External noise resetting opamp integrator

[Galaxyrise]

I've breadboarded a differential integrator as on http://www.ti.com/lit/an/sboa092a/sboa092a.pdf page 59.  The integrator is relative to the negative supply of the opamp. It works great.

I have a really old fluorescent desk lap with an "on" button and an "off" button. Hold the on button for a while, let go to start the lamp.  Press the off button to extinguish the lamp.

When I press the off button, the integrator drops to 2.something V if it was currently outputting more than that.  Not quite as low as it can go (which is around 0.5.)  A 100k resistor and the op-amp input leave one leg of the capacitor at a very noisy spot, but I was surprised that it had such a strong effect!  Other noise sources (turning the lamp on, power cycling my other desk lamp, which is even noisier) don't cause the circuit to hiccup at all, so I suspect there are things happening that I don't understand. So far, I've tried putting ferrite beads on the resistor legs, but I haven't got very many other ideas... I think I don't know enough about what's wrong to do more than try random things.

Any tips how to mitigate this? 

[c4757p]

I don't like how that circuit puts one end of each capacitor at a high impedance point. I'd rather use a differential amplifier to subtract the signals, then a standard integrator to integrate the output. Design it so that the side of the op amp not connected to the capacitor (noninverting if it's an inverting integrator, inverting otherwise) is at low impedance, and then use that low impedance node to drive a guard ring around the high impedance (capacitor-connected) node.

When breadboarding, try soldering the wires together and running them through the air. Leakage in the breadboard is a bit high.

Breadboarded circuits in general are more susceptible to interference. Do you have a nice, big decoupling capacitor on the board? God knows what the leads to the PSU are picking up...

[Galaxyrise]

Thanks for the input!

I have a 10u electrolytic on the breadboard where the power connects, and another 10u across the opamp itself.  The problem might be worse without them, it's hard to say.  Is 10u not big enough?  I didn't arrive at 10u via any process other than "I have a pile of these."

Do you mean bending that pin of the IC so it's off the breadboard and then soldering to it? Or take that IC off the breadboard and soldering to all its pins?

Having a leading difference amp would probably make it easier to adjust the gain of the integrator, too, since getting small integrator gains is a pain. But it means a pretty good instrumentation amplifier in this case.  Hmm... it does give me some ideas to play around with!

Do guard rings even make sense on a solderless breadboard? Just wrap wire around things? Ultimately the circuit will be soldered up and in a metal box. I'll look into guard rings for that build. I wonder if I should just avoid using that desk lamp until then and trust the final build won't be susceptible.