4-wire short - 2020

[coppercone2]

the instrument rack they tested it on is a bit suspicious, i suspect there could be a draft there from the instrument under it possibly, and its unknown what alloy its made of, and how the banana plugs are attached to the copper

[tooki]

* 18 mm long, 4 mm gold plated brass banana plugs with twisted springs from Stabuli (MultiConnect) with 0,2 mOhm contact resistance(springs are made from rigid hard-drawn copper alloy sheet)

FYI, that’s “Stäubli” (pronounced ‘shtoyblee’), formerly Multi-Contact. Nice choice, I REALLY like their products.

One modification I've made on banana pins to be sure there is good contact is to slide the springy sleeve thingy to the end where the wire connects and solder the sleeve just at that one end. It is important to slide the sleeve to one end before soldering or the sleeve may not compress properly. I found with some pins that may be of lesser quality or contaminated that there wasn't the best contact between the sleeve and the inner pin. I first noticed this on some leads that were rated 20A (and had 254 strand wire) That I was testing at rated load that had slightly higher drop than I expected. If I were using these leads with a power supply I'd use remote sensing to compensate for any drops but I wanted to see how the leads reacted to carrying 20A.

In this application and with good quality gold plated pins like those shown, and basically no current flow, this probably isn't a real problem.

Don't do this to a Stäubli plug. These aren’t el-cheapo plugs, in fact, they’re ultrapremium ones that make Pomona look like trash. (And I say that as someone who likes Pomona products!) They absolutely do not need that to make good contact. In bad plugs it’s needed to overcome excessive tolerances and awful plating. Stäubli plugs have excellent plating and insanely good tolerances. The spring “basket” (as they call it in German) fits very snugly onto its shaft, requiring a bit of force to move.

Also, wouldn’t solder potentially cause more thermal EMF problems, since it introduces another metal to the mix?

[Cerebus]

Also, wouldn’t solder potentially cause more thermal EMF problems, since it introduces another metal to the mix?

Not really, think of each dissimilar metal to metal contact as a little temperature dependant battery, then draw a circuit of the metal/solder/metal contact - two temperature dependant batteries back to back with close thermal coupling => no potential difference across the whole assembly, the two dissimilar metal interfaces cancel each other out. Still, like you, I don't like the whole idea on principle.

[tooki]

Also, wouldn’t solder potentially cause more thermal EMF problems, since it introduces another metal to the mix?

Not really, think of each dissimilar metal to metal contact as a little temperature dependant battery, then draw a circuit of the metal/solder/metal contact - two temperature dependant batteries back to back with close thermal coupling => no potential difference across the whole assembly, the two dissimilar metal interfaces cancel each other out. Still, like you, I don't like the whole idea on principle.

Ohhhh, gotcha, that makes complete sense!

[mawyatt]

Hi,

I am going to share some measurements that I made using an IET LOM-510A Micro-Ohmmeter.

Link: https://www.ietlabs.com/lom-510.html

This may be one of the best instruments available for measuring very small resistances.

The 4W-shorts are directional. The best results are obtained if you connected

Source + to Source - with a thick trace

Sense+ to Sense- with a thick trace

a short, wide trace connecting the two together.

I am going to call this the preferred orientation. The other orientation I will call non-preferred.

Keithley Design

Preferred Orientation

(Attachment Link)


This reads 0.000 m

This make sense, you are measuring a very short, very wide track.

Non-Preferred Orientation

(Attachment Link)


This measures 20.23 m

You are measure a long, thin track.

Universal Design (Keithley Style)

Preferred orientation

(Attachment Link)


This is a very wide, very short track.

Non-Preferred Orientation

(Attachment Link)


28.17 m\$\Omega\$. This makes sense, the track is a little longer than the design above.

Felixd Design

Preferred orientation

(Attachment Link)


4 u (4 Micro Ohms)

Non-Preferred Orientation

(Attachment Link)


1.389 m

The 'X' pattern is not as sensitive to the orientation as the Keithley 'H' style design. But, it is not as good as the Keithley 'H' in the preferred orientation.

Note:

The Datron 1281 requires a different 'H' design than the HP 3457A. The 'H' has to be rotated 90^o.

Regards,
Jay_Diddy_B

That's some really good 4-wire short readings!!

I get ~50 micro-ohms with a KS34465A using the resistor calibrator rotated shown above your post.

Best,

[DavidKo]

After many thoughts I have created version with 2 distances and various connecting wire width.

If someone interested the KiCad pcb files and gerber files from production to https://github.com/konarik/4-wire-short.

