Fluke 87-V HighRes values are different...?

[Amaruk]

Hi,
The other day when I was checking my 5uA DC current standard with a Fluke 87-V I noticed something really strange. When I compared the high-resolution values with the standard resolution values things didn't add up. I know that going into HighRes mode on this DMM does not make it more accurate, it just gives you more resolution which is fine with me. But what I did not expect to see was that the values did not agree when rounded. I then ran an experiment where I connected some of my DMM's in series with a Keithley 236 (set to a current source). The table in the attached image file shows my results. As you can see the high-resolution values have an offset of about 0.15uA when compared to the Keysight 34461A readings. However, in standard mode (one decimal less) the Fluke 87-V values are spot on...What could be causing this discrepancy? Are the Fluke 87-V calibrated separately in the standard mode and high-resolution mode? Thanks in advance!

[AVGresponding]

The calibration steps don't include any for the Hi-Res modes, and there is no change in the input resistance/burden voltage, so it must be an artefact of the ADC implementation. Mine behaves similarly when sourcing 5-199uA from my Agilent U1401B.

Worth pointing out that it is in spec, and that the LSD uncertainty count is x10 in Hi-Res mode, and that the nA that the LSD represent are well down into the noise for this meter.

User manual: https://dam-assets.fluke.com/s3fs-public/80v_____umeng0200.pdf?tKDGTic.KN0dP9_UJVtSyLsuYWEUp3SY
Calibration manual: https://dam-assets.fluke.com/s3fs-public/8xv_____cmeng0100.pdf

[Amaruk]

Thank you! Yes, I am not concerned about being in or out of spec. It was just so confusing to see that while it is not more accurate in HighRes mode, it is actually LESS accurate than in normal mode... Good to know! Still love this meter!

From a user's perspective it would seem logical that the HighRes mode are the internal values and that they get properly rounded before being shown in normal mode. But that does not seem to be the case here which caught me by surprise. Thanks again for sharing your insights!!

[AVGresponding]

Interestingly, if you manually set the 6000uA range, then go to Hi-Res mode, it gives a different reading to the 600uA range to the tune of 200-300nA for a 10uA nominal reading, when technically it should be identical.
It does it in both polarities.

I really think it must be how the firmware configures/uses the ADC.

[Amaruk]

Interesting! Will have to try that too. As you can see in my images, I used auto-range on all DMM's. Thanks for sharing!

[Fungus]

I like the way it comes up with a negative value in hi-res mode where it had a positive value in normal mode. 

[bdunham7]

I like the way it comes up with a negative value in hi-res mode where it had a positive value in normal mode. 

The 20K/6K thing has always seemed weird to me and seeing the behavior here is even weirder, in spec or not.  Clearly this is an offset issue, but I've no idea exactly how the 6K mode and 20K mode differ overall.  The 6K mode was able to read spot-on, so in this case at least, the 20K mode doesn't seem to be worth paying extra for!

[BeBuLamar]

I use several 87-5 but I don't think I ever switched it to 20K mode.

[Caliaxy]

I like the way it comes up with a negative value in hi-res mode where it had a positive value in normal mode. 

Not sure how exactly the hi-res mode works, but keep in mind that 0.1uA on the 600uA range represents an absolute value of ~0.0167% of that range. On the lowest range of this meter...

I'm actually quite impressed that it really reads 0.1 in standard mode in the first place.

[J-R]

I recently snagged a faulty Keithley 197A that just needed some TLC, and as luck would have it was calibrated only about 1 year ago.  This thread seemed like a good way to take it for a spin.  Results attached.

[TMM]

I thought this was a pretty well known issue. All the 87V I've come across have a slightly different offset in low res vs high res and it can drift over time, you just don't notice the drift as much on 6000 count mode as it only effects the LSD.

With nothing connected, one of my 87V reads 0.3uA in low res and 0.17uA in high res, so it's more accurate in high res... If you want LSD accuracy, high res and use of the rel function is your best bet.

keep in mind that 0.1uA on the 600uA range represents an absolute value of ~0.0167% of that range. On the lowest range of this meter..

0.0083% of the ADC range because the range is from -600uA to +600uA

[Amaruk]

I recently snagged a faulty Keithley 197A that just needed some TLC, and as luck would have it was calibrated only about 1 year ago.  This thread seemed like a good way to take it for a spin.  Results attached.

