Once again, wanted to find these posts and threads by a title search rather than digging for them.
TMC is a derivative/improvement on Cherry's Miller inclusive compensation (or Miller output inclusive) that was first described and tested by Peter Baxandall and refined and promoted by Edmund Stuart on here. Baxandall did not publish his work, rather his wife, after his death, forwarded his notes on Douglas Self's book to him and they included the yet unnamed concept of TMC. Self has posted the TMC part of those notes here: The Baxandall Papers: Transitional Miller compensation
And they are published in full in Linear Audio's "Baxandall and Self on Audio Power". The book is a collection of previously published papers by Baxandall and Self with his notes to Self at the end. Cherry's Miller inclusive compensation was used in very early OP amps of the 1970s and it is reported to work less well in power amps where the output stage significantly slows down in saturation. Post by Edmund Stuart on Cherry's work: Bob Cordell's Power amplifier book - Page 87 - diyAudio
TMC acts as Miller inclusive at audio frequencies but transitions to normal Miller compensation at HF where stability is important providing the best of both worlds.
Much of the detailed analysis and even the naming/promotion of TMC by Edmund Stuart happened on this forum and I wanted to use this thread to collect some of the better links.
Controversy started over TMC when some (michaelkiwanuka) claimed that TMC and TPC were essentially the same thing. I at first did not believe this, BUT some of our best people here jcx, megajocke and others backed up the claim. Megajocke's analysis here has convinced me with regard to small signal analysis but I believe that there are potential differences when the output stage saturates and the parameters change. Here is Megajocke's excellent analysis: Bob Cordell's Power amplifier book - Page 122 - diyAudio
Several argued that TMC was first order compensation as is normal Miller Comp but Megajocke pointed out that that was with regard to the global loop and that if one probes inside the local TMC loop there are other stability issues to consider. jcx argues that it is mostly _not_ a first order system: Bob Cordell's Power amplifier book - Page 110 - diyAudio
Another good post by megajocke with a response from Bob: Bob Cordell's Power amplifier book - Page 125 - diyAudio
One of megajocke's simulations: http://www.diyaudio.com/forums/soli...lls-power-amplifier-book-117.html#post2411182
Simulation from Bob showing that it is possible to choose compensation values in TPC to get the same distortion performance as TMC, he uses a simple Blameless topology for the example: http://www.diyaudio.com/forums/soli...lls-power-amplifier-book-119.html#post2413197
Post by Bob of his MOSFET amp with the EC removed and TMC added, he mentions the excellent distortion results. On closer inspection this is actually a MIC/TMC hybrid, very interesting, also note the driven cascode: http://www.diyaudio.com/forums/soli...ells-power-amplifier-book-86.html#post2391212
Wish I had time to read the entire thread and carefully link all of the best posts and simulations but there is just not enough time in the day - this is a start.
I'm inclined to use TMC, perhaps with gentle lead compensation in a new design but it is important to observe the changes in stability when the output saturates and with capacitive loads. I'd want to see if TPC is better stability wise when tuned for similar distortion. I was firmly in the TMC camp but megajocke's and jcx's posts have made it clear that they are similar but my opinion is perhaps not from a large signal perspective.
TMC is a derivative/improvement on Cherry's Miller inclusive compensation (or Miller output inclusive) that was first described and tested by Peter Baxandall and refined and promoted by Edmund Stuart on here. Baxandall did not publish his work, rather his wife, after his death, forwarded his notes on Douglas Self's book to him and they included the yet unnamed concept of TMC. Self has posted the TMC part of those notes here: The Baxandall Papers: Transitional Miller compensation
And they are published in full in Linear Audio's "Baxandall and Self on Audio Power". The book is a collection of previously published papers by Baxandall and Self with his notes to Self at the end. Cherry's Miller inclusive compensation was used in very early OP amps of the 1970s and it is reported to work less well in power amps where the output stage significantly slows down in saturation. Post by Edmund Stuart on Cherry's work: Bob Cordell's Power amplifier book - Page 87 - diyAudio
TMC acts as Miller inclusive at audio frequencies but transitions to normal Miller compensation at HF where stability is important providing the best of both worlds.