Some pictures below

[bdunham7]

After many thoughts I have created version with 2 distances and various connecting wire width.

If someone interested the KiCad pcb files and gerber files from production to https://github.com/konarik/4-wire-short.

Thanks for posting that.  What banana plugs did you use?

[DavidKo]

I have used https://za.rs-online.com/web/p/banana-connectors/2263095

In my case I have ordered them from Germany https://www.buerklin.com/en/4-mm-Panel-plug-Screw-connection-0-5-mm%C2%B2-4-mm/p/12F168

I already find some flaws/possible improvements:

• distance is not marked on the PCB
• Yokogawa uses different distance between posts (probably 20mm)
• connections need to be cut with a hack saw - the holes to them should be added
• maybe it is better to pick one width of middle wire and use various spacing between holes to have one universal PCB

[sequoia]

If someone interested the KiCad pcb files and gerber files from production to https://github.com/konarik/4-wire-short.

Thanks! 

Btw, I tested uploading the zip file from github to JLCPCB, and from the preview (and 3D gerber viewer on the size), it looks as if the 3x3 panel won't be routed?
Only showing holes for the banana plugs....  Is that to be expected?

[DavidKo]

If someone interested the KiCad pcb files and gerber files from production to https://github.com/konarik/4-wire-short.

Thanks! 

Btw, I tested uploading the zip file from github to JLCPCB, and from the preview (and 3D gerber viewer on the size), it looks as if the 3x3 panel won't be routed?
Only showing holes for the banana plugs....  Is that to be expected?

In jlcpcb 3D viewer you should see thin milled shape, but the scrap material is still there. I have not found what I'm doing wrong. In theory the scrap material shoud not be shown.

Final result you can see on the picture in my post https://www.eevblog.com/forum/projects/4-wire-short-2020/msg3821228/#msg3821228.

[jonpaul]

Bonjour a tous: Just found this post now, fascinating reading.

I think this iis a bit overwrought, and wonder  any PCB at all? Why the 4 wire layout?

My experience leads to a  simple plane of copper.

Also  Both DCV (2 term) and resistance (4 term) should be checked.


Experiment: Using a new Keysight 34465A    warm  up 2 hrs , ACAL. 

I have a variety of shorting plugs,   some copper foil,   gold plated foil and  gold plated bananas,  (eg from  Gen Rad standard resistor).

With the gold bananas bare copper foil across all 4 connectors, OR 4 mm wide copper braid no plating,

About same with gold plated copper sheet or HP CAL shorting from 4195A. \\

DCV 200 nV

4 wire R ~500...800  uOhm

Good enough for me,  the meter seems in CAL and in SPEC.

problem solved, NO fancy 4 wire PCB just a stupid sheet of copper foil and 4 holes.

Your thoughts and critique?

Bon Soiree,


Jon

PS: The fancy bananas seem to have a very tight fit into the 34465A input, hard to remove.

 

[tooki]

Bonjour a tous: Just found this post now, fascinating reading.

I think this iis a bit overwrought, and wonder  any PCB at all? Why the 4 wire layout?

Because of thermal EMF. See the attached PDF.

[jonpaul]

Tooki: Fine paper on thermal EMF. I do not have a level of equipment like the Fluke 8508A of Mr Kidd , just the Keysight  34465A

I believe that  Thicker copper (than a PCB) gives more isothermal system.

Found a Copper plate 44 x 76 mm ,  3.175 mm thick,  drilled for gold bananas, and added gold foil between the bananas  and the plate.

The 34465A is approaching 1 uV as it stabilizes.

Finally I wonder why this thread is not in the Metrology forum?

With Kind Regards,



Jon

[tooki]

I believe that  Thicker copper (than a PCB) gives more isothermal system.

This concept has been proposed, and the response why it's a bad idea, earlier in this thread:

Why not just get a thick copper block, then drill and tap 4 holes in it to make the short, without the PCB?

If you use thermal shielding compound (its like wet tissues), you can braze the banana jacks.

I think it would be better because then you don't need to buy a PCB to make it. It should work as one piece and spread heat better also, from touching it, then you can add bakelite handles to make 'handling' insulators. I think clamping on epoxy is always going to be less stable then solid metal, I am just not sure about how thermal EMF looks like, but if its built heavy, it should equalize really well?

@Coppercone2
That would work as a low resistance, but due to its mass it would never reach thermal equilibrium, and would induce a large thermal emf.
[screenshot from the same Fluke article]

[jonpaul]

Very interesting discussion,

 I am getting 1..2 uV with the plate.

