What kind of wire and connect do you use for 3458 measure of 7-10V references?

[notfaded1]

I'm curious what gauge and types of wire you use.  Also what connectors?  I have some solid copper blades and am wondering what gauge and type of tire you use with these for precision 7-10V reference measurements?  I'm talking fluke and other LTZ1000 type references and measured with HP3458A.  I'm looking at some mil spec wire.  Some come in multiple conductor shielded versions which seem to make sense.  I've even found some 5 conductor versions which would seem to be great for 4 wire measurements with ground for case to case connections.  Do you want solid core or stranded?  How large of gauge of the wires do I really need?  I'd like to use some wire that can be between 20-24 AWG because it'll work with avionics type crimped machined male and female pins as well in case I want to also use them for hooking up precision frequency references like some of my Rubidiums.  Is that a good gauge range also for precision voltage and resistance measurements?

Bill

[CalMachine]

Here is what I use and have had good results with. 

If it's a cable that will be swapped often, I use gold plated copper or bare copper spade connectors (crimped only).  As for the cable itself... I use this type cable cable that is used in Fluke's precision cables that are supplied with 5720A's and the like.  Part number 5440-7003.
https://catalog.belden.com/index.cfm?event=pd&p=PF_8719

If it's a cable that I make the connection and just 'set it and forget it', like in my volt array at home.  I use this type of cable with no connector applied.  Make sure the ends are clean and twisted.
https://catalog.belden.com/index.cfm?event=pd&p=PF_5500FE

[notfaded1]

If it's a cable that will be swapped often, I use gold plated copper or bare copper spade connectors (crimped only).  As for the cable itself... I use this type cable cable that is used in Fluke's precision cables that are supplied with 5720A's and the like.  Part number 5440-7003.
https://catalog.belden.com/index.cfm?event=pd&p=PF_8719

If it's a cable that I make the connection and just 'set it and forget it', like in my volt array at home.  I use this type of cable with no connector applied.  Make sure the ends are clean and twisted.
https://catalog.belden.com/index.cfm?event=pd&p=PF_5500FE

This looks good... The second wire would work with the avionics crimp pins as well.  The copper spades I have are solid copper and can handle thicker gauge wire like the first the listed.  If it's good enough for Fluke... how can you argue with that!  I also noticed that both of those Belden wire are I think Concentric stranded wire. Each layer of the stranding has all strands in the same direction and position.  I have a few new LTZ1000 references and a Fluke 732A I'm putting new batteries in and replacing the caps I can reach.  I've used Beldon wire for network connectivity for years and have had good luck with their datatwist and mediatwist wire.  I even though of using the same wire I use for 10G copper network connections but thought something else might be better.

Bill

[essele]

Can I hijack this thread with a slightly different (but related) question:

What is the best wire to use for "internal interconnects" in high precision references? For example you might have a soldered connection from a reference board going to a (probably soldered) binding post inside the enclosure.

I notice a lot of people use some form of twisted pair (which makes sense), so far I've mostly been using some silicone test wire, mainly because it's pure copper rather than tinned, but it's too flexible to twist sensibly.

So the main questions:

1. Is pure copper better in this case, or does the fact that it's soldered negate any benefit over tinned?
2. My reference is in a highly temperature stable oven (temp still to be firmed up, but likely 40ish C) but obviously the binding post isn't ... does that impact choices?

Thanks.

[CalMachine]

Can I hijack this thread with a slightly different (but related) question:

What is the best wire to use for "internal interconnects" in high precision references? For example you might have a soldered connection from a reference board going to a (probably soldered) binding post inside the enclosure.

I notice a lot of people use some form of twisted pair (which makes sense), so far I've mostly been using some silicone test wire, mainly because it's pure copper rather than tinned, but it's too flexible to twist sensibly.

So the main questions:

1. Is pure copper better in this case, or does the fact that it's soldered negate any benefit over tinned?
2. My reference is in a highly temperature stable oven (temp still to be firmed up, but likely 40ish C) but obviously the binding post isn't ... does that impact choices?

Thanks.

