Documentation for uCurrent Gold

[JDW]

I've not yet purchased a uCurrent Gold and am now browsing this forum to see what the caveats are.  I see some basic info here:

http://www.eevblog.com/projects/ucurrent/

But where is the complete documentation? 

I've read that at least one person added diodes for greater protection, so it's pertinent that documentation explains what pitfalls to avoid when using the uCurrent.  Surely someone must have put together some documentation on the uCurrent by now, but I am not finding it.  Where is the documentation?

Thank you.

[BroMarduk]

There is a link to the original Alternate Zone article here: http://alternatezone.com/electronics/ucurrent/uCurrentArticle.pdf that was linked to from the eevblog site.  It does cover a lot of the internal workings.  Or were you looking for something more specifically published as a user guide?   I'd be interested if that exists too...

(Link to the diode protection forum thread in case others were interested: https://www.eevblog.com/forum/crowd-funded-projects/current-gold-on-kickstarter/msg448208/#msg448208

[guenthert]

Wished I'd seen this earlier.  Not sure about the diodes, but the modification to the on-off-short switch I should have done too.  I was earlier already wondering why the uCurrent should be on, when the connection is shorted.  That doesn't make any sense.    Unfortunately, as it is designed, the input is not shorted in the off position (why, oh why?).    While testing a current supply (with a hearty compliance voltage), I managed to thoroughly fry my uCurrent which I forgot to be in the circuit (turned off)  - yes, cracking sound of bursting SMT resistors and all. 
  I tried replacing the broken resistors, but it seems at least one of the OpAmps didn't take the abuse lightly either -- FUBAR.

[MLXXXp]

Unfortunately, as it is designed, the input is not shorted in the off position (why, oh why?).

When I modified mine, I could have shorted the inputs when off, instead of using back to back diodes, but I though there might be cases where an "off" position that didn't affect the burden voltage (within the measurement range) would be useful. My other mod turns off the unit in the "short" position, so I can select either "off with shorted input" or "off with burden voltage".

[JDW]

With all this talk of critically important mods, it would seem only reasonable that there be a fire sale to clear out existing uCurrent Gold inventory so as to release an updated model with the aforementioned mods!  For truly, who wants to pay nearly $100 (shipping included) for a device that we might accidentally fry very quickly?  I know some of you out there have never experienced that, but the rest of us are not electronics gods -- we are but flawed mortals who err.

[MLXXXp]

For truly, who wants to pay nearly $100 (shipping included) for a device that we might accidentally fry very quickly?

A primary purpose of the µCurrent to is provide current measurements with a very low burden voltage. It's difficult and/or expensive to add circuitry which would properly protect the inputs, without increasing the burden voltage, at least on the higher current ranges.

If your situation doesn't require a low burden, then you could either add external protection to the µCurrent or use other equipment with the desired built in protection.

On the nA range, a forward biased diode for DC, or back to back diodes for AC, across the inputs, plus a fuse in series, would probably suffice. The diode's impedance below its forward voltage conduction point would be high enough, compared to the 10k shunt, to not affect the accuracy of the readings. If the current gets too high, the voltage across the shunt would exceed the diode's forward voltage point, causing the diode to conduct and the fuse to blow if the fuse's rating is exceeded.

On the nA range, the fuse's resistance would likely be low compared to the 10k shunt, so it also wouldn't affect the burden voltage much. On the other ranges, the fuse's resistance could affect the burden.

[JDW]

Question...

uCurrent Gold specs say this about Burden Voltage:

    3 Current ranges:
    +/- 0-1250mA (20µV / mA burden voltage typical) * switch contact resistance plays a role here. 10uV due to the shunt resistor.
    +/- 0-1250µA (10µV / uA burden voltage)
    +/- 0-1250nA (10µV / nA burden voltage)

Fluke 8845A 6.5-digit multimeter (>US$1k device) specs:



Take Note: 100uA range is <0.015 V burden voltage, and Resolution for 6.5 digits is 100pA

So in terms of the Burden Voltage spec, it would appear that my Fluke meter has the uCurrent Gold beat, hands down.  But even with the probes disconnected I see 23nA on the display of my Fluke 8845A.  And when I try to take low current readings, I get lots of fluctionations in the nA range.  I originally thought the uCurrent might benefit me, but looking at the specs for Burden Voltage, I don't see how.  Do you?

[JDW]

...I also use an 8846A (particularly, a Keithley rebrand, DMM4050), but for its price, if it benefits even only a single project, it is worth all the money you've paid for it.

The current measurement features of my 8845A and your 8846A should be exactly the same.  How well can you measure nA?  Can you measure something down to 1nA or lower with good accuracy?  I am not able too, which is what led me to uCurrent and this forum in the first place.

[JDW]

For me, low current usually mean a few hundreds nA. I've never had a chance to do nA measurements.

I made time today to do the following nA measurements, which include uA measurements for comparison:



I do not have a uCurrent to repeat the same test.  As such, it would be appreciated if those of you with (a) a uCurrent gold, (b) 10pcs of 10M-ohm (5% tolerance) resistors, and (c) a benchtop power supply could do the same test and let us know your Percent Error.  I am curious if your US$100 device is better, equal or worse that this >$1000 Fluke Meter.