Much of the detailed analysis and even the naming/promotion of TMC by Edmund Stuart happened on this forum and I wanted to use this thread to collect some of the better links.
Controversy started over TMC when some (michaelkiwanuka) claimed that TMC and TPC were essentially the same thing. I at first did not believe this, BUT some of our best people here jcx, megajocke and others backed up the claim. Megajocke's analysis here has convinced me with regard to small signal analysis but I believe that there are potential differences when the output stage saturates and the parameters change. Here is Megajocke's excellent analysis: Bob Cordell's Power amplifier book - Page 122 - diyAudio
Several argued that TMC was first order compensation as is normal Miller Comp but Megajocke pointed out that that was with regard to the global loop and that if one probes inside the local TMC loop there are other stability issues to consider. jcx argues that it is mostly _not_ a first order system: Bob Cordell's Power amplifier book - Page 110 - diyAudio
Another good post by megajocke with a response from Bob: Bob Cordell's Power amplifier book - Page 125 - diyAudio
One of megajocke's simulations: http://www.diyaudio.com/forums/soli...lls-power-amplifier-book-117.html#post2411182
Simulation from Bob showing that it is possible to choose compensation values in TPC to get the same distortion performance as TMC, he uses a simple Blameless topology for the example: http://www.diyaudio.com/forums/soli...lls-power-amplifier-book-119.html#post2413197
Post by Bob of his MOSFET amp with the EC removed and TMC added, he mentions the excellent distortion results. On closer inspection this is actually a MIC/TMC hybrid, very interesting, also note the driven cascode: http://www.diyaudio.com/forums/soli...ells-power-amplifier-book-86.html#post2391212
Wish I had time to read the entire thread and carefully link all of the best posts and simulations but there is just not enough time in the day - this is a start.
I'm inclined to use TMC, perhaps with gentle lead compensation in a new design but it is important to observe the changes in stability when the output saturates and with capacitive loads. I'd want to see if TPC is better stability wise when tuned for similar distortion. I was firmly in the TMC camp but megajocke's and jcx's posts have made it clear that they are similar but my opinion is perhaps not from a large signal perspective.
Last edited:
I know it's more than a year old. But when actually understand the circuit with sophisticated simulations and experiments. TMC is a way better solution in every way than TPC. The main and possibly the only difference is TMC preserve phase margin very very well, almost adding only Aol without affecting phase margin at all. Where it's quite opposite of the TPC. Surely at most part to obtain the same Aol TPC is much better than plain miller comp. But compared to TMC, TPC is almost as bad as miller comp.
Innovative move may be, potentially combination of TPC + TMC. Mainly TMC with low value resistor (200-1000) and adding a little bit TPC to boost a bit AOL further more(only if the VAS stage does have that much gain to begin with) with high resistance (1000-10k). And adding resistor in series of the inner cap of TMC boosts phase margin at zero crossing with a bit phase bandwitch sacrifice.
Innovative move may be, potentially combination of TPC + TMC. Mainly TMC with low value resistor (200-1000) and adding a little bit TPC to boost a bit AOL further more(only if the VAS stage does have that much gain to begin with) with high resistance (1000-10k). And adding resistor in series of the inner cap of TMC boosts phase margin at zero crossing with a bit phase bandwitch sacrifice.
Wrong on all counts. All other things being equal, TMC and TPC are perfectly equivalent when it comes to stability margins vs. the available loop gain. From a compensation perspective, only the compensation order (2 for either TPC or TMC) matters, the rest is all on the amplifier open loop bandwidth and gain.
Meantime, this Patent was found, then yours truly was able to prove mathematically the equivalence based on wye-delta transformations: TMC is exactly equivalent to TPC, plus a lead compensation, exactly how PB2 empirically discovered (and Megajocke proved independently, using more of a mathematical brute force approach). It's buried somewhere on this forum...
TMC or TPC, it's just a matter of preference, as so many other things in Audio (VFA vs. CFA, etc...).
Last edited:
You are wise now.
Please, analyze OITPC compensation by Dadod. OITPC is TCP + Cherry compensation.
I don’t know how to do that, I don’t think it’s possible, analytically, and honestly I don’t care much about.
