It should not be a VFA vs CFA contest . . . Why?
The interest in CFAs isn't completely academic, is it? The idea is to ultimately get better/cheaper amplifiers? That means you need to do some comparisons, preferably apples-apples.
jan
So we are done with the concepts and ready to compare now? Until it is explained, understood and designed, how can one jump into discussions on comparisons and compensation/stability issues here without a proper circuit?
CMA have evolved with CC and that opened a whole new field in itself. If you want to compare where this has all lead and a modern circuit... use SWOPA to compare. And, take his input seriously... look up his recommended reading source for the fine print.
Thx-RNMarsh
CMA have evolved with CC and that opened a whole new field in itself. If you want to compare where this has all lead and a modern circuit... use SWOPA to compare. And, take his input seriously... look up his recommended reading source for the fine print.
Thx-RNMarsh
Else, I predict that this tread will lead nowhere but will take years to get there 🙂.
jan
Was it even somewhere to begin with.?.😕
Was it even somewhere to begin with.?.😕
Different strokes for different folks...
LTP IPS vs 'the rest of the world which we might loosely call CFA amps'.
I want to go much further .. at least to the level of #499 with its simple topology but advanced compensation and SOA performance.
I want something very specific from this thread ... how to take simple CFAs to this level. Maybe its not possible with simple. But I'm still waiting for a good reason to use Diamond inputs.
Messrs Zan & Cordell have explained an important objection to CFAs (actually symmetrical topologies which CFAs seem to require) More difficult to implement fancy comp. and may not take full advantage of them.
I've brought up one advantage .. potentially better noise .. but this is lost with Diamond inputs.
I note Bonsai runs much smaller IPS currents than my VAS examples .. perhaps relying on the much vaunted 'current on demand' behaviour of CFA IPS for slew. This also gives smaller IPS gm which might help stability.
The only pearl I get from the (tiny) comp. discussion is perhaps I should play with MIC. Thanks to Bob & Bonsai for this. The more complex examples i dun unnerstan en suspak de not reel TMC etc
I've already conceded the use of evil Triples as the common OPS though I don't think it will highlight important differences in IPS/VAS as much as simple EF2.
How about a CFA guru explaining the advantages of turning my simple CFA to Diamond i/p? Waly's description was good but still insufficient incentive for a power amp .. and he's not even a CFA fan 🙂
____________
Mr. Marsh, Waly, you are right that this isn't really a discussion about CFAs ... but I'm just glad MikeK hasn't put his semantic pedantic oar in to stir the discussion into liquid BS.
It is really a discussion about LTP IPS vs 'the rest of the world which we might loosely call CFA amps'. At the moment we are trying to sort out a 'best' for 'rest of the world'.
_____________
We'll probably never decide 'VFA vs CFA' ... but if we come up with better & simpler CFA topologies, we will have increased the store of useful human knowledge.
How the IPS/VAS deals with 'real life' wonky OPS is perhaps it's most important characteristic. My #499 circuit has SOA performance because the IPS/VAS is integrated with the wonky OPS. Separating the IPS/VAS from the OPS turns it into the Blameless which has well defined performance limits that are detailed by Self.Once you have a clear idea of the CFA/VFA relative merits, you then connect the OPS. If the twonky load then swamps any advantages of one of the topologies, your conclusion is that none has an advantage in real life. If not, you have a clear winner.
In the first case you can also try to eliminate the destructive effect of the OPS by using a buffer between OPS and the CFA/VFA front end.
I want to go much further .. at least to the level of #499 with its simple topology but advanced compensation and SOA performance.
I want something very specific from this thread ... how to take simple CFAs to this level. Maybe its not possible with simple. But I'm still waiting for a good reason to use Diamond inputs.
Messrs Zan & Cordell have explained an important objection to CFAs (actually symmetrical topologies which CFAs seem to require) More difficult to implement fancy comp. and may not take full advantage of them.
I've brought up one advantage .. potentially better noise .. but this is lost with Diamond inputs.
I note Bonsai runs much smaller IPS currents than my VAS examples .. perhaps relying on the much vaunted 'current on demand' behaviour of CFA IPS for slew. This also gives smaller IPS gm which might help stability.
The only pearl I get from the (tiny) comp. discussion is perhaps I should play with MIC. Thanks to Bob & Bonsai for this. The more complex examples i dun unnerstan en suspak de not reel TMC etc
I've already conceded the use of evil Triples as the common OPS though I don't think it will highlight important differences in IPS/VAS as much as simple EF2.
