I suspect we have different understandings of what Miller Input Inclusive Compensation is and how it works. A quick example at hand, check Mr. Cordell error correction amplifier front end, a pure VFA that has 300V/uS slew rate.
Mr. Cordell said:
No, I know what MIC is.
Let me repeat: the TIS input current in a CFA can be an order of magnitude higher than a VFA.
Let me repeat: the TIS input current in a CFA can be an order of magnitude higher than a VFA.
Ok, and why should I care about this amazing property of a CFA?
Isn't 300V/uS slew rate in a VFA with MIC enough? Then perhaps the Stochino VFA amp at about 500V/uS slew rate (not MIC, though) is fast enough for you?
That's not the point. We again seem to be in a CFA vs VFA argument again when there is absolutely no need. There are a bunch of guys on this thread that want to discuss the technicalities and explore the topology, and is think we've seen some great progress over the last few months. And then there are a group of detractors that seem to feel threatened by CFA. What's the problem?
There are no rules that say bandwidth, SR and THD (and any other specs you care to throw into the soup) have to be at a certain level. To claim an amp with 1 ppm THD and 20 V/ us is better than one showing 10 ppm and 500 V/us is ludicrous. By whose measure? It's only an opinion.
There are no rules that say bandwidth, SR and THD (and any other specs you care to throw into the soup) have to be at a certain level. To claim an amp with 1 ppm THD and 20 V/ us is better than one showing 10 ppm and 500 V/us is ludicrous. By whose measure? It's only an opinion.
Waly just earned a vacation in the bin. And posting comments and then deleting them doesn't make it acceptable.
Member
Joined 2009
Paid Member
There have been some interesting posts. I am wondering if there is enough thought about the 'system' level aspects. For example, very high slew rates may be interesting for video op-amps that have to drive cable capacitance but I'm not sure I understand what level is relevant for audio when most of us slap an inductor on the output of the amplifier for stability reasons.
my 2c - I think it's because some people dislike it when there are posters who are not able to present a balanced approach, can not 'accept points from both sides of the table', insist that one topology is better than another and are skilled at English and presenting an argument strongly. It's typical of forum discussions and can make for a nice lively discussion. But when people feel compelled to argue that one approach is best because they have a narrow focus on such an approach for their own amplifier designs - they appear to have a conflict of interest. I've been there myself, and now I realize that this has clouded my own approach to amplifier design at various times. I am still learning, but so far don't see any inherent advantage of one topology over another if designed properly.
Last edited:
I prefer to see it as having another power tool in the workshop. 😉
I've built both topologies over the last few years and I enjoy listening to all my amps. That's what's important at thread of the day!
I've built both topologies over the last few years and I enjoy listening to all my amps. That's what's important at thread of the day!
Last edited:
Part of the thread origin was to discuss CFA and compare it to VFA, the concept of the CFA expansion i-v characteristic has been discussed only in terms of the potential increased slew rate.
Lately I thought the slew rate message was, its potential use as a metric for a yet unidentified mechanism, but now it appears to be back as a direct performance parameter.
I still fail to see the benefit of large slew rate margin if the operating conditions never call on it. What is the percentage of idle current called upon for even the worst case loads/transients (or even pure class-b output)?
On the other hand the actual slope of the iv input characteristics are very different even for small signal and yet that does not get discussed.
If nothing else its just fun to compare the dc characteristics Ivfa versus Icfa versus input (especially slopes) and Verror (vfa) versus Verror (cfa).
Hope this helps
-Antonio
Lately I thought the slew rate message was, its potential use as a metric for a yet unidentified mechanism, but now it appears to be back as a direct performance parameter.
I still fail to see the benefit of large slew rate margin if the operating conditions never call on it. What is the percentage of idle current called upon for even the worst case loads/transients (or even pure class-b output)?
On the other hand the actual slope of the iv input characteristics are very different even for small signal and yet that does not get discussed.
If nothing else its just fun to compare the dc characteristics Ivfa versus Icfa versus input (especially slopes) and Verror (vfa) versus Verror (cfa).
Hope this helps
-Antonio
Member
Joined 2009
Paid Member
I agree with you Antonio. Once you have enough slew rate there is little benefit in more unless you are driving a high capacitance at high frequencies - which not only requires high slew rate but stresses the amplifier considerably. I think there are amplifiers which have poor slew rate but they may still sound good to some people, so for me it's enough to know that both VFA and CFA can provide sufficient slew rate if designed properly for the application.
The I-V curve is much more interesting thing to look at. Of course I would say that - since I started a thread on that topic ! http://www.diyaudio.com/forums/solid-state/238124-choosing-feedback-error-amplifier.html
What I do note is that a symmetric CFA is quite different from an asymmetric CFA in terms of distortion profile whereas the LTP shows fewer differences if it is properly balanced.
Aslo the feedback factor will tell you how relevant these differences may turn out to be. With high feedback factors it may make little difference because the differential signal is tiny and the differences between CFA and VFA become much smaller in terms of I-V.
