Do you have experience with low DC current on the Fluke 8846A or Agilent 3446xA?

[MaxwellFaraday]

I am evaluating several battery powered modules that have long quiescent periods and communicate via bluetooth low energy or wifi when an external event occurs, say, a button push or the proximity of a BLE enabled phone.  The Amazon Dash button and any of the iBeacon devices are good examples.  The chip manufacturers (Nordic, TI, Broadcom, etc.) make some exciting claims about quiescent current consumption, but obviously the other components and overall design chosen for these modules (passives, power mgmt, etc.) have a significant impact on battery life.

My problem is that I have no way to evaluate low DC current consumption over time -- mostly DC microamps and nanoamps and maybe(?) high picoamps.  I can measure instantaneous microamps using any of several handheld DMMs, but each of these meters has lived a hard life in the field and I don’t have significant confidence in their accuracy and repeatability.  Plus, none of them can do any sort of logging.  Most of my other work has been with mains voltages and current.

I am considering either the Fluke 8846A or one of the  Agilent/Keysight 3446xA benchtop meters.  I have read through the specs for each meter and reviewed every eevblog post that mentions either of these meters (there are quite few).  Both are beautiful meters and I’d be lucky to have either of them.  I have been unable to find any information that gives me confidence in the low DC current accuracy of these devices.  Fluke lists an appealing 100pA resolution, but the word resolution is often misused and slippery.  The Keysight DC current ranges seem higher, but it’s hard to tell.  I understand burden voltage as it applies to user-selected current ranges, but I don’t know what cleverness the meter manufacturers use for circuit- or software-switched current ranges.

I would love to hear from you if you have experience with either the Fluke 8846A or Keysight 34461A in the low DC current ranges.  Also, I have never owned a logging DMM and don’t really know what options I might need to produce simple data dumps that would end up in a Google spreadsheet.  Maybe the USB option is sufficient.  I don’t think I need a fancy logging option, but it might be a good opportunity to gussy up my smarts.  I can code, if that makes a difference.

I have convinced myself that I need a benchtop meter if for no other reason than it will live a life of relative comfort indoors and not be thrashed in the field.  I may be better off with a lesser benchtop meter and a uCurrent device.  Keithley makes some beautiful low amperage measurement devices, but I’d rather avoid a task-specific device and I can’t spend much more than the ~$1,000+ I’ve got allocated for the 8846A or the 34461A.  I am pretty blind about Keithley’s offerings.

Thanks in advance.

[nctnico]

A while ago I choose the 34461A because of the graphic LCD display. The ability to make trend charts on the meter itself and then save as an image to USB have been helpfull to document stability over a couple of hours without needing external tools.

[Dr. Frank]

For such applications, you may chose the 34465A, plus digitizing option, because of the resolution to 1pA, and high dynamic range, I.e 1uA, in this case.

In combination with DCI digitizing at up to 50kHz, also short events can be detected.

From my experience, I can recommend the 34465A. Sorry, no experience with the other instruments for comparison.

Frank

[MaxwellFaraday]

Frank -- Thank you for heads-up on the 34465A.  I reviewed the Keysight spec sheet and it says:  "The ability to measure very low current, 1-microamp range with pA resolution, allows you to make measurements on very low power devices."

http://literature.cdn.keysight.com/litweb/pdf/5991-1983EN.pdf?id=2318052

Resolution on a 6-digit meter with a 1-microamp range will have a pA of resolution.  3 digits for nA and 3 digits for pA.  Is that correct?  That's my understanding, anyway.  A 100-microamp 6-digit meter (say, the 34461A) should then have a 100-pA resolution.  Can I safely assume that these meters are "accurate" to their "resolution?"  I realize that manufacturers are sometimes lose with their usage of accuracy, resolution, repeatability, etc.  I also realize that measurement accessories (leads, probes, etc.) must have some special requirements.

I apologize for the naive questions.  Most of my meter requirements have fallen within 1-sigma of most meters' abilities.  Thanks for the heads-up on the digitizing option.  Noodling that.
 

[6thimage]

On the 34465A's 100 uA range it can display up to 119.9999 uA, so the resolution is 0.0001 uA or 100 pA. On the 10 uA the resolution will be 10 pA, and on 1 uA the resolution is 1 pA.

It is very important to consider the burden voltage, with microcontrollers in particular, I have found it is best to start in the 100 mA or 1 A range (depending on micro) and to disable autoranging. Then lower the range to make a reading - if the range is too low (with the burden voltage too high) the microcontroller won't start and can give you some very strange results (mainly, changing range will drastically change the measured value).