For this I would probably use something similar to the Belden 5500FE, but slightly larger gauge.  Say 20 gauge or so, and just strip out the internal conductors.  Make the solder connections to the CCA and then attach ring terminals on the end meant for binding posts.  Sandwich rings between the 2 nuts that are provided on the golden Low Thermal binding posts.  I would definitely choose Low Thermal's binding posts in any LTZ reference applications...  The pomona 3770's seem to have a problem with the plastic cracking.

[niner_007]

Here is what I use and have had good results with. 

If it's a cable that will be swapped often, I use gold plated copper or bare copper spade connectors (crimped only).  As for the cable itself... I use this type cable cable that is used in Fluke's precision cables that are supplied with 5720A's and the like.  Part number 5440-7003.
https://catalog.belden.com/index.cfm?event=pd&p=PF_8719

If it's a cable that I make the connection and just 'set it and forget it', like in my volt array at home.  I use this type of cable with no connector applied.  Make sure the ends are clean and twisted.
https://catalog.belden.com/index.cfm?event=pd&p=PF_5500FE

Where do you order these from in small quantity (or length)?

[CalMachine]

Here is what I use and have had good results with. 

If it's a cable that will be swapped often, I use gold plated copper or bare copper spade connectors (crimped only).  As for the cable itself... I use this type cable cable that is used in Fluke's precision cables that are supplied with 5720A's and the like.  Part number 5440-7003.
https://catalog.belden.com/index.cfm?event=pd&p=PF_8719

If it's a cable that I make the connection and just 'set it and forget it', like in my volt array at home.  I use this type of cable with no connector applied.  Make sure the ends are clean and twisted.
https://catalog.belden.com/index.cfm?event=pd&p=PF_5500FE

Where do you order these from in small quantity (or length)?

Ebay usually works for me

[Dr. Frank]

Sorry for the bad quality of the picture..

For decades, I used multi contact cables LK4-B, with PVC isolation, and gold plated LS4 plugs, which were cold-welded to the Cu cables, and KT4-B isolation grip. They were 25cm and 50 cm long, and give low thermo electric voltages on the order of 1..2µV. The plugs can be stacked.
Disadvantage are the bad isolation of PVC, therefore cables can't be twisted for Ohm measurements, no shield, and normally can be used only up to 60V 

Recently, I got metrology grade PTFE cable, I think it is Belkin, 22 AWG Cu, silver plated.
I assembled them myself, by crimping gold plated spades, fitting the 3458A, or soldering these LS4 plugs. All cables have a shield, very high isolation of >10¹⁴ Ohm, and withstand > 1kV.
I have several sets for DCV and 2W/ 4W Ohm measurements.
The crimped versions show a few 100nV e.m.f., the plug version again around 1µV.

Additional hint: DCV measurements of e.g. 7 /10V references have a systematic problem, i.e. to cancel the e.m.f. because you can't short the source.
The metrological trick is to take a set of 16 samples, then another set with reversed polarity at the D.U.T.
The 3458A, especially when calibrated for zero and 10V gain, has extremely good linearity also at reversed input. Therefore, the difference of both values divided  by two gives quite precisely the true voltage of the D.U.T. w/o e.m.f. , and the sum divided by two gives the e.m.f.

That procedure gives reproducible results also for the lower grade cables, therefore it's more important than the cables used.

Inside my references, I use hard-lacquered Cu wire of I think 200µm diameter, not twisted.

Frank

[Kleinstein]

Reversing the source does not need a high linearity. It is more like measuring with both polarities not only compensates for offset but also for some of the INL error (the even powers if expressed as a polynominal). So a not so good meter can take even more advantage of this.
However one can not get the offset any more, as the sum divided by 2 gives the thermal EMF error plus the meters turn over error.

[IanJ]

Hi all,

For calibration I use LEC1 4-Wire Low EMF Cable (shielded PTFE) with gold spades at one end and gold banana's at the other.

For quick testing I use a home made silver plated over copper / PTFE / gold banana's twisted pair un-shielded cable.

All crimped or screwed, no solder.

Ian.

[splin]

Ideally you would not use any multi-core cable - so that excludes all those ready made cables, twisted pairs, CAT 5 etc. From page 21 of the ADA4523 datasheet:

Even junctions of copper wire from different manufacturers can generate thermal EMFs of 200 nV/°C

Each core in a cable will come from seperate reels and there is no guarantee that the reels will come from the same manufacture let alone from the same batch of copper. Therefore you should make your cables from a single piece of wire cut in half. Screening might be a bit tricky though!

