Not interested in this wretched email anymore. There has been,as far as I know, no suggestion that it contains any important technical insight. Give it a rest, eh?
Compensation in a feedback amplifier is normally considered to mean stabilising the main feedback loop. It is not expected to also suppress any parasitic instabilities. While the compensation used may well have some interaction with the liability to parasitics, to say that anything that suppresses any sort of oscillation counts as compensation is an unhelpful misuse of language, in my humble opinion.
You will not be disappointed!
The emphasis now is more on obtaining the optimal slewrate.
A lot more on feedback is planned.
A whole new chapter on that is in the works.
I should make it clear that I have your book and follow these recommendations.
Yes, you probably did not have a copy of Self's book handy but this is what I made reference to, a CFP that drives an Emitter Follower, what I called a CFP + EF.
I think that the CFP used as a driver pair is at least OK and, as D.Self wrote, shows "promise". So my question was on the use of a MOSFET in a CFP driver.
So, just as a CFP seems better sense as a pre-driver/driver than an output
does the rationale for a MOSFET alter as a CFP driver rather than an output?
And I repeat my support that TMC should be used as the default abbreviation. It is a reference, we don't complain that Self or Cordell are inadequate just because there could be someone else of the same surname.
Best wishes
David
Doug,
The history is simple. You were late to the party on TMC and you should give credit where credit is due. In this case, that is to Edmond (Baxandall's private musings notwithstanding). You can't, and aren't, first on everything, nor are the rest of us.
My suggestion for your sixth edition is to be more diligent and generous in giving credit to others and to understand that doing so does not reduce the fine quality of the texts that you write.
A case in point is Figure 8.43 in your Fifth Edition. It is nearly identical to one that I published and explained 30 years ago. An attribution there would have been in line with common practice in professional publications.
On the matter of compensation, you need to understand that there is more than one loop in most amplifiers that may need some form of compensation. A myopic focus on just the global loop is naive and it leads to many amplifiers that may have inexplicably poor sound quality because there may be instabilities internal to the main loop that are excited under certain program conditions. I'm surprized that you seem not to have picked up on this.
When I employed Miller Input Compensation in my MOSFET amplifier in 1983 I recognized that the wideband compensation loop needed compensation. Similarly, even simple Miller compensation can at times need compensation. The more transistors you have in a loop, the more excess phase can build up and the likelihood of needing local compensation increases. This can be the case when a Darlington is used for the VAS and the VAS collector is very lightly loaded, as when driving a Triple. The matter can be even more problematic if in addition the VAS is cascoded. This is why sometimes a shunt C or shunt series R-C is needed for stability at the VAS collector.
It is possible that this issue contributes in part to the difficulty that some have experienced with oscillations in Triples. I don't know. See my comments and suggestions for Stabilizing a Triple in my book on pages 207 and 208, and Figure 10.15.
You might also look at Figures 9.4 and 9.5 in "Designing Audio Power Amplifiers" to see that TPC by itself can sometimes benefit from a bit of further compensation to avoid a very large peak in the open loop gain in the audio band. I called this bridged T compensation (BTC).
Of course, there are many other cases where parasitic oscillations can occur, and whose taming is not what one might call "compensation". I would not call the use of a base stopper resistor compensation, for example.
Cheers,
Bob
Pardon a naive question but, why not? It modifies the open loop response (low pass filter) to provide stability against feedback (albeit unintended positive feedback).
Sorry to poop your party, but in this particular case the credit should go to Kunio Seki and Hitachi. I don't think that re-baptizing (to "TMC") something that was patented 35 years ago (1978) deserves any special credit.
The 1978 patent was brought to your attention several times so far, but for whatever reasons you don't seem to grasp this. Mr. Self is as much entitled to baptize this (again, 1978 patented) compensation technique as anybody else, provided he's quoting the original source. Which, BTW, you failed to do in your book (or admit it ever after).
I never said they were. In fact I said the exact opposite.
I must, to my infinite regret, disagree. If we are speaking here of either standard Miller compensation or Output-Inclusive Compensation, then in both cases the inner loop is a simple capacitor around the VAS at high frequencies. I don't see that shunt 33pF capacitors are going to have much effect on that.
I am not going to pretend that the action of these small shunt capacitors is rigorously grounded in theory, because it isn't. I suspect that they prevent the VAS output impedance from going inductive at very high frequencies. This fits in with my practical observations, but as to the theory, I doubt very much if parasitic oscillation is susceptible to either calculation or simulation.
The shunt capacitors have been found to be effective with both an EF-VAS (with added emitter-follower) and a bootstrapped cascode VAS, in large quantity production over many years.
As I explained, this is not the case. The purpose of Miller compensation is to stabilise the feedback loop. While susceptability to parasitic oscillation may be affected by the amount of compensation applied, parasitic suppression is not its purpose.