One thing I know is that there is no free lunch; the stability vs. loop gain trade is unavoidable, any attempt to go around is an illusion (usually simulation driven). The “secret sauce” of the OITPC compensation is the unusually high ULGF, not the compensation network itself. The only thing that matters when it comes to the compensation network is the order N (N the number of poles and N-1 the number of zeroes to bring back the phase). Miller is N=1, TPC and TMC are N=2, Cherry NDFL is N=3. OITPC is also N=2, hence it does not offer anything that TPC or TMC can’t do, if one takes care to create a double pole (an easy exercise).
In case anybody wonders about the above description, it’s foundation is in two facts a) the maximum feedback theorem and b) audio amplifiers are minimum phase systems, that is, the phase is not independent, but a function of the gain Hilbert transform (many other properties exist).
One thing I know is that there is no free lunch; the stability vs. loop gain trade is unavoidable, any attempt to go around is an illusion (usually simulation driven). The “secret sauce” of the OITPC compensation is the unusually high ULGF, not the compensation network itself. The only thing that matters when it comes to the compensation network is the order N (N the number of poles and N-1 the number of zeroes to bring back the phase). Miller is N=1, TPC and TMC are N=2, Cherry NDFL is N=3. OITPC is also N=2, hence it does not offer anything that TPC or TMC can’t do, if one takes care to create a double pole (an easy exercise).
In case anybody wonders about the above description, it’s foundation is in two facts a) the maximum feedback theorem and b) audio amplifiers are minimum phase systems, that is, the phase is not independent, but a function of the gain Hilbert transform (many other properties exist).
For small signal analyze, I agree with you.
Maybe someone can analyze large signal property.
I'm fairly certain that page was authored by Edmond Stuart on this site. You should be able to right click, then print, and choose to .pdf file, not sure if the formatting will be right.
Edmond and Ovidu (syn08 here) did the PGP amp many years ago which led to YAP:
YAP Schematics
This important quote is relevant to this discussion:
"The frequency compensation is Transient Miller Compensation (short: TMC) based. It's not the right time and place to get into a TMC debate, the subject was already beaten to death; enough to know it's a compensation technique derived from the two pole compensation (short: TPC) method and patented in the late '70s. It was eventually analyzed by Baxandall and lately revived by E. Stuart on DIYAudio (which also baptized it as "TMC") and also by Bob Cordell in his latest book. In this thread, it was mathematically proved (and confirmed by simulations) that TMC is exactly equivalent to a standard TPC plus a lead-lag compensation. As such, TMC does not, in any way shape or form, offer any performance improvement over TPC, neither in the loop gain or phase margin departments. It's only a clever way to split the same avalable loop gain across both the OPS and the VAS."
From this YAP page:
YAP Front End
This page from the YAP provides some excellent discussion on implementation issues,
triple stability, SOA, speed up cap - very nice work:
The YAP Output Stage
Edmond and Ovidu (syn08 here) did the PGP amp many years ago which led to YAP:
YAP Schematics
This important quote is relevant to this discussion:
"The frequency compensation is Transient Miller Compensation (short: TMC) based. It's not the right time and place to get into a TMC debate, the subject was already beaten to death; enough to know it's a compensation technique derived from the two pole compensation (short: TPC) method and patented in the late '70s. It was eventually analyzed by Baxandall and lately revived by E. Stuart on DIYAudio (which also baptized it as "TMC") and also by Bob Cordell in his latest book. In this thread, it was mathematically proved (and confirmed by simulations) that TMC is exactly equivalent to a standard TPC plus a lead-lag compensation. As such, TMC does not, in any way shape or form, offer any performance improvement over TPC, neither in the loop gain or phase margin departments. It's only a clever way to split the same avalable loop gain across both the OPS and the VAS."
From this YAP page:
YAP Front End
This page from the YAP provides some excellent discussion on implementation issues,
triple stability, SOA, speed up cap - very nice work:
The YAP Output Stage
The Baxandall papers show clearly the relation of parasitic capacitances, high (output-) impedances and high amplification of the given and mostly adopted amplifier topology, that of the input stage, voltage ampifier stage and current amplifier stage.
