In Reply to: Re: PCB problems please help posted by Darren on 11/28/01 at 11:12 AM:
There's an application note on the National Semiconductor website, AN-214 "Working with high impedance op-amps" which describes the technique for guarding high impedance nodes at an op-amp's inputs from leakage currents on the surface of the PCB. Another technique (mostly useful for prototyping or as a way of testing whether you have a leakage problem) is 'air-wiring', where you bend all component leads belonging to the node up in the air, and solder them together - thus isolating the node from the PCB. Obviously, this was a more practical technique with through-hole components, but it can be done with SMT parts.
But somehow, I doubt that is your problem. Leakage problems tend to show up when the board gets cold, and attracts condensation from the air, whereas I guess your drift is happening in a more benign lab environment.
What I meant about the D.C. current path for the LM324 inputs is that the op-amp inputs go to the base of the input transistors, and with a bipolar op-amp (like the LM324) there is a small bias current coming out of those pins. For the LM324, this current is typically 45nA. If one input was only connected to a capacitor, but to no resistor, the circuit might work for a while until the bias current from the op-amp charges that capacitor to the point where the input voltage gets to the limit of the common mode range. But, from what you said about the connections to the LM324 inputs, this doesn't seem to be the problem.
All this leaves us with input offset drift + high D.C. gain - but with no obvious heat source! I'm not sure what to suggest, other than you've got to work back from the op-amp output which is stuck at zero to track down the problem. What are the input voltages when the output is stuck at zero? One input or the other is likely to have moved away from the expected value. If possible, use a DVM with a high imput resistance, i.e. >> than the usual 10Megaohms to check this. Oh, and the other thing is - are you sure that nothing changed (other than package styles) when you redesigned the board for SMT? (When all else fails, check the obvious!)
Hope this helps,