I don't think Prof Leach used TMC. This is quite a new technique popularized by Edmond Stuart (Douglas Self also had some papers from Baxandall showing he was also thinking about something like this shortly before he passed).
The Leach used TPC, which is is just as effective in reducing distortion, but the mechanism is different. With TPC, you keep the loop gain high up until HF, and then you drop it off at 40 dB per decade, but before you get to the ULGF, you reduce the slope to 20 dB per decade. This pulls the phase margin back up again and the result is the amp is stable.
Google Harry Dymond 2 pole compensation for a definitive expose - it's a great technique.
😎
The Leach used TPC, which is is just as effective in reducing distortion, but the mechanism is different. With TPC, you keep the loop gain high up until HF, and then you drop it off at 40 dB per decade, but before you get to the ULGF, you reduce the slope to 20 dB per decade. This pulls the phase margin back up again and the result is the amp is stable.
Google Harry Dymond 2 pole compensation for a definitive expose - it's a great technique.
😎
"TMC" is a 2-pole technique - with some adjustment of part values the loop gain can be identical to 2-pole with the R to ac gnd when TMC is measured "inside" the loop formed by the bootstrap R, measuring the loop gain seen by the output stage - which is what matters for stability
and I recall that a long expired US patent was eventually found
and there's Baxandall: http://www.douglas-self.com/ampins/baxandall/Baxandall papers 34.jpg
and I recall that a long expired US patent was eventually found
and there's Baxandall: http://www.douglas-self.com/ampins/baxandall/Baxandall papers 34.jpg
O-stripper
""Super KISS" seems to beat "Miibtech" " Low distortion was newer the target. for that I think traditional GFB is better, The idea was to have two loops, one strong gain of one loop around the speaker drive, and one loop that then drives the outer loop. Again i don't see the purpose of getting the back emf and time delayed micro-phonic signals from the speaker into the input-stage of my amplifier, I also like to keep the problem from switching concealed in the OPS. My total data may suffer from this, but I have a strong feel that what I loose is gained in other areas.
""Super KISS" seems to beat "Miibtech" " Low distortion was newer the target. for that I think traditional GFB is better, The idea was to have two loops, one strong gain of one loop around the speaker drive, and one loop that then drives the outer loop. Again i don't see the purpose of getting the back emf and time delayed micro-phonic signals from the speaker into the input-stage of my amplifier, I also like to keep the problem from switching concealed in the OPS. My total data may suffer from this, but I have a strong feel that what I loose is gained in other areas.
I spent dozens of hours, used sims of 3 independent tests of loop gain/stability in the Cordell threads to try to help people understand the essentially 2-pole character of TMC - that TMC involves the same stability consequences as using 2-pole compensation
the "single pole" global loop gain appearance of TMC measured "the usual way" with the test probe in series with the global feedback only gives misleading tuning of the compensation, dangerously small stability margin
the "single pole" global loop gain appearance of TMC measured "the usual way" with the test probe in series with the global feedback only gives misleading tuning of the compensation, dangerously small stability margin
MiiB
is scary coming from someone billing themselves as an Engineer, Audio Designer - you do actually use your education in audio design?
the point of getting measured errors back to your input differencing stage is to have feedback reduce them
linearizing the input stage for audio frequency errors is simple - use local degeneration, high loop gain - it is easy to achieve -120 dB gm linearity at the input diff stage, for the input error levels seen with the amp's full scale output, audio frequency range, with actual commercial distribution recorded music signal levels - even after loudspeaker distortion adding harmonics, IMD
hi Edmund - I should say TMC can lead to dangerously small stability margin if you "tune" the compensation by eyeball, using step overshoot as you would with a single dominant pole compensated amp
if you measure correctly you can design 2-pole or TMC to safe stability margins - in fact 2-pole can have overshoot in the step response that isn't a direct indication of lack of stability margin either
is scary coming from someone billing themselves as an Engineer, Audio Designer - you do actually use your education in audio design?
the point of getting measured errors back to your input differencing stage is to have feedback reduce them
linearizing the input stage for audio frequency errors is simple - use local degeneration, high loop gain - it is easy to achieve -120 dB gm linearity at the input diff stage, for the input error levels seen with the amp's full scale output, audio frequency range, with actual commercial distribution recorded music signal levels - even after loudspeaker distortion adding harmonics, IMD
hi Edmund - I should say TMC can lead to dangerously small stability margin if you "tune" the compensation by eyeball, using step overshoot as you would with a single dominant pole compensated amp
if you measure correctly you can design 2-pole or TMC to safe stability margins - in fact 2-pole can have overshoot in the step response that isn't a direct indication of lack of stability margin either
Last edited:
A good reason is that it is a very good way to have them corrected to a large degree.