The aim to become less dependent of the somewhat conflicting properties of the voltage gain stage to eliminate the frequency-related problems caused by the miller capacitance and high output impedance ('resistance') to compensate these unwanted frequency responses or oscillations, by making a connection to a next low (lower) output impedence stage, one arrives at an even more complex circuit which has to be calculated with accuracy and possible adjustments and tuning afterwards before a reliable amplifier has been realised.
Studying the circuits in the various references, I noticed high complexity, countless components and hypnotising plots and graphs.
Not exactly preferred I guess by commercial enterprises (save the niche market) or down to earth experimenting diy-enthusiasts by the simple argument of the wide tolerances of most components other than precision 1% resistors. Capacitances +/- 20% are an usual yield when buying them, bjt's ranges of beta with a factor of two, jfet's spreading is notorious.
The results however reached by this subtle architecture are impressive anyway. But it turns my thoughts towards a different direction, to an approach which differs considerably from this classic (conservative) topology and is in its fundamental 'design-nature' less or not susceptible to millercaps and it burdens.
One might think of triode tubes, which do amplify (significant less) but with an output impedance in the kilo-ohms instead of mega-ohms. No millercap-compensating circuitry known, overall feedback modest and sounding awesome though less 'precise' as laboratory equipment is.
What would happen with these TMC and TPC solutions if we were able to transform the usual voltage amplifier stage with a high-beta bjt to a modified stage with less amplification but also with a substantial lower output impedance. Would all problems being solved then and would these TMC/TPC additions disappear?
I am curious if someone would dare to simulate such a variation.
The aim to become less dependent of the somewhat conflicting properties of the voltage gain stage to eliminate the frequency-related problems caused by the miller capacitance and high output impedance ('resistance') to compensate these unwanted frequency responses or oscillations, by making a connection to a next low (lower) output impedence stage, one arrives at an even more complex circuit which has to be calculated with accuracy and possible adjustments and tuning afterwards before a reliable amplifier has been realised.
Studying the circuits in the various references, I noticed high complexity, countless components and hypnotising plots and graphs.
Not exactly preferred I guess by commercial enterprises (save the niche market) or down to earth experimenting diy-enthusiasts by the simple argument of the wide tolerances of most components other than precision 1% resistors. Capacitances +/- 20% are an usual yield when buying them, bjt's ranges of beta with a factor of two, jfet's spreading is notorious.
The results however reached by this subtle architecture are impressive anyway. But it turns my thoughts towards a different direction, to an approach which differs considerably from this classic (conservative) topology and is in its fundamental 'design-nature' less or not susceptible to millercaps and it burdens.
One might think of triode tubes, which do amplify (significant less) but with an output impedance in the kilo-ohms instead of mega-ohms. No millercap-compensating circuitry known, overall feedback modest and sounding awesome though less 'precise' as laboratory equipment is.
What would happen with these TMC and TPC solutions if we were able to transform the usual voltage amplifier stage with a high-beta bjt to a modified stage with less amplification but also with a substantial lower output impedance. Would all problems being solved then and would these TMC/TPC additions disappear?
I am curious if someone would dare to simulate such a variation.
Last edited:
The authors are quite frank about the issues with an average home builder building the PGP, yes it is overly complex but it is just an example of what can be done. I believe that they learned a lot and have come a long way so that a simpler topology might come close enough to the PGP to make the complexity unnecessary. In fact they call their older designs outdated, still we can learn from their work.
Bob Cordell simulated a slightly modified "Blameless" design with TMC and obtained excellent, good enough IMO, results. This is what is of more interest to me. Bob's old MOSFET design with a bipolar output stage triple and TMC is also of interest to me - not too complicated.
Should have included a link to Bob's Blameless simulation comparing TMC and TPC: https://www.diyaudio.com/forums/sol...lls-power-amplifier-book-119.html#post2413197
Link to Bob's revision of his old MOSFET amp to employ TMC rather than error correction in the output stage and obtain similar results with less complexity: https://www.diyaudio.com/forums/sol...ells-power-amplifier-book-86.html#post2391212
I thought that I included these links in the first post of the thread.
Bob Cordell simulated a slightly modified "Blameless" design with TMC and obtained excellent, good enough IMO, results. This is what is of more interest to me. Bob's old MOSFET design with a bipolar output stage triple and TMC is also of interest to me - not too complicated.
