Current feedback: impedance and phase issue

On the other hand, CFB and VFB are clearly defined as working principles of the feedback mechanism inside the black boxes, by those who design and manufacture this stuff.

Yes, the manufacturer's terminology is not based on the quantity that is sampled (which is always voltage) but the way in which is it fed back (voltage / current).

Thus the CFA is one where the output voltage is sampled but fed back as a current into the low-impedance inverting input whereas VFA samples the voltage but feeds back a scaled version of itself instead of a current.
The point is that for a CFA the open-loop gain Aol of the amp is adjustable via the network impedance at the inverting input, whereas for a VFA the network impedance does not affect Aol.
It may not be completely correct to say this in case of the voltage feedback amplifier as what matters is the net loop gain (Aol & network) and not Aol itself. In case you meant to say that the value of Rf doesn't affect slew-rate etc. for a VFB amplifier, that's true, yes.


1715743978718.png



What counts is the implementation of the actual circuit.
Agree, the topology becomes irrelevant as soon as there's a VFB amplifier that could mimic CFB behaviour (say using cascaded feedback loops).
 

Attachments

  • 1715743618877.png
    1715743618877.png
    8.9 KB · Views: 13
Yes and no, degeneration of the CFB emitters amount to degeneration of a differential emitters in some way, the result
is the same since in both cases this will limit the OLG and max frequency.
The difference is that with a VFB you can have no degeneration at all and this will provide a higher OLG contrary to the CFB
where a minimal degeneration is mandatory.
Isn't that a more a property of the input stage topology?
VFA's typically have an LTP input (with or without degeneration) but not necessarily so.
LT1363:
1715756067472.png

Which is a VFA but based on a buffered CFA with fixed degeneration and fixed Aol to have a universal and (almost) unity gain stable OpAmp.

Obviously, we cannot make a CFA with an LTP input but we can make a VFA with a non-LTP input.
 
Here is one example of my CFA amp, and the same vesion changed to VFA just adding high imput impedance buffer before - input.
 

Attachments

  • 100W CFA MOSFET -1-1-sch.jpg
    100W CFA MOSFET -1-1-sch.jpg
    287.7 KB · Views: 86
  • 100W CFA MOSFET -1-2-bal_highimp-sch.jpg
    100W CFA MOSFET -1-2-bal_highimp-sch.jpg
    384.9 KB · Views: 85
  • Like
Reactions: 1 user
The difference between current feedback and voltage feedback for audio amplifiers is a non issue. Voltage feedback amplifiers for audio applications are superior to their current feedback peers.
are you starting another war? this topic has been tackled to death in previous years, resulting to some members getting banned in the process...i suggest you search out those threads....
 
  • Like
Reactions: 1 users
Isn't that a more a property of the input stage topology?
VFA's typically have an LTP input (with or without degeneration) but not necessarily so.
LT1363:
View attachment 1310314
Which is a VFA but based on a buffered CFA with fixed degeneration and fixed Aol to have a universal and (almost) unity gain stable OpAmp.

Obviously, we cannot make a CFA with an LTP input but we can make a VFA with a non-LTP input.

That s a VFA in all cases since those so called CFAs should be more accurately branded as high Imax IPS VFAs.

In a differential the IPS Imax is limited by a current source, say 2mA total current, so the max current in a branch cant exceed
2mA.

Now in a so called CFA if there s 2mA flowing in a degeneration resistance that is say 50R when there s no signal then if the base of the IPS transistor is excited with just 1V this will increase the current to 20mA, that s the fundamental difference with a classic
VFA, the IPS Imax can be unlimited and so will be the bandwith and slew rate since SR = (Imax x Av)/Cin, Av being the VAS gain and Cin its input capacitance.

Now if you use a differential with a 20mA current source you ll get the same bandwith and slew rate as a pseudo CFA that is limited to 20mA Imax IPS, the higher CCS current in the differential wont make it a CFA, and neither is the so called CFA a real one, at the end of the loop it s still the quantity of sampled output voltage that will decide for the NFB ratio in both cases and set a voltage gain, not a current gain.
 
