I wouldnt say a few, more like all of them, trend started in the 80s and continued to their latest designs.
True enough about the VAS output impedance
Not quite clear what sort of non-linear interaction you are talking about here.
I assume you mean stopper resistors (circa 5 Ohm) directly in series with the output device bases. These are not a good idea, though they appear to be endemic in Japanse hifi design; I assume the idea is to suppress local oscillations. Unfortunately they directly degrade LSN, worsening the linearity into sub-8 Ohm loads.
I have never found it necessary to use these in an amplifier design. I have used 100 Ohm resistors in front of drivers in the past, but in general I wouldn't do it now.
I found that as the base stopper resistor was increased from zero there was an improvement in stability, but only to a point, then it deteriorates. Normally one expects a simple trade-off - more resistance at the cost of frequency response or whatever.
So this optimum was a surprise to me.
Yes.
You mention in your book that the usual 100 nF Zobel capacitor was such a convenient round number that you were suspicious.
I felt the same about your 100 ohm driver stopper! Especially because there was no analysis or discussion of how this value was arrived at.
So this, in part, prompted the research on transistor base stopper resistors - optimization, allocation between the drivers and the outputs and so on.
Dennis Feucht analyses emitter followers in his book but the matter is not yet completely clear to me, and CFP drivers need even more work.
Best wishes
David
Last edited:
I would think that non-linear base current in the output stage (with a nice linear output voltage) would be converted to non-linear voltage by any resistance in the base circuit.
That would mean that one optimization would be for minimum base (stopper) resistance as well as minimum (pre)driver output resistance, no?
Jan
That would mean that one optimization would be for minimum base (stopper) resistance as well as minimum (pre)driver output resistance, no?
Jan
I probably did not write very clearly.
I meant that increase in the stopper resistor had a non-monotonic effect, there was a peak at a certain value and not just "more resistance = better damped response".
So I meant "non-linear" in that sense, just that "non-monotonic" sounded a bit mathematical, scares some people
I did not consider the distortion of the base current. Sorry for any confusion, it was more of a brief note.
Best wishes
David
Stability Margin - "TMC" simply isn't safely designed with an eye only on the overshoot, relying on the "single pole" appearance of the compensated VAS+output, using the standard Miller dominant pole heuristics
you really need to design to the same level of sophistication, with the same concerns as you would use with 2-pole
because "TMC" applies 2-pole feedback around the output stage
then it nests that in a 2nd "single pole" loop from the output
you really need to design to the same level of sophistication, with the same concerns as you would use with 2-pole
because "TMC" applies 2-pole feedback around the output stage
then it nests that in a 2nd "single pole" loop from the output
Last edited:
I thought TMC was the name some people use for OIC. What distinction are you making here?
Please, see below for an example of OIC (from E.M. Cherry, J. Audio Eng. Vol. 30, No. 5, 1982 May). In this case the compensation capacitor is placed between the VAS-input and OPS-output.
TMC is different: There is no direct (pure) capacitive path between the VAS-input and OPS-output. TMC also behaves differently: At AF it behaves roughly the same as OIC, at HF however, it behaves more like ordinary Miller compensation. Because of the transition between these two modes, it's called (at least on this forum) Transitional Miller Compensation. BTW, at first instance, I also believed that TMC was just a 1-pole system, but JCX et al have convinced me that it is indeed a 2-pole system.
Cheers, E.
TMC is different: There is no direct (pure) capacitive path between the VAS-input and OPS-output. TMC also behaves differently: At AF it behaves roughly the same as OIC, at HF however, it behaves more like ordinary Miller compensation. Because of the transition between these two modes, it's called (at least on this forum) Transitional Miller Compensation. BTW, at first instance, I also believed that TMC was just a 1-pole system, but JCX et al have convinced me that it is indeed a 2-pole system.
Cheers, E.
Attachments
Last edited:
Since we are back on compensation...
I did point out the distinction between between "TMC" and "OIC" back in post #336. Edmund has repeated it above.
The difference is fundamental, please can we use consistent nomenclature?
Second point is that a second order system will not inevitably have a peak, this is entirely dependent on the specific case.
So, while a peak implies a second or more order system, the lack of a peak does not mean much.
Note that there are two peaks sometimes talked about for TPC.
One is a low frequency peak in the open loop plot at around 1kHz to 10kHz.
This is shown in your book on p. 219 of the 5th edition and, equivalently in a Return Ratio plot in Cordell's on p.179.
This is harmless, as you discuss, and I propose it may even be helpful.
Harry Dymond's JAES article on TPC does not predict this accurately at all, and it is not obvious to me exactly how it works.
There is also often a peak in the Closed loop plot at MHz frequencies.
It seems this is not an inevitable consequence of TPC but it is quite common.
I suspect this is the result of an interaction between typical values of TPC compensation and practical, non ideal values for transistor capacitance and the like.
There has been considerable concern expressed about "peaks" in TPC, often without clarification. This one is the more serious issue so I assume it is the one people usually mean.
Neither peak is entirely clear, do you or Edmond, JCX, Waly et al have any ideas on this?
Best wishes
David
I did point out the distinction between between "TMC" and "OIC" back in post #336. Edmund has repeated it above.
The difference is fundamental, please can we use consistent nomenclature?
Second point is that a second order system will not inevitably have a peak, this is entirely dependent on the specific case.
So, while a peak implies a second or more order system, the lack of a peak does not mean much.
Note that there are two peaks sometimes talked about for TPC.
One is a low frequency peak in the open loop plot at around 1kHz to 10kHz.
This is shown in your book on p. 219 of the 5th edition and, equivalently in a Return Ratio plot in Cordell's on p.179.
This is harmless, as you discuss, and I propose it may even be helpful.
Harry Dymond's JAES article on TPC does not predict this accurately at all, and it is not obvious to me exactly how it works.
There is also often a peak in the Closed loop plot at MHz frequencies.
It seems this is not an inevitable consequence of TPC but it is quite common.
I suspect this is the result of an interaction between typical values of TPC compensation and practical, non ideal values for transistor capacitance and the like.
There has been considerable concern expressed about "peaks" in TPC, often without clarification. This one is the more serious issue so I assume it is the one people usually mean.
Neither peak is entirely clear, do you or Edmond, JCX, Waly et al have any ideas on this?
Best wishes
David
Last edited:
David, I think that this second peak is coming from the EF3 output stage and can be compensated there. This peak can be seen even in EF3 OPS in isolation, I did not simulate other types of an OPS, but EF2 which does not show that peak.
BR Damir
Thanks Damir, I see this peak most clearly in EF3 too, it can occur even in EF2 but much less severe.
I just skimmed a lot of posts about the claimed benefits of TMC over TPC and the lack of a peak for TMC was mentioned.
I am not sure of the full history, it was before my time, and so I wondered, like you, if perhaps a little roll-off in the TMC Return Ratio happens to compensate for an OPS peak, partly by coincidence.
Of course one tunes the amplifier compensation in a simulation with an OPS so it's not just luck, but I suspect it's not fundamental.
Best wishes
David
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.