[XFDDesign]

Dont also confuse resolution with accuracy.

On 100uA DC range, your certainty within 1 year is ±(0.05% of reading + 0.025% of range), so if you're reading 1µA, it could be 1.026uA or 974.5nA. In this case your 1pA of resolution isn't giving you much.

[zlymex]

It's not that difficult measuring low current as to use an expensive bench multimeters.
I had hand-held DMMs such as Fluke 289, Hioki 4282 and MetraHit 30M, they all have 0.01uA resolution or better, and all with logging ability.
hand-held DMMs have advantages of less interference because they are isolated with the mains.
Dave has a very good uCurrent meter. https://www.eevblog.com/forum/projects/ucurrent-v2-0/
I made one my self which can be 0.1pA full scale at reasonably accuracy.

[Dr. Frank]

Frank -- Thank you for heads-up on the 34465A.  I reviewed the Keysight spec sheet and it says:  "The ability to measure very low current, 1-microamp range with pA resolution, allows you to make measurements on very low power devices."

http://literature.cdn.keysight.com/litweb/pdf/5991-1983EN.pdf?id=2318052

Resolution on a 6-digit meter with a 1-microamp range will have a pA of resolution.  3 digits for nA and 3 digits for pA.  Is that correct?  That's my understanding, anyway.  A 100-microamp 6-digit meter (say, the 34461A) should then have a 100-pA resolution.  Can I safely assume that these meters are "accurate" to their "resolution?"  I realize that manufacturers are sometimes lose with their usage of accuracy, resolution, repeatability, etc.  I also realize that measurement accessories (leads, probes, etc.) must have some special requirements.

I apologize for the naive questions.  Most of my meter requirements have fallen within 1-sigma of most meters' abilities.  Thanks for the heads-up on the digitizing option.  Noodling that.
 

Your ideas about the resolutions are correct, but ONLY in the steady state, i.e. at NPLC 100.

The accuracy is NOT the same as resolution, the specification indicates of course less accuracy than corresponding resolution, in terms of (error of reading + error of range).
Resolution is about 1 ppm (NPLC 100), but accuracy is in the ballpark of 100pm.
I think, accuracy is not the crucial parameter for this application.

For fast changing signals, you also have to sample the current accordingly faster, which will reduce the resolution of your measurement!
At fast sampling rates, the system noise of the DMMs might also increase.

The 8846A and the 34461A are similar instruments, also concerning speed vs. resolution.
For both, the resolution will go down, from 6 1/2 digits at 100 NPLC to 4 1/2 digits at 1000 samples/sec (300µs aperture).
The 34465A is much better in this aspect, still delivering 5 1/2 digits at that speed.

Here are the resolution tables, from the 3446xA and 8846A manuals:






Rem.: The 34465A and 34470A both deliver more digits over the interfaces (GPIB, USB, LAN), than on the display, e.g. 0.1ppm resolution at NPLC10.
I have measured , that there is no difference between the 34465A and 470A in this aspect, so the resolution of the '470A might apply.

You have to figure out on your own, how fast your DUT might be, so what sampling rate is needed, and what dynamic is needed, which is critical in these low power systems, as the ratio between idle and power on state usually is extremely high.


Btw.: Comparing prices, the Fluke 8846A is about 1600€, and the 34461A is 1120€ only.
The 34465A is better / faster (5kSa/s at 200µs aperture) than both other instruments, and costs 1426€, plus 400€ if you need the (faster) digitizing option (up to 20µs/50kSa/s).

The proposal, to use handheld DMMs, or Daves µCurrent, will not work at all, as an instrument with fast sampling capability (< msec aperture / speed) AND high resolution is needed.
Handheld DMMs are not suitable, as they  have 10 rdgs/sec at most.
So they simply average out any short current spikes.

Frank

[stmdude]

I work with devices such as these.  And, yes, measuring the power-consumption of them is tricky, mainly because of two things:
1. The range of their current-draw is _very_ large. Typical idle at ~1.2uA, and peaks of ~10mA. A typical 6.5 digit DMM will need to range-switch to measure both levels with any accuracy
2. The time the device spends _not_ sleeping is very brief. As in, >=0.1mS. Meaning, a sample rate of 50KHz gives you very poor temporal resolution.

_Maybe_ a datalogging 8.5 digit DMM would suffice if left sampling long enough, so statistics would take care of the temporal resolution..

However, this is how we've approached it:
We've split the problem into two parts.
1. How can a SW developer determine if their code made things better or worse?
2. How long will our battery last?