In practice most cables are probably fine but you should test them to be sure.

[View[+]Finder]

When verifying, calibrating or adjusting precision DVM's, I use coaxial cable terminated on each end with a Pomona polarized double banana. I chose the coax based on published specs as well as how it was affected by being touched or moved. The goal is to avoid stray EMF from other stuff in the lab and piezo-electric effects from the insulation. Under the best of circumstances--other equipment shut off, no drafts and lab temperature constant to+/- 1 degree Celsius--calibration of DCV to 6.5 from 8.5 transfer standards is my norm.

I have resistance transfer standards as well and generally use a Keysight Kelvin 4-wire rig or a pair of the coax's, depending on whether the resistance standard has banana jacks or screw terminals.

Dave gets the credit for the coax wire setup; one of the first things I learned from his videos.

Cheers,
Donal

[NWerner]

Hello Frank,

Could you explain, what you mean by "drilling" in the context of cables?

Thanks in advance
Norman

[Dr. Frank]

Sure, Norman,
the first picture shows twisting of cables for a 2W Ohm measurement, HP3458A to Fluke 5450A, to suppress common mode noise.
Especially high Ohm measurements (100kOhm and higher) suffer from injected noise, as the resistances get higher and the test currents lower.
This twisting creates high errors, due to leakage currents of the PVC isolation.
10KOhm will be about 1ppm low, and 10MOhm 1000ppm.

The 2nd picture shows a 4W Ohm measurement, but twisted S+ / I+  and S- / I-.
This will give no such leakage error, but insufficient noise reduction.

Therefore a 4W + shield PTFE cable is best for that purpose.
Inside, all 4 cables are twisted, but there is no such leakage current.

The achieved StD values using these new PTFE cables are noticeable better.

Frank

[branadic]

As Frank already pointed out, twisted and shielded PTFE wire is good for measuring voltage references.
In case of resistors this is not the optimal solution though, even when measuring high ohmic resistors, as even the PTB recently had to learn the hard way in their K2 comparison.

PTFE is on a microscopic scale like a sponge, forming traps for charges, which results in a piezoelectric effect, thus the cable acts like a microphone. If PTFE cable is moved or bend, this creates noise. To get rid of that, low noise triax cable is used. It slightly differs to normal triax cable, as it has an additional conductive layer, as visible here:

https://www.koax24.de/produktinfos/triaxialkabel/uebersicht/001101.html
https://www.koax24.de/produktinfos/triaxialkabel/uebersicht/g-02330-ht.html

-branadic-

[Dr. Frank]

Hi branadic 
The triaxial cables have PTFE as well, so what's the point?
Usually, they have an advantage, only when you actively drive the 3rd conductor.
But this can't be done with the usual DMMs, as the 3458A?

And please describe the key aspects of this K2 comparison.. was this done for GOhm resistors?
Any reference link?

Thanks!

Frank

[branadic]

Hi Frank,

you are right, triax cable is also available with PTFE as dielectric, but also with other materials such as PE (see second link above). However, PTFE is good on leakage. Triax cable indeed only makes sense with a driven guard and such mimic could be generated with some additional circuitry, though I'm not sure if this is the case for 3458A.

To give you an example, when we build capacitive sensors with a large distance between readout circuit and sensor element, we create a guard signal with an external low capacitive input, fast opamp out of the signal to be measured itself. The guard has to be on the same potential as the signal and has to follow any change the signal does, so that no difference in potential between them occures. This way the signal is perfectly shielded from the ambient. In combination with low noise triax, with the inner braid used as driven guard and the conductive coating of the dielectric, charges are caught by the guard and doen't influence the signal to be measured. The outer braid is used as Lo or ground.

In the K2 comparison the resistor measured was 10 M . There was an interesting talk about it at the "Metrologie Tage" in Kassel by the PTB, though I didn't had the chance to talk to them after that talk. Unfortunately I don't have anything I can share with the public.

