I don't know how to do that . 😱
I have just noted "cause and effect" both in simulation and
on the 2 real EF3's I own and listen to.
One is 18 years old the other is 32 years old , decades of abuse.
If someone would light the way ,I already have the EF3 in isolation (simulated)
and have run bandwidth and THD tests on it.
By running my various VFA and CFA IPS's close to instability , I have
concluded that my EF3 is as stable ,( if not more ) than a typical
EF2 .
Various inductance's and other simulated "abuses" as Bob pointed out
in the "trouble with triples" thread have failed to make the OPS unstable.
PS - by making the simulated amp actually oscillate I found I cound raise that
frequency much higher than ULGF (OPS component choices) , the full simulated amp would no longer oscillate -
even with no compensation - crude , but it worked.
The resulting simulation ended up representing the real OEM OPS configuration.
OS
Last edited:
Your not unwashed. You are a super good engineer, Bob. And, a gentleman. It's great you have the time to give your experience and knowledge to others and to learn from your comments and book. What I do not understand is this--- CFA has been written about and described elsewhere..... IEEE, patents, books/literature etc. For a detailed analysis into the deepest and most difficult questions, I try the literature and IC manfr'ers literature.
THx-Richard
Last edited:
The added pole of the inverting transistor in a LTP makes all the feedback phase difference ?
DIY vs. OEM is just as much a factor as CFA vs.VFA .
Just seen member Junie's NAD372 CFA amp , no precision construction (sloppy).
A "sloppy" CFA might just match a typical DIY VFA. Use a superior
topology to save a few $$$ ??
OS
Reliability , sonic character , and specs.
Example - VFA designed with killer PSRR so a cheap PS can
be used .
Most "advancements" are driven by $$$. 🙁
OS
I don't see any reliability difference between a VFA and a CFA. It depends on phase margin, component's power margins and quality of components and build.
Sonic ? More complicated... too much factors involved.
Specs ? We know the differences.
Of course "Most "advancements" are driven by $$$". But those $$$ can be invested in various ways: time spend on design and study (included the one dedicated to find the best components for sonic) , price of components (like litycs qualities). But good engineering is more accurate topology's choices (SMPS or linear, regulated or not etc...) and talent in design than pure force of the $$$. Don't you think ?
Bimo are you saying that total configuration sounds bad or each of those "features" sounds bad in your experience?
Obviously 99% of CFA designs are symmetrical and a lot use Cascode VAS etc.
No reliability difference in topologies ,but in sourcing (no "badcaps" in DIYA).
$$$ are invested in advertising /hype -buzzwords "hyper current drive"...
The exceptions are High-end and DIYA... where what you say is true (er).
Here we do pay for it- $$$ . A 3000$ marantz CFA with toriod/40kuf PS,
and a sanken powered EF3 (like mine). A unit of this caliber IS the optimum
investment ratio.
Fancy rounded enclosure
and a white silk glove to unpack the wooden crate (genesis stealth).
Audiophile's are perhaps 10% of the market.
The NX and what I hope to make will "shame" the NAD ... it has
rave reviews in spite of it's mediocre construction.
OS
Extra gain does not cause extra phase shift.
It is the roll off rate that cause the phase shift so there is no reason why extra gain has to be a problem if it is rolled off correctly.
Don't mean to be pedantic about this but it is important to my point that the phase is a function of the gain curve shape.
This function is exactly the same for "CFA" and VFA if they are both minimum phase, which is fairly accurate.
Near the OPS pole there may be noticeable mixed phase behaviour but it will be from the OPS and hence practically identical for both "CFA" and VFA.
So the same gain curve, to have the same distortion reduction, will have exactly the same stability and PM.
So I don't immediately see where the "CFA" benefit lies.
Of course a VFA with sufficiently low gain can also push the OPS pole below the intercept.
Hence my earlier question about whether there is any benefit of a "CFA" over a comparable low gain VFA.
Seems only possible if the gain is redistributed within the amp.
That is my current area of research so I am not dogmatic about my objections.
I notice Bob has entered our discussion too, multi-quote if that makes it easier for you.
Best wishes
David
Last edited:
I'll just throw up some sources from time to time. Bob, Dave, etal let me know if you found your answer in one of these: If not, I'll keep trying --
www.ieee.li/pdf/viewgraphs/current_feedback_vs_voltage_feedback_amplifiers.pdf
www.dspace.thapar.edu/dspace/bitstream/10266/1306/3/1306.pdf
Thx-RNMarsh
Last edited:
Mr. Marsh, the TI ieee presentation is excellent. Summarizes much of what we've discussed here clearly & accurately and has some good pointers on how to chase evil oscillation.
