TMC supposed to provide NFB for the OPS at low frequency. But the gain of OPS is slightly less than unity. That means loop gain is 0dB or below. How does that reduce distortion?
Also how do I calculate the two capacitors and the resistor value? I know the two capacitors in series should give the normal miller dominant pole compensation, but how do you proportion the two capacitor? How do you arrive with the resistor value?
Thanks
Also how do I calculate the two capacitors and the resistor value? I know the two capacitors in series should give the normal miller dominant pole compensation, but how do you proportion the two capacitor? How do you arrive with the resistor value?
Thanks
TMC is a variation on 2-pole compensation
2-pole compensations work by freeing up some of the lower frequency VAS gain that is otherwise tied up in the single pole local feedback of conventional Miller compensation
so a requirement for their effective use is for the VAS to have lots of “excess” gain at lower frequency
this usually means “beta enhanced”/buffered VAS, possibly Cascode, equal attention to CCS
and even the highest unloaded open loop gain VAS can be loaded too heavily, shunting away the gain – so 2-EF BJT or direct MOSFET drive are likely too heavy VAS load to give enough excess gain to make 2-pole compensations effective
simplest is to just double a working Miller comp C's value and series 2 - good enough for quick eval, only a few dB more gain can be had with optimization
the R value has to be selected with an eye on stability criteria – you have an unavoidable trade off between how much gain is shifted to outer loops vs phase margin at the overall loop gain intercept frequency
and you have to properly measure - “inside” the TMC added loop, cutting all feedback around the output stage with the stability test probe
optimizing, ratioing the C, moving the 2-pole breakpoint frequency, overall amp gain intercept is a bit complicated and very dependent on output stage and load interactions too
a seductive problem with TMC is that it gives a "single pole" apparent response if measured by cutting only the global feedback loop, and it is tempting to tune by closed loop step response overshoot just like conventional Miller
but this doesn't reveal the true stability of TMC compensated amps - it is the same as conventional 2-pole compensation and has to be designed with the same care and understanding of the trade off from Bode's Integral relations - no free lunch applies
2-pole compensations work by freeing up some of the lower frequency VAS gain that is otherwise tied up in the single pole local feedback of conventional Miller compensation
so a requirement for their effective use is for the VAS to have lots of “excess” gain at lower frequency
this usually means “beta enhanced”/buffered VAS, possibly Cascode, equal attention to CCS
and even the highest unloaded open loop gain VAS can be loaded too heavily, shunting away the gain – so 2-EF BJT or direct MOSFET drive are likely too heavy VAS load to give enough excess gain to make 2-pole compensations effective
simplest is to just double a working Miller comp C's value and series 2 - good enough for quick eval, only a few dB more gain can be had with optimization
the R value has to be selected with an eye on stability criteria – you have an unavoidable trade off between how much gain is shifted to outer loops vs phase margin at the overall loop gain intercept frequency
and you have to properly measure - “inside” the TMC added loop, cutting all feedback around the output stage with the stability test probe
optimizing, ratioing the C, moving the 2-pole breakpoint frequency, overall amp gain intercept is a bit complicated and very dependent on output stage and load interactions too
a seductive problem with TMC is that it gives a "single pole" apparent response if measured by cutting only the global feedback loop, and it is tempting to tune by closed loop step response overshoot just like conventional Miller
but this doesn't reveal the true stability of TMC compensated amps - it is the same as conventional 2-pole compensation and has to be designed with the same care and understanding of the trade off from Bode's Integral relations - no free lunch applies
Last edited:
Hi,
I know what TPC is, and TMC is, and
they are not the same thing at all.
TMC or Transitional Miller compensation
is still single pole compensation, not TPC.
Trying to combine TMC and TPC seems a very
interesting idea, but handwaving doesn't help.
rgds, sreten.
TPC = two pole compensation
TMC = including the output in SPC
I know what TPC is, and TMC is, and
they are not the same thing at all.
TMC or Transitional Miller compensation
is still single pole compensation, not TPC.
Trying to combine TMC and TPC seems a very
interesting idea, but handwaving doesn't help.
rgds, sreten.
