Keysight B2980A series electrometer/picoammeter review/teardown

[Kirill V.]

Stunts in the measurement were known in the previous era. Then what are the advantages of this particular electrometer over its predecessors? It can demonstrate an unsurpassed level of accuracy and sensitivity with the direct measurement method?
And it is very strange that with such a high sensitivity it is not possible to connect the source as close as possible to the input amplifier...

[guenthert]

How is it useful to use the resolution of this electrometer? I'm interested in a specific example where you really need a resolution of 10 aA without accuracy

  Accuracy of the meter is in most cases irrelevant.  Who cares whether a current is 2.02fA or 2.03fA?  Chances are the instrument is attached to a sensor measuring some other physical property and the whole system will have to be calibrated in place and probably frequently.

[TiN]

Relative measurements also don't care about accuracy. E.g. if electrometer used as a detector to find null between two current sources. Resolution and noise floor is important in this case, accuracy is not.

[thm_w]

https://xdevs.com/doc/HP_Agilent_Keysight/B2987A/img/bem_ehigh_1.jpg

Wonder what is the deal here, three resistors are through hole mount, then there is the one white resistor surface mounted with no isolation routing. Maybe hand selected and solder after the fact?

[Kirill V.]

This device is called a meter and not an indicator or null-meter. Therefore, it is necessary to know its accuracy, it is a formality of Metrology.
If you can't offer specific examples, I'll come up with an example specifically. I am a consumer and I need to measure 50 aA current. I read the data sheet of this meter and see that the manufacturer promises accurate and reliable measurement of currents with a resolution of 10 aA. Since they do not explain anything, I conclude that we are talking about a direct measurement method: source-cable-meter - look at the display and see the result. Can this electrometer really meet my needs in reality? And what abs. accuracy can I expect when measuring attoampere currents?

[Kirill V.]

Well, no one could answer. Now I am interested in the following question: the manufacturer loudly declares that this electrometer can measure resistance up to 10^15 Ohms. I know that back in the 60s, devices were able to measure up to 10^18 Ohms. In other words, Keysight is proud of the achievements of the last century?

[TiN]

Your posts more like statement, rather than a questions. You could surely see B2980A specifications and determine accuracy and reliability of the measurements by yourself in few minutes and simple excel spreadsheet. There is everything one need in specs to do so. If you need accurate measurement of 10 aA source, you are in territory of metrology field, not commercial instrumentation. You can buy something specifically designed for that, if your project have such needs.

Keysight probably proud that anyone can buy commercial instrument that you can put on the bench and measure high resistances/low currents without need of whole lab filled with vacuum equipment, ultraclean chambers, triple shielded boxes and sapphire insulators, like you often needed back in 60s.

[Kirill V.]

This means that they are betting on Commerce and not on the best technical characteristics. I am trying to understand where the technique of electrometric measurements has gone today.
American companies are actively taking over the market and I am trying to understand what they offer.

[Kleinstein]

The test with the cable seem to have a less stable temperature. There is also quite some drift residual settling in the current. This adds to the low frequency noise.  Extra capacitance at the input can also add to the noise - so a long cable does not really help, even of very high quality.

There is a chance the drift seen with the cable could be from dielectric absorption (either in the cable or meter irself, with lust less waiting time before).

[bsw_m]

So in ranges 2nA and higher B2985A really faster, that V7-49. But in pA ranges V7-49 is faster: 1pA range settling time 1s, 10pA and higher settling time is 100ms.
This is not a totally lie, these are marketing flaws 

[Kleinstein]

As far as I can see the B2985A has higher ADC resolution. So one a fixed range one can get higher resolution or one could use a higher range to get a comparable resolution. To make use of higher ADc resolution one may have to add intentional filtering to reduce the noise and this adds to the settling, though it could probably be done better. In most real world application one has slow settling of the DUT anyway. One you are looking for fA's and below it often takes many minutes to hours for the DUT to settle.

There if often settling with more than one time constant: a fast part from classical RC time constants that may be fast (seconds to 0.1 - 1%) and than a slower process than may take minutes to hours to get down to below 0.01 % settling.

The "conventional" electrometer mentioned by HP is likely a Keithley meter.  A Rusian meter would be a exotic one on the US market.

[Trax]

Can the variant without the battery be upgraded later on to add battery support?
And with regard to low noise is the battery powered one really better?

I want to measure a low ion current in my setup so I'll be working at the low end of the range.

