The circuit shown works great if operated from a single power supply. But I'd prefer to operate it as shown, with a separate supply to the thermoelectric cooler - which requires a high current - and a common ground. When I try that, the circuit goes goofy and fails to switch the MOSFET cleanly. It will work, just not well.

I don't understand what's wrong with using a separate power source to the Drain-Source circuit of a MOSFET used this way. What am I missing?

Some details:
The 12 v. power source is an old computer power supply.
The 10v source is a car battery (with a bad cell, I guess).
The LM35 devices give a voltage that is proportional to temperature, eg. 250mv = 25.0°C.
The TEC is a thermoelectric cooler, connected thru a heatsink to the fan. It turns on when called for by a higher temperature at IC1 than at IC2.
The fan turns on at the same time, but goes off some time later as determined by the RC time constant of C1 and R8.
R3 is 10M, not the 1M shown

 

Do you have a 100nf cap across the comparator supply pins? Also adding a 10ohm resistor between the + rail and the comparator supply pin may help, the capacitor needs to stay as close to the comparator as possible.

 

I don't understand what's wrong with using a separate power source to the Drain-Source circuit of a MOSFET used this way. What am I missing?

You would need a separate wire directly from the source of the MOSFET back to the 10V battery negative terminal.

This is to prevent voltage drop from affecting your circuit operation.

 

You would need a separate wire directly from the source of the MOSFET back to the 10V battery negative terminal.

This is to prevent voltage drop from affecting your circuit operation.

I woke up this morning with this exact issue - and solution - on my mind. The current breadboard wiring is absolutely not up to a 5-10A current without voltage drop. It's wired exactly as shown in your addition to my drawing, but the wires are long and thin. Duh.

 

Do you have a 100nf cap across the comparator supply pins? Also adding a 10ohm resistor between the + rail and the comparator supply pin may help, the capacitor needs to stay as close to the comparator as possible.

So far (at breadboard) I don't have a bypass cap, and it seems to work OK. The LM339 datasheet says: "It is usually unnecessary to use a bypass capacitor across the power supply line." But I may add a cap when I go to PCB.