How about a CFA guru explaining the advantages of turning my simple CFA to Diamond i/p? Waly's description was good but still insufficient incentive for a power amp .. and he's not even a CFA fan 🙂
____________
Mr. Marsh, Waly, you are right that this isn't really a discussion about CFAs ... but I'm just glad MikeK hasn't put his semantic pedantic oar in to stir the discussion into liquid BS.
It is really a discussion about LTP IPS vs 'the rest of the world which we might loosely call CFA amps'. At the moment we are trying to sort out a 'best' for 'rest of the world'.
_____________
We'll probably never decide 'VFA vs CFA' ... but if we come up with better & simpler CFA topologies, we will have increased the store of useful human knowledge.
And here i thought page #1, line #1 defined what this forum was to be about. Silly me.
A better and maybe simpler CFA is good, though. Good enough.
I am curious why no one wants to ask Scott about this subject as he/AD designs both plus the more current (sic) C-C circuits and combinations. What's that all about?! Curious minds want to know.
Thx-RNMarsh
A better and maybe simpler CFA is good, though. Good enough.
I am curious why no one wants to ask Scott about this subject as he/AD designs both plus the more current (sic) C-C circuits and combinations. What's that all about?! Curious minds want to know.
Thx-RNMarsh
Well, some people have exited the thread, some cant help singing their own praises and others go to 'war' over the subject. Just nuts IMV.
Re Roy Grosser, I simmed the cross coupled front end and posted it up on the forums a few months ago. IIRC I got very good distortion and SR. I used the circuit concept straight out of Walt Kestlers app note.
However, for a discrete DIY amp, it was fairly complicated - ie a lot of devices.
Re Roy Grosser, I simmed the cross coupled front end and posted it up on the forums a few months ago. IIRC I got very good distortion and SR. I used the circuit concept straight out of Walt Kestlers app note.
However, for a discrete DIY amp, it was fairly complicated - ie a lot of devices.
Kgrlee wrote "Messrs Zan & Cordell have explained an important objection to CFAs (actually symmetrical topologies which CFAs seem to require) More difficult to implement fancy comp. and may not take full advantage of them."
I've posted up two or three symmetrical CFA circuits that get 10 ppm or below 20 kHz THD that use TPC or Alex comp, and there are practical designs running with Alex comp. There is no reason why you cannot get high performance. If you are looking for sub 1 ppm - good luck in any practical implementation unless you are prepared to spend a lot of time with some good test gear. And, what's it worth when everything else is adding orders of magnitude higher distortion?
I've posted up two or three symmetrical CFA circuits that get 10 ppm or below 20 kHz THD that use TPC or Alex comp, and there are practical designs running with Alex comp. There is no reason why you cannot get high performance. If you are looking for sub 1 ppm - good luck in any practical implementation unless you are prepared to spend a lot of time with some good test gear. And, what's it worth when everything else is adding orders of magnitude higher distortion?
Hi Guys
I'm not certain that the goal of the thread was clearly defined.
There are many "proven" CFA designs around, some simple, some complex. Personally, I think placing an arbitrary cap to the number of devices is pointless until "the best design to evolve" has been arrived at - then simplify it to the minimum parts count, or see what happens when you try.
Designing by committee is about the most difficult thing our kind of monkey can attempt. Each of us wants to show our relevance and ultimately that may not happen. People are impatient to get to the end result, which is normal given the speed of the web and discussions that can take place here. The problem is that there is no real feeling of "teaching", rather a lot of "preaching" - like this... haha...
Andrew has good designs posted on his site, and is alluding to others he is working on. Others, like LC, have their own pet design and find it difficult to see the merit in alternatives, lest it degrades their own status or the relevance of their creation.
Then there is the aspect of sims. So many here are fluent with LT Spice or other simulation software, and with the actual engineering principles behind it all. The "playing" one can do in the simulator answers Qs without a dialog being required, and without the public face of ignorance - not knowing; asking a question.Then the sim wars start. (At least they aren't real).
I have the greatest respect for Mr. Marsh, but it seems like he is trying to direct the discussion and hold back getting to a conclusion. Just an impression, I'm certain that is not his intention. But all the 99%ers - me included - want to see an "end" circuit and a discussion of why it is the way it is; what compromises were made? Why was one kind of CCS used instead of another? Could the circuit be made asymmetric and still perform well? Andrew provides that info in his own site offerings, and it is really great that he does. I'm not sure the same can happen in a free for all.
