uCurrent accuracy issue with AAA batteries?

[alexwhittemore]

I just started using my uCurrent in anger again lately for a microcontroller project, but kept leaving it on and killing the battery. In hopes of having a little more leeway, I soldered in a 3xAAA battery holder per the suggestion on the product page. Except something seems to have gone screwy now! I admit, I hadn't checked accuracy before adding the battery holder, but it seemed right on all the ranges, and certainly never far enough out to arouse suspicion. My main target is measuring a ~190uA sleep mode in my project.

Just today, I went to take some more measurements, and nothing seemed quite right, so I started testing the uCurrent itself. With a couple handheld meters in series, a bench supply, and a decade box, I dialed in 5mA and 5uA to test the 10ohm and 10mOhm ranges. With the 3xAAA power source, I measured 4.15uA at 5.05uA actual, for 17.8% (!) error. I also measured 5.28mA at 5mA actual, for 5.6%. So clearly, something is QUITE wrong.

But then when I took those out and switched back to a coin cell, things get WAY better: 4.96mA at 5mA actual (0.8% error) and 5.06uA at 5.02uA actual (0.79% error). Of course, both are a far cry from the .1 and .05% typical accuracy on the spec, so I'm worried I've killed something.

For giggles, I measured the shunt resistors themselves with an LCR meter - 11mOhm and 10.009 Ohm. The first is within the error of the meter, so I have to assume the resistor is functional and accurate. 10.009 is .09% out, but that's within .1%, so I also assume that resistor is fine.

So then, any thoughts on why the error goes to hell with the 3xAAA power source, or on why the error seems to be out of spec now that I'm back to a coin cell?

Of course, I'd really like to just save this as a curiosity for later and get a new one in the mean time... but they're out of stock

[Kleinstein]

3 new AAA cells might have slightly higher voltage. The OP is only specified up to 5.5 V or so - 3 new alkalines might come close to that. It might help to have a diode in series, thus giving an extra level of reverse polarity protection. Nor sure if the diode might already be there.

It is still odd to have the values change that much. If at all I would expect an extra offset and maybe a changed voltage range to allow slightly higher output, when using the 3 AAAs.

[alexwhittemore]

It is still odd to have the values change that much. If at all I would expect an extra offset and maybe a changed voltage range to allow slightly higher output, when using the 3 AAAs.

That was the idea. Within acceptable limits, I wouldn't expect the calibration to change whatsoever.

[Kleinstein]

One possibility might be the OP oscillating for some reason (e.g. to much capacitive load - despite of 100 Ohms series resistor). Also interference with RF signals (like a mobile phone) could cause trouble - not sure how it changes with a higher supply.

[alexwhittemore]

This is an interesting point, so I took some noise measurements. Sure enough: w/ coin cell, measuring a short, I see ~30mVpp noise, ~6mV rms, and that goes up to ~75mVpp/9mVrms with the AAA batteries. Except that the scope measuring its shorted lead is 6mVrms, so that's about the noise floor of the range.

I think what I'm getting here is that the unit is simply not in spec anymore.