If you read the Cherry articles closely you will find some major and important assumptions that are somewhat obscured when buried in the maths.
"Pure Cherry" works well in this purely idealized case but more realistic assumptions don't favour it.
The optimum starts to look more complex, not a rational transfer function, so can only be approximated by real circuitry.
TPC is a reasonable first approximation.
TMC looks a little better but I plan to study this more.
The extra complexity is not bastardization but a response to reality.
Best wishes
David
Last edited:
Yes, they are quite similar. Exact optimization once non-idealities are included is not simple to me. I would be happy to receive a PM if you have an analysis beyond what you have already posted.
Best wishes
David
???
I'm VERY disappointed to read this under the Douglas Self signature
An closed loop amp is an high gain open loop amplifier where the distortions generated by the amplifier are re-injected in the input, in phase opposition, in order to cancel themselves. Right ? Basic servo.
Any distortions added to the output signal before it is substracted will add distortion. Both in the level and time domain. Right ? (Like the little ones generated by the feedback resistances, + the high frequencies phase turns due to the feedback resistances+ the parasitic capacitances of the input stage ).
Well, if we use a VFA in a inverting configuration, we are in this best situation. But, if we use the non inverting configuration, the feedback signal will have to across the - input stage BEFORE to be subtracted. This active stage will ADD its own distortion and delay to the feedback signals.
We add errors to the errors we want to cancel, we do not cancel them.
Let push this to absurd: this - input stage cancel exactly the distortion of the amplifier, including the ones generated in the + input stage ? You had suppressed the amplifier distortion from the feedback signal: the feedback will not cancel any distortion, just reduce the gain.
Am-i clear ?
He meant that the LTP itself does not generate 2'nd harmonic distortion, because the 2'nd harmonics of the two transistors cancel each other.???
I'm VERY disappointed to read this under the Douglas Self signature
CFA vs VFA
Now, about this old sea snake CFA vs VFA.
During all my audio professional life (half as an audio designer in hifi manufacturer companies, half as a sound engineer) i had preferred CFA against VFA. Not because some religion, but because it sounded better on my point of view.
I discovered-it when i realized, years after, that all the amplifiers i had chosen (blind, and ignoring their configuration and schematics) for my various recording studios were all CFAs. And the best sounding amps i designed since 1970 were all CFAs too.
When i realized this i tried to understand why.
First, one of the reason for miss-understandings and controversies is this stupid classification and name "current feedback".
V=RI. it is like Speed of exposure and diaf in photography, talking about light quantity.
Let's forget-it.
Well. Trying to understand better the behaviors of the both confiurations, i recently settled-up the same fast amplifier (hex power fets, Cascode VAS) in the two configurations: VFA, CFA. With the same currents on each stage, or in optimizing each stage for the better possible performances in the two configurations. Spend a lot of time to test, measure, listen, analyze.
Conclusion: CFA behave, measure and sound better to my ears in all situations with this amp. in bandwidth, H and IM, slew rate and... stability margin.Whatever i tried.
What is the most obvious is to look INSIDE the feedback loop, at the input of the VAS, just after feedback is subtracted to the input signal.(See attached)
The bandwidth of the mixed signal is flat till a certain point, then the level increase up to the upper limit of the amp. it gives a good idea of the distortion increases with frequency. In CFA, the corner frequency is several octaves higher than VFA.
Listening to the amp in the two configurations, the difference is obvious: Basses drier with a better separation and details, more space between instruments, little details in CFA, that you do not hear with VFA, more fluid and natural trebles, less aggressive. A general feeling of "ease".
So, Doug, i understand you can be afraid to enter in this very controversial VFA vs CFA subject (i don't know why it is-it), but you cannot deliberately forget to talk about CFAs in your book, whatever your convictions, and, because you know your influence, you have some kind of responsibilities
As a reader, i would consider a book about amplifiers which do not talk about CFA and errors correction (an other interesting subject) as... incomplete and, so, irrelevant.
To add a word about the interest presented by a high slew rate, i would just say that, in a servo loop, faster the correction is applied, less errors, or errors rejected at a higher frequency. So obvious.
After all, what we call a current feedback configuration is just to use the same input active device in two different way: common emitor for the signal, common base (which is fast) for the feedback. And you can keep a differential input stage as well, if you want to keep high common mode rejection, or symmetrical inputs.
I will just add that, in all analog mixing desks i had modified, there is a great improvement, in the mixing bus input stages, to swap to current feed-back OPAs.
While, with VFA, the sonic quality (due to bandwich ?) reduce itself with the number of tracks plugged in the mixes bus, with CFA, the sound definition remain unchanged, using two tracks or 48. I believe it is because in CFA, bandwitch is not overall gain dependent, and small signals similar to high signals , looking at square waves.
Jut my two cents, i will not argue any more about all this subject here, just i wanted to share my experience and point of view. And, please, apologies for my poor English.
Now, about this old sea snake CFA vs VFA.
During all my audio professional life (half as an audio designer in hifi manufacturer companies, half as a sound engineer) i had preferred CFA against VFA. Not because some religion, but because it sounded better on my point of view.
I discovered-it when i realized, years after, that all the amplifiers i had chosen (blind, and ignoring their configuration and schematics) for my various recording studios were all CFAs. And the best sounding amps i designed since 1970 were all CFAs too.
When i realized this i tried to understand why.
First, one of the reason for miss-understandings and controversies is this stupid classification and name "current feedback".
V=RI. it is like Speed of exposure and diaf in photography, talking about light quantity.
Let's forget-it.
