Help with amplifier compensation

[Boscoe]

So I'm looking into making this op amp based amplifier with the LTC6090-5 with thoughts to sticking it inside another loop controlled by a higher performance lower voltage amplifier.

Problem is, I'm not sure how to get it stable, it oscillates a little at the crossover point especially when I load the output with some capacitance.

I'm looking for advice on resources to help me understand the methods needed to make this amplifier stable and ideal for putting into another loop in the future. I have tried capacitors around the op amp and the global feedback but they only make it worse. I have tried bode plots but they look okay to me.

Thanks
Boscoe

 

[jaycee]

First things first - put a real 5mA current source in.
Second, put a stopper resistor on the +ve input of the opamp. 1K will do for starters.

You're probably going to need a combination of techniques. CFP output stages on their own are a pain in the **** - so you'll probably want a capacitor across base/collector on at least one half of the output - probably the lower half.

Then combos of capacitance across the feedback resistor, and R/C feedback from the opamp output to the -ve input

 

[U130421]

Put an inductor of 1Gh in series with R11 to see the open loop response. For the closed loop gain you need , look the phase shift ,it should not exceed 180°, or else it oscillate. Applying capacitors here and there you adjust it to have lower than 180° what we call phase margin.

 

[jan.didden]

Good advice.
Starting with a cap across R11 to cancel the inv input capacitance, say 100pF. If the oscillation persists but changes in frequency, put 1k in series with this 100pF.

Jan

 

[odnaizutra]

good way out

 

[Mark Tillotson]

A cap is needed across the bias Vbe multiplier so both drivers are driven with low ac impedance - otherwise the opamp lacks authority on the negative half-cycle.


Check the bias current isn't zero - that will definitely cause issues at cross over.



Capacitive load will usually destabilize an amp without an output inductor - so add an output inductor of a few µH between the 4 ohm load and the feedback take-off point.

 

[rayma]

Do you really want a gain of only x5?

 

[djoffe]

please post model files for your transistors.

 

[Boscoe]

First things first - put a real 5mA current source in.
Second, put a stopper resistor on the +ve input of the opamp. 1K will do for starters.

You're probably going to need a combination of techniques. CFP output stages on their own are a pain in the **** - so you'll probably want a capacitor across base/collector on at least one half of the output - probably the lower half.

Then combos of capacitance across the feedback resistor, and R/C feedback from the opamp output to the -ve input

So I implemented my own current sources, I used one at the top , one at the bottom to lower the dissipation of the op amp. The capacitors only made the oscillation worse. The resistor did nothing.

Good advice.
Starting with a cap across R11 to cancel the inv input capacitance, say 100pF. If the oscillation persists but changes in frequency, put 1k in series with this 100pF.

Jan

100pF made it into an oscillator, any value made things dramatically worse. The resistor did nothing unfortunately.

A cap is needed across the bias Vbe multiplier so both drivers are driven with low ac impedance - otherwise the opamp lacks authority on the negative half-cycle.


Check the bias current isn't zero - that will definitely cause issues at cross over.



Capacitive load will usually destabilize an amp without an output inductor - so add an output inductor of a few µH between the 4 ohm load and the feedback take-off point.

This capacitor really helped but did not eliminate the oscillation. The bias is 100mA. The capacitor on the output was a test of stability.

Do you really want a gain of only x5?

I want to keep the closed loop gain as low as possible for lowest distortion. I can provide gain with some gain from an op amp with basically no distortion.

 

[hitsware]

Use LTC6090

 

[Mark Tillotson]

The LTC6090-5 should be stable at gains of 5 or more, and will have higher slew rate than the LTC6090.


But that's without the output stage in the loop. Making the high frequency feedback more directly from the opamp's own output will improve the phase margin, typically this is a capacitor from LTC6090's output direct to the inverting input.


Increasing the gain of the stage will also help as the amount of feedback is reduced in proportion, increasing the gain margin.

 

[djoffe]

If you shunt R10 with 10K in series with 150 pF, you will add phase margin out of band. It shouldn't much effect the in-band distortion, but should help stability. If indeed that's useful, then you can experiment with smaller C's.