Kelvin Probes

[trilerian]

I have a GwInstek GOM -804 milliohm meter.  I have the kelvin clips that came with it.  But the clips are difficult to use sometimes, especially since the main reason I got this was to measure the stator resistance of bldc RC motors.  Our "governing" body uses the 802 as the official test meter.  Anyway, I am looking for a set of probes to use instead of clips.  When I search online there are a few results over $100 for a set.  I just want to make sure if I spend that much they would work.  Also, would there be a big difference if I just used two sets of probes with maybe a 3d printed holder to hold them at the right angle so they almost touched?

Thanks

[bdunham7]

Yes, 4W probes with the wires connected at the tips work, but there is always a small error due to the resistance of that last little section of tip.  There are a variety of configurations, but this is an interesting type I haven't seen in person.  They appear to actually be 4W all the way to the connection.

https://www.aliexpress.us/item/3256804254982399.html

[trilerian]

Yes, 4W probes with the wires connected at the tips work, but there is always a small error due to the resistance of that last little section of tip.  There are a variety of configurations, but this is an interesting type I haven't seen in person.  They appear to actually be 4W all the way to the connection.

https://www.aliexpress.us/item/3256804254982399.html

Well those are cheap enough to give them a try.  Thanks and ordered.  I will measure a few motors with my clip leads and then these and post the results when they get here.  About 2 weeks.

[J-R]

I thought it would be good to clarify a couple things:

There are products out there being sold as "Kelvin Probes" where the force and sense are connected as close to the tip as possible, but technically and practically they are not Kelvin or 4-wire.

I've performed some tests in the past where the force connection was RIGHT AT the tip of the sense.  But even in this extreme case, there are still only two physical connections to the DUT, and this method introduces significant error.  So I don't see the point, especially considering how trivial it is to simply split out the connections.  That is really all that is necessary.  No fancy leads, just ensure that the force and sense are touching the DUT at different locations to ensure negligible current flows via the sense contact point.  (Force should be outboard of sense, although the impact of this being reversed is typically negligible.)

But these "Retractable 4-Wire" probes would essentially be considered a true Kelvin connection.  The inner cylinder is spring loaded and insulated from the outer portion.  They appear to be based on a Hioki product.  There is also another design with two separate spring loaded tips.  High force currents could be an issue with both designs.

[alm]

Two other "true Kelvin" probes I used is these Parrot Kelvin probes (I don't love the mechanical design) and the Keithley 5805 Kelvin probes. The Keithley probes are much better built with sharper tips, but probably expensive new. And the two points are kind of far apart, so they are not ideal for use on PCB traces. But they are great for bigger parts.

[bdunham7]

I've performed some tests in the past where the force connection was RIGHT AT the tip of the sense.  But even in this extreme case, there are still only two physical connections to the DUT, and this method introduces significant error.  So I don't see the point, especially considering how trivial it is to simply split out the connections. 

What level of error are considering to be significant here--DMM-level measurements or micro-ohms?  The residual resistance of good 2X4W probes will be under a milliohm.  Even a 'true' Kelvin connection can suffer from contact resistance and thermal voltage issues at very small signal levels like the ones that you get from the small test currents of a normal DMM.  But the real reason the 2-probe solution is still relevant and useful is for probing small components where you just can't get any true Kelvin connection--even the coaxial probes I posted are probably much too large to work on small SMD devices and traces.

[mawyatt]

Where the actual Force and Sense point terminates is critical for accurate repeatable milliohm levels. One reason we've developed an LCR meter SMD Fixture capable of accurate and repeatable low Z measurements, since experience with such proved difficult.

The concept with the SMD Low Z Fixture is to allow the DUT to make the Force and Sense connection, rather than the probe. The SMD DUT ends are conductive and solderable, and these "ends" create the connection.

Here's some images of the SMD Low Z Fixture (and the LCR DC Bias Adapter in background). The concept should be independent of the surface contact resistance, at least at the higher level, maybe not so much at the microscopic level tho, as the Force and Sense are terminated as the DUT connects the "split" surface contacts (see images, zoom in).

Later we plan on getting another PCB with gold plating to improve surface contacts and slight layout change (if interested let us know). This has proven quite good as shown, but does require a little fiddling with the DUT position to get good surface contacts, which there are 4 of, thus Kelvin Type sensing!!