I suspect that CFA shows more of its colours when used in single ended input with moderate feedback that was popular with JLH.
The I-V curve is much more interesting thing to look at. Of course I would say that - since I started a thread on that topic ! http://www.diyaudio.com/forums/solid-state/238124-choosing-feedback-error-amplifier.html
What I do note is that a symmetric CFA is quite different from an asymmetric CFA in terms of distortion profile whereas the LTP shows fewer differences if it is properly balanced.
Aslo the feedback factor will tell you how relevant these differences may turn out to be. With high feedback factors it may make little difference because the differential signal is tiny and the differences between CFA and VFA become much smaller in terms of I-V.
I suspect that CFA shows more of its colours when used in single ended input with moderate feedback that was popular with JLH.
Last edited:
Any standard and convenient way for you, such as ССIF (19+20 kHz), IMD (11+12 kHz),MIM ( 9+10,05+20 kHz) or SMPTE/DIN which uses 60 Hz, 7 kHz and a 4:1 ratio (12 dB) .
Thank you.
Will do CCIF (because i'm Lazy).While i prefer DIN in real life, because it stresses more the PSU and is more about what we fear in musical reproduction, omho.
If you want to explore more on this, on your side, files are available for both versions.
Forget the numbers, my new prototype amp has 20db fb around the input stage/vas and 8db feedback from the OPS, it`s CFB and it sounds great.. FB factors could be changed in the final version, maybe the result will be ngfb amp, who knows.... It`s a silly to talk about that kind of things, it takes time to listen what sounds the best, simulations are interesting but only if you know what to simulate. Start to build and LISTEN!
Please, can somebody good in LTSPICE verify this ? I did-it twice, but i don't believe in what i see.
It is .four 1k, FFT of 19+20Khz at 0.6V input each. Start time 1mS, stop time 5mS.
In green CFA, in blue VFA.
ASC files are there: http://www.esperado.fr/fr/temp/compare_vssa/files.zip
It is .four 1k, FFT of 19+20Khz at 0.6V input each. Start time 1mS, stop time 5mS.
In green CFA, in blue VFA.
ASC files are there: http://www.esperado.fr/fr/temp/compare_vssa/files.zip
Attachments
Last edited:
Member
Joined 2009
Paid Member
I'm not quite sure what drives these differences. I decided to try the same thing (10kHz + 11kHz large signal swing) with my SKA GB150 file and compared the VFA with CFA version - it appears in this case that the CFA version came out worse, but I wasn't unhappy with either of them. Maybe the details of the spice simulation settings are important, the number of cycles in the FFT or it depends on whether you do 10 +11 or 11 + 10 etc. - somebody more knowledgeable than me will be needed.
Thanks, Bigun. I just closed my LTspice, reopened-it, tried with 1mS start and 6mS stop, same results...
BTW, can-you point-me to your files ? I ignored you made such a comparison too. i'm interested.
BTW, can-you point-me to your files ? I ignored you made such a comparison too. i'm interested.
Member
Joined 2009
Paid Member
Bear in mind that I'm not trying to peel back the onion on this CFA vs VFA thing, the two versions I'm playing with are both designed to be practical solutions, so not everything is held identical. I've found that the idle current is more important than whether I use CFA or VFA 😉
I did this IMD comparison again this morning...
It was an other bug in LTspice*... (less and less fear in fairy tails simulators)
Got more consistent results (see attached, juxtaposed images with a little horizontal offset to can read both)
Green is CFA, red is VFA.
Conclusion: Nothing really different between them...
* I had generators in PS rails to simulate ripples. They were both to 0V. One was at 100Hz, the other at 1MHz. Changed them both to 'none', shorting them, and it changes everything.
It was an other bug in LTspice*... (less and less fear in fairy tails simulators)
Got more consistent results (see attached, juxtaposed images with a little horizontal offset to can read both)
Green is CFA, red is VFA.
Conclusion: Nothing really different between them...
* I had generators in PS rails to simulate ripples. They were both to 0V. One was at 100Hz, the other at 1MHz. Changed them both to 'none', shorting them, and it changes everything.
Attachments
Last edited:
Member
Joined 2009
Paid Member
That matches my general results too from looking at the SKA amplifier.
I suspect that for resistive loads, providing the OLG / CLG profiles between CFA and VFA are matched there isn't going to be much to choose between them regarding distortion.
I don't think there is going to be much of a difference between CFA and VFA in any significant way that affects the performance of a well designed amplifier for audio applications.
However, when the feedback factors are low enough that the accuracy and I-V curve of the error amplifier become relevant there may be some important differences (dynamic performance, TIM etc.).
I suspect that for resistive loads, providing the OLG / CLG profiles between CFA and VFA are matched there isn't going to be much to choose between them regarding distortion.
I don't think there is going to be much of a difference between CFA and VFA in any significant way that affects the performance of a well designed amplifier for audio applications.
However, when the feedback factors are low enough that the accuracy and I-V curve of the error amplifier become relevant there may be some important differences (dynamic performance, TIM etc.).