Inverting buffer with op-amps

I know that it's easy to make a unity gain buffer with an op-amp (as a voltage follower): simulate this circuit – Schematic created using CircuitLab

I also know that it's easy to make an inverting buffer with an op-amp (an inverting amplifier with $R_1 = R_2$): simulate this circuit

However, the accuracy of this inverting amplifier depends on the precision of $R_1$ and $R_2$ - if they're not closely matched, the output will be a bit different from $-V_{in}$.

Is there a way of making an inverting buffer with an op-amp that doesn't depend on the precision of these resistors, like the voltage follower? Is it a better idea to get higher precision resistors?

No, there is no way to make an inverting buffer with just an op-amp that does not depend on the resistor values. You can get resistors with very fine accuracy and stability (at an equally impressive price) or you can get networks with matched (in value and in temperature coefficient) where the absolute accuracy may not be so impressive but the ratio is tightly controlled.

There is a way to invert a signal without accurate resistors- the so-called flying capacitor method, but it's fairly complex and resistors are a better solution for most situations down to ppm level accuracy.

• Great - this covers exactly the amount of real world practical stuff I wanted. – Greg d'Eon Jan 9 '15 at 14:41
• So the question I've been trying to figure out is: why not? Since R1 and R2 can be any value, why can't that value be zero? (I'm sure there's a reason, just not found it yet) – jgalak Dec 23 '17 at 20:13
• @jgalak The gain is -(R1/R2). Aside from the mathematical improbability of 0/0, for any positive number the percentage error in either R1 or R2 shows up directly in the gain error. So there is no advantage to very small (or very large) values in terms of ideal error. Very small values load the op-amp and introduce additional issues. Very large values cause additional errors due to bias current and leakage and maybe Johnson noise. – Spehro Pefhany Dec 23 '17 at 20:29
• But for a non-inverting amplifier configuration, the formula is also G=1+(R2/R1), yet in the buffer configuration 0/0 works fine. So why is it ok there and not with an inverting buffer? – jgalak Dec 24 '17 at 0:51
• Huh. And I just got it. 0/inf trends to 0. 0/0 doesn't trend anywhere. Ok, never mind, got it! Thanks anyway! – jgalak Dec 24 '17 at 0:53

One possibility. They used to make a few specialized op-amps that had differential outputs, a positive and a negative, likely for driving differential line pairs. I've never used one and I don't recall the part numbers. But I assume if you were to connect the positive output up as a voltage follower the negative output would be equally negative.

Note that even a voltage follower op-amp arrangement is not perfect. There are internal gain spec's and small offsets, while small, can results in outputs that are not perfect one to one of the inputs.

• This is a pretty clever idea, but I'm not sure that it would be useful in practice (mainly availability of parts) – Greg d'Eon Jan 11 '15 at 14:32

Adding to Nedd's answer, the precision unity gain differential amplifier ICs you are looking for are of the type INA105, DRV134, THAT1240 etc. There are several qualities, makes and manufacturers and of course a varying price range. The circuit in question would be the one below, from the INA105 datasheet. Practical experience often deviates from textbook ideal particularly in the analogue domain. If you're using good quality resistors from the same manufacturer and the same batch the accuracy between individual resistors is better than the quoted tolerance of the nominal value.

If you are wanting a unity gain inverting op amp in practice using a general purpose device like an LM324, just calculate the resistor values to give calculated gain G = -1.009 For example, Rin = 218k, Rf=220k. This will give you actual measured gain of G = -1 with your voltmeter.

I've never used anything more exotic than an LM324, i've used this part to interface to sensors requiring microvolt levels of accuracy with no problems making a rock steady output. Your difficulties begin trying to get your microcontroller's AD converter stable.