  • Like
Reactions: 2 users
Agree, the topology becomes irrelevant as soon as there's a VFB amplifier that could mimic CFB behaviour (say using cascaded feedback loops).

The following uses VFAs to obtain gain - bandwidth properties like those of a CFA. The resistor R311 is parameter swept to vary the overall voltage gain from 0dB to 20dB. It is fairly easy to see that the bandwidth is the same across all gains.

1715773823556.png

1715774005363.png
 
Yep, it retains one property of a CFA which is constant BW...
This property is said to greatly improve the step / pulse response, which is what makes people pick a CFA, despite its drawbacks. If not, correct me.

....when you scale the resistors by k and the capacitor by 1/k the response does not change, unlike a "real" CFA.
Come on, how often do these change values, especially in a real amplifier whose design has been finalised ? And, there're also the other differences like DC accuracy, inverting character etc. of which some could actually be improvements over the CFA. One might also want to add a buffer (as initially mentioned by the OP) to obtain the high input impedance of VFAs.

Yes, that is not a 'real' CFA but something that can provide similar performance for people who don't like / want 'real' CFAs. What I've drawn is rudimentary, and the design needs to be worked out in detail, on a case by case basis.
 
This property is said to greatly improve the step / pulse response, which is what makes people pick a CFA, despite its drawbacks. If not, correct me.
Non-slewing step/pulse follows mag and phase frequency response and vice versa, it the same thing just displayed differently.
A constant BW at various gain settings is nice feature for a versatile measurement amplifiers but for audio I see no real advantage and the amps run fixed gain anyway, usually.

When talking about slewing behavior (a rare and mostly irrelevant corner case in real life), CFA (or a VFA with a CFA core) tend to slew faster and recover earlier from slewing because the input-stage differential output current is not coming from a fixed low current source and hence does not saturate, as described by:
In a differential the IPS Imax is limited by a current source, say 2mA total current, so the max current in a branch cant exceed
2mA.



Come on, how often do these change values, especially in a real amplifier whose design has been finalised ? And, there're also the other differences like DC accuracy, inverting character etc. of which some could actually be improvements over the CFA. One might also want to add a buffer (as initially mentioned by the OP) to obtain the high input impedance of VFAs.

Yes, that is not a 'real' CFA but something that can provide similar performance for people who don't like / want 'real' CFAs. What I've drawn is rudimentary, and the design needs to be worked out in detail, on a case by case basis.
I'm not trying to judge anything here, rather my point was, from a behavioral viewpoint on amplifier types seen as a 5-terminal black box (In+, In-, Out, Vs+, Vs-), the CFA differs from the VFA in that the In- network impedance affects open-loop gain (giving one more degree of design freedom) whereas for a VFA it does not. That's all, basically, and it is what I think qualifies as the main criterion.
Both have (different) strong and weak points, and both can be implemented in good or bad ways.
 
CFA (or a VFA with a CFA core) tend to slew faster and recover earlier from slewing because the input-stage differential output current is not coming from a fixed low current source and hence does not saturate,

A constant BW at various gain settings is nice feature for a versatile measurement amplifiers but for audio I see no real advantage and the amps run fixed gain anyway, usually.

As you've yourself mentioned, the slew current (that charges the compensation capacitor) is obtained from the final output (through Rf) which would ensure proportionally larger slew currents for larger outputs. This is also the case in #28, resulting in gain-bandwidth characteristics similar to that of CFAs.

However, in audio, even when the gain is fixed, there could be sudden percussive or explosive sounds of much higher amplitude, especially in high dynamic range professional applications. In such cases, this (seemingly unlimited) slew-rate should definitely help the amplifier follow the input signal in a more accurate way, giving lower error and distortion etc.

Even in a CFA (or similar) with fixed gain, the settling time for different step inputs (say 0.1V vs 1V) would remain comparable due to the increased slew current for the 1V case. Thus, the gain need not change, as the slew current itself is variable. It just climbs better (vs. VFAs ), if I understand correctly.

