More current source ppms

[Terry Bites]

https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1029145/faq-how-to-achieve-0-01-constant-current-accuracy-for-dc-power-supply?HQS=sys-ind-tm-precisionb-exexnl-ta-ElectronicDesign-wwe&DCM=yes&dclid=CJaxh7vx1vMCFeenUQoddQgLiA

[David Hess]

So 100 ppm accuracy over some combination of line, load, time, and temperature?  Was this a problem which needed solving with a better instrumentation amplifier?

I think Texas Instruments was looking for a way to showcase their INA188 solving a problem through brute force which was previously solved with a different circuit topology.

[guymo]

I think Texas Instruments was looking for a way to showcase their INA188 solving a problem through brute force which was previously solved with a different circuit topology.

As a relative beginner I am intrigued by this: what alternate circuit topology would be the standard / better solution? I've seen this exact topology in a lot of places recently (looking at SMU schematics for example) but perhaps I'm overlooking some key differences.

[David Hess]

I think Texas Instruments was looking for a way to showcase their INA188 solving a problem through brute force which was previously solved with a different circuit topology.

As a relative beginner I am intrigued by this: what alternate circuit topology would be the standard / better solution? I've seen this exact topology in a lot of places recently (looking at SMU schematics for example) but perhaps I'm overlooking some key differences.

The common alternative is to move the error amplifier for the current control loop to the current shunt, so no instrumentation amplifier is required.  Now the only limit for the common mode rejection is the native common mode rejection of the operational amplifier itself, and no resistors need to be matched.  Another advantage is that with only one stage inside the current control loop, frequency compensation is considerably easier.

The disadvantage is that now the control voltage for the constant current must be referenced to the current shunt, which means either using an instrumentation amplifier to shift it from common to the voltage at the current shunt, or converting the voltage to a current and then using that.  The later is very common and again, does not require any resistor matching for good common mode rejection.  If an instrumentation amplifier is used, then it is *outside* the control loop so it does not complicate the frequency compensation.

Another disadvantage is that the operating range of the current control operation amplifier must include the entire output voltage range, but this can be handled with either a floating supply, which is common in older designs, or through bootstrapping the supply voltage to that part of the circuit.  If your power supply only operates up to 20 to 30 volts, then this is not a problem.

An SMU is more of a special case since it will likely have a larger output voltage range than a power supply, and will likely support bipolar current and voltage, so a differential or instrumentation amplifier based design might make more sense, but I would still try to do without it.  Designing a precision voltage to current converter (transconductance amplifier) without the disadvantages of a difference amplifier is an interesting challenge, but that excludes the Howland current pump which again requires matched resistors.  I am not sure how I would handle this but I have seen some interesting implementations.

[guymo]

Many thanks for this -- plenty to think about here. It has me looking inside my bench supply to see how it handles current limiting.

[Terry Bites]

I'm not seling it as a good thing. But noteworthy as a power supply design.
Just how good can we make a current source?
Everything conspires to lower your ouput impedance.
Can we get any speed out of it?

I'm trying to resist the urge to look too deeply down the black hole these days.

[David Hess]

Current sources can be made very good, but most power supplies do not usually require that kind of precision.  Source measurement units are much better but sure get complicated.