Horsefeathers.
You change the gain block of the so called CFA as the open loop gain depends directly of the feedback network.[/QUOTE]
Different from a VFA, right?
Andrew, you don't have to admit CFA=VFA, and neither do I. You already admitted everything it is to admit about: "two separate topologies...". Glad you departed from the (wrong, in my opinion) idea that CFAs are anything else but a clever circuit topology, with special properties.
Please don't (once again) conflate small signal and large signal properties of any circuit topology. I was discussing mostly the small signal properties of CFAs and VFAs, and proved that they can be mathematically deduced using the same general feedback formalism that is serial input, shunt output.
COD is a large signal property and has almost nothing to do with the feedback configuration. As mentioned multiple times, COD can be rather easily implemented in a VFA topology (be it your next long tail pair VFA, or any other input stage configuration. It is true that COD comes natural to CFAs, that some COD solutions for the VFAs are not necessary easily integrable, and, given an idle power budget, the CFA COD certainly cannot be surpassed by any VFA COD trick.
CFA is Unbuffered VFA, CFA specificities are that open loop behaviour depend of feedback network and one pole less than VFA.
Not true. A CFA has two poles, just like a VFA. Ignore the IPS pole, because it is usually very high for both topologies. The other two poles are the TIS/TAS and the OPS (which is usually the one that you have to take care of with compensation).
A CFA is not an unbuffered VFA. The operating principles are fundamentally different as Chris and numerous others have tried to explain. Current into the 2nd stage is steered from the LTP current source in a VFA. In a CFA it is controlled primarily by the value of Rf.
You've totally lost me on this.
I've made my understanding of the two topologies clear and am quite aware of the differences between small signal and large signal behaviour. That you have derived the small signal properties using two port analysis is great but not new - I think its been done numerous times previously in various applications notes and academic papers.
I don't see CFA's as anything special - I agree its a clever topology that simply does some things in its native state (slew rate and bandwidth primarily) a bit better than VFA, but there it ends. I have never claimed anything else (I've designed and built both . . . )
Inverting input impedance. Great !
It is not the criteria.
Don't understand what you say about DC current flow.
Let me remind you that you sent a MDIT implementation sim that didn't work at all, and you didn't know it didn't work.
Try to make something working, and test it.
Ian, thanks for looking into this. The results are worrisome for our general use of LTSpice.
Regardless of the transistor spice models we use, the Ebers-Moll/Early equations and the Hybrid Pi model support the existence of the small signal effect I was discussing. It would surprise me if the non-ideality of these were so extreme as to make the effect itself questionable, but what appear to be the relevant issues are the values of various parameters and the extent to which the effect occurs.
Then you need to be clearer about what you're talking about. You would tell me what it isn't, but not what it is?
Then I don't understand your problems with my original comment.
Let me remind you that I corrected it by adding 1 Gohm resistors across the current sources and you have yet to reply to it.
Sorry, in the message quoted above you shifted liberally from small signal behaviour ("increment the gain...") to large signal behaviour (square wave response and COD), hence my confusion.
If you agree that CFAs are circuit topologies with series input, shunt output feedback, with some specific small and large signal properties, then indeed there is nothing more that you need to accept. I can't stop noting that your position regarding CFAs has softened considerably since this crazy debate started and I can only hope I contributed a little to this 😀.
My only "problem" is with prof. Franco Closure statement:
which is, to me, wrong. I went through the references he provided and I was unable to find anything supporting "two-port techniques to manipulate a CFA circuit into a series-shunt configuration (...) fails on other vital issues, such as the correct representation of the type of feedback actually taking place" but then again, I'm not in the position to debate with a renowned authority, so I'd rather leave it there, and agree to disagree on this.
The GBW independence in CFA is small signal behaviour. But using a swuare wave to show gain peaking and response anomalies is also done in the small signal regime.
COD is large signal - agreed - but I was simply pointing out to (Herve IIRC) that low closed loop gain 'COD like' behaviour in a VFA is nothing to do with CFA COD and everything to do with a sub-optimally compensated amplifier that will oscillate.
"If you agree that CFAs are circuit topologies with series input, shunt output feedback, with some specific small and large signal properties" - yes, I agree.
However you should emphasize you can do the same with a VFA because you are talking here about the canonical feedback forms which apply to both CFA and VFA. This is a point that kicked off this whole thread because people (forr et al) confused 'current output' with 'current feedback' - and I pointed out waaay back in the forst CFA vs VFA thread that they were not the same and the canonical feedback forms make this very clear.
The CFA trick is however that its possesses COD (large signal behaviour) which a VFA in its classic guise (LTP, MC compensated, no tricks) does not.
