The constant BW property of CFA is only established with one specific type of feedback network as I've tried to explain in an earlier post here in this thread.
The constraints are :
1 - circuit gain is established by varying the bottom leg of the divider (Rg) and keeping Rf constant. This network impedance introduces an ecactly counteracting OLG increase when closed loop gain is increased this way.
2 - Rf is selected such that it maximises BW and stability in unity gain (or for any choosen minimum gain). This leads to a netwok impedance that isn't too low compared to the internal node impedance so that the compensating OLG gain increase isn't compromised at higher circuit gains.
For me, the notion of VFA being buffered CFA holds true in any aspect I can think of.
Last edited:
At the time, you appreciated the comments on the principle of the circuit I wrote on a french forum. It is a very smart circuit.
There is a resistor connected from the low impedance inverting input of the AD844 and the output of the power stage. Any voltage difference between them will induce a current across the resistor, delivered by the emitters of the inverting input transistors. The error voltage is converted to current.
Only diminished from the base current, the collector currents are then reflected by current mirrors, and made available at a high impedance output. They are the injected ahead in the circuit in positive feedback.
The inputs and outputs of a current conveyors are named Y for the high impedance input, X for the low impedance input/output (it is the inverting input if used in a feedback circuit), and Z for the high impedance output.
A few days ago, I saw that the whole circuit of a current conveyor can be called a transductor, meaning that it acts as a combined NPN/PNP transistor.
When looking at Lidgey and Hayatech's article published in Electronics World March 2000, "Are current conveyors finally coming out of age ?", one meets many voltage to current and current to voltage conversions. All under the control of bipolar transistors which are transconductance devices (voltage to current converters). It helps to understand the indeniable role of Vbe voltages in "current feedback".
Last edited:
I agree with that in a topological sense, but would like to note that that buffer does not just increase the inv input input impedance. It also changes the basic functioning of the stage: the feedback current now no longer is the output current of the inv input device.
With the added buffer, the inv input signal now controls a voltage sensitive node (the B of the buffer device), while the input device output current now comes from the tail current source and no longer from the feedback network.
So that innocent 'buffer' does a lot more than just buffer, it changes completely the workings of the thing.
Jan
3 Mohms??
That sounds extremely high. I would have said it allows a transimpedance gain of not more than 100'000, i.e. 1V per 10uA, injected at the neg input, the current input port 😛
That would mean a Tz impedance of 100kOhms
https://accelconf.web.cern.ch/accelconf/a04/PAPERS/TUP13002.PDF seems to agree.
This is almost becoming a universal subjective result. I mentioned this a couple years ago and wondered what it was about CMA amps which seems to always get highest marks in sound quality.
I had also noted that among many brands, thier top model(s) would be CMA designs while lower models would be VFB type. Not a coincidence, I thought. Other's had been listening, also.
This lead to a lot of new high slew rate amp designs and other CMA features on another forum here at DIYaudio. To compare CMA and VMA's
So, I now ask you, why do you, Jan, think CFB/CMA "sounds" so neutral and effortless?
THx-RNMarsh
Last edited:
The constraints are simple: a CFA has approximately constant bandwidth as Rg is varied and Rf is held constant as long as Rg || Rf > re (inverting input resistance). Try this in simulation or bench measurements of any commercially available CFA. You won't see this with a "buffered CFA". For confirmation of this for a simple discrete CFA, see part 1 of Current Feedback: Fake News or the Real Deal? | audioXpress
Analogously, a parked car is just a car with the parking brakes on. Oh yeah, except that if I disengage the brakes, the car can actually move.
The Tz impedance is 3 MOhms, with R input being 50 ohms the gain is 60,000. This is what I intended to convey as being stated on page 12 of the AD844 data sheet
Of course, because Rg doesn't change the open loop gain as in a CFA.
But it is easy to do in a VFA, simply change also the emitter degeneration resistors of the differential pair and you have the same result.
Last edited:
Changing emitter degeneration resistors doesn't change closed loop gain. Changing Rg in a CFA and a VFA does.