Well. Trying to understand better the behaviors of the both confiurations, i recently settled-up the same fast amplifier (hex power fets, Cascode VAS) in the two configurations: VFA, CFA. With the same currents on each stage, or in optimizing each stage for the better possible performances in the two configurations. Spend a lot of time to test, measure, listen, analyze.
Conclusion: CFA behave, measure and sound better to my ears in all situations with this amp. in bandwidth, H and IM, slew rate and... stability margin.Whatever i tried.
What is the most obvious is to look INSIDE the feedback loop, at the input of the VAS, just after feedback is subtracted to the input signal.(See attached)
The bandwidth of the mixed signal is flat till a certain point, then the level increase up to the upper limit of the amp. it gives a good idea of the distortion increases with frequency. In CFA, the corner frequency is several octaves higher than VFA.
Listening to the amp in the two configurations, the difference is obvious: Basses drier with a better separation and details, more space between instruments, little details in CFA, that you do not hear with VFA, more fluid and natural trebles, less aggressive. A general feeling of "ease".
So, Doug, i understand you can be afraid to enter in this very controversial VFA vs CFA subject (i don't know why it is-it), but you cannot deliberately forget to talk about CFAs in your book, whatever your convictions, and, because you know your influence, you have some kind of responsibilities
As a reader, i would consider a book about amplifiers which do not talk about CFA and errors correction (an other interesting subject) as... incomplete and, so, irrelevant.
To add a word about the interest presented by a high slew rate, i would just say that, in a servo loop, faster the correction is applied, less errors, or errors rejected at a higher frequency. So obvious.
After all, what we call a current feedback configuration is just to use the same input active device in two different way: common emitor for the signal, common base (which is fast) for the feedback. And you can keep a differential input stage as well, if you want to keep high common mode rejection, or symmetrical inputs.
I will just add that, in all analog mixing desks i had modified, there is a great improvement, in the mixing bus input stages, to swap to current feed-back OPAs.
While, with VFA, the sonic quality (due to bandwich ?) reduce itself with the number of tracks plugged in the mixes bus, with CFA, the sound definition remain unchanged, using two tracks or 48. I believe it is because in CFA, bandwitch is not overall gain dependent, and small signals similar to high signals , looking at square waves.
Jut my two cents, i will not argue any more about all this subject here, just i wanted to share my experience and point of view. And, please, apologies for my poor English.
Attachments
This is true for the signal if you use-it as a symmetrical input, false when you consider feedback. That i tried to lighten.
Nonsense. The two are not the same, and TPC is clearly inferior to TMC.
Again, sweeping statements with no proof.
Heaven forbid that I would sink to using TPC on one of my designs.
Inutterable nonsense. I didn't say they are the same.
And TPC is NOT inferior to "TMC" at all. I proved this a long time ago on this forum.
You obviously missed some late discussions on this topic.
Let me put it this way: show me a TMC compensated amp and I'll give you back a TPC compensated version with exactly the same performance metric (loop gain, phase margin, etc...)
It was mathematically proven that TMC is equivalent to TPC plus a lead lag phase correction. Yes, TMC could be considered as a more simple and straightforward 2nd order compensation method compared to TPC, but to claim that TMC is superior in any other way shape or form, that's totally incorrect.
I assume you refer to the work by the esteemed "Megajocke"?
This is a very ingenious analysis but not a mathematical proof.
It is based on approximations such as that the IPS has infinite bandwidth and no phase shift.
Probably more important is that it considers only transfer function and not what Bode called "sensitivity" (I think it would be better called desensitivity since it is the factor by which the sensitivity is reduced).
Usually the Return Ratio and the (de)sensitivity are equal but this is not inevitable, which is precisely why Bode created the two distinct terms.
I think they are not equal in this case and that TMC can do better on this metric.
But I would like to learn more. Perhaps after criticism of Mike's repeated assertions without evidence, now Bonsai can contribute his own analysis?
Best wishes
David
A number of people noticed great improvement when CF opamps are used for summing amplifiers in mixing consoles.
I assume you refer to the work by the esteemed "Megajocke"?
This is a very ingenious analysis but not a mathematical proof.
It is based on approximations such as that the IPS has infinite bandwidth and no phase shift.
Probably more important is that it considers only transfer function and not what Bode called "sensitivity" (I think it would be better called desensitivity since it is the factor by which the sensitivity is reduced).
Usually the Return Ratio and the (de)sensitivity are equal but this is not inevitable, which is precisely why Bode created the two distinct terms.
I think they are not equal in this case and that TMC can do better on this metric.
But I would like to learn more. Perhaps after criticism of Mike's repeated assertions without evidence, now Bonsai can contribute his own analysis?
Best wishes
David
Why? Why? Surely I can also just go on pounding away, saying that I am right and then I will have proven my point! Or, do I really have to get busy on the simulator?
A monologue with the occasional word in bold does not constitute proof, as of course we are seeing in that other discussion.
Now, repeat after me
TPC does not equal TMC
TMC is better than TPC
VFA does not equal CFA
CFA is great for audio
He who knows not and knows not that he knows not is a ......
A Canadian good friend of mine (but not so esteemed here) reached exactly the same results using a star-triangle transformation of the TMC feedback network. Now, if that's not mathematical proof, then I don't know what that is.
In fact, my offer to provide a TPC version with exactly the same performance metric is based on his work. Calculating the equivalent TPC network RC values takes about 5 minutes.
A star-delta transform seemed more robust to me too, so that's what I used myself.
But our results, or perhaps interpretations, seem to differ a little. Can you post your friend's results?
A worked example would help too, can you post one?
Best wishes
David
PS. Nice to read a post with a hint of actual substantiation.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.