U=R*I.... so they are connected, You can't have one without the other. Almost all audio circuits have voltage sensing, (like most amplifier that drive speakers are voltage driving devices, with low output impedance), but it may well have current feedback, that is solely dependent on the impedance of the node where you inject it. whether that node works primarily on current modulation or on voltage modulation.
Interesting enough, you are exactly right. The LF closed loop gain expression is exactly the same for a CFA as for a VFA and, assuming the open loop gain is large enough, depends only on the R6/R36. The big difference appears when you consider how the bandwidth varies with the closed loop gain. Here, CFAs and VFAs behave fundamentally different. The constant closed loop GBW in CFAs is a direct consequence of the open loop crossover frequency modulation by the feedback network loading at the input node, as discussed above. Using the same type of analysis you can easily determine why, while pretty standard in VFAs, a phase lag cap correction in CFAs is a big stability nono.
You don't even know what a virtual ground is and its role in the analysis of electronic circuits.
No, you did NOT quote Dr. Cherry correctly. Clearly, from what you've said, you didn't even understand the little that you read. See below:
http://www.diyaudio.com/forums/soli...k-how-do-i-see-difference-74.html#post3435888
" The constant closed loop GBW in CFAs is a direct consequence of the open loop crossover frequency modulation by the feedback network loading at the input node, as discussed above. Using the same type of analysis you can easily determine why, while pretty standard in VFAs, a phase lag cap correction in CFAs is a big stability nono."
I prefer to replace input node loading you describe above with modulation of the TIS current. Indeed, if you change the total value of the feedback network in a CFA, but keep the ratio the same, the bandwidth changes and as I noted way back in this thread, this is not the case with VFA. It was my contention, based on this, that the two were therefore fundamentally different topologies. If you raise the total feedback resistance in a CFA to the point where the -3db bandwidth is similar to that of a VFA, well then maybe you don't have a CFA anymore, but a sub optimal VFA, but then why would you do that anyway.
Now Michael, lets keep it civil, else I will have to leave again.
I prefer to replace input node loading you describe above with modulation of the TIS current. Indeed, if you change the total value of the feedback network in a CFA, but keep the ratio the same, the bandwidth changes and as I noted way back in this thread, this is not the case with VFA. It was my contention, based on this, that the two were therefore fundamentally different topologies. If you raise the total feedback resistance in a CFA to the point where the -3db bandwidth is similar to that of a VFA, well then maybe you don't have a CFA anymore, but a sub optimal VFA, but then why would you do that anyway.
Now Michael, lets keep it civil, else I will have to leave again.
I prefer to replace input node loading you describe above with modulation of the TIS current. Indeed, if you change the total value of the feedback network in a CFA, but keep the ratio the same, the bandwidth changes and as I noted way back in this thread, this is not the case with VFA. It was my contention, based on this, that the two were therefore fundamentally different topologies. If you raise the total feedback resistance in a CFA to the point where the -3db bandwidth is similar to that of a VFA, well then maybe you don't have a CFA anymore, but a sub optimal VFA, but then why would you do that anyway.
Now Michael, lets keep it civil, else I will have to leave again.
No.
Yes. Nothing like a zero in the feedback path of a WB amplifying device to get some oscillation going . . .
Whatever fancies you, the result must be the same since circuits are deterministic. In fact I would agree that a current based analysis is more natural here (based on the empirical observation that the impedance at the inverting node is low).
What I tried to prove is that this discussion is entirely semantic. Michael's stance is "there are no CFAs". Michael is right (better said, he's not wrong), given that the so-called CFAs can be analyzed and their properties rendered using entirely and only the VFA analysis methodology (which is what I did).
However, CFA topologies have certain properties that makes them special and desirable in certain applications. These properties certainly deserve a particular designation and based on the alternative analysis method (that does not make use of the hardcore control theory methodology which, by the way, when it comes to electronic circuitry, is today mostly of an academic interest), CFA was chosen. This is certainly not a marketing ploy (like "zero feedback amplifier") and confirms once more what I've said earlier: feedback theory is only a convenient analysis instrument. A circuit is a circuit, that is a circuit, and to determine the transfer functions one can make use of feedback theory or, if he's not familiar with, the Mason rule or brute force Kirchoff - the results will be the same.
I would choose as more appropriate "VFA with CF topology" but certainly that's not gonna be a catcher
.The output is a virtual ground from a Thevenin perspective; you have to set voltage sources to zero when establishing the impedance of the feedback network from the perspective of the inverting port.
The output buffer is a near ideal voltage source for Thevenin purposes and must, therefore, be made a virtual ground, which makes the feedback resistors appear in parallel as far as the inverting port of your so-called "CFA" is concerned.
Last edited:
Bonsai,
I think I understand what you are describing, but modulation of the TIS current may not be as useful a description as noting that the feedback network sets the open loop gain. How you use this gain is up to you.
As the feedback network impedance increases the open loop gain of the system reduces to the point where the closed loop gain is significantly reduced, but the actual linearity at the negative input emitters node is continually increasing (i.e. all the available transistor gain is local across the emitter followers).
As I see it, for audio if there is any value in such a circuit it would be this resulting input stage linearity, but of course this would require either someway of making up the open loop gain loss or not needing it in the first place.
edit: the phase lead capacitor problem is just another example of increasing the open loop gain as a function of the feedback impedance.
Thanks
-Antonio
Last edited:
Maybe not, but it is certainly useful to prove the potential (very) high slew rate of CFAs. That is, the "current on demand" property.
Waly,
Agreed, but "slew rate" for precision audio amps just really bugs me.
For the same bias conditions surpassing an LTP would require the input pair to enter class ab operation. I would think the amp is poorly designed if this ever happened, and I would be very interested in seeing some of what is going on with other nodes while such slewing was occurring. I do believe this can be a very useful attribute but with no benefit in audio. I would think comparing delta bias current at the fatest required output voltage change to be more indicative of a properly biased amp.
Thanks
-Antonio
Agreed, but "slew rate" for precision audio amps just really bugs me.
For the same bias conditions surpassing an LTP would require the input pair to enter class ab operation. I would think the amp is poorly designed if this ever happened, and I would be very interested in seeing some of what is going on with other nodes while such slewing was occurring. I do believe this can be a very useful attribute but with no benefit in audio. I would think comparing delta bias current at the fatest required output voltage change to be more indicative of a properly biased amp.
Thanks
-Antonio
Slew rate: Nothings free.
Don't forget CFAs, rather, understand their subtleties as one day a requirement will come along where such a cfa will be superior.
My 2 cents.
Thanks
-Antonio
I doubt there will ever be a need for needlessly high slew rates from "CFAs" in audio applications.
Michael, based on my previous comments I think its clear that slew rate was not likely to be the parameter of need.
But I thank you for leaving the door ever so slightly open.
Thanks
-Antonio
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.