why remove the ground plane under the input pins of an opamp?

[apchar]

I've read in several sources that you should etch away the ground plane under the input and output pins of an opamp to avoid 'capacitive coupling to ground." This doesn't make sense for a lot of reasons, especially in wideband surface mount circuits.

1. The entire signal trace has capacitive coupling to ground. Why would that last bit to the input pin be any different? In fact the opposite seems true. Coupling to ground is a lot better than coupling between pins. Pin-to-pin coupling is far worse without a ground plane directly underneath to confine the field lines. Pin-to-pin coupling will cause instabilities. The worst thing capacitive coupling to ground might do is cut into your bandwidth.

2. A signal trace without a ground plane is just a loose wire, with all the inductance of a loose wire near the worst possible place - the juncture of the feedback line & input of the opamp (for an inverting amp.) Stray inductance is an invitation to instability.

3. I'm from the RF world and ground planes everywhere are critical, especially in microstrip (signal trace over a ground plane) circuits. That's how you control the impedance of the line, by balancing the lines capacitance with it's inductance. I know you don't usually worry about impedance matching at the input of the op amp but at least the mismatch is real (resistive.) Removing the last bit of ground plane will only make the situation worse by introducing a little reactance.

Everything about this rule is counter-intuitive.

So why do it?

 

[crutschow]

Well, intuition doesn't always work. The reason for removing the ground plane is simple. Any capacitance at the inverting node (summing junction) of an op amp creates peaking in the high frequency response which must be compensated by a cap across the feedback resistor, which reduces the available bandwidth. If you have a Spice simulator you can easily demonstrate that by adding a few tens picofarads (the higher the circuit resistances, the greater the effect) to the summing junction in an inverting op amp configuration.

Thus, if you want maximum bandwidth, you want to minimize anything that creates capacitance at this node, including the ground plane. Any components connected to the summing junction should be placed as close as possible to the summing junction, with short trances, and no ground plane under the op amp inverting pin any trances to that pin and any component ends connected to the summing junction.

The small inductance increase in the short traces at the summing junction due to removal of the ground plane will have little effect on the op amp performance at the usually op amp frequencies.

Now, if you are not interested in maximum bandwidth from the op amp circuit, then there may be valid reasons to maintain the ground plane, (shielding, crosstalk, etc.).

 

[apchar]

right you are. I spiced it and see what you mean. I still find it counter-intuitive that an RC filter at the input would cause an increase in gain (at the knee.)
Thanks for the reply.

 

[crutschow]

right you are. I spiced it and see what you mean. I still find it counter-intuitive that an RC filter at the input would cause an increase in gain (at the knee.)
Thanks for the reply.

Your just need to direct your intuition to the right spot. With a little thought experiment you can see how the peaking happens.

In an inverting amp the feedback resistor has to supply a current equal to the input current at any frequency. At the frequency where the stray capacitance starts to shunt a significant portion of the input current to ground, the output voltage will have to increase to now supply both the input current and the current being shunted to ground. Thus you get a rise in the AC output response with frequency. For usual values of stray capacitance this often occurs near the normal roll-off frequency of the op amp so this will look like a peak in the response. If it's a high frequency op amp this can cause oscillations rather then just peaking.