Thank you.

[JDW]

Thank you for your time in testing.

For the sake of our readers who may not understand, the "SMU" you refer to is your Keithley 2410 which is defined as a high-impedance multimeter that has a built-in low-noise Power Supply.  (In comparison, the Fluke 8845A used in my test is high-impedance but lacks a built-in Power Supply.)  The documentation for the 2410 shows it can, in theory, measure from 50pA to 1.05A, whereas my Fluke 8845A can, in theory, measure only as low as 100pA when using the 100uA range setting.

In your test, rather than using resistors, you set your meter to generate a 10nA reference current, then used that same meter to measure it, and you repeated for 20nA, and so on, up to 200nA.  You then said your meter gives you an error of only 1%+200pA.  You then tested the uCurrent and found its error to be up to 11%.

I myself have no means to generate a stable and trustworthy "reference current" and so in my test I used resistance values (see my earlier post).  And you can see that I am getting huge errors in the nA range, but small errors in the uA range.  I am therefore curious if this is some problem with my Fluke 8845A, or a problem with using resistances in the test?  To know the latter, I would be curious to know what kind of current measurements you get using the same resistance values I did.

Without being forced to buy a new meter (my Fluke works fine otherwise), I am seeking to know how to get the most accurate nA measurements possible.  Do I really need a uCurrent, or is there something I need to do to adjust my Fluke meter?

Thank you.

[JDW]

...nA measurement is a tricky business.

That brings me back to square one. 

I originally discovered uCurrent and ultimately this forum because I began Googling the issue of nA measurements due to my Fluke 8845A not giving me reliable nA measurements.  And yet my Fluke works fine otherwise!  So I have been trying to determine if buying a uCurrent is worthwhile.  At this stage, I honestly do not know.

In your case, it seems that there is no need for a uCurrent insofar as your meter is giving you highly accurate readings (relative to my Percent Error, and relative to your own uCurrent). 

I'm not sure what to do at this point.  And it seems no one else has an opinion, seeing that only you and I are in this discussion.   

[JDW]

...modify your uCurrent and increase its burden resistance... The preset one is 10K... If you mod the sensing resistor to 100K, then you get 10mV for the same current...

I haven't purchased a uCurrent yet, mainly because I have been trying to ascertain if it would be a benefit to me at all in light of the fact I have a Fluke 8845A.  But thank you for that suggestion about changing the uCurrent's 10K resistor to 100K.

I found the following post in this forum which speaks about AC riding on the measurement cables and negatively impact nA measurements using the uCurrent:

https://www.eevblog.com/forum/testgear/ucurrent-not-acurate-on-na-range/msg882779/#msg882779

Could that have contributed to the 11% error you cited?

[JDW]

Didn't bother to hook up a scope. Will see if clipping happens or not and let you know tomorrow.

I appreciate that.

In the meantime, I am curious how much the resistor TOLERANCE will affect the accuracy of the uCurrent Gold.  (I still don't have a uCurrent Gold to test, but I may purchase one at some point.)  Right now Dave uses +/- 0.05% resistors (10ppm) for R2 & R9.  You suggested a change of R2 from the current 10K value to 100K.  I was able to find one SMD resistor that has a +/- 0.01% (5ppm) spec here:

http://venkel.com/part/UPTF2512-2W-P-1003UT

Do you feel that the difference in tolerance between 0.05% and 0.01% for the replacement 100K resistor be largely irrelevant for the nA measurements we are talking about?

[JDW]

Under that condition, when you read the output of the uCurrent with your Voltmeter (set to DC), what is the measurement?  (How close is it to the 100nA?)

[JDW]

...today the DMM reads higher than set point, this time 11% higher... it is using its default 10PLC.

Which means you measured 111nA instead of the 100nA you are sourcing, tested with your DMM set to 10PLC (and I assume your DMM was set to measure DC voltage, not AC).  Curious what voltage value you get when trying the AC setting.  If it is noticeably different than when using the DC setting, that would imply noise is impacting the reading.

[JDW]

I will report back on AC+DC voltage plug AC voltage only.

Thank you.

How much will the uCurrent Gold's resistor TOLERANCE will affect nA measurement accuracy? 

Dave uses +/- 0.05% resistors (10ppm) for R2 & R9.  You suggested a change of R2 from the current 10K value to 100K (so as to get better accuracy for nA measurements).  I was able to find one SMD resistor that has a +/- 0.01% (5ppm) spec here:

http://venkel.com/part/UPTF2512-2W-P-1003UT

Do you think the difference in tolerance between 0.05% & 0.01% for a replacement 100K resistor be largely irrelevant when taking nA measurements?  (My guess is that it would be irrelevant if we are fighting this much AC noise.) 

[JDW]

...uCurrent is not really meant to used to measure tens of nA.  Changing 10k to 100k will help, but don't expect offset voltage will magically go away. Also if you can, increase C3 and C4 to suppress parasitic oscillation.

C3 & C4 are 10pF now.  What larger value do you propose?

uCurrent Gold SCHEMATIC:

[JDW]

Just to clarify...

