yourself in EW+WW, Feb, 1994, p.137 (C1=100pf and C2=1nF) shows that for optimal results C2 must be larger than C1.
Cheers,
E.
Hi Doug,
Thanks again for posting these Baxandall papers. As always, his work is an enjoyable and insightful read. I have now had a chance to read them carefully and have a couple of comments, but they may spark some controversy.
Although the circuit he is proposing is topologically identical to TMC, I hesitate to say that he actually came up with TMC as we know it here at DIYaudio. Rather, he was focusing on a way of stabilizing Cherry's output-inclusive compensation that just happened to be this topology. I know that I am making a fine, possibly nitpicking distinction here, and I am not trying to take anything away from Peter. Let me explain why I say this.
He has C2 and C1 in series from collector to base of the VAS and mentions that the effective Cdom at high frequencies is their series combination. So far, so good. But look at his suggested value for R (the resistor from the output of the amplifier to the junction of the capacitors. It is only 68 ohms. This is VERY low for TMC compensation as I think we know it and have been practicing it.
If he set HF Cdom to be 100pF and used C1=C2=200pF, the rough "transition frequency" would be on the order of 15 MHz. I agree that where the transition frequency lies is a matter of degree, but I think the transition we have practiced for TMC is at a much lower frequency, well below the point in frequency where the output stage becomes brain dead.
Note also that by the time his transition has occurred, the VAS collector is heavily loaded with an impedance on the order of 100 ohms (albeit bootstrapped a bit if the output stage still has any functionality at all). Is the VAS really something that we would still consider acting as a Miller integrator under these conditions?
It seems clear that in TMC as we know it, the VAS is still very much functioning as a healthy Miller integrator at the frequency where the transition of its feedback pickoff occurs.
His use of such a low resistance feeding back to what can also be considered the input of the non-inverting output stage could also be troublesome at very high frequencies, insofar as a parasitic oscillation localized to the output stage. As you know, near-unity feedback around an emitter follower at very high frequencies can cause parasitic oscillations.
My second comment is that my reading of his papers does not suggest that he built and measured an amplifier using this technique. This can also be a matter of interpretation, and you know his style better, but his discussion of this circuit comes across to me as being very speculative.
Anyway, although many may not agree with the impressions I have shared here, I thought it useful to share them. I will be the very first to agree that this all is a matter of degree.
Cheers,
Bob
Yes, lets keep the attitude here less emotional Edmond. Consider this a warning. Play the ball not the man.
ETMC, 'devastatingly effective'
Here a few more 'very characteristic Edmond ideas', and they are 'devastatingly effective'.
http://www.diyaudio.com/forums/soli...erview-negative-feedback-301.html#post2352970 (notice the disclamer, please)
and here:
http://www.diyaudio.com/forums/soli...erview-negative-feedback-302.html#post2354463
and here:
http://www.diyaudio.com/forums/soli...erview-negative-feedback-304.html#post2357325 (notice the credits, please)
Hi Douglas,
............
It's also a very characteristic Edmond idea.
............
Here a few more 'very characteristic Edmond ideas', and they are 'devastatingly effective'.
http://www.diyaudio.com/forums/soli...erview-negative-feedback-301.html#post2352970 (notice the disclamer, please)
and here:
http://www.diyaudio.com/forums/soli...erview-negative-feedback-302.html#post2354463
and here:
http://www.diyaudio.com/forums/soli...erview-negative-feedback-304.html#post2357325 (notice the credits, please)
Last edited:
I found the Baxandall papers to be very "comforting" and easy to understand .. he must of been a great teacher. His observations are very similar to what I have experienced upon adoption.
His idea to use a trimmer as the main TMC R is fun to play with It seems (as I listen to my TMC amp) that one more distortion mechanism is "licked".
To make the operation of TMC very noticable both to my ears and on the scope , I underbias to exaggerate the effects (audible and observable). With any "R" below 150 , the distortion becomes massive as I approach class B (under 20mA per OP). For this reason , I've settled on either 390 or 470. With these values (100p/390p - 390R) I can hear NO change subjectively until I actually reach class B and even at that point it is subtle. "Devastatingly effective" is an understatement. In normal operation I must say that class A has nothing on an amp compensated in this manner.