Very interesting results! Thanks for sharing! Your results confirm what I am seeing on my end with the 87-V: the standard resolution value is not just a rounded value of the high-resolution value for small currents. In your case, though, the accuracy is actually higher in the high-resolution mode. Perfect!

To dig a bit deeper into this I also added my Fluke 87-III to the mix. As you can see in the results that are added to this post, the 87-iii does not seem to have this issue. Its values match nicely!

I should also say that what I am observing here is related to small currents on the uA range. I also tested for small voltages on the mV range and neither my 87-V nor my 87-iii show rounding issues with those measurements. Attached is a picture that shows my results when all the DMM's are connected in parallell to a 5mV voltage source (87-iii and 87-V are in HighRes mode).

[Amaruk]

I thought this was a pretty well known issue. All the 87V I've come across have a slightly different offset in low res vs high res and it can drift over time, you just don't notice the drift as much on 6000 count mode as it only effects the LSD.

With nothing connected, one of my 87V reads 0.3uA in low res and 0.17uA in high res, so it's more accurate in high res... If you want LSD accuracy, high res and use of the rel function is your best bet.

Interesting. I was not aware of this. Great idea about using the relative function in situations like this! Thanks for sharing!!

[J-R]

To clarify, I didn't use relative mode on the Fluke 87-V in any of my tests.  Using relative mode on my specific sample yields about a -0.20uA offset throughout the readings so isn't usable.

I obtained this 87-V as a new in box return and it came with a calibration certificate which stated 330.0uA for an "as left" value.

Same story on the 287, but with a value of 500.00uA.

No data for the Keithley 197A but after some additional tests I would say it is between 0.001uA and 0.003uA low from 0 to 5uA.  Running on up to 200uA, it ultimately is about 0.025uA low, which is still well within spec: 199.785uA to 200.215uA.

Was surprised to find my Keysight U1461A has a "hidden" 0.0001uA display and the digits appear to be meaningful.  Just discovered this later today so no extensive tests yet. 

[AVGresponding]

Someone in the TEA thread mentioned it might be thermal drift, so I decided to see if that is so. TLDR; it isn't.

I took two sets of measurements at a t_A of 19°C and 26°C. I measured at 1uA, 2uA, 4uA, 10uA, 100uA, and 199uA.

Aside from the readings drifting up by the best part of 1uA and getting a bit noisier, the offset remains the same.

Then I discovered something which might start to shed a bit of light on the matter, and I'm not ashamed to say, entirely by accident.

I was taking the leads out and switching the meter and calibrator off, and purely by chance I left the F87 on until after I removed the leads... and the offset almost disappeared!

A quick fiddle about reveals that without the banana plug (once I saw there was an effect I used a plug on its own rather than a lead) the meter reads 0.2uA in 600uA range, 0.07uA with HR, and 0uA in 6000uA range, -0.1uA with HR.

With the plug in, the readings are 0.0, -0.14, 0, -0.3 respectively.

So it looks a bit like the lead detection system is causing at least a part of the issue, maybe they thought to compensate for it in normal res mode but forgot to do so for HR? Here's the circuit diagram, see what you think.

[Amaruk]

Wow - that is quite the discovery! I will have to take a closer look at that too. I'll let you know what I find out on my end. Good find!

[Amaruk]

An update: I can now confirm the finding first reported by AVGresponding. Great job figuring this out!!! 

Here is a short video that shows my Fluke 87V with and without leads connected in both standard and high-resolution mode on the uA range. As you can see in the standard resolution mode the meter goes to 0.0uA as soon as you connect the leads as one would expect (they are not connected to anything in the other end). However, in the high-resolution mode that is not the case. My meter goes to about -0.15uA which is the offset I observed when I ran test measurements in the 0.1-100uA as reported above.


Link to video:

[J-R]

Connect the leads through a decade resistor box instead of a short/DUT and run through the tests again.  Try 100-500 Ohms.  I got some interesting results. 

[EEVblog]

An update: I can now confirm the finding first reported by AVGresponding. Great job figuring this out!!! 