You sourced 100nA of current and then measured it with a stock (unmodified) uCurrent Gold using the 8846A in Volts DC Mode, measuring 100mV (which means "100nA") and you saw an error of only +/-1nA? 

You then switched your 8846A to AC mode and measured 170mA (170nA)?  Is that correct?

And you also used your 8846A when set to 100uA mode to measure that 100nA directly and got an accurate reading too?  Correct?  If so, I am not sure what is wrong with my 8845A.  However, I don't have the means to source accurate current like you do.  Like I said, I am using large resistor values to limit current, and perhaps that is simply something the 8845/8846 cannot handle.

[FrankBuss]

Test setup: SMU sourcing 100nA, while uCurrent is set to 10K mode, output of uCurrent is grounded with scope's ground lead, scope used here is a Keysight MSOX6004A, bandwidth is unlimited, using 700MHz 10:1 kit probe, low frequency calibrated.
Result: 500mV p-p 37.5kHz distorted sine wave noise, plus 1500mV p-p mains frequency spikes.

See this posting:

https://www.eevblog.com/forum/testgear/problem-with-ucurrent-gold/

I solved the problem with a capacitor at the input. Of course, not a solution if you want to measure AC, but works for my applications (mostly power consumption of microcontrollers and other DC things). Without it, you have to really be careful with long wires and feedback of the output to the input.

[JDW]

...8846A can handle nA measurements, and is more accurate than uCurrent, without oscillating nor need of warming up. I don't have a 8845A, so I can't test it. I think I can find a DMM4020 somewhere in the lab, that should also be a Fluke rebrand, likely 8845A. I can test on that one.

The problem is that our test cases are not the same.  You are sourcing current.  I cannot do that so I am using large resistors to ground and inserting my 8845A probes between the resistor and ground.  That is how I created this data:



It could be that your 8846A would give you the same Percent Error as I discovered, if you repeated my test case using resistors instead of sourcing current.  But that is only a guess.  (I did not have a single 100M-ohm resistor, which is why I used 10pcs of 10M-ohm instead.  As you can see from my data, I measured nA and then measured uA.  The uA test was more accurate, however it uses a smaller resistance value and higher voltage.)

[JDW]

Dear uCurrent users,

If you have 10pcs of 10M-ohm resistors, could you please conduct the same nA current test I did (see results in my previous post) and then post your results?

I don't have a uCurrent and that is why am seeking your assistance.  I want to know if my $1000+ Fluke 8845A is more or less accurate than the uCurrent.  But the only way to know is for me and you to conduct the same exact test.  blueskill kindly conducted tests in this thread, but he used a Current Source which I do not have access to.  Therefore, the only way to confirm my findings is if we all conduct the same exact test.  That means using 10pcs of 10M-ohm resistors.

I look forward to your kind replies.  Thank you!

[JDW]

The dearth of replies indicates that none of you possess 10M-ohm resistors.
I guess they are hard to find.   

[FrankBuss]

Actually I didn't have 10 meg resistors, but bought some from eBay. I can't measure the whole resistance, but measured the individual resistors and the sum was 100.31 meg ohm. Now I have a 10 meg to 100 meg kind of variable resistor:



With 1 V the uCurrent shows 9.83 mV, with 2 V it is 19.82 mV and with 3V it is 29.8 mV. That's without my capacitor. If I carefully avoid output cables coupling to input cables, I can see the same result without the input capacitor (i.e. unmodified uCurrent). All measured with a BM257s, and not on a breadboard to avoid additional parallel resistance and getting too much noise because of the long traces in the breadboard.

[JDW]

Bless you for repeating my 10M-ohm x 10pcs test, FrankBuss!

That seems to prove the uCurrent, even without any hacks and without your capacitor is extremely accurate for nA measurements down to 9nA!

My US$1k+ Fluke 8845A doesn't hold a candle to that accuracy.  I actually wrote to Fluke last week.  I sent them the same "nA Measurement" data I posted in this thread.  Today they replied as follows:

Well looking at the spec at that range - percent of measurement (reading) + percent of range  that percent of range alone will add 25nA of error on the 100uA range.  So that is going to make it very shaky down there.   So I would say - no - not quite accurate enough for what you are trying to do.  Unfortunately that next jump in accuracy is an exponentially larger price tag..

So it would seem quite safe and reasonable for me to purchase a uCurrent.

Many thanks!

[FrankBuss]

Well looking at the spec at that range - percent of measurement (reading) + percent of range  that percent of range alone will add 25nA of error on the 100uA range.  So that is going to make it very shaky down there.   So I would say - no - not quite accurate enough for what you are trying to do.  Unfortunately that next jump in accuracy is an exponentially larger price tag..

So it would seem quite safe and reasonable for me to purchase a uCurrent.

Looking at your measurements, this makes sense. If you subtract 20 nA for your last two measurement series as an offset error, then it looks like it gets quite accurate. But I don't understand why the 100 uA range has errors in the nA range, doesn't make sense. Can you measure the 20 nA with shorted leads, too? This would allow you to measure the offset.

Of course, buying a uCurrent solves this, too, without any manual offset calculations and reasonable accuracy.