Terrific! I still owe you thanks for the cascoded JFET idea, which I'll try if the BJT-based idea you inspired falls flat.
Thanks much, friend!
Last edited:
"My" voltage feedback has nothing to do with the ratio of Early effect in the input stage current generation.
CFA need low impedance feedback network connected to inverting input to work.
As voltage sources are low impedance, and current sources are high impedance, I claim that CFA behaviour is closer to voltage feedback than current feedback.
That is an evidence : you cannot have gain on a CFA with a current feedback as it need low impedance on inverting input.
That said, the fact that the world calls low-impedance inverting amps CFA does not stop me from sleeping.
What are your definition of voltage and current feedbacks ?
I think it is almost always the case as the feedback increase the inverting input impedance, except in inverter mode.
This is an non relevant argument.
1) you point a particular input stage schematics
2) even if it is always the case, your conclusion is wrong : in a differential long tail pair with current mirror, current in one side is equal to the difference of the total diff current and the other diff side, I never heared that diff pair is current driven.
Guys, this is a very interesting and inspiring thread.
It is inspiring to us, eternal beginners, to see Masters of the Art still discussing about how thinks do work...though some decide to put their trust on mathematical models, which are descriptive rather that explanatory tools.
Ah! What a delight to read these paragraphs. Excellent style.
I wonder... IF Faraday, Heaviside, Thomson, Tesla, Steinmetz were honorable members of our beloved Forum, what would they have to say about our opinions...
(I'm just beginning to read those heavy books )
Keep up the spirit of cooperation!
(This is not a forum for competition).
Best wishes to all.
M.
It is inspiring to us, eternal beginners, to see Masters of the Art still discussing about how thinks do work...though some decide to put their trust on mathematical models, which are descriptive rather that explanatory tools.
Ah! What a delight to read these paragraphs. Excellent style.
I wonder... IF Faraday, Heaviside, Thomson, Tesla, Steinmetz were honorable members of our beloved Forum, what would they have to say about our opinions...
(I'm just beginning to read those heavy books
)Keep up the spirit of cooperation!
(This is not a forum for competition).
Best wishes to all.
M.
I agree that an ideal controlled current source used as a feedback element in a CFA won't make the circuit work as we'd like. And I understand your reasoning. But I do not agree that this leads to your conclusion.
An excellent question! I puzzled over this in part 2 of Current Feedback: Fake News or the Real Deal? | audioXpress .
One answer is to measure the voltage and the current gains inside the loop. The one with the least gain has the most feedback, and it can be said to be the predominant form in the circuit. Hans Polak pointed out a good article that discusses this: https://www.edn.com/design/analog/44...back-amplifier
Another answer is historical. Some amplifiers drive a load in series with a small ground-referenced sense resistor. In the limit as the sense resistor goes to zero, the voltage across it becomes proportional to the load current. This signal is fed back to the inverting input. Even though no current is fed back to anything in this design, this has been referred to as a current feedback amplifier!
My preference is to start by noting that the signal output of all input stages, CFA or VFA, is a collector current. We can ask where this current comes from. In a classical VFA, it comes from a (non-ideal of course) "constant" current source within the input stage, and the voltages applied to the bases steer it to one transistor or the other; no current is being fed back from anywhere. But in the CFA, it comes from the emitter current which is fed back through the feedback network to and through the input stage: current feedback.
Yes. But I am referring to the comparison between the open loop impedance looking into the feedback network and the same looking into the inverting input. In my article (and I think in the EDN one cited above), the equality of these two things can be shown to be the tipping point between predominantly voltage and predominantly current feedback. If the network's is the larger, we have predominantly current feedback; if the inverting input's is, predominantly voltage feedback.
I quite disagree. It is surely relevant that if the output current of an input stage is fed back to it through a feedback network, from an etymological viewpoint alone, we are justified in calling it current feed back. And...
1) This is true for all classical CFAs, not just a particular one.
2) Now that's what I call irrelevant. In a long tailed differential pair, the input stage current is NOT fed back from anything; it comes from a constant current source within the input stage.
...though some decide to put their trust on mathematical models, which are descriptive rather that explanatory tools.
I'm intrigued - how does one explain transistors non-mathematically? What shall we trust in place of such models? Are you referring to semiconductor physics?
According to this point of view, the CFA name does not refer anymore to a topology but to a relation between two impedances.
I handled the relation between the impedances of the inverting input and the feedback network post # 1189 where it can be seen that the emitter current (and then the output current of the first stage) is not affected by the impedance of the feedback network. Question : why is it so constant ?
I think that the inverting input of a CFA can be interpreted as a voltage controlled voltage source (VCVS).
Interesting.
So we are discussing since aprroximatively 1 billion years, exchanging 50 billions messages and we are not able to agree on what we call voltage or current feedback ?
I am not sure we can find a more stupid situation elsewhere than in a hifi forum.
I like your first answer. VFA or CFA depends on the output load and you can choose your preferred input to make the calculation.
In any case, with this criteria, the amp which I attached the scheme is a VFA.
2nd answer describes an amplifier which output is a current, so with this definition the attached scheme is still a VFA.
I understood the third was your preferred definition.
So your definition is : a CFA is an amplifier where the input stage output current comes from the output stage via the feedback network and inverting input.
What does become your definition if we add an emitter follower to the input stage ?
Fine.
So my attached amp is still a VFA.
Network Z is 9.5 ohm
Inverting input Z is 10.91 ohm
Attachments
Can't we say after so many postings that the ongoing discussion on the semantics about what the industry resp. calls VFA and CFA doesn't brings us any further.
You can look at it from different angles and because Amps cannot flow without Volts none of both alone can be made responsible for the difference between the two.
