Electronic load precision

[ifonlyeverything]

I'm watching Dave's tear-down of the BK 8601 as inspiration for my own electronic load project, and it appears this device is current sensing with 0.05R resistors and doing the measured current feedback loop with NE5532 op amps.

https://youtu.be/5dNxzzuMTB8?si=5PTiDtJkumnjEm-q&t=730





The device claims +/- 0.05% + 0.05% full scale error over the 0-6A range with 0.1 mA resolution. So for a commanded current of 1 mA we have an error of +/- 3 mA, still pretty good. How are they achieving this accuracy?

NE5532 input offset is ~5 mV. The device is current sensing with 0.05R resistors. There are 10 parallel FETs. (1 mA / 10) * 0.05R = 5 microvolts. Or let's just guess that they're disabling the other 9 FETs at the low end of the range for better precision, that's still only 50 microvolts which is 100x the input offset voltage.

[tszaboo]

Those are not the shunts that give you the accuracy.
https://www.youtube.com/watch?v=5dNxzzuMTB8&t=730s

These are the shunts, those wiggly-wire things.
The small shunts are only used for current sharing in those FETs, and the opamp is to drive the FETs, they are not used for precision measurements.

[ifonlyeverything]

Those are not the shunts that give you the accuracy.
https://www.youtube.com/watch?v=5dNxzzuMTB8&t=730s

These are the shunts, those wiggly-wire things.
The small shunts are only used for current sharing in those FETs, and the opamp is to drive the FETs, they are not used for precision measurements.

Are you sure? I assumed the 4 wire shunts (and maybe the 5W, 1R resistor) were used to measure the total current that flows through all 10 FETs. This would make sense because they have a separate specification for "readback current".

At any rate these wire shunts are likely <0.05 Ohms resistance, making the measured voltage drop even lower for a commanded current in the single milliamp range. How are they reliably and accurately sensing microvolts in the feedback loop?

[ArdWar]

I'm not sure about 5532 specifically, but input offset while can have quite wide spread tend to be quite stable over time and temperature. Once you do initial calibration (and maybe implement autozero) you can get 10s to 100s of uV performance.

[tszaboo]

Those are not the shunts that give you the accuracy.
https://www.youtube.com/watch?v=5dNxzzuMTB8&t=730s

These are the shunts, those wiggly-wire things.
The small shunts are only used for current sharing in those FETs, and the opamp is to drive the FETs, they are not used for precision measurements.

Are you sure? I assumed the 4 wire shunts (and maybe the 5W, 1R resistor) were used to measure the total current that flows through all 10 FETs. This would make sense because they have a separate specification for "readback current".

At any rate these wire shunts are likely <0.05 Ohms resistance, making the measured voltage drop even lower for a commanded current in the single milliamp range. How are they reliably and accurately sensing microvolts in the feedback loop?

You don't need very high accuracy in the feedback loop, because it is in a closed digital loop. They also have 4 shunts for the output, and can range switch. They probably also can selectively use only ie. 1 FET for low currents.

[mtwieg]

I'm certain those large metal element shunts are for measuring the total load current. The smaller axial resistors near each FET are for balancing. It's likely that each FET has its own current regulation feedback loop with its own op amp, and all are slaved to another feedback controller which uses the metal element shunts.

Hard to estimate the accuracy/precision of the shunts at a glance, depends on their construction and of course on how well the cooling works. I'm guessing they're positioned in front of the intake fan in order to prioritize keeping them at ambient. But if you were to put a heat source in front of the intake for this, then its accuracy would suffer quite a bit. So watch where you put your soldering station...

[ArdWar]

Yeah, it being an inner loop is much more likely. It doesn't need to be accurate, just fast. The accuracy is provided by separate outer loop, or in some low cost design digitally from readback value.

[ifonlyeverything]

Is there an example of this inner/outer loop? I'm having trouble understanding how this would work.

[RandomUser]

Something like this?
I had the same question before from Daves Video and this question /the outer loop actually solves the issue with my own load project I think