 Fluke paper mentions bare Cu wire as low mass and low thermal.

 I have some #24ga Bell System unplated, solid Cu insulated PVC wire, I will make another test with just the bare Cu and then the bare Cu wire on the gold bananas.


Jon

[DavidKo]

But this thread is about 4 wire short, not 2 wire like on your picture.

I do not want to say that you are wrong. I have had the same idea at the start (take Cu plate, drill 4 holes and put bananas), but than I have read more about that and changed my mind. For sure you can use bare copper wire like is mentioned in HP 3458A manual.

[Cerebus]

The point is being missed here. Yes, a simple wire short, of low thermal mass, is best. It is however completely useless with a meter with shrouded sockets instead of binding posts - e.g. HP 34401A. It is for those style of meters that these PCB 4 wire shorts are designed.

[Shonky]

Bit of a thread dig. I took @felixd 's design and made some "improvements"

- There were a number of "bonus" (i.e. duplicate) tracks some of which threw up some DRC warnings. One was a track right in the middle of one of the holes
- I made the tracks a bit smaller including the link track
- I changed the soldermask to expose the link tracks so when getting ENIG finish, the whole track is ENIG. There's not a huge point in getting ENIG otherwise. You're just going gold to gold from the banana plugs on the pad then to the copper track then back to gold on the other pads. Yes this should largely cancel out any thermal EMF but better to just stay gold all the way. This necessitated moving the silkscreen

Pictures from my JLCPCB order.




Update PCB file and gerbers attached too.

I have a couple of spares that I can share basically at my cost (about $7 + say A$1 to post in Australia), if anyone wants one. Or just order your own from JLC or whoever.

PM me if you want one.

[coppercone2]

if you really wanna be precise don't you need a tellurium copper PCB and tellurium copper plugs? preferably welded on.

harder to make though. Best one would probobly be to weld 4x plugs to a copper plate, put it upside down in a fixture so the plugs are underneath, then grind it thin, glue it to a panel, then use a press to cut out the shape. and glue it to a thicker substrate 

no interconnects then.

preferable equipment is a custom vacuum clamp grinding block, cbn abrasive wheel, micro plasma or micro tig torch, helium box, custom space rated bakelite panels, space rated epoxy. Not sure about cutting it into shape, I guess a custom die, or maybe a fast sharp router is good enough.

[Shonky]

Impressive sarcasm.

[coppercone2]

not really someone is probobly going to need the stability eventually?

the other thread had super conductor shunt talk, that means copper shunts on superconductive substrates for 100 nanoohms

[donlisms]

Frank Wenner's article (a hundred years ago) on four terminal conductors has a whole lot of good theory that applies directly to these shorting devices, though that wasn't its original purpose. It shows why some of the intuitive approaches won't work, and provides background for how to do it right.  (It also walks right up to the door of tetrajunctions, without really going through.)

The other thing that always bugs me around these discussions is the lack of good explanations of Seebeck and thermal EMF.  The potential difference(s) are within conductors that have different temperatures at their ends.  A junction between the two conductors merely provides a place for the current to flow between them, and generally, a place to make one end of each conductor the same temperature.  The difference in the two potentials appears between the other (non-joined) ends.

It is not the junction that "generates" the voltage, as seems to be commonly thought.  It took me nearly forever to dig up good explanations of this; there are so many explanations, even from otherwise authoritative sources, that say, or imply, that the voltage "comes from" the junction - the point of contact between the two metals.  Having a more correct explanation helps deal with the issue more effectively, e.g. why plating has no effect, and other ideas. 

[coppercone2]

A junction is what everyone talks about because there is documentation that supports this, i.e. the voltage generated by different types of solder joints, and the use of specialty solders to reduce the effect. That is identical structures will generate different voltages depending on what solder you use, according to the documentation.

And a standard thermocouple works by measuring the junction temperature,and different wires make different voltages.


I don't see how you could get around talking about the junction? short and long thermocouples seem to work pretty much the same other then noise pick up. so do different wire gauges.

[donlisms]

I suppose the main point is that the junction is not where the voltage "comes from".  The potential is (pretty much) the same in both sides of the junction - the thermocouple voltage does not appear there.  It's a different way if thinking about things than what some writers present. 

[coppercone2]

but its important to think about the joint because its the parameter that you can change to reduce the effect.

the plating increases drift by a order of magnitude

https://www.edn.com/wp-content/uploads/5.3.01_Minimizing_thermocouples_maintains_20-bit_DAC_precision.pdf