However the thapar.edu link seems to be dead.
"Extra gain does not cause extra phase shift."
Perhaps unfortunate phrasing on my part. The point I am making is that with the much higher open loop and hence loop gains, you have to deal with the phase shift associated with it. Interestingly the OL UG intercept on a lot of VFA's is higher than CFA's (because of the higher gain of course). But, you have to compensate for that.
"It is the roll off rate that cause the phase shift so there is no reason why extra gain has to be a problem if it is rolled off correctly."
That's not the point I'm making. You have to set a VFA ULGF at somewhere between 1 and 3 MHz no matter how you roll the gain off ( MC, MIC, TPC etc). You do not have to do that in a classic CFA.
"Don't mean to be pedantic about this but it is important to my point that the phase is a function of the gain curve shape."
No issues with this from my side
"This function is exactly the same for "CFA" and VFA if they are both minimum phase, which is fairly accurate. "
See my point above
"Near the OPS pole there may be noticeable mixed phase behaviour but it will be from the OPS and hence practically identical for both "CFA" and VFA."
But in a CFA you don't have the extra 30 to 40 dB you have in a VFA and that's the point I am making.
"So the same gain curve, to have the same distortion reduction, will have exactly the same stability and PM."
The trick is to comp both designs for the same PM at say 1 MHz. If you do this, you will see that a CFA has lower loop gains, but wider BW and higher ULGF. As remArked earlier, this can give higher loop gains at HF. Granted this is for classic MC, but it's to my point that you can close the CFA loop at higher frequencies.
"So I don't immediately see where the "CFA" benefit lies. "
I doubt there will be agreement on which is better. My interest is in understanding why CFA ULGF can be higher, and why some people have claimed 4 or 5 MHz and still be stable. It's about peeling the onion.
Perhaps unfortunate phrasing on my part. The point I am making is that with the much higher open loop and hence loop gains, you have to deal with the phase shift associated with it. Interestingly the OL UG intercept on a lot of VFA's is higher than CFA's (because of the higher gain of course). But, you have to compensate for that.
"It is the roll off rate that cause the phase shift so there is no reason why extra gain has to be a problem if it is rolled off correctly."
That's not the point I'm making. You have to set a VFA ULGF at somewhere between 1 and 3 MHz no matter how you roll the gain off ( MC, MIC, TPC etc). You do not have to do that in a classic CFA.
"Don't mean to be pedantic about this but it is important to my point that the phase is a function of the gain curve shape."
No issues with this from my side
"This function is exactly the same for "CFA" and VFA if they are both minimum phase, which is fairly accurate. "
See my point above
"Near the OPS pole there may be noticeable mixed phase behaviour but it will be from the OPS and hence practically identical for both "CFA" and VFA."
But in a CFA you don't have the extra 30 to 40 dB you have in a VFA and that's the point I am making.
"So the same gain curve, to have the same distortion reduction, will have exactly the same stability and PM."
The trick is to comp both designs for the same PM at say 1 MHz. If you do this, you will see that a CFA has lower loop gains, but wider BW and higher ULGF. As remArked earlier, this can give higher loop gains at HF. Granted this is for classic MC, but it's to my point that you can close the CFA loop at higher frequencies.
"So I don't immediately see where the "CFA" benefit lies. "
I doubt there will be agreement on which is better. My interest is in understanding why CFA ULGF can be higher, and why some people have claimed 4 or 5 MHz and still be stable. It's about peeling the onion.
Cool ... OPA861 = marantz ma-9s1 (nx ips + symmetrical CM + hawksford).
Same specs ...I got the marantz to 750V/uS vs. 900 claimed for the opa.
Link 2 doesn't work ??
oooh .. first document is almost a CFA "bible" !! I like...
OS
Last edited:
Yes. I didn't mean that "CFA" had no possible benefits.
Just that they are not clear to me in this aspect.
Perhaps a small increase in ULGF is possible but I can't see how it can be sufficient to make up for the lower gain.
To make a major increase in ULGF would seem to require improvement to the OPS.
I do think the typical OPS is far from optimised, not sure what the practical limit is.
I look forward to your updated paper and plots rather than we both repeat ourselves at this point.
Best wishes
David
Last edited:
