Author Topic: µCurrent and noise  (Read 4380 times)

0 Members and 1 Guest are viewing this topic.

Offline elcaron

  • Contributor
  • Posts: 21
  • Country: de
µCurrent and noise
« on: January 09, 2017, 12:55:50 pm »
Hi, I recently encountered some µCurrents. I build one myself on a Seedstudio-produced board using solder paste and a hot-air rework station and had two built by the electronics guy in our institute - one on a CNC-milled board and one on another piece from the Seed order I donated. He handsoldered them. All parts where purchased from Digikey in two separate orders according to the order numbers from the BOM.

Now, just using a Rigol precision power supply and a multimeter (both the Fluke in the institute and my OWON at home), and a MOhm resistor, I find all of them to be very accurate. Like, spot on. The drama begins as soon as I hook up an oscilloscope to measure actual time varying changes (like on an HM-10 BT module, which I would like to measure while it sends iBeacon messages every 500ms).  No matter if I use the 40k€ Rohde&Schwarz in the institude, or the hacked Rigol DS1054Z at home, the noise is extreme. On the latter, using the included 10x probes directly on the binding posts, I got 60mV peak-to-peak, high-freqency noise with shorted (via banana cable) current input binding posts on the other side.

So what is the deal here? Am I doing something fundamentally wrong? Is that an antenna effect on the input cable? Do I have to take extrem precautions with EMI? Are all my units defective? Something seems to be really wrong here, because as it is, it certainly cannot do what the advertisement tells me, especially after the bandwidth upgrade revision.
I did notice that the noise substantially changed with the length of the BNC cable attached to oscilloscope (we used one of those rigid dual-banana to BNC adapters, which gladly perfectly fit on the output posts) and the input impedance (the R&S can chose between 1MOhm and 10Ohm(?)), but no combination was really usable.

Any help?
« Last Edit: January 09, 2017, 01:29:51 pm by elcaron »
 

Offline vealmike

  • Regular Contributor
  • *
  • Posts: 192
  • Country: gb
Re: µCurrent and noise
« Reply #1 on: January 09, 2017, 02:24:10 pm »
Not sure exactly what you're trying to measure, or how.

Please post a schematic, indicate your measurement point and take a picture of your probing technique.

Your measurement could be real, or it could be pickup.
 

Offline elcaron

  • Contributor
  • Posts: 21
  • Country: de
Re: µCurrent and noise
« Reply #2 on: January 09, 2017, 03:05:57 pm »
There is not much of a schematic to post. I am connecting the input terminals of the µCurrent, so I should measure 0 current and do not have to care about noise from the a power source.
I am attaching the 10x probes (be it from the 40k R&S or the Rigol) to the output posts. My DMM measures 0mV, but my oscilloscope goes crazy with 30mV high-freq noise.

If that really needs a picture, I will have to postpone it until Saturday.
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1114
  • Country: fi
Re: µCurrent and noise
« Reply #3 on: January 09, 2017, 05:51:07 pm »
It's actually very helpful if you draw what exactly you're hooking up and what your result is, since it is difficult to know just what it is you're doing.

[...] I am connecting the input terminals of the µCurrent, so I should measure 0 current and do not have to care about noise from the a power source. [...]

Already here I first thought I understood and then again I have no idea what you hooked up to the uCurrent because you don't mention it.

Anyway, the scope alone already measures far more bandwidth than the uCurrent has (version 1 had 2kHz bandwidth ?) so your scope contributes a lot of noise. Additionally, a DMM has somewhere around 400Hz...20kHz of bandwidth so you will see less noise already due to this compared to the scope.

Also don't use a scope probe for this, use banana to 50 Ohm coax so that the scope is on 1x attenuation. Enable the scope's the bandwidth limit. You're measuring a small voltage and you don't want to attenuate the signal with a probe. If you insist on using a 1x probe then use the shortest ground path available (some kind of spiral ground clip) to avoid the ground lead inductance from picking up unwanted signals.

Then I don't understand what you're hooking up to the input terminals to see 0 current, I am assuming there is something with high impedance? This may pick up a number of things that are on your desk.
« Last Edit: January 09, 2017, 05:53:39 pm by Neganur »
 

Offline elcaron

  • Contributor
  • Posts: 21
  • Country: de
Re: µCurrent and noise
« Reply #4 on: January 09, 2017, 06:19:59 pm »
For the test where I measured the noise with 30mV amplitude, I did not connect ANYTHING. I put both current terminal at the same potential by connecting them with a cable, minimal impedance, so 0 current should flow. One end of the cable in one terminal, the other end in the other terminal. Picture attached for clarification.
I did, however, measure similar amounts of noise in cases where I e.g. measured a battery powered HM-10.

Also, while I did the 30mA noise test with the times 10x oscilloscope probes, I also tried this with a BNC cable. The cable length seemed to have an influence on the noise, but even a 15cm cable didn't give usable results.

I will try to reproduce this on the R&S oscilloscope tomorrow and post pictures.
 

Offline guenthert

  • Frequent Contributor
  • **
  • Posts: 357
  • Country: us
Re: µCurrent and noise
« Reply #5 on: January 10, 2017, 02:48:10 am »
For the test where I measured the noise with 30mV amplitude, I did not connect ANYTHING. I put both current terminal at the same potential by connecting them with a cable, minimal impedance, so 0 current should flow.
Err, no.  You equalized (or nearly so) their potential, but for that a current flows (you have now a loop through which a changing magnetic field will induce a current).  In order to assure there is no current flowing, you need to maximize the resistance, i.e. keep the input open.

The uCurrent uses a moderately large input resistance of 10kOhm in the nanoamp range, which seems sufficient to pick up EMI.  For measurements of very small currents you need to minimize EMI.  One way to do so is to put the source *and the uCurrent* in a shielded enclosure or have at least a shielded (or twisted) cable from the source to the uCurrent (and if there is bad EMI in your environment also from the uCurrent to the measurement instrument).

Also keep in mind that the uCurrent's lo output is on the same potential as the lo input.  Not a problem if a handheld DMM is attached, but your oscilloscope lo (GND) is connected to shield which is typically (for oscilloscopes) on (or close to) earth potential.  Depending on how your input is connected, you might easily create a dreaded ground loop.
« Last Edit: January 10, 2017, 05:09:19 am by guenthert »
 

Offline JonM

  • Supporter
  • ****
  • Posts: 150
  • Country: us
Re: µCurrent and noise
« Reply #6 on: January 10, 2017, 03:54:51 am »
I did a quick test of a official Australian uCurrent Gold connected to a Keysight DSOX3024T. I am seeing 15-20 mV p-p noise on the output for all combinations of current terminals shorted, current terminals open, and short switch engaged. There is one exception: with current terminals open on the nA scale I see 150 mV p-p of mostly 100 kHz "signal". I am doing this in a fairly quiet EMI environment, only a couple of computers and one monitor running nearby.

However, just before pressing "Post", I see that the 100 kHz noise is coming from a switching power supply that is supplying a LED bar under my desk. So, as usual, it pays to look for EMI, and grounding / shielding issues.

Jon
 
 

Offline elcaron

  • Contributor
  • Posts: 21
  • Country: de
Re: µCurrent and noise
« Reply #7 on: January 10, 2017, 07:55:40 am »
10/15mV is something I could agree with under the right test conditions. So, I feel that with all this "up to 1pA" AND "300kHz", this something that should be mentioned somewhere. Or is there a trick how I could measure anything even in the low (n/u/m)A range at even a few dozens or hundred Hz with this kind of noise? Regarding EMI, we did this in an environment where electrophysiology recordings are regularly performed. It really cannot be that bad ...
 

Offline vealmike

  • Regular Contributor
  • *
  • Posts: 192
  • Country: gb
Re: µCurrent and noise
« Reply #8 on: January 10, 2017, 01:26:51 pm »
For the test where I measured the noise with 30mV amplitude, I did not connect ANYTHING.

OK, thanks for the picture of the uCurrent (I had no idea).

PROBLEM 1/
Looking at your front end, the two banana plugs coupled at the top with a wire link AND the traces on the PCB are forming an inductor, or if you prefer a loop aerial.

The inductance is directly proportional to loop area. And that looks like a lot of area to me.

As an alternating magnetic field interacts with the inductor, current will flow in the inductor.

More loop area = more inductance = more (noise) current.