I thank all involved , stop the bickering .. enjoy the pure amplification.
PS .. Edmond ... "You the man" , I gave you and Baxandall the credit in my thread , one may of discovered it , but you refined / shared it.
His idea to use a trimmer as the main TMC R is fun to play with
It seems (as I listen to my TMC amp) that one more distortion mechanism is "licked".To make the operation of TMC very noticable both to my ears and on the scope , I underbias to exaggerate the effects (audible and observable). With any "R" below 150 , the distortion becomes massive as I approach class B (under 20mA per OP). For this reason , I've settled on either 390 or 470. With these values (100p/390p - 390R) I can hear NO change subjectively until I actually reach class B and even at that point it is subtle. "Devastatingly effective" is an understatement. In normal operation I must say that class A has nothing on an amp compensated in this manner.
I thank all involved , stop the bickering .. enjoy the pure amplification.
PS .. Edmond ... "You the man" , I gave you and Baxandall the credit in my thread , one may of discovered it , but you refined / shared it.
Last edited:
Hi Bob
Whether he invented it or not is certainly obscure, and at this date may not be possible to determine. I think it is however clear that his description on p35 is that of TMC.
I have to say the phrase "Cherry's output-inclusive compensation" grates on me a little, because it sort of implies that it works, and I think we all agree that it really, really, doesn't, in a tweeter-melting sort of way.
I wholly agree. I never even tried that value. I treated the scheme as a variant of 2-pole compensation (which I gather you call TPC here?) and I began with 4K7. Peter says "68R might be used" which could be evidence that he never got around to practical experiments. On the other hand he says on p30 "Some months ago I devoted much thought and experiment to this problem..." I always found Peter to be quite precise about words, and I have always taken that statement at its face value. Seems unlikely we'll ever know for sure.
The history of TMC
Hi OS,
Thank you for your kind words. As so many times, also this invention was 'in the air'.
Baxandall and I invented it independently from each other, though, apparently, he did it a few years before me. OTOH, that 68 Ohms resistor puzzles me, as it's far too low. Perhaps this was the reason (as reported to me) that D. Self -at first instance was- was not impressed by his invention.
BTW, I'm glad to hear about your positive experiences with TMC.
Cheers,
Edmond.
[snip]
PS .. Edmond ... "You the man" , I gave you and Baxandall the credit in my thread , one may of discovered it , but you refined / shared it.
Hi OS,
Thank you for your kind words. As so many times, also this invention was 'in the air'.
Baxandall and I invented it independently from each other, though, apparently, he did it a few years before me. OTOH, that 68 Ohms resistor puzzles me, as it's far too low. Perhaps this was the reason (as reported to me) that D. Self -at first instance was- was not impressed by his invention.
BTW, I'm glad to hear about your positive experiences with TMC.
Cheers,
Edmond.
Positive is the key word. Everybody has definitely contributed. I am building a Self "blameless" (linn ?? - refined and explained by Doug in his very good book). Adding to that is H. Deans psychoacoustical application of a bootstrapped current source (actually a early RCA idea). Throw TMC on top of that and you have perfect amp that is both musical and flirts with ppm distortion levels.
Of course this is too narrow an application for TMC. Add it to a high loop gain balanced single end topology (hitachi - beta enhanced) and the simulator "sings" with ultra-low THD and almost no H5 and up. Maybe even a class A line level stage using TMC ??
Funny how you sometimes come to a conclusion only to find someone else has also come to the same one independent of yours.
The 68R (under 150R) does have an undesirable effect , the transition to even a "hard distortion" is evident at low AB bias. I would not even rule out "tweeter melting" at higher power levels. I checked for outright oscillation - none. Maybe , as Bob mentioned .. parasitic oscillation ??? This is where the real amp differed from the expected simulation. On the other end of the spectrum , TMC seems to be scalable. At 680R and above the effects are more and more subtle until there is no improvement. I am even going to try smaller C1/2 combos and adjust for my UG point by increasing input degen. (silver mica's are expensive) ... of course then I could increase the "R" (1K)
I can try all this out in minutes/hours as I am modular. No more guessing or total simulator reliance.
