(Solved: Not sane) Sanitycheck: MHz DC current measurement with a uA to mA range

[mh262361]

Hello,
I have been looking around for a while and it seems I need a sanity check. My target measurement range includes single digit uA to ideally around 100mA. While there are many solutions for this range, the problem comes with the sample rate, I would like to sample at around 50 MHz while the solutions I found either got up to only 5MHz or started with a resolution of mA. The problem of switching ranges is not that urgent.

It seems standard bench DMMs are more aimed at the KHz area, like the Keysight N6705c. With accurate measurement ranges in the pico range its also overkill and the wrong tool for the job. Then there are DC current probes like the N2820A with 3MHz at 50uA resolution or the TCP0020 with 50MHz at a 1mA resolution, but nothing in between. High resolution probes just dont seem to be sampling quicker then 5MHz.

So I looked into a normal oscilloscope combined with a shunt resistor where at least the osci can easily reach samplerates of 100MHz and up, even with something cheap like the Rigol. However, with their 8bit ADC and a high inaccuracy of up to 3% i wont reach anywhere near the necessary accuracy. The pinned Osci thread (absolutly amazing work btw) lists the USB Osci PicoScope with (up to) 200Mhz and (up to) 16 bit. So in theory that might work? https://www.picotech.com/oscilloscope/5000/flexible-resolution-oscilloscope

While I wont reach the entire range with one shunt without looking into voltage drop compensation (with the ucurrent is also operating in the kHz range), looking at smaller ranges with this samplerate would already be a start.

But, since there are apparently no other solutions in this area, it sounds a bit too good to be true. It seems I am missing some underlying information why sampeling in the uA range with a MHz samplerate just isnt done anywhere? Is there just no market or is there some underlying issues that makes this attempt senseless?

For a bit of background, I am looking into energy consumption of energy-limited constrained devices and am wondering how accurate models could be, since tracing is possible in MHz and up, which for something like a cortex m4 means per cpu zycle.

Thanks a bunch for any input.

[Marco]

A high speed high bitrate scope or digitizer would certainly be part of it, but not a solution in and of itself. You need 100 nA range RMS noise current. So the current shunt has to be 100+ Ohm (give or take). That presents an obvious problem when the current goes to 100 mA.

The easiest solution is to go to a transimpedance architecture, which can swallow that 10V at the cost of a little power consumption. Is that an option?

PS. the only people I see doing stuff like this are the sidechannel attack researchers, but they are often a secretive lot.

[nctnico]

This question seems to pop-up regulary. There are at least two forum members who sell solutions for measuring low currents up to high frequencies.

[exe]

What is the problem are you trying to solve?

[mh262361]

Thanks a bunch for the answers, that helped alot.

@ Marco The range of the shunt resistor is certainly a problem if I want to cover the entire range, but i was more worried that the high speed was somehow introducing problems i am not aware of. And thank you for the tip with the transimpedance architecture and especially the sidechannel attack aspect. I didnt realize they ofcourse needed such equipment as well, its a great idea to ask around if such hardware already exists in the vicinity, even if the papers from the field turn out to be close lipped. I really cant say how much that hint helped.

@nctnico Thanks for the heads up, i will give the search another go.

@exe I am looking into energy consumption modeling, and while I will likely get there with a KHz samplerate, it would be a shame if it was possible to match the frequency of the DUT and I didnt do it. Even if for nothing else then to take a deeper look into instruction level models in the future.

[mh262361]

Just an update for anyone looking into the same field, I ended up on a Red Pitaya

https://redpitaya.readthedocs.io/en/latest/developerGuide/125-10/vs.html

edit: If you are thinking ordering one, look at the impedance of the 16bit version, 50 Ohm vs 1MOhm in the others

[EEEnthusiast]

Why would you want to sample the current at 50MHz. Generally the high frequency currents to any component needs to be provided by the decoupling capacitors such that the current bandwidth as seen by a battery or power supply is restricted to few 100KHz. Unless you are trying to do some differential power analysis on crypto hardware after removing the decaps.

[Hydron]

Keysight do a (simultaneous) dual-range type current probe which may be of interest, though it only has 3MHz of bandwidth (on the low gain channel, less on the other): https://www.keysight.com/en/pd-2229629-pn-N2820A/3-mhz-50ua-high-sensitivity-ac-dc-current-probe-2-ch?cc=GB&lc=eng
Bet it costs a lot though, and obviously needs a matching KS scope.

[Marco]

Personally I'd get the new 16 bit one, 50 Ohm will be easier to interface with.

You won't get nearly those 16 bits though ... I think you'll have to do something tricky like making a low pass current source (something like an active inductor) so you can digitize the high frequency component of the current separately. The 100 Ohm shunt is near the lower theoretical limit for noise to resolve a uA signal, 1K is probably more realistic and that really becomes a problem when the uA signal is superimposed on 100 mA.

Multi-range current measurement doesn't solve your problem, what you want is true >16 bit precision at 50 MHz at 100 Ohm resistor equivalent noise floor ... and that's not going to happen. Separating out the high frequency component of the current (easier said than done) is probably your best bet.

[mh262361]

@ EEEnthusiastUnfortunately @Marco Thank you very much for giving me that heads up. My CS background, and the lack of an EE background, are unfortunately showing. I meant the samplerate not the actual frequency (so MSa/s instead of MHz), sorry for the confusion and wasting peoples time.

Thank you all, I think i already found a solution, just wanted to give anyone stumbling over this thread in the future a headsup for the cheaper alternative to the picoscope.

[Marco]

The SNR of the original Pitaya is maybe 10 bits worth over the full bandwidth ... so resolving uA signals at 50 MHz bandwidth is not going to work. Not at 5 MHz either.

[mh262361]

Thanks, I stepped back from the initial range requirement after realizing there was no way around a shunt. It would have been a nice to have. I am now planing to use this setup to look for the existence of short spikes in comparison to measurements in the KSa/s region.