BTW one thing we found out with this fixture was Zero Ohm SMDs ain't even close to 0 ohms, even the 2512 types

We know, we know, RTF Data Sheet

Best,

[J-R]

Yes, agreed, the error can be just below 1m Ohm with some extra care, but around 5m Ohm was the raw error in my tests.  To even get below 1m Ohm you have to zero leads using consistent contact points, such as zeroing out on the same side of the DUT first.  Compare this to a true 4-wire setup where you can be 100 times better with basically no drama.

The entire point of 4-wire is to avoid these errors at low resistances, so that is what I meant with my "significant' statement.

But from another post where I used the 1mA precision source on the DMMCheck Plus to measure it's own 100 Ohm resistor, I saw a 4m Ohm raw error when not using a real 4-wire setup.  It was easily visible and reproducible on my 6.5 digit DMM.  Trying to zero that out wasn't practical due to the noise.

With regard to the thermal effects, they can be visible in certain conditions but essentially 99% of the time both connections generate a similar voltage but in opposite directions, so they cancel out.  The risk is far higher with the "fake" 4-wire probes, lower with the "true" 4-wire probes and around zero with a Kelvin clip because the portion you grip is far away from the connections and you also aren't going to be holding it as much.

There are some products on the market with spring-loaded pins that could help in some situations.  I'd probably use those for the sense and make the force connections by hand during the measurement, especially at higher currents.

[alm]

The contact resistance between the tip of pseudo-Kelvin probes and the surface also depends a lot on the surface. If the surface is rusty or otherwise has a high contact resistance, the error from only having a single point of contact will increase. I'd say the contact between the probe tip and DUT is the least reproducible part, so the part that most needs to be four wire for a reliable measurement.

Regarding offset voltages, decent milli/microohmmeters will have solutions for compensating this (AC excitation, switching polarity, turning off the current source), and so will the better high-resolution DMMs (often called something like offset compensated ohms or true ohms).

[kloetpatra]

Yes, 4W probes with the wires connected at the tips work, but there is always a small error due to the resistance of that last little section of tip.  There are a variety of configurations, but this is an interesting type I haven't seen in person.  They appear to actually be 4W all the way to the connection.

https://www.aliexpress.us/item/3256804254982399.html

I bought those tips and had a closer look. The quality seems good. Probe handle is silicone overmold. The cable is PVC, but it's more flexible than cheap PVC probes. I have done short circuit measurements on a copper ground plane and got reliable contacts between multiple measurements within +- 0.1 mOhm. It's a true 4W-Probe. The spring tip is not connected to the outer contact ring. The tip material seems to be pure copper with gold plating (I scratched it). It looks like the spring probe can be unscrewed for replacement, however I could not unscrew it and I did not want to damage it. The insulation of the spring contact seems to be PTFE.

Attached are some pictures for size comparison and details.

-- Edit:
I doubt the spring tips are replaceable. However I found another product where they really are replaceable:
https://www.aliexpress.us/item/3256804254551420.html

[alm]

I doubt the spring tips are replaceable. However I found another product where they really are replaceable:
https://www.aliexpress.us/item/3256804254551420.html

If this price is accurate for a pair of replacement tips (they look like the 2mm version), then I'm not sure if I'm interested..

[ajb]

Unfortunately not a product you can buy, but (former?) forum member robrenz made an amazing set of probes with an extremely fine split point, so they can be used just like regular probes.  The tips are made from split carbide pins, bonded together with a thin insulating layer of epoxy and ground to an extremely sharp pyramid point.  He demonstrates how effective the tips are at 33:33 in this video: https://youtu.be/0wbC8fOKswM?t=2018

He has a series of videos on the whole system of kelvin clips and probes he made that is an absolute tour de force of mechanical design and miniature manufacturing and well worth watching if you're into that sort of thing.

https://youtu.be/wwgO6Lg4RZM
https://youtu.be/0wbC8fOKswM
https://youtu.be/coqgyv4FPGY

[F4]

Keep your lead ends clean, I shorted a new set of Kelvin leads and got a .019 ohm reading. Cleaned the tips with methanol, the shorted reading went down to .003 ohms. Quite a difference.