I completely agree with the final point that indicates an extra degree of freedom for the designer.
 
Even in a CFA (or similar) with fixed gain, the settling time for different step inputs (say 0.1V vs 1V) would remain comparable due to the increased slew current for the 1V case. Thus, the gain need not change, as the slew current itself is variable. It just climbs better (vs. VFAs ), if I understand correctly.

I see no real advantage and the amps run fixed gain anyway, usually.

The following is the earlier circuit with a fixed gain of 10V/V, when fed with a 0.1V pulse followed by a 1V pulse. As expected, the voltage across the compensation capacitor shows comparable rise / fall and settling times for both pulses, confirming the slew current scaling action that occurs within.

1715935134920.png


1715935190260.png
 
Last edited:
When talking about slewing behavior (a rare and mostly irrelevant corner case in real life),...
Now, you could say the above and assume that the VFA bias current doesn't saturate, but won't a CFA designed using the same transistor process be faster than the VFA (unless the bias current is very high) and why won't this higher slew rate be beneficial, performance-wise ?
 
Now, you could say the above and assume that the VFA bias current doesn't saturate, but won't a CFA designed using the same transistor process be faster than the VFA (unless the bias current is very high) and why won't this higher slew rate be beneficial, performance-wise ?
Well yes, I think its safe to say that a CFA tends to be generally faster than a same-technology LTP-based VFA. But one can buffer the CFA's -IN and have a fast VFA, from the outside view... and which is what we can find in some designs, discrete and monolithic ICs. You can also always use a precision VFA as the master in a composite and let the CFA do the heavy lifting.
 
So far all good points about the functionality, impedance and gain-bandwidth product are discussed. But what about the PSSR of these buffered CFAs? Is the PSSR of this topology at par with that of a VFA? A point to highlight here is, with the buffered inputs at the front end of these CFAs, the input stage looks like a long tail pair but without the CCS. How does it impact the PSSR? It would be great if someone shares a PoV. Thanks in advance.
 
OK, here are some simulated data of the balanced amp from post #23.
I never built 100W balanced version but only 200W, it sounds great.
First one is PSRR, second one is CMRR.
Damir
 

Attachments

  • 100W CFA MOSFET -1-2-balancedin-PSRR.jpg
    100W CFA MOSFET -1-2-balancedin-PSRR.jpg
    129.6 KB · Views: 18
  • 100W CFA MOSFET -1-2-balancedin-CMRR.jpg
    100W CFA MOSFET -1-2-balancedin-CMRR.jpg
    256.1 KB · Views: 18
  • 100W CFA MOSFET -1-2-balancedin-noise.jpg
    100W CFA MOSFET -1-2-balancedin-noise.jpg
    163.6 KB · Views: 18
  • 100W CFA MOSFET -1-2-balancedin-FFT1k-1W-balanced-drive.jpg
    100W CFA MOSFET -1-2-balancedin-FFT1k-1W-balanced-drive.jpg
    174.7 KB · Views: 10
  • 100W CFA MOSFET -1-2-balancedin-FFT1k-100W-balanced-drive.jpg
    100W CFA MOSFET -1-2-balancedin-FFT1k-100W-balanced-drive.jpg
    183.9 KB · Views: 9
  • 100W CFA MOSFET -1-2-balancedin-FFT20k-1W-balanced-drive.jpg
    100W CFA MOSFET -1-2-balancedin-FFT20k-1W-balanced-drive.jpg
    174.7 KB · Views: 11
  • 100W CFA MOSFET -1-2-balancedin-FFT20k-100W-balanced-drive.jpg
    100W CFA MOSFET -1-2-balancedin-FFT20k-100W-balanced-drive.jpg
    185 KB · Views: 13
  • 100W CFA MOSFET -1-2-balancedin-outputimpedance.jpg
    100W CFA MOSFET -1-2-balancedin-outputimpedance.jpg
    209.6 KB · Views: 12
Last edited:
  • Like
Reactions: 1 user