But at the same time there are very distinct differences between CFA and VFA, so it makes absolute sense that two different names are being used.
Given the fact that those two names have already been chosen for us, we better live with them, so what, they're just names !
Therefore I would like to propose to reverse the procedure which of both names applies to a specific component or circuit design, without further worrying about the semantics.
The way to classify a VFA or a CFA could very well be by applying R. D. Middlebrook’s Double Injection Technique or DIT for short.
With this DIT technique a very good discrimination can be made between the two topologies by comparing the Fr. Resp. of both Current Transfer Gain Ti and Voltage Transfer Gain Tv to the Loop Gain T.
For a correct classification. there should be an order of magnitude between Ti and Tv.
The one coming close to T, whether Ti or Tv determines the circuit to be classified as CFA or VFA.
When however Ti and Tv are close to one another it's a mix of VFA and CFA and not deserving to get any of both names.
In # I have already shown what it looks like when LT1395 CFA and a LT6202 VFA are both given the DIT treatment, but I will show the images just another time to show what I'm talking about and give it another try.
I have also included an LTSpice model, giving the option to ecperiment yourself with DIT.
So my adage would be: forget the semantics and welcome the CFA as a thing of its own as against the VFA.
Hans
You can look at it from different angles and because Amps cannot flow without Volts none of both alone can be made responsible for the difference between the two.
But at the same time there are very distinct differences between CFA and VFA, so it makes absolute sense that two different names are being used.
Given the fact that those two names have already been chosen for us, we better live with them, so what, they're just names !
Therefore I would like to propose to reverse the procedure which of both names applies to a specific component or circuit design, without further worrying about the semantics.
The way to classify a VFA or a CFA could very well be by applying R. D. Middlebrook’s Double Injection Technique or DIT for short.
With this DIT technique a very good discrimination can be made between the two topologies by comparing the Fr. Resp. of both Current Transfer Gain Ti and Voltage Transfer Gain Tv to the Loop Gain T.
For a correct classification. there should be an order of magnitude between Ti and Tv.
The one coming close to T, whether Ti or Tv determines the circuit to be classified as CFA or VFA.
When however Ti and Tv are close to one another it's a mix of VFA and CFA and not deserving to get any of both names.
In # I have already shown what it looks like when LT1395 CFA and a LT6202 VFA are both given the DIT treatment, but I will show the images just another time to show what I'm talking about and give it another try.
I have also included an LTSpice model, giving the option to ecperiment yourself with DIT.
So my adage would be: forget the semantics and welcome the CFA as a thing of its own as against the VFA.
Hans
Attachments
Question for all, take 15V rails, your choice of resistors and capacitors, and nothing but say 3904's and 3906's. To be somewhat arbitrary pick a useful task like a non-inverting 20dB line stage driving 600 Ohms from a 20K volume control and make the metric the distortion floor of a standard multi-tone signal at 2V rms.
Show me all the posts here that would give some true insight into accomplishing this.
Show me all the posts here that would give some true insight into accomplishing this.
When you connect a resistor R between your emitter follower and the minus input stage, you will have a VFA.
Hans
Your proposal has the advantage of being backed up by something we can measure. At the very least, even without the order of magnitude qualification, this is a way to determine whether voltage or current feedback predominates at the inverting input of any amplifier. (Note that while the total loop gain is the same everywhere in a given circuit, the balance between current and voltage loop gains varies depending on the point in the circuit that the measurement is being made.)
However, this does not allow us to distinguish between topologies. CFA manufacturers' recommended feedback network component values for certain gains actually lead to cases where voltage feedback predominates. And even if this didn't happen, it would still be possible to achieve VF by using suitably low-valued components. So then I could change "topologies" by simply changing a component value. I don't find this very satisfying.
My personal preference for discriminating between topologies is to ask a simple question: what is the source of the signal current in the amplifier input stage? If it is a current source which is part of the input stage, then we have a VFA; if it is the feedback network, it's a CFA.
When it comes to modern double diamond pairs biased by current sources and interconnected by a resistor between their emitters, I would have to say that by the above rule, these are enhanced VFAs.
That does not matter that much as long as the whole audio range is covered, AP has a set they use. That was not the point, this discussion has diverged quite far from doing actual design.
I'm sure you're not trying very hard. Although I suspect we can quibble about the "1 billion years" timeframe and "50 billion messages", it took the concerted efforts and contributions before I was born of a lot of bright people to bring us to this point.
How?
... and a voltage, unless your load and sense resistors are dead shorts.
Another interesting question. I'd have to see the circuit. First, how would you bias your emitter follower? If with a constant current source, then on a voltage stroke in one direction, the original input stage would benefit from the "slewing current on demand" feature of the CFA. But in the other direction, it would have the VFA characteristic of being limited by the current source current. This is not a circuit I'd seriously consider using. Perhaps you'd care to propose something useful for discussion?
Still, dwelling on your question, since both the original input stage and the modified new one would presumably see a feedback network whose impedance is very much lower than their own, Middlebrook would say that voltage feedback predominates in this circuit.
I would say that the circuit topology is CFA. Certainly, it retains the CFA feature of "slewing current on demand". However, I don't believe it would have the "constant bandwidth" feature of the CFA as you lowered the value of R8. So it's kind of a combination of the two, I'd guess. Again, I'd question the usefulness of a circuit like this.
One more thing to consider: The CFA topology enjoys "slewing current on demand" even when Middlebrook says that voltage feedback predominates. Middlebrook alone is not a good criterion for distinguishing between a CFA and a VFA.
You need to distinguish between small signal vs large signal analysis this at times has been very sloppy in this thread.
Hans,
my emitter follower is located at the output of the input stage, so that the output current of the input stage is not the current coming from the output stage.