Regarding the complimentary designs, too many devices may be one issue for some, but it is often the case that most circuits found are do-able in chip form where matching is possible. That matching is main impediment to super CFA performance - capacitances, gains, linearities - and the sensitivity to such matching probably escalates as performance gets harder to measure.
Have fun
Kevin O'Connor
I'm not certain that the goal of the thread was clearly defined.
There are many "proven" CFA designs around, some simple, some complex. Personally, I think placing an arbitrary cap to the number of devices is pointless until "the best design to evolve" has been arrived at - then simplify it to the minimum parts count, or see what happens when you try.
Designing by committee is about the most difficult thing our kind of monkey can attempt. Each of us wants to show our relevance and ultimately that may not happen. People are impatient to get to the end result, which is normal given the speed of the web and discussions that can take place here. The problem is that there is no real feeling of "teaching", rather a lot of "preaching" - like this... haha...
Andrew has good designs posted on his site, and is alluding to others he is working on. Others, like LC, have their own pet design and find it difficult to see the merit in alternatives, lest it degrades their own status or the relevance of their creation.
Then there is the aspect of sims. So many here are fluent with LT Spice or other simulation software, and with the actual engineering principles behind it all. The "playing" one can do in the simulator answers Qs without a dialog being required, and without the public face of ignorance - not knowing; asking a question.Then the sim wars start. (At least they aren't real).
I have the greatest respect for Mr. Marsh, but it seems like he is trying to direct the discussion and hold back getting to a conclusion. Just an impression, I'm certain that is not his intention. But all the 99%ers - me included - want to see an "end" circuit and a discussion of why it is the way it is; what compromises were made? Why was one kind of CCS used instead of another? Could the circuit be made asymmetric and still perform well? Andrew provides that info in his own site offerings, and it is really great that he does. I'm not sure the same can happen in a free for all.
Regarding the complimentary designs, too many devices may be one issue for some, but it is often the case that most circuits found are do-able in chip form where matching is possible. That matching is main impediment to super CFA performance - capacitances, gains, linearities - and the sensitivity to such matching probably escalates as performance gets harder to measure.
Have fun
Kevin O'Connor
Last edited:
I've posted up two or three symmetrical CFA circuits that get 10 ppm or below 20 kHz THD that use TPC or Alex comp, and there are practical designs running with Alex comp. There is no reason why you cannot get high performance.
I certainly agree. There are many many many examples if one wants to look here or else where. Build them. test them. measure them. Listen to them. Compare them. SIM them.
Of course, I was trying to hold back until all the info I wanted to put out that would lead by the nose to why newer SOTA designs work so well.... maybe better. I demonstrated, with a simple example, a strategy of cancellation which all but solves many matching issues. And there are many such ways. THEN, let go and see what happens. However, most of even basic simple CMA designs do very well as Bonsai points out (again).
Kick the tires. Look it over. Feel your way thru. What ever your own internal process is. Most of all have fun with the creative process. I look forward to what will be developed. Perhaps an appreciation for the pro's and con's or some new tricks. But on your own time. I've got some tires of my own to go kick. I'll check in later.
Thx-RNMarsh
Last edited:
Andrew, I'm in fact at this moment playing with your qx-Amp_6 and its clever TPMIC. [*] I'm indebted to your generosity which has been the most instructive & useful so far.I've posted up two or three symmetrical CFA circuits that get 10 ppm or below 20 kHz THD that use TPC or Alex comp, and there are practical designs running with Alex comp. There is no reason why you cannot get high performance.
I have no doubt that sub 10VAS ppm @20kHz THD is possible cos a version of my #500 design with enhanced VAS already achieves this with plain Miller. I'm hoping better compensation schemes will improve on this without too much added complexity but haven't been successful so far.
Kevin, my cap on active devices is self imposed and no one is obliged to follow my quirks.
But what I'm looking for is some explanation of the need for more complexity.
eg no one, except for evil VFA fan Waly, has said anything useful about the advantages of Diamond IPS. IMHO, this is far more fundamental than arguing over cascoding.
I'm not promoting any topology. What I've stated is what I want out of this discussion. I expect my favoured topology to change as pearls of wisdom proceed out of the mouths of the CFA (& VFA) gurus.
So more CFA examples but with explanations of why they are better than simpler or existing topologies.
