Come on, having the feeling to be treated like a child is no excuse to behave like this.
Approx 3 years ago this question was raised by me..... It started another forum....See Slew Master CFA vs VFA .... forum Lines 9964, 9965, 9968.
If you look at the topology of the top models of the top brands... they are CFA. Look at Marantz for example and others. Their top of the line is CFA while lower ones are VFA. Hmmm.
Must just all be another one of them coincidences.
THx-RNMarsh
If you look at the topology of the top models of the top brands... they are CFA. Look at Marantz for example and others. Their top of the line is CFA while lower ones are VFA. Hmmm.
Must just all be another one of them coincidences.
THx-RNMarsh
Last edited:
Not the authority, just the price. And would it have a one or two sentence definition, or long dissertations subject to the probably endless interpretations of it that our herd of cats would offer?
Behold, the modern day Diogenes, raising a lamp in search of an honest (and authoritative and succinct) definition of "feedback."
Anyone got it?
I see that you keep on mentioning this book. I'm uploading the pages that you are quoting from so that others can see what you are talking about and can see if it's relevant or not. Not only having your words for it and see you brag about your own superiority.
I also include a page about current feedback in current amplifiers from the same book.
In my opinion you are confusing the whole discussion with your nonsense.
As I have said before I think that the document I uploaded in post #1033 is a lot more relevant when it comes to everything related to audio amplifiers and different techniques used when designing electronic circuits.
You keep saying to people that dont agree with you things like: you dont know what you are talking about.
To be honest I think it's you that don't know what you are talking about.
I'll put on the popcorn waiting for your respons and you and your cheerleaders telling me that I'm and idiot that don't understand what I'm thalking about and that every sane person must understand that CFA/CMA Current Mode or whatever you like to call it is superior to everything else.
BTW: I think Marantz are using a Hypex nCore amp in the top model.
Cheers
Many thanks to forr for this. At first glance, it's not going to offer us a succinct definition of feedback, but I'll keep looking through it.
References such as the following would probably be more fruitful:
IEEE Standards Dictionary: Glossary of Terms & Definitions (CDROM) - Standards
(Definitions)
I understand this subject you guys are debating. what annoys me is that several people here who really do know what they are talking about and the many many books on the subject as well, get ignored or debated with arguments that are false. What is the point of doing that? To understand the CMA principles and how it does what it does? if one wants to understand the principles of CMA... there are many good books on it. And several here have been very clear in explaining it. Maybe even the one you suggested. There have been other forum discussions at DIYAudio about it as well.
Now, I am not saying one is absolutely always better audibly than the other... i am saying many people think so and many manufacturers do as well. So, my question is Why is that?
THx-RNMarsh
Last edited:
Reading the debates at EDN on the topic, I see that my approach is very similar to some others which mention the not null impedance of the inverting input and the degenarative resistive action on its emitters.
I discovered that Aylward, Feucht and Cherry (I know Cherry since his NDFL amp but did not yet find a reference to him on the present topic) are not kind toward the CFA concept. These designers certainly do not merit the hard and unuseful words about the CFA concept oponents seen at the begining of some articles.
The criticized point in the mostly admitted analysis of CFA is the imperfect buffering action of the input stage. In the circuit proposed by Franco, there is some voltage across Rg which would disappear with an ideal buffer.
If I remember correctly, Jung said something like "CFA designers try to get the lowest ouput impedance" (I lost the reference).
The consequence of the imperfection is that with a realistic, non ideal, input stage, we can't keep the input stage transconductance out of the game.
This is the crucial point of the whole theory.
However in the absence of perfection, we can admit that the voltage across Rg in Franco's circuit is insignificant. But, then, we have to agree on a value for the treshold of insignificance.
I think Feucht is most pertinent when he writes :
https://www.edn.com/electronics-blo...nsistor-?utm_source=Aspencore&utm_medium=EDN:
[CFA, VFA]Those names have more to do with how the circuit is analyzed rather than what it is.