PROBLEM 2.
Someone has stated that this device has an input resistance of 10K.
Well, if you want to know your noise voltage at the input you won't be surprised to hear that V=IR, where I is the inductor current and R is 10K.
Noise voltage at the output is simply input noise multiplied by system gain.


PROBLEMs 3 & 4.
Are exactly the same as problems 1 & 2. But this time you're making a big loop with your 'scope and it's GND lead. The 'scope may have an input resistance in the meg ohm range, meaning smaller currents give more noise.
This is why I asked for a photo of your probing technique.

PROBLEM 5.
Mains GND loops. Your 'scope GND is mains GND. What other kit do you have hooked together that has it's own mains GND? Same principal as above, but with much lower impedances.

PROBLEM 6.
System noise.
The PCB will pick up noise. The IC will generate noise. Your PSU may have noise that the IC couples into it's output.


Problems 1 - 2 are telling you that your uCurrent isn't that suitable for high precision low current measurements. Or at least not as you have it configured.

Problems 3 - 5 are measurement issues. The measured noise isn't real.

Problem 6 is real unavoidable system noise.
 

Offline guenthert

  • Frequent Contributor
  • **
  • Posts: 357
  • Country: us
Re: µCurrent and noise
« Reply #9 on: January 10, 2017, 10:15:30 pm »
For the heck of it, I measured the output of the uCurrent Gold using an AM502 amplifier connected to an oscilloscope.  Discharging a 9V battery (puts out some 9.6V) through a 1GOhm (nominal) resistor, all within an aluminium case, yields some 9.63mV (my Keithley 617 (*), also in line, claims 9.69nA) with some 13mV noise peak-to-peak wide bandwidth (DC to 1MHz).  The AM502 offers (and that's the reason I put it there) a band pass filter.  With the upper frequency (-3dB point) reduced to 100kHz, the noise level is down to some 10mVpp, at 3kHz it's 2mVpp and at 100Hz it's 0.8mVpp.  The AM502 and oscilloscope combo have a noise floor of 0.4mVpp in this configuration (amplification by 100, wideband).

One could also look up the noise characteristics of the OpAmps deployed in the uCurrent (I'm in 'my lab' right now and can't be bothered).


*) None of my instruments is calibrated, for what I can tell, they are within stated long term (1 year usually) specification.
 

Offline elcaron

  • Contributor
  • Posts: 21
  • Country: de
Re: µCurrent and noise
« Reply #10 on: January 11, 2017, 10:37:27 am »
Thanks for checking, guys. So we actually are talking about around 10 mVpp noise. I think this is an issue, since the ranges hardly overlap. I'll consider ditching the intended ranges and replace the shunt resistors to something that does not try to span 12 orders of magnitude (if you include the 1pA resolution that is advertised for a 5 1/2 digit DMM) but allows me to do sensible measurements e.g. in the µA + lower mA range.

I mean, it's a nice idea to have 1mV=1n/µ/mA, but at that cost, I don't really know who this is targeted at. Who needs to read nAs with a Professional Precision Current Adapter for Multimeters, the only low cost industry solution, but cannot be bothered to do some simple calculations?
« Last Edit: January 11, 2017, 10:40:24 am by elcaron »
 

Offline guenthert

  • Frequent Contributor
  • **
  • Posts: 357
  • Country: us
Re: µCurrent and noise
« Reply #11 on: January 11, 2017, 06:00:30 pm »
Thanks for checking, guys. So we actually are talking about around 10 mVpp noise. I think this is an issue, since the ranges hardly overlap. I'll consider ditching the intended ranges and replace the shunt resistors to something that does not try to span 12 orders of magnitude (if you include the 1pA resolution that is advertised for a 5 1/2 digit DMM) but allows me to do sensible measurements e.g. in the µA + lower mA range.

I mean, it's a nice idea to have 1mV=1n/µ/mA, but at that cost, I don't really know who this is targeted at. Who needs to read nAs with a Professional Precision Current Adapter for Multimeters, the only low cost industry solution, but cannot be bothered to do some simple calculations?
I looked at the noise only in the nA range.  The uCurrent doesn't strike me as particularly noisy, the 13mVpp I measured was for the whole system.  Despite the aluminium box, some EMI might have crept in (someday I might get a 2nd bigger box for double shielded experiments ...).  Dave claims -90dbV for the uCurrent Gold -- I have neither equipment or environment to verify that, nor am I qualified to do so.