* [deleted: more rants about why its JLH & not MIC or Alexander 🙂 ]
... my own amps are intended for a dedicated multi-way active system, and in particular I intend to optimise each amp for its specific driver...
I think this is the best solution for getting the best sound: drive each speaker independently from V to I amp, with corrected characteristic to damp speakers' resonances.
Why do you say this? Have you tried it?I think this is the best solution for getting the best sound: drive each speaker independently from V to I amp, with corrected characteristic to damp speakers' resonances.
For the few gentlemen involved in this topic, and those who read it silently, the result of a long process of comparison.
Thanks Esperado!
Let's do a quick summary:
VFA topology or Voltage-Feedback Amplifier:
1.STRENGTHS:
- Low Distortion - Low Frequency;
- Low Input Voltage and Current;
- Input Symmetry;
- Noise
2.EXAMPLE USE:
Communications Systems
CFA topology or Current-Feedback Amplifier:
1.STRENGTHS:
- Low Distortion - High Frequency;
- Slew Rate;
- Bandwidth.
2. EXAMPLE USE:
ADC Driver
VFA topology or Voltage-Feedback Amplifier:
1.STRENGTHS:
- Low Distortion - Low Frequency;
- Low Input Voltage and Current;
- Input Symmetry;
- Noise
2.EXAMPLE USE:
Communications Systems
CFA topology or Current-Feedback Amplifier:
1.STRENGTHS:
- Low Distortion - High Frequency;
- Slew Rate;
- Bandwidth.
2. EXAMPLE USE:
ADC Driver
evil VFA fan Waly
Surprise, surprise, I am no VFA fan, and also no CFA fan. All I'm saying is that, as far as audio power amplifiers are concerned, there are no CFA's, but only VFAs with different input stages (LTPs vs. diamond buffer), each with it's advantages and disadvantages. There is no good reason to assert that any of these input stages are "more linear" or may provide more loop gain for NFB linearizing.
One may think that the even harmonics cancellation effects in LTPs is a advantage. That would be correct, only that those distortions are two or three orders of magnitude under what a real world output stage is able to. Also, the slew rates that the "current on demand" property of the diamond buffers are capable make no difference when it comes to audio reproduction. No audio source known to men, not even those designed for bats, are slewing more than a few V/uS, so 1000 V/uS for an amplifier is again a number only. Larger noise and offset in diamond buffers are also of no concern when it comes to audio, while the lower PSRR can be tamed by using current sources to bias the input stage.
Hi Guys
I think Waly's VFA - CFA issue comes down to the "current-on-demand" advantage inherent in CFA designs. Such a quality is easily made available to VFA designs, freeing them from at least one slew rate limitation.
As Bob Cordell demonstrated quite well with his mosfet amp, compensation and slew rate can be disentangled and made independent of each other. 300V/usec is pretty darned fast for audio. Sure, it's not 1kV or 3kV as is readily attained with CFA, but me-thinks it's possible for both approaches.
Similarly, attaining ppm THD is do-able with either approach as the many designs for both demonstrate. Any circuit with very high slew rate will have issues of layout regarding short current paths and stray capacitance.
Have fun
Kevin O'Connor
I think Waly's VFA - CFA issue comes down to the "current-on-demand" advantage inherent in CFA designs. Such a quality is easily made available to VFA designs, freeing them from at least one slew rate limitation.
As Bob Cordell demonstrated quite well with his mosfet amp, compensation and slew rate can be disentangled and made independent of each other. 300V/usec is pretty darned fast for audio. Sure, it's not 1kV or 3kV as is readily attained with CFA, but me-thinks it's possible for both approaches.
Similarly, attaining ppm THD is do-able with either approach as the many designs for both demonstrate. Any circuit with very high slew rate will have issues of layout regarding short current paths and stray capacitance.
Have fun
Kevin O'Connor
The required rate of increase depends on the voltage amplifier. With a small voltage amplifier for mobile phones do not need to have a large slew rate. It is enough to say about the bandwidth at full power which should not be narrower than the 70 ... 130 kHz. Then there is no need to calculate the slew rate.
best regards
Petr
best regards
Petr
For example at what the amplifier slew rate higher? at the operational amplifier with power supply +-10 V equal to the 10 V/us or in audio amplifier with power supply + -100 V and is equal to 50 V/us. Certainly in the operational amplifier, although the magnitude and lower than the audio frequency amplifier.
best regards
Petr
best regards
Petr
- Home
- Amplifiers
- Solid State
- CFA Topology Audio Amplifiers