Schmitt Trigger Comparator Output Not Snapping Cleanly, Output Varies 5V to 4.5V

Hi. I'm still having a lot of trouble with this circuit, which it supposed to take a varying input, and snap on and off at certain levels.

Unfortunately....! Careful adjustment of the pot at the switchover points allows various voltages at the output.
Going up (output going down), the output is variable between 5v and 4.5v. Below 4.5v it "snaps" down.
Something similar, but opposite, happens at the output-up transistion.

What am I doing wrong?  This shouldn't be rocket science!
Thanks,Graham

Hi Graham-Your schematic looks fine.  The things I can think of are: First, check the 100K feedback resistor and make sure it is actually going to the non-inverting (+) input, not the inverting (-) input.  Second, you don't show a bypass capacitor on the schematic.  Make sure you have a ceramic capacitor, maybe .1uF or so, connected very close to the chip between Vcc and ground. Without it the chip could be oscillating.You might also double-check pin 4 to make sure it is actually connected to ground, and pin 8 to make sure it has 5 volts on it.-Rick

Graham this may have something to do with using a comparator, which does not provide a high voltage output, only a transistor to ground.If Rick's suggestions don't sort it out, try the circuit with an op amp instead and see how you go.

One other thing- the LM393 is a dual comparator. Make sure the inputs of the other comparator are defined- connect one to gnd and one to 5V.You don't mention the input voltages it switches at- this might give another clue. The upper voltage is when the upper 10K has (100K+1K)  connected across it (open collector output-as if nothing is there). I make this to be 2.62V. The lower voltage would be where the lower 10K has 100K connected across it (the comparator grounds the  junction of the 100K resistor and the 1K resistor). I make it to be 2.38V.

Graham,I think the problem is that both legs of the comparator circuit are connected to the same voltage source. As Aubrey mentioned, if the input is 5V the upper threshold will be about 2.6 V and the lower will be about 2.4 V. You should be able to adjust your potentiometer to cause the circuit to switch.The problem is that the threshold voltage will change as the input changes. So when your input changes, the thresholds change. Since both sides are resistor dividers, they will tend to change at the same rate which causes the circuit to not switch.
The most reliable fix for this is to add a zener diode into the side with the potentiometer. If you expect the change in voltage to be rapid, you could also add some capacitance at the junction of the two 10k resistors.

I suspect that your LM393 is working correctly. Your "Careful adjustment of the pot at the switchover point" forces the LM393 to hover in its linear range, where the output voltage will be between high and low. The gradual output shift from 5.0V to 4.5V is analogous to the flipflop metastability that Max Maxfield discusses in his articles here on EEWeb.
As Rick Curl writes, be sure you have a .1 uF capacitor across the power and ground pins of your LM393. Look very carefully at the LM393 datasheet, where you will find significant details:1. The LM393 requires 5mV of overdrive to reliably switch.2. The supply voltage range is from 2V to 36V, so 5V is OK.3. If you are not using the other comparator in your LM393, ground +input and -input and output (This is specific to the LM393, and may not apply to other comparators). The output is NPN open-collector, so grounding the output is OK.4. The inputs are PNP Darlington differential pair, and the input common-mode range is from ground to Vsupply-2V. The common-mode voltage is the average of the voltages on the two inputs. Measure your 5V supply. If the supply is 5.0V, then the input common-mode range is from ground to 3.0V, not much above your designed threshold of 2.6V.I disagree in part with Elizabeth Simon. Yes, the threshold voltage changes as the input voltage changes. However, the threshold voltage changes in the opposite direction to the input voltage change, increasing the effect of the input change. This opposite-direction change is the intent of the Schmitt trigger, to give hysteresis. You report the output "snaps" below 4.5V. This is the positive feedback working.

Thanks for the thoughtful responses!I'll work my way through them, and report.Graham.
PS: Ms Simon, I take it you're suggesting a zener between the "-" input and the wiper of the pot? What sort of voltage, and which way round? Sorry to be dim, but what would it do? Thanks!

There a a few ways you could use a zener in this circuit. One is to put a 2V (or so) zener in place of the lower 22K resistor. The other is to use a higher voltage (3.3V or so) across the 10K resistor divider with a resistor between it an the 5V supply. You will have to adjust the resistor divider values to give the correct threshold.One thing to watch out for is that you have enough current through the zener for it to give you a reliable zener voltage. See my article on zener diodes for more information.
https://www.eeweb.com/profile/elizabeth-simon/articles/how-to-read-data-sheets-zener-diodes

Graham, I don't think you are doing anything wrong. Your LM393 is behaving as I expect it would.From the LM393 datasheet, the LM393 gain (when the output is far from saturation) is a minimum of 50 and typically 200. This is rather low gain for a comparator, and explains Graham's "the output is variable between 5V and 4.5V". As Graham adjusts the pot, the LM393 comes out of saturation, and the LM393 gain goes from zero (when saturated at 5V) gradually up to 20 (at the 4.5V point where the output "snaps"), and continues to rise and then fall until the snap finishes with the LM393 saturated at the other output state (at which point the gain is back at zero).

Graham, if you are connecting the LM393 output to a CMOS gate powered from the same 5V supply, all will work well. 5V to 4.2V are good logic high inputs to a 74HC04 inverter, and 0V to 0.9V are good logic lows, so your variable outputs above will always be seen as good 1's and 0's.

Graham - one of my suggestions above was to use an op-amp instead of a comparator. If you do that you won't  require the 1K pullup resistor on the output - opamps have push/pull or sink/source outputs.