Edit: ... as jcx so eloquently noted.
Jan
Aren't TPC snd TMC both technically two pole compensation schemes? Am I correct thinking simple Miller compensation is single pole compensation?
Jcx/jan. I completly understand this assumption, but correction is something you can do to things happening inside the control loop, external things from outside the loop is not corrected. So for back emf and mechanical microphonic pickup from inductive cables are not corrected by the feedback loop. For that we have the output impedance as the control mechanism. Current and voltage are not really connected in time, we control the voltage via feedback, current based issues are handled by damping factor, this current can be seen as an out of time error signal feed to the summing node of the feed back. This is what I try to avoid with the dual loop.
One thing I note by the two loops is the fact that distortion only doubles when you go from 1Khz 2v out to 20khz 40v out. With normal GFB that figure is 100 times.
Also when you have two Independant loops the stability margin is better as you work with fewer poles.
I do function as a speaker designer. This here is a hobby of mine. I am clearly on foreign turf, and I have no illusions of understanding all and everything. But if all just keep designing the amplifier over and over, then how do we ever move..??
One thing I note by the two loops is the fact that distortion only doubles when you go from 1Khz 2v out to 20khz 40v out. With normal GFB that figure is 100 times.
Also when you have two Independant loops the stability margin is better as you work with fewer poles.
I do function as a speaker designer. This here is a hobby of mine. I am clearly on foreign turf, and I have no illusions of understanding all and everything. But if all just keep designing the amplifier over and over, then how do we ever move..??
Last edited:
That is, at least, something we all have to ... understand ;-)
It gives a more innovative position and help to exchange nicely and interestedly the few things we all know, with are different for all of us. (Read my lips ;-)
I think you are right, trying to separate as much as possible the various evils with occur an amplifier, (even those coming from outside) and try to cure them as close as possible to their sources. It seem obvious than simpler is the error signal, better it is
MiiB, my education pushing me toward Jcx/jan rezoning, but you are right, maybe it is something there, try and see.
Yes they are corrected. Anything that would appear at the output and which is not in the input is corrected. EMI radiated into the PCB, power supply ripple intrusion, EMF from the speaker, all is corrected by nfb.
Jan
But to mitigate the additional polls in a standard GNFB amp you can reduce the ULGF. Then you gain stability. The interesting part would be to set up the GNFB amp to the same stability criteria as you "two loop" amp and then compare the THD etc.
My prediction would be that the spec's would be comparable. This is just my gut feeling and will probably be proved wrong...
Thank you for pointing this out. How do I miss these things?
Although, from sims I don't see the 5x reduction in THD with TMC vs TPC. I do see the benefit of the 2 pole nature of the TMC when comparing with single pole miller. TPC seems more effective but the square wave overshoot bothers me. The implementation of the rest of the amp must have significant effect.
The question in my head now is how the finer details of TMC compare to the "split" feedback method. Something to ponder me thinks...
Including HF components and the IM that the amp will product with them...
One of the advantages of output inductances.
(On my side, i use shielded Speakers cables and enclosures because i found a difference ;-)
One thing that puzzles me with amplifiers is why I get very different distortion results with loads, and with real speakers as loads the distortion is really bad. If feedback was the key to make load invariant amplifiers, then it would all be very easy. Frankly it's. not and I don't see feedback being to cure. Things are not so ideal as the theories tells us
Total. 31650
Total. 31650
One thing that puzzles me with amplifiers is why I get very different distortion results with loads, and with real speakers as loads the distortion is really bad. If feedback was the key to make load invariant amplifiers, then it would all be very easy. Frankly it's. not and I don't see feedback being the cure. Things are not so ideal as the theories tells us