Problem 1 is addressed by (semi) cheap Siglent SDM3055 Bench DMMs, left sampling for a while. This will give the developer a number, but it won't be a correct number. However, the developer can compare it to previous numbers they've gotten.
Also, it's very helpful to have code that can lock the device into certain modes (idle, RX, TX, etc), so you can profile those more closely.

As for problem 2, you first need to profile all the power-states the device can be in (idle, RX, TX, etc), and get _actual_ numbers for these with good resolution. We use 34410As here, but the 34461 should do the same job, if not slightly better.
After that, the device is hooked up to a scope measuring a current-shunt, so we can get good temporal resolution. We run the device through typical scenarios (bonding, etc), and measure the periods and ramp-times for all the transitions. Then the DMM numbers are applied (as the scopes Y-resolution is _way_ too poor for this), so we can get the actual "area under the graph", which would be our total consumption.

Yup, problem 2 takes a while to analyze.. 

Oh, and pro tip:  Profile your batteries _carefully_. The same massive range-difference does funny things to most (small) batteries when they're <40% full or so...

[Dr. Frank]

_Maybe_ a datalogging 8.5 digit DMM would suffice if left sampling long enough, so statistics would take care of the temporal resolution..

Nope!

I also use the 3458A, but the limitation of resolution (4 1/2 digits) versus speed (100kSa/sec) applies in the same manner.
The 3458A is not able to digitize current directly, DCV and Ohm only! (DCI: max. 2300Sa/ sec, at 4 1/2 digits)

Therefore, 34465A and 470A are better suited, and may deliver more resolution under these circumstances, especially when using averaging techniques.

The digitizing option also allows for Level Triggering, which will help on that.

You may also use this feature to trigger on current pulses, so to mitigate the dynamics problem.

Frank

[MaxwellFaraday]

Thank you all so very, very much for your thoughtful and generous comments.  I learned more from this dialog than I have from all of the whitepapers, spec sheets, and videos that I have consumed over the last two days.  If I've said it once ..... we are truly in the golden age of anonymous generosity. 

Keysight seems to suggest the double DMM approach recommended here by stmdude.  http://www.farnell.com/datasheets/1898970.pdf

It's hard to imagine that the 34465a isn't the obvious choice at $1400.  It's only $300 more than the 34461a ($1100).  And it's $1,500 less than the 34470a ($2900).  I need to do some noodling on the DIG option.  There are other threads covering DIG, so I will seek info there first.

The other conclusion that I reached is that I need to become familiar with the Keithley product line.  The seem to make some terrific instruments, but have probably spent marketing and advertising money towards engineering.  Not so with Fluke and Agilent/Keysight.

[stmdude]

The other conclusion that I reached is that I need to become familiar with the Keithley product line.  The seem to make some terrific instruments, but have probably spent marketing and advertising money towards engineering.  Not so with Fluke and Agilent/Keysight.

Keithley is on my list of "equipment to look into" as well. Just haven't gotten around to it.

What really peaked my interest was what "trevwhite" said to me in this post:
https://www.eevblog.com/forum/projects/pax-instruments-multilogger/msg937849/#msg937849

"integral based current measurement". Sounds like the ticket. Would want to test it first though.

[HighVoltage]

Here is an example of a low current measurement with my 34470A.
It works very well and stable in this range

 

[nctnico]

The other conclusion that I reached is that I need to become familiar with the Keithley product line.  The seem to make some terrific instruments, but have probably spent marketing and advertising money towards engineering.  Not so with Fluke and Agilent/Keysight.

I think it is the other way around. Keithley is Tektronix nowadays which is Danaher which is all about milking a product for as long as they can and then some longer.

[stmdude]

Here is an example of a low current measurement with my 34470A.
It works very well and stable in this range

Yea, that part is "easy".  It's the high dynamic range of the current-consumption that's the problem though, combined with the "slow" sample-rate of most DMMs.

Add some 10mA spikes in there, lasting a a millisecond or less. Compare the value you get on the DMM, with what you should have gotten.

[Dr. Frank]

It's hard to imagine that the 34465a isn't the obvious choice at $1400.  It's only $300 more than the 34461a ($1100).  And it's $1,500 less than the 34470a ($2900).  I need to do some noodling on the DIG option.  There are other threads covering DIG, so I will seek info there first.

I've demonstrated, that the 34465A is equivalent to the '470A in most aspects: https://www.eevblog.com/forum/testgear/keysight's-new-34465a-(6-5-digit)-and-34470a-(7-5-digit)-bench-multimeters/msg889215/#msg889215

Therefore, for your application, 34465A is sufficient, especially when doing digitizing DCI measurements.

Frank