Gain Blockwith Variable Loop Gain

[Bonsai]

Edmond - ok - I agree no point in starting futile arguments. I just want to see for myself.

JCX- I unticked the compression boxes in the control tab - I assume thats the the same as applying the directive you refere to?

KSTR - are you refering to my cct or the blamles s amp? my circuits CFA - miller cap makes it oscillate. To improve phase margin, the recomended technique is to apply a filter to the front end. I found triming the VAS cap also helped (sim'd with square wave input).

Traderbam - good point, but my only concern is that the input would have to be attenuated at the same time in order to keep the output level constant (it would be less useful to trim the output through a pot becuase this would raise the output impedance and there is also the output saturation issue to deal with.

[KSTR]

Originally posted by Bonsai
KSTR - are you refering to my cct or the blamles s amp? my circuits CFA - miller cap makes it oscillate.

I meant your circuit, and increasing C1 (the miller cap IMO, forming the dominant pole) from 22pF to 100pF. But it wasn't a full detail analysis with the Tian loop gain probe, only a quick plot of Vout/(Vin+ - Vin-).

- Klaus

[traderbam]

C1 isn't strictly a Miller cap, though...what I see is a transconductance amp driving an RC shunt to gnd, then a voltage buffer to the o/p.

I predict this circuit will be prone to instability without proper miller compensation.

What if you made an ordinary circuit, forget the symmetry thing, with an ordinary miller VAS? Then you could easily vary the OL gain by having a variable R across the miller cap (with dc blocker). This would have the effect of turning the VAS into a variable low pass filter.

IOW, a transconductance, differential input stage feeding an integrator with variable corner frequency, then a unity gain buffer to the output.

[KSTR]

Originally posted by traderbam
C1 isn't strictly a Miller cap, though...what I see is a transconductance amp driving an RC shunt to gnd, then a voltage buffer to the o/p.

You're right, it's not what should be called a Miller cap the way that term is normally used... although it does the same thing, forming the dominant pole together with that load R, and setting bandwidth.

I predict this circuit will be prone to instability without proper miller compensation.

In the sim it is stable unless loaded with capacitance, and shows the typical CFA attributes; when feedback R is reduced the integrator cap must be increased -- that's what I mistakingly named "gain" in a previous post, forgetting the CFA aspects for a moment...

- Klaus

[traderbam]

Klaus, I didn't mean to be nit-picky...I know you know what you are talking about. Just trying to avoid confusion of the passer-by.
Regarding stability of simulations correlating with stability of real circuits. Ha, how much time have you got?

[Bonsai]

This is a small signal amp - so the output would be fed via an isolating resistor - e.g. 20 to 50 Ohms. This would remove any problems with load capacitance - I should have shown that in the circuit for completeness.

[traderbam]

If you are interested, there is another way to do what I think you are trying to do. See conceptual circuit below. Might be simpler to implement and will be very stable. That is to use a wide-band differentiator (U1) feeding an op-amp (U2). Varying R4 will vary the feedback factor without changing the GBP. U1 can be a transconductance amp too, or a discrete subtractor like you have in your current design. U1 needs to have a GBP greater than U2.
I would have thought a couple of decent, off the shelf amps would do.

[Bonsai]

Thats neat traderbam. It can be done with discrete op-amps and IC types.