Subject to the simple constraints I mentioned, VFA bandwidth is inversely proportional to closed loop gain; CFA bandwidth is not.
Sorry to interject into your thread. Just one thought.
A transistor in free space is not identified by manufacturers as a common base amplifier for obvious reasons that other connections exist. Yet CFA's and VFA's are identified as feedback amplifiers. I am guessing the argument is that the Tz resistance is so high (intended infinite) that these devices cannot be used in absence of feedback, hence a legitimate position.
Yet by that argument, manufacturer's of CFA's and VFA's that provide access to the Tz resistance node and that thereupon permit reduction of gain to zero cannot reasonably be called CFA's or VFA's, rather perhaps CMA's and VMA's
I don't say that.
I said that changing Rg in a CFA change the open loop gain, it is why the bandwidth doesn't change when you increase the closed loop gain in a CFA by lowering Rg.
To have the same behaviour with a VFA you lower Rg AND the emitter resistors.
Come to think of it, subject to the same constraint, VFAs might have the same bandwidth characteristics as CFAs. But there are good practical reasons that you don't see VFAs with Rg || Rf greater than inverting input impedances.
It is still true, however, that classic VFAs are limited by their DC bias currents in slewing internal compensation capacitor voltages, whereas CFAs can add feedback network current to that task. This is true regardless of the value of Rf || Rg.
It is still true, however, that classic VFAs are limited by their DC bias currents in slewing internal compensation capacitor voltages, whereas CFAs can add feedback network current to that task. This is true regardless of the value of Rf || Rg.
I didn't say you did. I was making a different point.
So you agree that VFAs and CFAs have different properties.
I never disagreed on this point.
VFA topology isolate the feedback network from the open loop with the price of additional buffer and additional pole.
That doesn't make very different animals.
Data sheet.
The 10^5 in the article you quoted is not open loop but determined by the load R's on the Tz terminals (fig 2 I guess).
Jan
Attachments
Last edited:
Some of those in this thread have focused like a laser on the fact that ic = gm x vbe. This has perhaps led to ignoring another fact: ic = alpha ie, where alpha is a dimensionless parameter slightly less than 1.
Fortunately, there exist well-accepted small signal transistor models that reflect each of these principles. In the attachment, nodal current equations of closed-loop CFA schematics are written by inspection and then solved for ic. The results are reproduced here:
Note how difficult it is to simplify the coefficient of vbe in comparison to the fact the coefficient of ie is approximately 1. For this reason alone, I prefer the ie formulation. Add to that the fact that ie = ignd - io, and we can now formally mathematically confirm the extent to which input stage current is in part output stage current.
Clearly, part of output stage current feeds input stage current in a CFA, just as part of output stage voltage drives one of the inputs of the VFA input stage.
In both cases, there is a feed "back".
In a straight forward application of language, VFAs experience voltage feedback and CFAs current feedback.
If you still contest this claim, you must believe that there is a flaw in this logic. It should be easy for you to point it out.
Will you?
Fortunately, there exist well-accepted small signal transistor models that reflect each of these principles. In the attachment, nodal current equations of closed-loop CFA schematics are written by inspection and then solved for ic. The results are reproduced here:
Note how difficult it is to simplify the coefficient of vbe in comparison to the fact the coefficient of ie is approximately 1. For this reason alone, I prefer the ie formulation. Add to that the fact that ie = ignd - io, and we can now formally mathematically confirm the extent to which input stage current is in part output stage current.
Clearly, part of output stage current feeds input stage current in a CFA, just as part of output stage voltage drives one of the inputs of the VFA input stage.
In both cases, there is a feed "back".
In a straight forward application of language, VFAs experience voltage feedback and CFAs current feedback.
If you still contest this claim, you must believe that there is a flaw in this logic. It should be easy for you to point it out.
Will you?
This is true only if the differences in bandwidth and slew rate characteristics are not "very different". I won't argue with you about the meaning of the word "very".