[Kleinstein]

With a ion current one usually has a well defined ground (the vacuum chamber). There should be little difference from a mains powered and battery powered meter here.
Battery power would be a thing with a isolated setup with no low impedance ground connection.

[Xandinator]

That may as well be my imagination fooling me but below 1 PLC and 2 nA range (using the food tray "technique") it seems to give readings that are slightly more stable. Apart from that, the day the battery sees its emancipation out of the quite-nice-to-have category is still to come. Plus, the empty compartment can conveniently serve as a storage for the triax cables, no more running back and fiddling around...

[maxwell3e10]

I just noticed this old post, very nice pictures! TiN, did you have a chance to test it in detail? Keysight B2900 SMUs of the same vintage are rather disappointing. So I am curious if this electrometer exceeds its specs like old HP gear, or barely meets them under ideal conditions.

[TiN]

Didn't use it much, because not long time after I got B2987A chance presented itself to acquire much more capable Keithley 6430 setup for low currents.
But I do have a project soon that will be relying on very low current detection, so it would be perfect chance to compare 2987A with 6430 and perhaps some other pA gear  (have K6485 and HP 4142B). If you have something particular that interests you, I can set up and log something, B2987A is unused anyway.

Out of latest data I used these for was triax connector leakage test with -20V bias.



DUT in question

[bsw_m]

If you have something particular that interests you, I can set up and log something, B2987A is unused anyway.

1. 100MOhm resistance standard (not an ordinary resistor), preferably placed in a thermostat
2. Stable voltage source 10V.
Place 100MOhm standard between voltage source and electrometer input.
Write log measurement from B2987A within about 3..5 hours. Ideally if you also record the measurement log from the analog output (TIA output) of the B2987A
Why such measurements, I will tell a little later.

Out of latest data I used these for was triax connector leakage test with 20V bias.

I had one question, why, with a positive bias, the current went into the negative region. Why this happened, I have some thoughts. Here I just want to focus on this.

[fcb]

I'd be interested in seeing if the B2987A source/voltage output is truly bipolar (can it generate a waveform with both + and - voltages?) when running in 1000V (i.e. >20V) mode.

I suspect that in 20V mode it can seamlessly generate stepped waveforms from +ve to -ve etc..  but in 1000V mode it has to rearrange the output relays?

[eplpwr]

I suspect that in 20V mode it can seamlessly generate stepped waveforms from +ve to -ve etc..  but in 1000V mode it has to rearrange the output relays?

Yes, that's correct. Table 3-4 in the User's Guide (B2980-90010) shows three different modes for the voltage source: Bipolar 20V, +1000V, or -1000V.

[fcb]

I suspect that in 20V mode it can seamlessly generate stepped waveforms from +ve to -ve etc..  but in 1000V mode it has to rearrange the output relays?

Yes, that's correct. Table 3-4 in the User's Guide (B2980-90010) shows three different modes for the voltage source: Bipolar 20V, +1000V, or -1000V.

Thank you.  Must have missed it.

[TiN]

1000V is fixed polarity only, yes.

Got this setup running, if I understood what bsw_m suggested. Will let it run overnight, then can try with B2987A on battery power if it affects anything (don't expect so on high current like 100nA).

RAW datalog live page.



Connection between source and standard is done with triax. Guard of triax is driven by same 10V with shorting BNC adapter at calibrator.
Then BNC coax with adapters running from standard to B2987A current input. Common LO is connected between 3458A LO and source LO.

[bsw_m]

Thank you very much, Ilya. I received quite interesting information.
By the way, as I understand it, B2987A, apparently has some kind of correction? Since the results processed by him are more stable than the output of TIA.
As for repeating the measurement from the battery, I think there will be no differences, you should not waste time.

[TiN]

I don't think there is any corrections. Analog output is not output of TIA but generated separately from digitized code by separate DAC, unless I mistake something. I have also 1G DIY standard and will try that on weekend.

[maxwell3e10]

Didn't use it much, because not long time after I got B2987A chance presented itself to acquire much more capable Keithley 6430 setup for low currents.

What makes Keithley 6430 superior to B2987A ?

Out of latest data I used these for was triax connector leakage test with -20V bias.

The noise level seems comparable. But why does B2987A take much longer to settle?

[bsw_m]

Analog output is not output of TIA but generated separately from digitized code by separate DAC ...

It is a pity that there is no direct exit from TIA.

I have also 1G DIY standard and will try that on weekend.

Thanks!