Why aren't these instrumentation amps working for linearly driving a generic MOSFET?

Question

If you've read my previous posts, I would say now that I've pretty much given up on the idea of parallelizing op-amps. I've stumbled onto instrumentation op-amps and decided they were good enough for my purpose.

So here I am trying out different ones that are in LTSpice's standard library.

With their respective \$R_g\$, I've set the gain to 2 for all of them. Yet none was producing 12 V at \$R_{load}\$.

LT1167 produced 19.965992V, AD8422 produced 2.6913893V, and AD8221 produced 19.916515V.

Making the assumption that all in-amps in the market have unity gain if \$R_g\$ is left out, that is exactly what I did and only the AD8422 produced some amount of difference: 19.965992V, 1.0792856V, and 19.916515V, respectively.

Answer 1

1. None of these chips are designed to operate on a +/-24V supply.

2. Some of the circuits have 0V between VDD and VSS, so will not work.

3. All of the circuits use a topology that requires high gain to function. You want to use an op-amp, not an in-amp. The in-amps will, at best, offer 1k or 10k gain. That's lousy compared to most any op-amp, even a 741.

   In fact, all of the circuits you show have gain a couple orders of magnitude lower than the Philbrick K2-W op-amp. And that op-amp used just four triodes to give open-loop gain in excess of 10,000.

To get those circuits to work, replace the in-amp with an op-amp, and power them from a recommended operating voltage.

I've stumbled onto instrumentation op-amps & decided they were good enough for my purpose.

You're misunderstanding something, since in-amps always underperform single op-amps in terms of DC specs, noise, distortion, etc. For any in-amp you choose, you can always find an op-amp that will perform better if you don't need the differential input function. And the circuit you're showing does not need differential inputs.

The question is a classic XY problem: You're showing a Rube Goldberg-esque solution to a problem with a very simple solution. Please update the question to show the actual problem you are trying to solve. Not some approximation. Just tell us what you want to do.

Also, please understand that precision circuits do not work on solderless breadboards. If you need ultimate in DC precision and low noise, you'll be designing PC boards for every prototype. You can get excellent in-amps and op-amps to perform like crap in solderless breadboards, or poorly designed solder-ful protoboards.

And also: it takes quite a bit of experience to be able to derive any measures of "precision" from SPICE simulations. It requires deep understanding of the various non-ideal aspects of the circuit being modeled, and the ability to model those non-idealities adequately so that the SPICE output will have sufficient resemblance to bench measurements.

Basically, high precision isn't something you just simulate or prototype without some prior experience. Otherwise, unfortunately, it'll be a case of "garbage in, garbage out" whether you're putting together physical circuits, or their SPICE models.