Create exact inverted signal

Question

I need to create a differential signal using low power single ended op amps(not a fully differential op amp).

Obviously an inverting op amp gets the signal inverted BUT the use of inexact resistors produces a differential error that is not acceptable. I do not have the equipment or quality components to get decent results.

Is there an op amp topology(can use more than one op amp) that can take a signal and produce an inverted signal that is nearly perfect and doesn't dependent on external component values?

For example, a non-inverting buffer is "ideal" as far as I'm concerned. Since there are no external components the gain is exact. Unfortunately the inverting buffer requires resistors which create a variable gain(either due to tolerance, temperature, or whatever).

Basically the circuit topology I'm using is one op amp for gain/drive and another taking it's output to invert the signal. This produces a differential signal but the inverting op amp doesn't produce an "exact" copy which causes problems. (Here the issue, again, comes from the external gain resistors not being exact)

Maybe there is some way to combine the two stages in to one in some ingenious way to get a fully differential output?

Also, the output impedance of the signal to the op amps is a not very high or very low(around 1k-10k) will effect the accuracy of the inverting op amp configuration.

(I'd use a fully diff. op amp if I could find one for my applications... but they don't seem to exist. (low power < 500uA, high voltage(>= 36V)))

Answer 1

This is a kind of silly question, but not in a bad way, it can be used as a teachable moment.

The issue you have isn't so much that resistors can be poorly matched, it is that you're assuming op-amps are ideal. That is, "the only source of error is from the resistors, so if I get rid of those I will reach nirvana".

Even if you get matched resistor made from un-obtainium you'll not achieve your (unstated - or rather unparameterized) goal.

In reality all op-amps have issues with input current balancing, input voltage offsets, frequency dependant behaviour etc. Your solution must account for all of these.

And the parameters of your solution space will dictate what can be ignored and what can't be corrected. Hint -> this means you need to give more details as to what is needed, BW, etc.

One classical technique is to use a chopper stabilized amplifier which switches back and forth and self corrects for op-amp and in some topologies, even the resistor network non-linearity.

Another technique is to use switched cap techniques.

Answer 2

I don't think there is any way** that does not reduce to having a tightly matched pair of resistors somewhere.

Of course you can buy those precision-trimmed resistors as part of a chip and pretend they're not there (for example, use a dual instrumentation amplifier with G=1, inputs connected in inverse).

You can buy a pair of 0.05% ratio matched resistors inexpensively (eg. Susumu) or use a trimpot and adjust 0.1% resistors.

The LT5400 has resistors matched in ratio to 0.01% (but poor absolute tolerance). Or if your shoulders are broad and pockets deep, there is the exemplary Vishay Z-foil series that have 0.01% matching and superior TCR tracking.

** That doesn't involve precision transformers or flying capacitors.