Should have included a link to Bob's Blameless simulation comparing TMC and TPC: https://www.diyaudio.com/forums/sol...lls-power-amplifier-book-119.html#post2413197
Link to Bob's revision of his old MOSFET amp to employ TMC rather than error correction in the output stage and obtain similar results with less complexity: https://www.diyaudio.com/forums/sol...ells-power-amplifier-book-86.html#post2391212
I thought that I included these links in the first post of the thread.
As I am biased myself: I prefer simple designs such as by Hiraga. Millercaps are present there too, but only affecting local, openloop and closed loop performance when it is poorly build (eg 'long' twisted wires to the output stage, increasing capacative load on the voltage driver causing oscillations; yes, some people do this to prevent external noise interfering).
But in overall, low gain stage impedance (load resistance of 1-2 kohms), no bias circuit at all, no protection needed (no pops, no dc lability). But of low power output as well. Sufficient for home use, others prefer beefier menus.
After my current design exercise, a challange awaits with the classical, more powerfull design. Then this and referring topics are very usefull indeed! I might end up with some strewn capacitances in obscure but yet unknown design corners.
Hats off for Baxandall.
And thanks PB2.
But in overall, low gain stage impedance (load resistance of 1-2 kohms), no bias circuit at all, no protection needed (no pops, no dc lability). But of low power output as well. Sufficient for home use, others prefer beefier menus.
After my current design exercise, a challange awaits with the classical, more powerfull design. Then this and referring topics are very usefull indeed! I might end up with some strewn capacitances in obscure but yet unknown design corners.
Hats off for Baxandall.
And thanks PB2.
Totally agreed.
All variables are set in compromise values.
However, balancing and refining these might end in different topologies.
Lots of fun though!
This is another two pole compensation which is TPC + cherry or OITPC by Dadod:
https://www.diyaudio.com/forums/solid-state/338815-anistardi-peletuk-4.html#post5835054
Distortion is H2 dominant at all audio frequency.
https://www.diyaudio.com/forums/solid-state/338815-anistardi-peletuk-4.html#post5835054
Distortion is H2 dominant at all audio frequency.
OMG, I just did the math for TMC, at you are correct. For purposes of fixing errors in the output stage, the TMC and TPC are the same at the frequency of interest. They diverge at high frequencies as we get close the unity gain frequency of the amplifier, but at that point, it does not matter anymore.
Also I just simulated it to make sure my analysis is correct ... and it is.
This was quite a revealing analysis since from an overall loop gain perspective, they behave very differently: TPC gives a 2 pole roll-off and TMC is single pole with a small transition half way when you go from C1 to C1||C2 roll-off, but for the purpose of fixing output stage errors, they are both 2 pole roll-off.
Very interesting indeed. I guess this is one of those deceptive circuits, where the visualization makes you believe something, but once you do the analysis you realize is not true. Also, the fact that Bob explained it wrong in his book does not help.
Now, if the TMC and the TPC networks are identical, they perform the same at reducing output stage errors. Now the question left is, are they optimum when they are the same. Probably not since they affect the global loop differently. This is the next analysis to do... but first some reading to see what people have discovered so far, specially how the minor loop of the amplifier affected.
EDIT: I read megajocke's analysis... way too complicated. You don't need to do all that to see it.
Thanks again,
Sandro
Also I just simulated it to make sure my analysis is correct ... and it is.
This was quite a revealing analysis since from an overall loop gain perspective, they behave very differently: TPC gives a 2 pole roll-off and TMC is single pole with a small transition half way when you go from C1 to C1||C2 roll-off, but for the purpose of fixing output stage errors, they are both 2 pole roll-off.
Very interesting indeed. I guess this is one of those deceptive circuits, where the visualization makes you believe something, but once you do the analysis you realize is not true. Also, the fact that Bob explained it wrong in his book does not help.
Now, if the TMC and the TPC networks are identical, they perform the same at reducing output stage errors. Now the question left is, are they optimum when they are the same. Probably not since they affect the global loop differently. This is the next analysis to do... but first some reading to see what people have discovered so far, specially how the minor loop of the amplifier affected.
EDIT: I read megajocke's analysis... way too complicated. You don't need to do all that to see it.
Thanks again,
Sandro
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.