Once in a while you see a circuit that has an opamp with a humongous capacitor from its output to ground, not one of those cool C-Load National things, just an ordinary opamp. I've even seen this in Analog Devices and other "respectable" appnotes.
Today's problem is that I have one of those notoriously fickle National LM45 temperature sensors in a temperature controller, used as the thermocouple reference junction temp sensor.
I got lazy and connected it directly into the ADC mux, and the charge injection from the mux makes the LM45 go nuts for longer than the available settling time, a bit under a millisecond. So I wrote an ECO to add a 1K resistor to ground, which fixed some units and made others break into oscillation. Grrrrr.
So, not wanting to cut traces if I can avoid it, I was thinking of hanging a biggish tantalum cap from the LM45 output to ground, 47 uF maybe. If the resulting pole crushes the internal frequency compensation (if it actually has any, which is problematic) the whole mess ought to be stable. The charge injection thing will sure be fixed, and there won't be much oscillation at the output, whatever happens inside.
Incidentally, the resistors from the output emitter to ground add up to about 40k. Measured small-sig AC impedances of the unloaded LM45 are roughly
1 KHz 40 ohms? hard to tell on current setup 10k 200 r 20k 400 r 50k 1k 100k 2k 200k 3.4k