TPC = two pole compensation
TMC = including the output in SPC
Last edited:
the similarity in feedback factor around the output stage, the stability near equivalence is what I am emphasizing when I say they are "the same"
http://www.diyaudio.com/forums/soli...-6th-edition-douglas-self-36.html#post4017165 comes late in the discussion
here I point to another couple of my illustrative posts on TMC, its 2nd order compensation properties:
http://www.diyaudio.com/forums/soli...-6th-edition-douglas-self-36.html#post4017165 comes late in the discussion
here I point to another couple of my illustrative posts on TMC, its 2nd order compensation properties:
TMC is multi-path feedback, the extra feedback in low frequency is injected by the R resistor, not by the differential input stage! Thus, only feedback path will have 2nd order behavior, so that the positive input does not experience any overshoot. Positive input response is just like 1st order system! If every setting is the same, TMC is better than TPC.
again the discussions over many pages point out you can match TMC "single pole" closed loop forward gain appearance with TPC by adding a RC
but TMC is as "dangerous", reduces stability margin by the same amount as the equivalent TPC
this is a unavoidable consequence of Negative Feedback's "conservation laws" - Bode Integrals
you can't have "better" performance from any rearrangement/increase of feedback excess gain in negative feedback without paying that price
but TMC is as "dangerous", reduces stability margin by the same amount as the equivalent TPC
this is a unavoidable consequence of Negative Feedback's "conservation laws" - Bode Integrals
you can't have "better" performance from any rearrangement/increase of feedback excess gain in negative feedback without paying that price
Last edited:
Is there a simple spice file where one can simulate the two pole compensation or higher poles to see the realtime response before one implements. Rather than hit and trail method why can t we consider the spice way?
by following back the posts I quoted, linked? skimming back and forward the posts in those threads?
picking up keywords and user names for advanced forum search?
I do like to use Spice sim as more accessible to many for illustrating circuit principles
picking up keywords and user names for advanced forum search?
I do like to use Spice sim as more accessible to many for illustrating circuit principles
Last edited:
Sounds like there is no formula to calculate the value, just by try and true starting with the series combination of the two cap equal to the conventional miller dominant pole cap.
How about my first question, according to Cordell, the resistor and the cap at the input of the OPS serve as feedback loop for the OPS at low frequency. But the OPS has gain of less than unity. How does that help when you put NFB from output of the OPS back to it's input? In order to reduce distortion, you have to have loop gain more than unity.
How about my first question, according to Cordell, the resistor and the cap at the input of the OPS serve as feedback loop for the OPS at low frequency. But the OPS has gain of less than unity. How does that help when you put NFB from output of the OPS back to it's input? In order to reduce distortion, you have to have loop gain more than unity.
its the excess gain of the VAS that would otherwise be tied up in the single slope local Miller compensation that is redistributed by TPC, TMC
you do lose some with TMC if the output stage gain droops much below unity with load, we are usually interested in 20+ dB increased feedback at most audio frequencies
a loss of more than a dB or two in the output does limit TMC's effectiveness - TPC can be considered instead
and I did link to a start on equations - but by the time you model important parasitics and the interface load effects you have a pretty hopeless mess of equations
Spice models can be more realistic than human readable equation's abstractions
I just use general principles, approximations suggested by simplistic equations and play in sim
of course the design/analyze/measure loop should be closed by loop gain measurements of the built hardware
you do lose some with TMC if the output stage gain droops much below unity with load, we are usually interested in 20+ dB increased feedback at most audio frequencies
a loss of more than a dB or two in the output does limit TMC's effectiveness - TPC can be considered instead
and I did link to a start on equations - but by the time you model important parasitics and the interface load effects you have a pretty hopeless mess of equations
Spice models can be more realistic than human readable equation's abstractions
I just use general principles, approximations suggested by simplistic equations and play in sim
of course the design/analyze/measure loop should be closed by loop gain measurements of the built hardware
Last edited:
Meantime I got a better probe; built on Al=10,000nH/turns^2, 10:1 ratio, works great from 500Hz to 10MHz (input impedance limits the low frequency and the resonances limit the high frequency). For very low (<10Hz) the primary has to have 2.5H minimum, won't work properly over 500KHz.
Attachments
Actually, TMC is not as dangerous as TPC. For TMC, the extra feedback does not go through differential input stage, it is directly feedback to VAS. The deterioration of phase margin of TMC and TPC are different. TMC will be much better. That's why you could use high "transition" frequency in TMC. Nowadays, mordern power transistor becomes really fast in response. Differential input stage becomes the slowest stage in the system. If the extra feedback does not involve the input stage, you push the Unity Gain frequency even higher. 3MHz crossovers 0db, is not a problem with TMC.
Both TMC and TPC are using RC network to switch feedback path. Did anybody try to use LC instead of RC network? As my first calculation, the required inductor values is about within tens of uH. That is completely do-able!
Last edited:
stamp out inductors
Why for heaven's sake should we use an inductor, as we can do it with just one extra R and C?
Cheers, E.
Why for heaven's sake should we use an inductor, as we can do it with just one extra R and C?
Cheers, E.
- Status
- Not open for further replies.