How to stop op-amps from oscillating

[Solberg]

Hi

Does anyone know how i stop this circuit from oscillating?

I haven't bought the parts yet, so i cant test in real life.

It started oscillating when i added the x10 feedback op-amp (U2).

Any help Will be greatly appreciated.

[Jan Audio]

Add a 100pf or 220pf cap between the output and inverting input from the lower half.

[dietert1]

When you use a unity gain stable OpAmp it means you can feed the full output signal to the inverting input. If you feed 10x the output to the input it will oscillate. So you need to reduce loop gain at high frequencies to the unity gain situation by adding parallel to R4 a capacitor or a capacitor in series with up to 20 KOhm resistor.

Regards, Dieter

[Jan Audio]

Here is another topic :
https://www.eevblog.com/forum/projects/opamp-ringing/msg2459511/#msg2459511

It is called : opamp instability.

[Solberg]

Thank you all for the help.

I have a better understanding of the problem now. thanks

[schmitt trigger]

I see that you are building a constant current load.

If you plan to actually build this unit and not constrain yourself to a simulation, for best DC precision make sure that the grounding of the resistor R5 is routed as shown in my sketch.

[magic]

Is this really necessary? Why not divide the input signal to the upper opamp 10x?

[Zero999]

Is this really necessary? Why not divide the input signal to the upper opamp 10x?

Agreed. It still may require some additional compensation, but it's still much more sane to use an attenuator, rather than an amplifier.

EDIT: The attached schematic is wrong, see below.

[Mechatrommer]

you got it the other way around. the divider should be on +ve input of U1 dividing V3. OP wants to amplify the R1 voltage because he got 10X input (V3), so instead, he should divide the V3 1/10X.

[dietert1]

I understood the second amplifier was eventually meant for scaling the shunt signal up for measurement (e.g. ADC input). That amplifier stage doesn't need to be in the control loop. One can scale down the loop input signal by the same factor.

Regards, Dieter

[Zero999]

you got it the other way around. the divider should be on +ve input of U1 dividing V3. OP wants to amplify the R1 voltage because he got 10X input (V3), so instead, he should divide the V3 1/10X.

You're right. I had a total brain fart. I just moved the gain to the other op-amp. 

[MarkF]

A 1Ω sense resistor will get pretty hot.

Here's what I did with a 0.2Ω 3W sense resistor for up to 20V and 2.5A.  Actually, the limit is about 50W.
The voltage across the sense resistor will be 0V to 0.5V.

   

and here is another circuit:   eLoad.pdf (32.95 kB - downloaded 116 times.)
It is discussed here but I can't find the thread.

[David Hess]

You're right. I had a total brain fart. I just moved the gain to the other op-amp. 

Exactly, U2 was not doing anything useful to improve the ultimate performance of the circuit.

[magic]

When you use a unity gain stable OpAmp it means you can feed the full output signal to the inverting input. If you feed 10x the output to the input it will oscillate. So you need to reduce loop gain at high frequencies to the unity gain situation by adding parallel to R4 a capacitor or a capacitor in series with up to 20 KOhm resistor.

You are making U2 an integrator with 90° phase delay above the RC cutoff point. I'm not sure if that's a good idea even if gain falls with frequency too. The more common solution is what others showed: bypassing the U2 circuit entirely at high frequencies.

[dietert1]

Don't know what this comment shall be good for. Are you asking a question?
Please don't forget this is "Beginners", so i tried to explain a little what may be wrong, in case somebody wants to understand how to use multiple OpAmps in one control loop. Of course, the concept of unity gain stability is a limited approach, yet it is something to start with. In reality also the Mosfet adds phase shift and gain. A complete analysis would involve breaking the loop, connecting a generator to the input and plotting phase shift and gain as a function of frequency in order to define a stable solution.

Regards, Dieter

[magic]

Sorry, I somehow ended up assuming your solution would add significant phase lag through U2 at high frequencies. On second thought, it seems OK.