Current Feedback Opamps in simulation and practice.

[jan.didden]

Its unstable which is exactly how it should be and yet LTspice doesn't show a problem here.

Hi Mooly, most of these differences between sim and real world are caused by parasitics, in particular any package/socket parasitics on the inverting input. I would expect the sim to be unstable also with say 5pF from inv input to ground.
Did you try something like that?

jan

 

[jcx]

the 2nd order linear model for CFA isn't literally unstable with small C on inverting input to gnd - but the C does cause huge peaking

output to -in C is unstable in general - though you can still have fractional to low single digit pF (depending on everything as usual)

a big fail in freebie CFA spice models is lack of any attempt at getting high frequency output Z even in the ballpark


and PS currents, internal compensation rail coupling are often left out of VFA and CFA macromodels
so circuits with anything other than the lowest PS Z with AC=0 may have very different behavior than the sim

 

[Mooly]

Hi Jan,
Well the results are a wee bit strange. I've got the original file and tried a few things. Yes, I can provoke oscillation but it seems to need a large value of C adding and the result varies with the sim run time. Make it 10 or 100us with a suitable C and it oscillates after say 60us. Make the run time longer and there is just femto volts of noise.

 

[jan.didden]

Yeah, this is uncharted territory I guess, see also jcx's post above.
One would expect that the input pin/socket cap would be included in the model, don't know if it is though.

Jan

 

[Mooly]

Thanks Jan. I would agree with jcx that the model doesn't reflect the real device in ways like this.

I guess the morale of the story (as always I suppose) is to back the simulations up with real world builds and prove or disprove that they work as expected.

 

[31697B]

Thanks Jan. I would agree with jcx that the model doesn't reflect the real device in ways like this.

I guess the morale of the story (as always I suppose) is to back the simulations up with real world builds and prove or disprove that they work as expected.

Data from measured physical circuits can then be used to make the SIM more realistically accurate.


THx-RNmarsh