Given an LM393 with one comparator set up as non-inverting and the other set up as inverting, and both sharing a reference voltage and Vin, what can be done to prevent the outputs from jittering when Vin is very close to Vref?

I have added 3k pullups, and 10M of hysteresis on both comparators, but it didn't help.

Thanks.

 

Don't know exactly what your circuit looks like, but the reference may move around a bit when they switch. You could try a resistor between the reference and the input on each and the same with the input signal. Might also decouple the reference and the IC power.

 

I built the circuit without a schematic and was too lazy to draw it, but I realize it is needed.

 

I hate having the jitters!

After looking at your circuit I think you are making an adjustable window comparator... I bet the oscillation is coming from the LED turning on and loading the 9V supply... What is interesting about this is that I bet you see the oscillation on one comparator and not the other. So here is a few things to do

1. Make sure you have a decoupling cap on your ic.
2. Remove the leds and see if you still have the oscillation assuming you're seeing the oscillation on the scope.
3. Add filter caps from the trim pot wipers to ground to filter out high frequency noise
4. Reduce your feedback resistance to provide more hysteresis.

Note that in most comparator applications a precision reference is used... Check out the lt1009. It's a pretty common reference.

 

Since you already have loads on the open-collector outputs, you do not need R5 and R6.

 

Add a resistor (maybe 100K or more) between each pot's wiper and the junction of the respective 10M resistor and - input. With no resistor there the hysteresis will vary from little to none depending on the pot position.

 

Here is the final circuit. There is very little (seemingly no) oscillation, and the hysteresis is small, but sufficient. Thanks for the help.

 

You could also make a zener regulator somewhere below 9 volts and attach the pots to that, the point being to regulate the switching points. The Vcc on the chip can bounce a little as long as the inputs don't bounce.

 

Probably an idea to include a 10K resistor in series with the pot and Vcc (9V).
The common mode range for an LM393 only extends to Vcc -1.5V.