OPA355 with excessive and asymetric overshoot

[SparkingDog]

I am stumped. I have an VFA (OPA355) ckt configued with gain=+2. Rf=499ohms, Vs=+/-2.5V. I have done everything I can think of to minimize stray capacitance. The amp has good decoupling.
The amp is currently unloaded.

The Problem:
When I apply a 0.5V, 10Mhz square wave to the input, the output has a large overshoot (more than 50%); but only on the falling edge. What could cause this asymmetry?

As an experiment I changed the gain to +1. With a 499 ohm feedback resistor the overshoot remains but it disappears if I short across Rf. A 2pf Cap across Rf reduces the overshoot by half.

 

[FvM]

No comments without knowing the circuit layout.

 

[SparkingDog]

What in the layout could cause asymmetric overshoot?
I would be grateful even for speculation.

I believe I have been very careful in my layout. Tight placement of components; removal of planes beneath the output and feedback nodes; multiple decoupling caps (0.1uf & 1uf) between the supply pins. There is decoupling between the gnd and -V but this should not be relavent in the follower configuration.

Also I don't see any droop or ringing on the supplies.

 

[crutschow]

If you short across Rf and the overshoot disappears then it would appear to be due to stray capacitance at the summing junction. Probably to get that capacitance any lower you would have to mount the components and summing junction in the air. So the obvious solution is to experimentally add feedback capacitors until you find the minimum value that eliminates the overshoot and hope that doesn't reduce your desired frequency response too much.

Why it's asymmetrical I don't know, but it may be just the difference between the positive going and negative going drive characteristics of the output stage.

 

[SparkingDog]

This is an image of the layout if that helps.


- - - Updated - - -

This is the first design I have done with a bandwidth this high. I would appreciate any insight from those with more experience.

Is it typical that the stray capacitance cannot be reduced to where a feedback cap is unnecessary?
The datasheet gives no guidance to add capacitance; in fact it warns that even a few pf of capacitance on the output can cause problems with phase margin.

 

[albbg]

How did you measure the output signal ? If you use an oscilloscope probe with the usual alligator to connect the ground you could experience the bouncing due to the inductance of the ground wire.
If this is the case, do the measurement again taking the ground as close as possible to the point you take the signal, using a very short wire. Usually the probes are provided with many tip tools.

 

[FvM]

How did you measure the output signal ? If you use an oscilloscope probe with the usual alligator to connect the ground you could experience the bouncing due to the inductance of the ground wire.
If this is the case, do the measurement again taking the ground as close as possible to the point you take the signal, using a very short wire. Usually the probes are provided with many tip tools.

Only a wideband resistive probe or an active probe can be expected to reproduce > 100 MHz analog signals reliably. A 450 or 950 ohm series resistor connected to a coaxial cable (10:1 or 20:1 resistive divider) and directly soldered to the circuit is an unbeatable cheap and high performant probing method. In test boards, resistors and coaxial connectors can be provided by design.

I don't see a particular problem with the layout, but I would always place bypass capacitors to ground in the first level. The latest when you load the OP with a probe, the missing direct bypass to ground matters. Also common mode interferences on the supply planes aren't blocked.

You didn't tell yet about the input signal characteristic. I guess it's a high speed "digital" signal with fast edges? Although I've been using OPA355 and similar devices a lot, I'm not aware of the behaviour with fast edges that drive the OP into slew rate limitation. Thus I can't exclude that there's a systematical problem with these OPs. The datasheet pulse responses already show a slightly asymmetry in the about 10% overshoot. And unfortunately there's no related input waveform shown, so we don't know if it has an intentionally adjusted slew rate.

On the other hand, if you see the chance to improve the pulse response by adding a small capacitor to Rf, you should do that.

As a more general comment, VFB OPs have rarely perfect pulse response, if you can live with the finite input impedance of bipolar OPs, you should try CFB OPs, at least for comparison.

 

[crutschow]

.............................

Is it typical that the stray capacitance cannot be reduced to where a feedback cap is unnecessary?
The datasheet gives no guidance to add capacitance; in fact it warns that even a few pf of capacitance on the output can cause problems with phase margin.

It's very difficult to reduce the stray capacitance to where it has no effect. Even a few pF can be a problem. As noted by FvM, a current-feedback type op amp is less affected by stray capacitance and may work better if you are looking for the fastest risetime without overshoot.

The feedback capacitance you add does not adversely affect the phase margin. They are referring to any output capacitance to ground. To minimize the affect of output capacitance on the response you can add a small resistor is series with the output (outside the feedback loop).