The page you quote deals exclusively with stabilising an amplifier with Input-Inclusive compensation, or TIIMC, and I don't see it as relevant here. The use of small shunt capacitors from the VAS collector to ground (rather than the supply rails) to suppress parasitics appears to be effective with most forms of compensation, and is covered on page of the 5th edition of my book. (you do have my book, don't you?)
That peak is one that I looked at with some interest.
If I am correct then it is partly an artefact of your model shown in 9.4.
The IPS modelled as a simple transconductance can provide essentially unlimited gain. In a real circuit the gain continues to climb as the frequency drops until the IPS gain approaches its limit and then tends to flatten out without such an extreme peak. In other words, extra gain just lowers the frequency of the pole(s).
There was a similar issue in Harry Dymond's JAES article on TPC where his simplified analysis model predicted a Q for the peak much in excess of that revealed by the SPICE plots of his typical circuit. He points out the discrepancy in his update and review addenda after I raised it in a post.
The peak can become more prominent if the IPS has some "spare" gain because it is resistively loaded. IIRC JCX has pointed this out too.
Dennis Feucht has done a detailed analysis of TPC but it is not very intuitive.
Perhaps more material for the next edition?
Best wishes
David.
Addenda
And similarly for the VAS of course. I am not yet sure how they interact to drive the peak.
Best wishes
David
And similarly for the VAS of course. I am not yet sure how they interact to drive the peak.
Best wishes
David
I'd say not quite, although I agree that these semantical things can be a bit of a judgment call.
Many times a local situation can develop with as little as one transistor. Often the local parasitics around and within the transistor find a way to achieve an oscillator topology, often of the form of a Colpits or Hartley. Often the collector-base or gate-drain capacitance is involved. Sometimes the negative impedance seen at the base due to some capacitive loading in combination with the 6dB/octave beta roll-off in an emitter follower is involved. Base stopper resistors often kill these oscillation tendencies by damping the Q of the oscillatory elements or canceling negative resistance seen at the base. The inclusion of a small base stopper resistor also usually has very little influence on the open-loop gain. So for these types of measures, I'm more inclined not to call them compensation. Nevertheless, I agree it is a semantical judgment call.
Cheers,
Bob
But that negative resistance is a function of feedback (again, often within a single device), so I think I'll stick to my semantic view.
Feedback is feedback, whether as an explicit loop, unintended coupling, or intentional degeneration/regeneration. It exists in every real world circuit, regardless of the "philosophy" of the "designer." (Rule of thumb: avoid any box of gain designed by someone with a philosophy. Remember what Masha said to Max Renn in Videodrome.)
Feedback is feedback, whether as an explicit loop, unintended coupling, or intentional degeneration/regeneration. It exists in every real world circuit, regardless of the "philosophy" of the "designer." (Rule of thumb: avoid any box of gain designed by someone with a philosophy. Remember what Masha said to Max Renn in Videodrome.)
Good question. I'm guessing it would reduce the need for a base stopper resistor if a base stopper resistor was needed in the first place, but I don't recall trying it, so I'm not sure.
Cheers,
Bob
But that negative resistance is a function of feedback (again, often within a single device), so I think I'll stick to my semantic view.
Feedback is feedback, whether as an explicit loop, unintended coupling, or intentional degeneration/regeneration. It exists in every real world circuit, regardless of the "philosophy" of the "designer." (Rule of thumb: avoid any box of gain designed by someone with a philosophy. Remember what Masha said to Max Renn in Videodrome.)
I agree. Feedback is feedback, and it can take many forms. It can be extremely local or very global. A Triple without any global negative feedback in a no-NFB amplifier will still try to find a way to oscillate. Of course, one might argue that there is no such thing as a truly no-NFB amplifier anyway
.Cheers,
Bob
Edmond said:
So what? Besides, I said 'similar issues exists.....'
Happily, most of us do see the relevance. The point is that if the Miller loop encompasses more than just a single gain stage, no matter if it includes the input stage, driver, output stage, or whatsoever, this loop (it's a feedback loop!) tends to be unstable. It's not due to 'parasitics' (whatever that may mean), rather due to the number of gain stages involved. Therefore the Miller loop itself needs to be stabilized by means of local frequency compensation. Bob Cordell did it by means of a shunt RC network between the collectors of the IPS and Ed Cherry did it with two shunt caps at the TIS output. Although different circuitry, yet these compensating tricks are based on the same principle.
The use of small shunt capacitors from the VAS collector to ground (rather than the supply rails) to suppress parasitics appears to be effective with most forms of compensation, and is covered on page of the 5th edition of my book. (you do have my book, don't you?)
Yes, I do have your book. I've downloaded it, so I've paid the right price for it....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.