-branadic-

[notfaded1]

I decided to try a bunch of stuff.  The copper crimp spades I bought from someone in China that I've bought some LTZ1000 and related stuff from.  I actually got two type of the copper spades.  This is pic of one that didn't work as well as the other which were a little smaller and have only one dimension inside the space vs. two levels on these.  I also found some pure copper female jacks for the banana plugs to plug into which also unscrew so you can connect the copper spades to them.  I'd never seen these before.  I'm at work now but I'll post some pics on the LTZ1000 thread of them built into the shielding box the LTZ board is mounted inside of.  I also have some solid core copper wire i've tried twisting myself and using as well.  What's the thought between stranded vs. solid core?

Bill

[martinr33]

I like FEP-jacketed shielded twisted pair. PTFE is good (HP and Keithley use it for their nanovolt meter cables), but ETFE is thinner - good for making comparison cables that have to fit in the nanovolt meter connector.

I don't like Kapton. Kapton cables are very sensitive to temperature, and generate significant and hard to explain offsets.

As in, hold a cable in the middle with your hand and observe a couple of microvolts of differential voltage appear at the ends. 

[TiN]

PTFE cables are nice to have in lab , but not required for low-ish impedance sources like 10V DC references. Shielding and keeping ground loops (having few big car batteries in a lab around is great assistance) away is important however. 3458A nor other DMMs do not have active drive for guard, so its just grounding shields in single points to help reduce EMI/RFI-related issues.

Bare LTZ1000 circuit however is very sensitive to everything, so extra care always needed (no switchmode supplies, no wifi nearby, no LEDs with crap power converters, no airflow or turning gear on/off). Also you want low mass spades, so they get to stable temperatures of binding posts quickly, reducing settle time. And every time using bare copper connectors clean them well, copper oxide is no good for precision work.

I also vote categorically against using any kind of banana jacks at all. Spade lugs only. Fluke says the same thing, above shown connection of 5450A front bananas is no good and for best precision work only using rear 5-way binding posts is recommended. Hence my 5450A sits in a lab with rear side in front. Who need display and those buttons anyway? 

I use coaxial cable terminated on each end with a Pomona polarized double banana.

Worst possible combination for precision 7-10V DC measurements and 3458A. Do not attempt

[Dr. Frank]

I also vote categorically against using any kind of banana jacks at all. Spade lugs only. Fluke says the same thing, above shown connection of 5450A front bananas is no good and for best precision work only using rear 5-way binding posts is recommended. Hence my 5450A sits in a lab with rear side in front. Who need display and those buttons anyway? 

So, can you tell us, why, please?
W/o any profound reason, that sounds more like dogmatic arguments. 

Especially by using Offset Compensation for 4W Ohm measurements, all thermal offsets are cancelled, and the higher offsets are the only difference that I can spot, between banana jacks (low thermal ones, of course), and crimped spade lugs.

You're correct, when making high Ohm measurements at 2W and w/o OCOMP, but then there is still the possibility to reverse the leads.

It's really inconvenient to use spade lugs in some experimental setups, especially if you have the 5450A with the clamp jacks on its backside, and have to switch to other standard resistors with either plug or clamp elsewhere.

And what shall the people do when their high grade DMMs don't have clamp jacks, like the KEI 2002, 7510, KS 34470A, and alike?

Frank

[View[+]Finder]

Bare LTZ1000 circuit however is very sensitive to everything, so extra care always needed (no switchmode supplies, no wifi nearby, no LEDs with crap power converters, no airflow or turning gear on/off). Also you want low mass spades, so they get to stable temperatures of binding posts quickly, reducing settle time. And every time using bare copper connectors clean them well, copper oxide is no good for precision work.

I use coaxial cable terminated on each end with a Pomona polarized double banana.

Worst possible combination for precision 7-10V DC measurements and 3458A. Do not attempt

The question was posed as "what do you use" but it is helpful to learn other's opinions as to best practice. Some folks don't have the option of the environment TIN describes, free of WiFi, drafts, etc. What seems most helpful for readers is a discussion that incorporates the desired level of precision into advice on methods. Perhaps some readers are just trying to avoid making a mistake while others push the envelope on precision. For the latter, Keithley Low Level Measurements Handbook_7Ed, is an excellent resource. My objective is to help put good quality used equipment to use and keep it out of landfills. To that end, I would prefer to have "as built" performance, however I will accept just getting close to the specification.

Cheers,
Donal