Why doesn't my op amp circuit produce output as low as the simulation?

Question

I'm working on a design to test/qualify vintage AT power supplies. I want to test all (+5v, +12v, -5v, -12v) rails. I want to be able to measure from -20% to +10%. I've got a DAC (built into my microcontroller for now) which can measure from 0v to 3.3v. To build in safety margins I want the outputs to be +/- 10% from that or 0.3v to 3.0v.

I'm intentionally stretching my skills a bit: they're strong with digital circuits but less so in analog. I believe I should be able to use an op amp here, and a quad op amp chip will match the four rails I want nicely. The primary reference I'm using is "Op Amp Gain and Offset Page" which can mostly automatically suggest the exact circuits I need to set up, and the resistor values to use. For three rails, this has worked great! I've even (LTSpice simulated and) breadboarded the results and with only minor tweaks I get exactly what I'm hoping for.

Except with the +5v rail. For that, it suggests case 2 and the case 2 page is broken. If I download and fix it, I get completely broken results (all solutions have huge errors -- maybe I didn't fix it perfectly?). If I instead use Designing Gain and Offset in Thirty Seconds I get the same m and b, and a resistor network. That's (I'm using a +5v reference, though I could be flexible a bit there if it helps)

completely wrong, per simulation? Output is way too high, above the input rather than below. (This is probably a hint that I'm "doing something wrong", but I don't know what.) Somehow I figured out that "case 8" from the earlier-linked page seems to work, so I tried that and it gives:

An acceptable-looking solution. And the simulation

matches perfectly. But when I breadboard it (edit: with GND at 0v and V+ at 12v) an input at about 4.2v produces about 0.7v output and lower inputs never produce a lower output. I'm using potentiometers for my resistors, and a DIP LM324N quad-op-amp. These same components have worked great, producing the expected output range, for my other three input ranges (and their matching appropriate circuits).

Am I running into a real-world op amp limitation that the simulation does not cover? On a hunch, I repeated the "case 8" solution with an output range from 1.0 to 3.0v -- that works as expected on the breadboard, and seems to be a workaround I could select. But I'd like to use the full output range down to 0.3v.

Answer 1

The LM324 cannot sink current greater than (typically) about 50uA without some substantial output voltage (approaching a diode drop) and your topology requires it to do so. It would have to sink Vref/(Rf+Rg) to get the output to zero.

You can make the LM324 outputs range from 1V to 3V and it should work with a 5V supply.

Answer 2

Am I running into a real-world op amp limitation that the simulation does not cover?

The op-amp you chose probably needs to have a small negative rail so that the output can adequately reach 0 volts.

Even so-called rail-to-rail op-amps struggle to get closer than a few tens of milli volts and, regular op-amp outputs usually can't get within a volt of the power rails.

You also need to ensure that the input voltage range is within the power supply range of the op-amp and, in particularly, for the LM324, it's output limits at around 3.5 volts on a 5 volt positive supply.