[J-R]

Keep your lead ends clean, I shorted a new set of Kelvin leads and got a .019 ohm reading. Cleaned the tips with methanol, the shorted reading went down to .003 ohms. Quite a difference.

I don't see how this would make a difference when making a real resistance measurement.  The entire point of the Kelvin/4-wire connection method is to essentially eliminate error due to contact and lead resistance.  Put a resistance substitution box into the force and sense parts of your circuit and start cranking up the values.  You'll see what is actually required to impact the measurement.

Also, when clipping alligator-style Kelvin clips to each other, make sure that the sense half of each clip is on the outside, with the force on the inside.  This may explain the observed discrepancy.

[F4]

A shorted 4 wire does read a resistance value. My assumption is that I cleaned a layer of surface something from the leads and changed the baseline resistance reading. Something similar was mentioned earlier in the thread. I think my readings are correct. I measured the relatively high resistance several times, did the cleaning and repeatably got the lower result.

[J-R]

Yes, I understand you are seeing this, and I ran some quick tests last night with my 34461A where I observed a 10mOhm difference (negative, actually) depending on how the Kelvin clips were clipped together.  I was just sharing my suspicions in that regard.

HOWEVER, this means nothing!!  When using the Kelvin method to make an actual resistance measurement of a physical object, the contact and lead resistance of the force and sense simply do not impact the measurement (up to hundreds of Ohms, in most cases).

So I still have to push back and say that cleaning your Kelvin/4-wire leads is just not a thing, by design.  Thanks to Lord Kelvin.


Today I ran some more experiments using my Keithley 2010 which has some relatively impressive 4-wire capabilities and actually the only valid numbers I got were with the sense terminals shorted directly together.  Every other arrangement gave negative numbers.  As far as the test equipment is concerned, the lowest (valid) resistance value is obtained with the lowest possible voltage on the sense terminals.  In this configuration, no current is flowing, so resistance is undefined (R=V/I, can't divide by zero).  On my K2010, this value was 0.018mOhms. Clipping the Kelvin clips to a gold-plated shorting bar (almost touching each other), I got 0.034mOhms.  A 1-inch piece of 10AWG copper wire was 0.104mOhms.  Subtracting 0.018mOhms from that results in 0.086mOhms which is amazingly close to the published value of 0.083mOhms per inch.

[tautech]

I have a GwInstek GOM -804 milliohm meter.  I have the kelvin clips that came with it.

A shorted 4 wire does read a resistance value. My assumption is that I cleaned a layer of surface something from the leads and changed the baseline resistance reading. Something similar was mentioned earlier in the thread. I think my readings are correct. I measured the relatively high resistance several times, did the cleaning and repeatably got the lower result.

Use the 'Measurement of REL, Actual and % Value' features this DMM supports.

[J-R]

I believe F4 has a Keithley 2001, so in that case he shouldn't use the REL feature along with the shorting methods he's presented so far because that will introduce a significant error (at least 3mOhms).

[F4]

I believe F4 has a Keithley 2001, so in that case he shouldn't use the REL feature along with the shorting methods he's presented so far because that will introduce a significant error (at least 3mOhms).

Is there a zeroing method?  Yes, I have a 2001.

[J-R]

4-wire shorting plug: https://www.tequipment.net/Keithley/8620/General-Accessories/
There are DIY and other cheaper options available.

I labelled my Kelvin clips with F and S and ran through more tests with the Keithley 2010 (offset compensation enabled):
S-F-F-S  =  -0.001xxx
S-F-S-F  =  -0.001xxx
F-S-S-F  =  -0.001xxx
S-S-F-F  =  0.000019 (100 average)

Also, to clarify, these are the readings based on the configuration on the clip side.  If I reverse the polarity of the sense at the DMM side, then all those negative numbers are positive, but still invalid.  If you use REL to null that reading out, you will be WAY off when making a real measurement.  With the Keysight 34461A, I got similar results, -0.001x on the first three and and a tiny positive reading on the last.

So I think the short answer is that you normally shouldn't zero out your 4-wire setup, but if you really want to hit the volt-nut territory, you can use that one specific arrangement and perhaps null a TINY offset that might actually be due to a need for a calibration adjustment of the DC zero.

Also, a quick photo showing you can really use anything when making a true Kelvin measurement.  No special leads or clips necessary.