I rarely use the nA range of the uCurrent, as I find measuring small currents (say less than a couple hundred nA) quite a challenge in the noisy environment I operate in.  For that I would use a picoammeter or electrometer, partly because of the Triax connector those use.  They don't deal well with rapidly changing signal levels though.

Measuring very small currents at higher frequencies will be double challenging due to the capacities of shielded cables.  I think you're leaving the territory the uCurrent was meant for (as far as I understand it, its claim to fame is the low burden voltage compared to many DMMs, which comes handy, if one wants to measure e.g. the supply current to a micro controller).
 

Offline elcaron

  • Contributor
  • Posts: 21
  • Country: de
Re: µCurrent and noise
« Reply #12 on: January 12, 2017, 10:08:49 am »
Well, I got similar problems when I tried to measure the 8mA of the HM-10 bluetooth module. But perhaps that ist my test equipment again. I'll post photos on the weekend.

The nanoamp range would have been interesting for impedance measurements of tetrodes. We are currently experimenting with gold plating them at 15nA (using the reference nano amp CC source from the ltc6082 with a bigger shunt). Afterwards, impedance would be measured at 1kHz, with a max current not much higher than that (expected impedance is between 30kOhm and 3MOhm).
 

Offline guenthert

  • Frequent Contributor
  • **
  • Posts: 357
  • Country: us
Re: µCurrent and noise
« Reply #13 on: January 12, 2017, 05:11:16 pm »
Well, one generic approach to dealing with noise is to employ a band-pass filter.  If your signal source is oscillating at about 1kHz, that should be the center frequency of that filter (ideally also the phase of the signal would be taken into account).  This can also be done digitally after the measurement has taken place.
 

Offline jolshefsky

  • Regular Contributor
  • *
  • Posts: 227
  • Country: us
    • Jason DoesItAll
Re: µCurrent and noise
« Reply #14 on: January 12, 2017, 06:04:59 pm »
I tried getting my scope to reproduce the results and I got the same thing: about 40mVpp. I thought I'd try cranking down the digital filter so with 500Hz, I was still getting a fair bit of noise albeit only 15mVpp. Here's the rub, though: if I simply disconnected and shorted the probe leads, I got nearly the same amount of noise (about 10mVpp). See attachments ... note the naming.
May your deeds return to you tenfold.
 

Offline mmagin

  • Frequent Contributor
  • **
  • Posts: 613
  • Country: us
Re: µCurrent and noise
« Reply #15 on: January 12, 2017, 06:06:02 pm »
On my actual uCurrent Gold (from Dave), input shorted with a banana jumper, output right into a piece of coax (bananas at one end, bnc at the other), I see about 12mv p/p on the scope, roughly the same on all three ranges.

That was with my main scope, a HP 54645A, with the 20 MHz BW limit engaged. I'd like to try again with the differential input of a Tek 7A22 in one of my Tek 7000 series scopes when I have time, I've had suspicious measurements using the modern-ish HP digital scope on low level stuff.
 

Offline vealmike

  • Regular Contributor
  • *
  • Posts: 192
  • Country: gb
Re: µCurrent and noise
« Reply #16 on: January 19, 2017, 07:54:03 am »
It occurred to me that you chaps might be interested in this:
http://www.google.com.mx/patents/US6985037
It's directly related to high side current measurement, not to the uCurrent. But the technique can be used for any small signal high side measurement.
 

Offline mmagin

  • Frequent Contributor
  • **
  • Posts: 613
  • Country: us
Re: µCurrent and noise
« Reply #17 on: January 23, 2017, 02:16:20 am »
Ok, well, with the 7A22 differential amplifier set to DC to 1 MHz, I see about 12 mV peak-peak.  With the uCurrent off, I have about 2 mV peak-peak crud on the scope, so I don't think the accuracy is any better than that. :)
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 7331
  • Country: de
Re: µCurrent and noise
« Reply #18 on: January 23, 2017, 09:15:09 pm »
The µCurrent is not designed for absolute lowest noise. If you need less noise for he low currents (e.g. µA range and below), one should use the trans-impedance amplifier way instead of shunt an amplifier. There is also some "noise" / switching background from the AZ OP used.

There is only little EMI filtering - so it can be sensitive to EMI form the environment.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf