Hmm, other than how you're deriving the reference voltage of the clamp for the VAS output, it looks very similar to this design to me.
Hi Glen,
I just thought of something I saw in simulations of clamping with a Blameless circuit. In my case, there was an LM385-1.2 setting the voltage between the bases of the input devices of the CFB output stage. This of course was loading the main VAS, which had a clamping circuit. When I first set up the sim, each clamp clamped the voltage on either side of the LM385-1.2. In hard clipping of the positive half-cycle, all the current of the constant-current VAS load went into the emitter of the clamping transistor. The transistor polarities were all opposite from yours, so it would be the negative half-cycle in your case. When this happened, the current that would ordinarily bias the LM385-1.2 was all diverted away into the clamp, so the voltage reference turned off. It had a 0.1 uF capacitor across it, and when it turned back on (which was slow), it caused a big glitch in the clipping of a 20 Hz signal. I fixed it by tying both the positive and negative clamps to the same side of the shunt reference. That way, when all the current of the constant-current VAS load went to the clamp, it was diverted through the reference such that the reference did not turn off.
The disadvantage of this was that the clamp reference voltages were slightly asymmetrical for symmetrical clipping. But I set the references with a couple of TL431's, so this just meant slightly different resistors in the two TL431 circuits.
I didn't see your exact voltage reference configuration specified (just ideal sources), so I thought I'd warn you about this potential problem. If I recall correctly, it didn't happen with clipping of 20 kHz signals, as the capacitor across the reference was enough to keep it on in that case. But with clipping of a 20 Hz signal, there's lots of time for the reference to turn off. I hope what I've described makes sense.
I just thought of something I saw in simulations of clamping with a Blameless circuit. In my case, there was an LM385-1.2 setting the voltage between the bases of the input devices of the CFB output stage. This of course was loading the main VAS, which had a clamping circuit. When I first set up the sim, each clamp clamped the voltage on either side of the LM385-1.2. In hard clipping of the positive half-cycle, all the current of the constant-current VAS load went into the emitter of the clamping transistor. The transistor polarities were all opposite from yours, so it would be the negative half-cycle in your case. When this happened, the current that would ordinarily bias the LM385-1.2 was all diverted away into the clamp, so the voltage reference turned off. It had a 0.1 uF capacitor across it, and when it turned back on (which was slow), it caused a big glitch in the clipping of a 20 Hz signal. I fixed it by tying both the positive and negative clamps to the same side of the shunt reference. That way, when all the current of the constant-current VAS load went to the clamp, it was diverted through the reference such that the reference did not turn off.
The disadvantage of this was that the clamp reference voltages were slightly asymmetrical for symmetrical clipping. But I set the references with a couple of TL431's, so this just meant slightly different resistors in the two TL431 circuits.
I didn't see your exact voltage reference configuration specified (just ideal sources), so I thought I'd warn you about this potential problem. If I recall correctly, it didn't happen with clipping of 20 kHz signals, as the capacitor across the reference was enough to keep it on in that case. But with clipping of a 20 Hz signal, there's lots of time for the reference to turn off. I hope what I've described makes sense.
Please let me know, where about this URL:
http://users.ece.gatech.edu/mleach/
Thank you in advance.
By the way, regarded negative feedback (NFB) this thread could be of intetest:
http://www.diyaudio.com/forums/soli...-nfb-negative-feedback-me-wrong-question.html
Just Google on Marshall Leach or on Double Barreled amplifier or on Leach Superamp and you will have a listing to get to Marshall's site. It is a very good site and there is a lot of good stuff there.
BTW, his observation about square waves into an HPF is a good one and true. In practice, the slower risetimes of most real program material will mitigate the peak amplitude. I saw a paper at the AES a long time ago point essentially this phenomena out. It was at a time when most people thought that if they bi-amp'd they would not need such a big amplifier for the tweeter channel. You really should, of course, if the tweeter sensitivity is the same as that of the woofer. In speakers designed specifically for bi-amping with active crossovers, however, the tweeter efficiency is often quite a bit higher than that of the woofer.
In my active speakers playing the Rickie Lee Jones Getto of My Mind track at realistic (but not painful) levels, I have seen brief 30V peak transients on the tweeter when there is an agressive snare drum thwack. This is an unusually well-recorded CD with very little compression.
Cheers,
Bob
Yes, M. Leach not only publish the instructions to build his amplifier
but he also take some time to explain the idea behind his amp.
Andy, it occurs to me that Edmond's arrangement here looks like a double-pole compensator with the resistor referred to the output instead of directly to ground.
Douglas Self has recently written about this arrangement in Jan Didden's publication (linear Audio volume 0). it's pity he didn't compare it with the equivalent double pole compensator.
P.S: Mikeks=michaelkiwanuka.
Member
Joined 2009
Paid Member
Interesting, I remember reading a comment from D. Self that he was looking at a compensation technique that partially included the output stage, that the results were very good but he was not yet prepared to divulge what he was up to.
Here's the quote:
".... I have recently done some work that shows it is possible to partly include the output stage in the Miller loop, and it does give significant advantages in normal operation. This is a rather exciting development, but still under study and I am not ready to disclose it..."
do you think these are related ?
Last edited:
Unfortunately, Doug did not mention explicitly in his article that the technique has been explored at length here (starting at least three years ago), and has generally been referred to as Transitional Miller Compensation (TMC), dubbed that by Edmond.
TMC and TPC, and their comparison is covered in my book, Designing Audio Power Amplifiers. The table of contents can be found at CordellAudio.com - Home. See Chapter 9, "Advanced forms of Feedback Compensation".
Cheers,
Bob
TMC
Indeed, no pros and cons compared to TPC. For example, differences in phase and step response, loading effects on the VAS output, etc.
There are even more pities:
- No clear explanation under which conditions TMC does work or doesn't work.
(he only says that the amp needs to be 'blameless', whatever that may mean)
- A totally wrong example (i.e. equal caps, instead of say 1:5), that even leads to wrong conclusions and subsequent irrelevant blah blah.
- There's one more serious omission, but, as it is so obvious, I don't want to go into details any further.
In this regards, Bob Cordell has written a far better chapter about TMC. See his book, pp. 182-183.
edit: posts have crossed.
Cheers,
E.
Indeed, no pros and cons compared to TPC. For example, differences in phase and step response, loading effects on the VAS output, etc.
There are even more pities:
- No clear explanation under which conditions TMC does work or doesn't work.
(he only says that the amp needs to be 'blameless', whatever that may mean)
- A totally wrong example (i.e. equal caps, instead of say 1:5), that even leads to wrong conclusions and subsequent irrelevant blah blah.
- There's one more serious omission, but, as it is so obvious, I don't want to go into details any further.
In this regards, Bob Cordell has written a far better chapter about TMC. See his book, pp. 182-183.
edit: posts have crossed.
Cheers,
E.
Last edited:
Indeed, no pros and cons compared to TPC. For example, differences in phase and step response, loading effects on the VAS output, etc.
There are even more pities:
- No clear explanation under which conditions TMC does work or doesn't work.
(he only says that the amp needs to be 'blameless', whatever that may mean)
- A totally wrong example (i.e. equal caps, instead of say 1:5), that even leads to wrong conclusions and subsequent irrelevant blah blah.
- There's one more serious omission, but, as it is so obvious, I don't want to go into details any further.
In this regards, Bob Cordell has written a far better chapter about TMC. See his book, pp. 182-183.
edit: posts have crossed.
Cheers,
E.
Edmond,
I tip my hat to you for having explained and popularized TMC years ago.
Cheers,
Bob
Thank you, Bob!
@Mike,
I'm glad you're back. I hope you will stir the pot again, just as in the past.
Cheers,
E.
Hi forr,
Yes, Baxandall did a lot. If anyone here has a specific reference in Baxandall's writings where he discussed or proposed TMC, please let me know.
Cheers,
Bob
Doug Self refers to it as 'Private Communication, 1995'. Personally, I've never seen any published writing by Baxandall on TMC.
jan didden
I want to have also that papers.
about the term
"Transitional Miller Compensation" (TMC) isn't to find a lot, only few URLs::
http://www.diy-audio-engineering.org/index.php?topic=6.0
post 29 about
http://www.diyaudio.com/forums/soli...feedback-how-realy-works-some-examples-3.html