Yeah the overall design is just improvised, yeah I just made a quick current source, (I'll be using the one in the DMM) but it was the problem. I had the top divider resistor R6 connected to 15V, not GND. I still don't see where all the current goes, I see some diodes inside the LM393. But the moment I put in certain DUT resistor's, that will make the LM393 sink all that extra current, and also my PSU is lowering the neg, rail, so when it would sink 20mA, it would be from -6.2V. I thought maybe all that current goes across the 470R I had between the inv. amp and LM393 Inv pin, so I tried a 4.7k, and same thing happens, and Vout from the inv. amp stays about 13.6V the whole time I was checking.
However, it seems to work ok when wired right, IDK if it function exactly as I think it should overall, (the last version with just op-amps, wasn't so good) but this one can also tell the difference between 1.2R and 2R no problem.
But I didn't want to cheat and look up a real schematic from a DMM, but yeah I'll look up what you're saying about Q3.
OK so the extra current is going across my DUT, I didn't realize that's what you meant, and that's why it happens when I hooked up the DUT.
Yes so for now I added a reverse biased diode on the base of Q3, I forgot that resistor will still be low on both ends when the 393 goes low. And I changed the hysteresis a bit, as with 510k I could get some near 1MHz oscillations when adjusting the pot for triggering on 1.2R. With 255k, I don't see any osc. any more, at least not for small DUT's.
I remembered about emitter degeneration resistors, and yeah the temperature runs away. So I used a basic current mirror with 5k emitter resistors on both. For a 4.7R, I get -4.7mV and ~ 4mV from the buffer.
I added a 555 circuit for the buzzer, and that gave me a huge noise problem on the buffer output, over 10Vpp at times. I'm not used to that, but after playing around with adding caps, and having a better current source, my ebay LM358 seems pretty good again. I went back to 510k Rf on the 393, and haven't had any issue. It seems to work down to about 2 ohms, and so far seems repeatable. So that's good enough for now.