Microcurrent measurement peculiarity

[alexwhittemore]

My end goal is to measure supply current of a microcontroller with best-as-possible dynamic range. In sleep, it draws ~10uA and peak, it's about 300mA, so of course, it'll be tricky to get that all in one go. But I figure catching that 300mA with 1mA resolution without browning-out should be doable with a microcurrent, and I'll set up a different measurement to quantify sleep mode. But in trying to get this on a scope without a savage amount of noise, I'd like to use a 1x probe rather than 10x. [As it happens, I think this doesn't much help my cause, but it's how I arrived at the problem below.]

I actually have 2 microcurrents, one modded to run from 3x 1.5V AAAs and a newer one just running off a normal coin cell. Call them old and new.

I've got a test circuit set up to figure out a repeatable DC measurement before I go adding the actual DUT. The test circuit is a 1V power supply, through a decade resistance box to dial in the current, through a microcurrent to ground. A multimeter (10Mohm input) is reading that microcurrent. I'm also trying to probe that microcurrent with a scope. I discovered that connecting the scope probe ground in parallel actually caused a noticeable measurement error due to the additional ground path, so I un-grounded the negative of the 1V bench supply to eliminate that. Now, when I plug in a 10x scope probe in parallel (10Mohm input impedance), the multimeter reading doesn't change appreciably (as you'd expect). But when I switch the probe to 1X (1Mohm impedance), then things get interesting. On the old (modded) microcurrent, nothing much happens. The output stays appreciably the same on the multimeter (and the scope measures what I'd expect, plus a whole bunch of mains noise which is annoying AF). On the new microcurrent, however, dropping the impedance on the output port to 1MOhm drastically changes the measurement - 20% or so at 50mV (reading of 50mA actual drops from 50mV to 40mV), and much more at lower currents (reading of 10mA actual drops from 10mV to 1mV).

Both are uCurrent Gold Rev 5, and on visual inspection, have identical populations.

[alexwhittemore]

I've run into this problem again - in this instance, the "new" uCurrent goes WAY out when I attach a second multimeter. Like, measuring a 99.95mA current on the 1mV/mA range, I get 99.94mV out. Attach a second meter (dropping the attached impedance from 10Meg to 5Meg) and that reading goes down to 90mV - a nearly 10% error. The "old" uCurrent with the 3xAAA mod doesn't have this problem - attach a second meter and the reading stays stable.

Anyone run into something similar?

[bugi]

At first I was thinking of suggesting rechecking that the scope input is not accidentally set to 50ohm internal termination (if it has such), but since you now get issues with just two DMMs, too, that doesn't sound likely... Assuming the uCurrents are indeed otherwise equal, except the power source, have you checked that coin cell state (i.e. it can push out enough current under "load", including uCurrent's internal loads). If you have third DMM, measure the voltage over the battery and if it droops during the issue. Although the voltage supervisor+LED section of the uCurrent would probably reveal supply problem, too, if such happens.

Also, while 10M input resistance is apparently the common thing in DMMs, have you checked both the DMMs that they actually have that 10M?  (Just in case one happens to be some chinese crap with who knows what, or has gotten damaged over time, or whatnot..)

Came to my mind (so I haven't even sanity checked this idea yet), due to the large range of current to be measured, have you considered log-converters? Would need a bit of tinkering, though.  The shown output would then not be directly readable as the current (would need to calculate the anti-log), but single setup could then measure that full range, with slightly reduced accuracy. Log-converters' bandwidth limits might be an issue, depending on how rapid changes you want to see.

[Gyro]

Anything to do with this... https://www.eevblog.com/forum/blog/eevblog-1057-current-murphy/?

The uCurrent design is sensitive to the manufacturer of the rail splitter and a batch of bad ones went out I think.

[alexwhittemore]

Anything to do with this... https://www.eevblog.com/forum/blog/eevblog-1057-current-murphy/?

The uCurrent design is sensitive to the manufacturer of the rail splitter and a batch of bad ones went out I think.

Sure looks like it, thanks for pointing me that way! The "bad" unit sure enough has a rail splitter marked 321S