Here is a short video that shows my Fluke 87V with and without leads connected in both standard and high-resolution mode on the uA range. As you can see in the standard resolution mode the meter goes to 0.0uA as soon as you connect the leads as one would expect (they are not connected to anything in the other end). However, in the high-resolution mode that is not the case. My meter goes to about -0.15uA which is the offset I observed when I ran test measurements in the 0.1-100uA as reported above.


Link to video:

My 87V shows 0.3uA on 6000 count mode, and 0.14uA on 20000 count. No change when input is shorted.

[BeBuLamar]

I thought this was a pretty well known issue. All the 87V I've come across have a slightly different offset in low res vs high res and it can drift over time, you just don't notice the drift as much on 6000 count mode as it only effects the LSD.

With nothing connected, one of my 87V reads 0.3uA in low res and 0.17uA in high res, so it's more accurate in high res... If you want LSD accuracy, high res and use of the rel function is your best bet.

keep in mind that 0.1uA on the 600uA range represents an absolute value of ~0.0167% of that range. On the lowest range of this meter..

0.0083% of the ADC range because the range is from -600uA to +600uA

I have 3 87 V here and they seem to agree with yours.

[AVGresponding]

Connect the leads through a decade resistor box instead of a short/DUT and run through the tests again.  Try 100-500 Ohms.  I got some interesting results. 

An update: I can now confirm the finding first reported by AVGresponding. Great job figuring this out!!! 

Here is a short video that shows my Fluke 87V with and without leads connected in both standard and high-resolution mode on the uA range. As you can see in the standard resolution mode the meter goes to 0.0uA as soon as you connect the leads as one would expect (they are not connected to anything in the other end). However, in the high-resolution mode that is not the case. My meter goes to about -0.15uA which is the offset I observed when I ran test measurements in the 0.1-100uA as reported above.


Link to video:

My 87V shows 0.3uA on 6000 count mode, and 0.14uA on 20000 count. No change when input is shorted.

You misunderstand me; I am not using a shorting link, I am just inserting a bare plug with no lead in order to connect the two halves of the uA socket that make up the lead detection system, I didn't connect anything to the common socket at all.

[Amaruk]

Good point regarding the shorting ports. I also did not short anything in my video as my two leads are not connected in the other end so the meter is seeing an open - with and without leads connected to the ports. But as you say, the magic happens when you insert the banana plug in the uA port. The COM port has nothing to do with this at all.

[EEVblog]

Connect the leads through a decade resistor box instead of a short/DUT and run through the tests again.  Try 100-500 Ohms.  I got some interesting results. 

An update: I can now confirm the finding first reported by AVGresponding. Great job figuring this out!!! 

Here is a short video that shows my Fluke 87V with and without leads connected in both standard and high-resolution mode on the uA range. As you can see in the standard resolution mode the meter goes to 0.0uA as soon as you connect the leads as one would expect (they are not connected to anything in the other end). However, in the high-resolution mode that is not the case. My meter goes to about -0.15uA which is the offset I observed when I ran test measurements in the 0.1-100uA as reported above.


Link to video:

My 87V shows 0.3uA on 6000 count mode, and 0.14uA on 20000 count. No change when input is shorted.

You misunderstand me; I am not using a shorting link, I am just inserting a bare plug with no lead in order to connect the two halves of the uA socket that make up the lead detection system, I didn't connect anything to the common socket at all.

Ah, got it.
Yeah, confirmed:

[EEVblog]

FYI, on the BM786 it's zero with no lead, and then when the lead is plugged in it just up for a second or two but then settles back down to zero.
On the 121GW with no lead there is an offset, but it drops to zero with the lead.

[J-R]

The offset in Hi Res mode on my 87-V jumps a bit between -0.02uA and -0.05uA.  (Naked plug in mAuA jack only.)

My somewhat cryptic comment about connecting a decade resistor box is pointing out that there is probably current flowing between the COM and mAuA jacks.  You can see this in the 100-500Ohm range, as the readings change quite a bit.
The best I could do to measure this at the moment was 1kOhms through the Keithley 197A which resulted in about -0.020uA and 3kOhms through the U1461A was -0.0007uA.  I did what I could by using shielded leads, swapping polarities at both ends, etc. but maybe someone with some metrology-grade equipment could jump in.