Understanding precision current measurement

[tmadness]

I have been thinking a lot about precision current measurement, typologies and what goes into a well designed system.
Specifically high side measurement, mostly around DC, perhaps expanding into single digit KHz. Say, 5.5 Digits resolution is the target.
A spec I have come across a lot is CMRR, especially for difference amp based typologies. CMRR tends to be the Achilles Heel of a difference amp based system, higher precision requires higher CMRR, ... requires highly matched resistors that have a exponential cost curve.
Yet I see these things everywhere in precision test equipment, Keithley SMUs, consequently djerickson's DIY-SMU are two examples.
Am I not getting something about CMRR/ Diff amps?
In my research I came across trans-conductance based typologies, eg  INA 290 or LTC6102. The LTC6102 has definitely peeked my interest, I believe that it should work adequately for my stated needs. But, I am a novice on this subject, am I missing something? what's the gotcha(s)?

EDIT:
After a bit more searching I found the schematic of 41420A's current stage (Thank you xDevs https://xdevs.com/fix/a4142b/ ).
This one has a difference amp but with with tweaks. A cursory analysis says CMRR is still an issue, this time determined by the mismatches in the op-amps and gain dividers, and overtime, separate drifts. I do not know the magnitude but is this better?

[voltsandjolts]

For low bandwidth current shunts amplifiers, auto-zero amps are worth a look, at least for low-resistance shunts where input current noise doesn't matter too much.

[Kleinstein]

The current ouput current sense amplifier need the high side to be at least some 5 V higher than the final ouput level.
Some of the types also don't wirk for very low currents, as they may have an interanal bias current in the nA range or possibly higher.
Usually the current measurement is also only 1 polarity. This can be a problem with low currents and superimposed hum.

The HP 41420A seens to have the first amplifier stage powered from the high side / output side. This reduces the demand on the CMR, the first stage does not see much common mode votlage.

[tmadness]

The LTC6102 has really low input current, the INA 290 not so much. The voltage requirements are bothersome, but I can deal with it. I remember the INA193 had a specific requirement that the shunt voltage not drop below a certain level as the amp is not linear in that range (18-22 mV I think). But the LTC6102 does not seem to have that specific requirement.

 voltsandjolts you point got be thinking about a topology like HP 41420A, but with zero drift amps. A system like this would have < 1 uV error I think. something to think about.

Also the revered HP 3485A has a a diff amp (https://xdevs.com/doc/HP_Agilent_Keysight/3458A/doc/3458A%20CLIP.pdf), maybe I should not be too hasty to dismiss the diff amp topology.  I don't understand how they are able to maintain common mode range though.

[Kleinstein]

For the current measurement it also makes a difference if this is about small currents (e.g. µA or pA range) or large currents (e.g. 1A or even more). The large currents have difficutlies with the shunt heat. The low currents fight with leakage. The CMRR is usualyl one of the smaller problems, except for special circuits. Current measurement can be over quite a large range.

The usualy DMMs have no real issue with the common mode voltage, as the shunts are with one side at ground. If at all they have a tiny bit of drop due to trace / wire resistance as common mode voltage. So the amplifier can be often just a simple one sided amplifier, not even need for a differential amplifier.

The LTC6102 still only works with positive currents, at least in the normal configuraion.

[tszaboo]

For very precise current measurement, forget those shunt amplifiers. They have a relatively large offset voltage, and bias current. That is not an issue if you want to measure "how much power goes into my X equipment", but they are not really suitable for very precise measurments.
For that, you need 4 wire shunt, RC common mode filter, and a standard 3 opamp instrumentation amplifier. Drive the ref pin of the INA with another opamp, or connect it to ground. Select a gain resistor that gives you the desired gain, use a thin film resistor for it at least.
Then if you have a working proof of concept like this, then you can start worrying about thermocouple effects, cooling the shunt, and so on.

[David Hess]

A spec I have come across a lot is CMRR, especially for difference amp based typologies. CMRR tends to be the Achilles Heel of a difference amp based system, higher precision requires higher CMRR, ... requires highly matched resistors that have a exponential cost curve.

Am I not getting something about CMRR/ Diff amps?

Nothing requires using an instrumentation or differential amplifier for current sensing.  An alternative makes the measurement at the current shunt with the amplifier so the common mode voltage does not change, and no resistor matching is necessarily.  If a ground referenced measurement is necessary, then the current sense amplifier can mirror the current at a lower level to drive a grounded resistor.  Then the matching of the two resistors creates only a gain error which is much easier to handle.

The disadvantage of this method is that it often requires a floating power supply to power the current sense amplifier but that may be a small price to pay for the improved performance and lack of a need for matched resistors and CMRR trimming.

[Doctorandus_P]

If you are thinking about different topologies, then think a bit about adding a small floating power supply, measure the current directly at the shunt and send data through some isolation barrier such as optocouplers.

Less stuff between the shunt and the ADC means less opportunities for measurement errors.