COD is large signal - agreed - but I was simply pointing out to (Herve IIRC) that low closed loop gain 'COD like' behaviour in a VFA is nothing to do with CFA COD and everything to do with a sub-optimally compensated amplifier that will oscillate.
"If you agree that CFAs are circuit topologies with series input, shunt output feedback, with some specific small and large signal properties" - yes, I agree.
However you should emphasize you can do the same with a VFA because you are talking here about the canonical feedback forms which apply to both CFA and VFA. This is a point that kicked off this whole thread because people (forr et al) confused 'current output' with 'current feedback' - and I pointed out waaay back in the forst CFA vs VFA thread that they were not the same and the canonical feedback forms make this very clear.
The CFA trick is however that its possesses COD (large signal behaviour) which a VFA in its classic guise (LTP, MC compensated, no tricks) does not.
Last edited:
Nope. Try to simulate in Spice the square wave response from an AC (which is linear small signal) simulation. You can only do a frequency response, but a frequency response peaking may or may not be related to a significant square wave (large signal) overshoot.
To determine the square wave response you need a Transient analysis in Spice, and that's a non linear, large signal, analysis.
There is a graph provided in the reference, the deviation is substantial above ~2V Vce.
Sure.
I have no problem with the not DC current electrons flowing.
Yes, and I have no doubt it works fine.
Attachments
Last edited:
I take it we are talking about the same thing here . . . small signal 1 to 2 volts, large signal 70 to 80% of the full rail swing. Nothing to do with AC or transient analysis. (You can always try a Bowles stability test if you want to combine the two - quite useful.)
Fact is the square wave response of an amplifier (small signal) can tell you an enormous amount about the system's stability and/or about its frequency response - think about an RIAA amplifier for example in the latter case where you can use square wave testing to quickly and accurately test for response conformance. In a straight amplifier, for audio applications, I'd expect low/very low overshoot and flat response within the audio band.
Small signal (1~2V) frequency response of a CFA will show GBW independence wrt to gain stepping. If it shows peaking, the designer better understand why and make sure its tolerable in the application. Nothing to do with COD - Herve made that claim IIRC - and there I agree large signal transient analysis is what is required.
Last edited:
No I don't think so. A small signal analysis (or simulation) assumes linear models for all devices, while a large signal analysis should consider nonlinear models for all devices (where applies).
A small signal analysis would not care if you apply a 1kV signal at the input, the result will be a scaled up version of the 1mV input.
I spoke of IPS pole, one less device (feedback network buffer), one less pole.
Why don't you see that ?
Firs amp schemes were single device input, with feedback on the -cathode, emitter, source- now named as CFA. VFA appears by adding a device to have a LTP or others, and a buffered feedback network.
As I said, for the purpose of a small signal analysis, the input signal level is irrelevant, specifying it ("1~2V") confuses me.
Perhaps what you want to infer is that a large signal analysis (using non linear model), with a input signal level -> 0, should converge to the small signal analysis.
BTW, "1~2V" is a pretty large signal for "small signal". The convention is to use signals that are <<kT/q at every node of the circuit. Could be ok in some cases, but I would not always count on that.
"Why don't you see that ?"
I could ask you the same thing Herve! We have been around and around with you and forr and you still cling onto your 'understandings' of how these things work that are manifestly wrong.
They don't behave the same.
They don't look the same.
Academics and engineers who have worked in the field for decades tell you they are not the same and you need to go back and re-evaluate your understanding
Your response? Ignore them and just repeat your dogma.
"Firs amp schemes were single device input, with feedback on the -cathode, emitter, source- now named as CFA. VFA appears by adding a device to have a LTP or others, and a buffered feedback network. "
Adding a 'buffer' to a single ended amplifier does not make it a VFA. Likewise, removing one half of an LTP does not magic up a CFA.
See CPaul and Hans Polak's analysis. You keep 'modifying' circuits to try to support your theory and it does not work.
I could ask you the same thing Herve! We have been around and around with you and forr and you still cling onto your 'understandings' of how these things work that are manifestly wrong.
They don't behave the same.
They don't look the same.
Academics and engineers who have worked in the field for decades tell you they are not the same and you need to go back and re-evaluate your understanding
Your response? Ignore them and just repeat your dogma.
"Firs amp schemes were single device input, with feedback on the -cathode, emitter, source- now named as CFA. VFA appears by adding a device to have a LTP or others, and a buffered feedback network. "
Adding a 'buffer' to a single ended amplifier does not make it a VFA. Likewise, removing one half of an LTP does not magic up a CFA.
See CPaul and Hans Polak's analysis. You keep 'modifying' circuits to try to support your theory and it does not work.