Last edited:
Does everyone understand how CFB works now? can we move on?
"As far as the dynamics go, you can view the entire CFA as a mere R-C network (R = R2, C =
Cc), where the resistor current (vI-vO)/R2 is conveyed to Cc not 'directly' via R2 itself, but 'indirectly' via the current mirrors. It is precisely because of the current mirrors that
a CFA is 'inherently fast'. Except for the C node, each node in the circuit exhibits a low
resistance (~1/gm), so the pole that each of these nodes forms with its own stray capacitance occurs at a much higher frequency than the pole associated with the C node; this, so long as R2 is much greater than 1/gm. (This situation is similar to the CMOS op amp of the folded-cascode type, whose open loop response is dominated by the pole of just one node.) There is no slew-rate limiting in an R-C network; likewise, thanks to its unique internal architecture, the CFA is virtually slew-rate free." --Franco 2015.
THx-RNMarsh
yep.
So many people want to simply replace currents with volts and think it's performance will be the same. They miss the unique characteristics of operating in current-mode.
As you all must know, as this CFA behavior inherently leads to very high speed circuits, the CFA has quickly advanced into RF apps. And with supply voltages dropping to 3.3v and 1.8 v , those circuits are all now current-mode operated.
Clue #1 ---> If you use high impedance circuitry as in VFA to understand CFA, you wont get it. CFA are relatively low Z circuitry. Notice that in the original CFA I published back in the 1970's there was no Z used higher than 1K. and currents were high compared to "opamp" VFA. Low Z values and higher operating currents automatically push the SR and BW higher as device and stray C's are charged quicker.
Also, the gain/BW doesnt change much with fb under Cur-Mode operating conditions. The main difference often pointed out.
THx-RNMarsh
"As far as the dynamics go, you can view the entire CFA as a mere R-C network (R = R2, C =
Cc), where the resistor current (vI-vO)/R2 is conveyed to Cc not 'directly' via R2 itself, but 'indirectly' via the current mirrors. It is precisely because of the current mirrors that
a CFA is 'inherently fast'. Except for the C node, each node in the circuit exhibits a low
resistance (~1/gm), so the pole that each of these nodes forms with its own stray capacitance occurs at a much higher frequency than the pole associated with the C node; this, so long as R2 is much greater than 1/gm. (This situation is similar to the CMOS op amp of the folded-cascode type, whose open loop response is dominated by the pole of just one node.) There is no slew-rate limiting in an R-C network; likewise, thanks to its unique internal architecture, the CFA is virtually slew-rate free." --Franco 2015.
THx-RNMarsh
Last edited:
Absolutely correct. Here is the transfer function, courtesy of ADI. You can see how small the effect of Ro is, since for audio frequencies, the denominator of the expression is a very small amount larger than 1, akin to the error due to the finite open loop gain of a VFA.
To which I say, so what? What's your point? The signal current flowing in the input stage collector is still almost exactly equal to and almost exactly is, the current flowing from outside of the input stage through its emitter. Therefore, we have current feedback.
As if that were not enough, as I showed in post 1028 to which you still have not responded, the transistor is hardly acting as an ideal transconductor. In that post's simulation, the resistive element R_early between the collector and emitter due to the Early effect carries half the current that the voltage-controlled current source does. The high loop gain drives vbe to such a small value that the effect of the transconductance is significantly reduced. But that same loop gain can do nothing to reduce vce, and so that resistive element becomes significant.
You can say that the output stage causes a current to flow through the feedback network which interacts with the transconductance and R_early to create a voltage v_e at the transistor emitter which interacts with the impedance of a current mirror input to cause a current to flow through R_early and add to another current which is the result of the interaction between v_e and a voltage at the non-inverting input driving a transconductance.
Or you can say that you have current feedback.
