Author Topic: 4 Wire Measurement Details (Keithley 2000)  (Read 2434 times)

0 Members and 1 Guest are viewing this topic.

Offline forrestc

  • Supporter
  • ****
  • Posts: 573
  • Country: us
4 Wire Measurement Details (Keithley 2000)
« on: September 29, 2017, 08:02:50 am »
I've recently had the need to add measurement of a low value (aka shunt) resistor to the automated test system.   Ran into a problem (which I ended up solving in a different way), but would like to discuss some finer points about the 4 wire measurument on DMMs in general and the Keithley 2000 in particular.

The background:   To support my customers' need for low-current (<10A) 100mV shunts, we've been soldering various current sense resistors onto a carrier board and providing them in the ampacity they need.  These are generally 0.01 ohm - 0.1 ohm resistors with 4 terminals on a tiny circuit board with some terminal blocks.   Because of the very low volume, and lack of decent suitable test equipment, I've never really been able to test these adequately.  Not sure if we've ever had a problem due to missassembly, but it's always bothered me with the lack of testing.  With a revision of the design, and the recent addition of a rack of automated test equipment, I figured it was time.

So envision this:   Keithley 2000 in 4 wire mode, wired through a switch matrix, then through a cable harness connected to the 4 wire shunt.  I figured in 4 wire mode, the keithley should be able to measure the resistance of the resistor with at least enough accuracy that I could verify that the correct resistor was populated (I.E. 0.01 ohm vs 0.02 ohm), and print the correct device label for the board.   Unfortunately, this didn't seem to work.   I would almost always get a reading which was too high.   I.E. A 0.01 ohm resistor would read 0.03 ohms, and the exact reading would vary quite a bit from read to read, but never near the 0.01 ohm reading.

I spent some time trying to figure out what was going on - my best guess is that the matrix/harness wasn't kind to the 4 wire instrument testing.   However, the lead resistance is well under the spec for the range (10% of range per lead (10 ohms), I'm at about 3 ohms).    I also was thinking about noise pickup which is my best guess at this point, just my experience with this matrix is that it is rather quiet and what noise is there tends to dissapear when long term averaging.   Since this is a DC measurement averaged over a full second (10PLC (60Hz), 10 averaging), I would have expected there not to be a noise problem.  I guess there could be some sort of EMF effect, or some other effect which creates a DC bias on the sense wires, but I'm not really sure what it would be.  I would also have expected that the keithley would use a high enough current source for this not to be a likely cause.    Which is sort of where I am and why I am posting...  This type of stuff bothers me since I'm obviously missing some piece of the puzzle.   

It has occured to me that I should see about figuring out how much current the current source that the keithley uses in the 4 wire measurement puts out.   If it is low enough that an 0.01 ohm resistor has so little voltage across it that EMF effects are likely, then that's where I should look.  Unfortunately, this info doesn't seem to be in a spec sheet, and I haven't had a chance yet to go try to measure it.

FWIW, I ended up 'solving' the problem by using the DC supply and the precision DC load (0.1%) which is in the rack to generate 0.25A through the resistor, and measure the voltage drop across it using the same keithley 2000.    I'm getting very good accuracy with this, but I would still like to understand what effect is causing the Keithley 2000 to not work in this environment so it doesn't bite me in the future.   So, any detailed escoteric discussions about the realities of 4 wire resistance measurement of low ohms would be helpful.
 

Offline The Soulman

  • Frequent Contributor
  • **
  • Posts: 697
  • Country: nl
  • The sky is the limit!
Re: 4 Wire Measurement Details (Keithley 2000)
« Reply #1 on: September 29, 2017, 08:49:22 am »
The k2000 100 Ohm test current is specified in the the datasheet, it's 1mA, so yeah not the best tool for this job.
https://www.tek.com/datasheet/broad-purpose-digital-multimeters/model-2000-6-1-2-digit-multimeter

Also have a look here:
https://www.eevblog.com/forum/metrology/low-ohm-precision-resistor-standard-and-testing/
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 2007
  • Country: ca
Re: 4 Wire Measurement Details (Keithley 2000)
« Reply #2 on: September 29, 2017, 01:43:53 pm »
First, I would suggest that directly measuring the resistance of the current shunt might not be the best test method, even if successful and repeatable. This is especially true if it is the only testing performed. But you want to understand why this failed, so...

The Keithley 2000 is a great meter in many respects, but it does have limitations, and the important one in this case is the lack of offset compensated ohms. A meter with offset compensated ohms will read the voltage across the resistor both with the current source turned on and turned off. The difference is used in the resistance calculation. In theory, with zero current through the resistor, it should have zero volts across it, but in practice that is not the case, mostly due to thermally generated offsets in the various connections (including inside the meter). This can make a big difference at the low end of any of the resistance ranges, but especially so in the lowest ranges, where the voltage read is very small due to the limited current used.

