Input filters for current sense amplifiers

Don’t you love it when things are consistent? I was reading about input filtering for current sense amplifiers. Compare and contrast figures from two different application notes:

The left one is from ST: How to filter the input of a high-side current sensing. The right one is from Maxim: Performance of Current-Sense Amplifiers With Input Series Resistors.

What’s funnier is that it seems as though the writer of one app note copied the other, since some of the language is more or less identical. But they swapped the “common mode” and “differential” labels for the two circuits…

I’ve more or less convinced myself that the ST one is right, and the copying and confusion was done by Maxim, but electronics is hard enough without this sort of mixup!

Yeah, Maxim got DM and CM swapped. Fig 2 does nothing for CM because CM will be the same on both sides of the cap.

This kind of filtering always introduces some error, depending on opamp bias current and resistor tolerances. Maybe better to filter in software if you can, or after the amplifier.

1 Like

Thanks. Software filtering sounds like a better idea. I know what I’m doing with software!

Keep in mind too that many Instrumentation Amps have good CMRR already. It really depends on what you are measuring and in what environment.

@JuliaTruchsess is right too, extra passives introduce other non-idealities into the system. For example, we worked on a Thermocouple sensor that was riding on 480V common mode. Learned the hardware what a Voltage coefficient of resistance is and why you should either have it matched or really, really low. = Search for Bulk metal foil resistors.

Caps are much much worse to deal with in odd scenarios.

The main reason that you would want to do this in hardware is to anti-alias.

Is there ever any reason why you would want to build a non-integrated instrumentation amplifier? You see circuits in some places showing the classic 3 op amp instrumentation amplifier, but matching the components to make the gain predictable seems either difficult or expensive or both. If all the matching-critical components are on the same die, presumably it’s much easier to match them, either by trimming or just intrinsically as a result of being made on the same piece of silicon. Or is that totally wrong, and it’s easier to match external components? (I’ve seen these matched resistor packages from Analog that are intended for this, but the prices for those are quite something. Here you can pay $27 for four resistors, for example:

I think what @LukeBeno was referring to is using a generic integrated IA as opposed to a “high side current monitor” IC, which is basically just an IA with a wide CMR and good CMRR.

1 Like

Yes, something like INA199.

When I was talking about matching, I was referring to any passive components that are tacked on for filtering, etc.

Generally speaking, monolithic instrumentation amps are significantly better than discretely built ones, especially for CMRR because the silicon process lends itself very well to component matching, temp matching, etc. Silicon for example has terrible absolute resistance tolerance but very good resistor to resistor matching.

1 Like

Watch out for nyquist. You cannot filter in SW alone, it will be downsampled

Also, the opamp can rectify CM noise so any filtering in SW can create errors