Attachments
Last edited:
No you don't. You just keep referencing and quoting sections of text from notable authorities - as if an erudite technical exposition on the operation of CF amplifiers, which dispels some of the technically-illiterate nonsense posted here, somehow, in some magical way, supports your subjective audiophile fantasies.
I gave enough hints months ago about how a low-noise, dual-utility mic/line pre-amp with a switched gain range of 40 - 60 dB can be assembled that essentially maintains a constant closed loop bandwidth over that full range (yes, really). That's one real world application in audio where a topological property of so-called "current feedback" can in reality actually be put to real, tangible use.
But no, zero interest there. Innovation and creativity around here is just about alluding to an understanding that in reality is non-existent and being a high-priest of audio sound.
Last edited:
:This is good fantasy. Very creative.... High-Priest of Audio.
In whose mind is that? Not mine.
Sorry to being the barer of news you dont like. Was Just sayin' its what people hear. And, its been rather consistent for a long time. This kinda reminds me of high DA vs low DA capacitors. Most people consistently choose low DA caps. But with just simple thd/Im... doesnt look like any significant difference. But listeners are lining up on one side or the other via what they hear.
My amp design criteria is similar to others and CMA meet them all so easily and even compensation is easier to apply to CMA.
THx-RNMarsh
Last edited:
This was post 946 on the difference the buffer on the CFA inverting input makes -- turning it into a VFA.
It follows from my post 887 which said the buffer on the inverting input only changes the amount of current that flows via this input so no fundamental change in workings occurs when the buffer is added.
In other words the CFA uses the same internal processing as the VFA. This leads me to believe that the CFA is just a variant of the VFA with the inverting input stage dropped off.
If you are still unsure try this thought experiment. Take the the LT1364 circuit
.and imagine that you can control the beta's (current gains) of the inverting buffer transistors to be anything in the range of say 100 (typical) down to zero (only base-emitter current or just a diode). If we reduce the betas to 10 we get more input current flowing. Is it now a VFA or a CFA? Repeat with betas of 1. Is it a VFA now or a CFA? Repeat with beta's of zero (effectively direct feed in). It is effectively now a CFA. Right? You can now remove the (non) buffer and you have a CFA..
So is there some transistor beta value where we can say the crossover occurred? If so, what justification is there on technical grounds like some 50% change in a figure-of-merit parameter.
The Ramus presentation mentions (see quote Post 981) that there is no significant difference between modern VFA (like the LT1364) and CFA speed and bandwidth for similar current and IC process. That leads me to say the CFA is just a variant of the VFA with more current flow into the inverting input than the buffered version.
Last edited:
Ian, I beg to differ. One large difference is that in a CFA the input stage output current comes from the -inp emitter, i.e. from the feedback network.
In a CFA the input stage output current comes from the tail current.
So, in addition to the difference in impedance level, there is a conceptual difference as well. A VFA is not 'just' a CFA with lower Zin at -inp!
Jan
Yes, like deep down below layers of denial I am really morbidly terrified that I will be roasted on a spit by Satan for all of eternity. No, seriously, I just roll my eyes - this isn't even news.
Exactly Jan, almost limitless current on demand provided for VAS by the output stage, not limited by VFA's CCS constant current. Current on demand needed to neutralize VAS parasitic capacitances. We can only fight capacitance with current, excess of it. What we get is high slew rate, translating into enormous A/us on the output, needed to force drivers to reproduce natural transients. In short.
Like closed-loop BW at a gain > 1? BTW Aol at some level for, let's call it the CFA configuration, is relatively meaningless. Fixing the gm at 500 Ohms or so also fixes the RTI noise at about 2X what it is at high CLG gain.
Optimizing each configuration (VFA/CFA) for an actual application rather than an intellectual exercise requires access to internal pins. At these speeds this is not practical or optimum.
Attachments
In open loop the current in the feedback network's Rg comes from the emitter, In closed loop, this current decreased a lot because of the bootstrapping effect of the current across Rf which coming from the output stage. The emitter current is controlled by a voltage which is the determing factor.