OS
"Cherry's" output inclusive compensation is used on a few commerical products - the TDA7293 MOSFET output chip amp and the LT1028 family op amp - despite any problems it may have with audio power output BJT
Cherry did claim any such problems were due to local feedback at the output device and at least implied that they could be corrected there
Cherry did claim any such problems were due to local feedback at the output device and at least implied that they could be corrected there
Hi Doug,
I don't think one should deny Cherry the use of the term output inclusive compensation just because it did not work under some conditions, perhaps many conditions. I would not use it, but that does not mean he can't claim name to it. Indeed, there are some who cannot make Miller compensations work
.Moreover, Cherry published it, Baxandall apparently never published it. That is a shame, of course.
"this problem" that I think Baxandall was referring to was the fact that Cherry's compensation was not reliably stable. He obviously did some experiments, but I just did not see any evidence in his papers of having put together a TMC amplifier.
Given his resistor value, it appears clear that his compensation was not behaviorally the same as the TMC that you implemented. What say you to my observation about the VAS no longer acting as a Miller integrator by the time Baxandall's transition occurs?
Cheers,
Bob
Hi guys, sorry for jumping in, very interesting discussions.
Has anybody around built and measured a TMC based power amplifier? Or is aware of smbody that did? It would be great to see some construction pictures, practical measurements, diagrams or any other objective comparative results regarding various compensation methods (Miller, TPC, output inclusive, TMC). I'm myself bored to death of synthetic simulated results, and ego driven arguments about who invented what.
Has anybody around built and measured a TMC based power amplifier? Or is aware of smbody that did? It would be great to see some construction pictures, practical measurements, diagrams or any other objective comparative results regarding various compensation methods (Miller, TPC, output inclusive, TMC). I'm myself bored to death of synthetic simulated results, and ego driven arguments about who invented what.
Well, I certainly have. See "Inclusive compensation & ultra-low distortion power amplifiers" in Jan Didden's Linear Audio journal, Volume 0:
Linear Audio | your tech audio resource Get your copy now!
I am not very familiar with this forum so I am not sure if anyone else has built one. I'm sure they'll soon tell you if they have.
I Built 2 in prototype , am listening now. One of my builders ,(SNG001) in Australia has a pair http://www.diyaudio.com/forums/solid-state/169590-mongrel-supersym-ii-17.html#post2377265
Tomorrow , I will have 4 more TMC creations (pix 1-below) , just burned the boards. 1 pair is the replacement for the working prototype (BX - "blameless" w/cascode and bootstrap current source) and the second (CX) is the untested beta enhanced balanced VAS (like a symasym). Both top 90db OLG. . As far as measuring them , a simple CRO will do. Observe Xover point while decreasing R or watch the correction (you can hear this) as you underbias. I can't afford an AP precision , but the improvement is SO apparent it almost seems like "cheating".
OS
OS
Attachments
Hi OS,
Indeed, good ol' bootstrapped current source! Although the GEB will certainly have objections against electrolytics (who cares), there's noting wrong with it. I'm also using these "poor man's" current sources, see: http://www.diyaudio.com/forums/solid-state/174218-rebirth-phoenix-7.html#post2392466 . Better, simpler and more output swing.
Hi Bob & OS,
As for the very low resistor value, I like to add the following: If the transition frequency is 1MHz (Baxandall mentioned 1 to 2 MHz), and R=68 Ohms, then C2=2.34nF. With such low impedance stuff between input and output of an OPS, no doubt that the performance will be affected. Indeed, in this sense it can hardly considered to be a genuine implementation of TMC. But there is more at stake. Such RC network is also a compensation technique in his own right. If my memory serves, Linear Technology had applied this technique in one of his op-amps, years ago. It was especially developed to drive capacitive loads.
Cheers,
E.