OPA2192: Gain=-1, step responce @Rf=10k or 1k

Sorry,

//Rf = 10k with 10pF

Rf = 1k without Cf

Hello Hiroshi,

Certainly, a 10 kilohm feedback resistor can be used with the OPA192, but using a value this high can result in the overshoot you are observing. Be sure that a bit of overshoot, and ringing is something the application really needs to be concerned about suggesting a circuit change.

A 10 kilohm resistor in conjunction with the Op amp input capacitance and any circuit stray capacitance introduces additional phase shift into the feedback loop and that reduces the phase margin. One effect of reduced phase margin is increased overshoot when the Op amp is subjected to a step, or impulse input. The lower the phase margin, the higher the overshoot.

Adding a capacitor of an appropriate value across the feedback resistor compensates for the reduction in the phase margin. If the capacitor is of just the right value the overshoot effect can be minimized by reducing the total phase shift. Alternately, the feedback resistor value along with the input resistor can be scaled down in value. Since the capacitance seen at the input node remains the same reducing the lower value resistors result in less phase shift. Reducing the feedback resistor to 1 kilohm certainly should produce less overshoot response compared to a 10 kilohm resistor.

It is a little difficult to know the PC board circuit capacitances and the exact input capacitance of the particular OPA2192. However, applying the OPA2192 simulation model will get you close to determining the optimum feedback capacitance. Based on the simulation results you provided it looks like the 10 pF might be a bit high and could likely be a smaller value.

Regards, Thomas

Precision Amplifiers Applications Engineering

Tom is right, 

Essentially, for loop gain purposes, the datasheet shows 8pF total C on the inverting node (Ccm+Cdm), with 10k R's that introduces a feedback pole at 1/(2pi5k*8pF) = 4MHz. That is pulling the phase down before LG=0dB crossover. 

Common problem, using the speedup cap idea from this link is usually effective - 8pF across the 10kohm feedback in this case, 

https://e2e.ti.com/blogs_/archives/b/thesignal/archive/2012/05/30/taming-the-oscillating-op-amp

Of course the real test here is to run a LG phase margin test, 

Easy to do, here is the original 10kohm with no feedback C, sense meter is rotated to report phase margin directly - 37degrees with no feedback C - the key thing is to add that input C at the summing junction from the datasheet 

And then adding an 8pF across the feedback R brings this back up to 76deg - pretty effective I would say, 

LG sim file here, 

3022.OPA192 LG test.TSC