A better DMM with offset compensated ohms, such as any of the other Keithley 2000 series (2010, 2001, 2002) would definitely help (I don't know off-hand which HPAK meters do this, but I know the 34401A does not). A sourcemeter (e.g. Keithley 2400 series) could also do an excellent job of measuring resistance at both extremes but is perhaps overkill unless you have another excuse for the high expense. Even better would be a dedicated milliohm meter. These use a much higher current that is typically pulsed into the load to help reduce heating of the DUT. The higher current means higher voltage drop, and much greater ease in achieving high precision and accuracy at very low resistances.

You can attempt a poor-man's offset compensated measurement by setting up your four-wire ohms reading, disconnecting the high side source lead (top right jack on the K2000), zeroing the reading (using the REL button), reconnecting the source lead, then taking the reading. This must be done in a short time frame, and the zeroing must be repeated for every single reading.

You will also need to examine the switch matrix to see if it is suitable. For accurate low level measurements like this you need specialized low thermal offset small signal relays and connectors, particularly on the sense connections. If you use generic power relays or generic switches on the sense connections you have set yourself up for a headache. Power relays require a minimum switching current frequently, to maintain their contacts. It sounds counter-intuitive but it's true. Otherwise oxides build up, presenting both high, unstable resistance, and excessive thermal EMF due to metal-oxide-to-metal connection. Small signal relays' contacts remain clean and usable even with virtually no current. They are often made of inoxidable metals or are mercury-wetted (though not so much anymore). Special low thermal EMF relays are designed for a small thermal EMF difference between pairs of contacts... and you must always use a pair of contacts in a single relay for the two sense connections!

You might want to read Keithley's "Low Level Measurements Handbook" if you haven't already. It is in its 7th edition now, and is perhaps the best source of information on this topic... if anyone knows of something better, post it, I want to read it!
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 7035
  • Country: de
Re: 4 Wire Measurement Details (Keithley 2000)
« Reply #3 on: September 29, 2017, 04:06:09 pm »
Wiring 4 terminal shunts in parallel is a little tricky. It usually needs extra resistors in the sense lines to get good averaging.
 

Offline forrestc

  • Supporter
  • ****
  • Posts: 573
  • Country: us
Re: 4 Wire Measurement Details (Keithley 2000)
« Reply #4 on: September 30, 2017, 01:42:27 am »
The k2000 100 Ohm test current is specified in the the datasheet, it's 1mA, so yeah not the best tool for this job.
https://www.tek.com/datasheet/broad-purpose-digital-multimeters/model-2000-6-1-2-digit-multimeter

Well there it is...  I looked at the spec sheet *a lot*, not sure how I missed that very obvious column labeled "TEST CURRENT OR BURDEN VOLTAGE".   Now I feel a bit silly saying it wasn't there....

1mA * 0.01 ohms = 10uV if I did the math correctly.   I could believe ~20uV of noise/EMF effects...  Especially through the matrix.

Also have a look here:
https://www.eevblog.com/forum/metrology/low-ohm-precision-resistor-standard-and-testing/

Will do, looks like a good thread.
 

Offline forrestc

  • Supporter
  • ****
  • Posts: 573
  • Country: us
Re: 4 Wire Measurement Details (Keithley 2000)
« Reply #5 on: September 30, 2017, 02:20:14 am »
First, I would suggest that directly measuring the resistance of the current shunt might not be the best test method, even if successful and repeatable. This is especially true if it is the only testing performed.

This was more of the 'try this and see if it works' method.   The way I ended up doing it is probably the best I can do with the available test equipment, but I definitely agree that a direct measurement isn't the right method, now I see the spec for test current....and:

The Keithley 2000 is a great meter in many respects, but it does have limitations, and the important one in this case is the lack of offset compensated ohms.

Now this makes more sense.    I figured something like this was the case, and after finding the spec for the test current, I was pretty confident that that was the reason.  Understanding that the 2000 doesn't automatically adjust for DC offsets, this makes perfect sense.   I had considered using the matrix to take a relative reading and subtract it (either in software or in the meter using the rel function), but I figured if I was getting those type of effects, that I couldn't be sure that disconnecting a lead via the matrix was the right thing to do since I couldn't be sure that the matrix was disconnecting the right part of the circuit.  I also didn't really have a good picture in my head (I do now, thanks) of what was really going on here...

You will also need to examine the switch matrix to see if it is suitable. For accurate low level measurements like this you need specialized low thermal offset small signal relays and connectors, particularly on the sense connections.

Yeah, not even close.  The matrix is full of low-signal relays (0.5A ratings), but not arranged in a way that is even going to come close to working for this application now I understand the limitations of the Keithley 2000.   I'm definitely mindful of the current limitations (too high and too low) across a relay.   This particular matrix gets a fair workout most days, and the 0.5A reed relays I use seem to be fairly immune to the low-current issues so mostly I'm worried about the almost-to-high currents causing premature contact failure.  I definitely know that it wasn't likely to work if we were talking very low signal and noise-sensitive measurements.   I was actually pleasantly surprised with the accuracy of reading a 10A shunt with 0.25A of sense current through it.

You might want to read Keithley's "Low Level Measurements Handbook" if you haven't already. It is in its 7th edition now, and is perhaps the best source of information on this topic... if anyone knows of something better, post it, I want to read it!

I'll hunt this down and read it.   
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf