Precision Resistor Box Project

[enut11]

Hi Guys
I have just finished my first Precision Resistor Box. Feedback appreciated.

The resistors were manufactured by Edwin Pettis (Ultrohm) who has been making high stability precision wirewound resistors for a very long time.

I purchased the following 0.01% units:

100R (supplied as +63 ppm)
1K (supplied as -27.2 ppm)
10K (supplied as +26.5 ppm)
100K (supplied as +43.7 ppm)
The supplied values exceed the specs as measured by the vendor.

The resistors were approx $10US each plus $13US for postage to Australia all of which I thought was quite reasonable.

With my box I also added placeholders for 10R and 1M for when I can get hold of these values for a reasonable price.

Box is hermetically sealed plastic 115x90x55mm from Jaycar (Australia).
Terminals are Chinese: http://www.ebay.com.au/itm/282227445109?_trksid=p2057872.m2749.l2649&ssPageName=STRK%3AMEBIDX%3AIT

The resistors are meant to be measured 4-wire although this is really only needed for the lower 3 values. 2-wire measurements are also possible and indeed the only way to measure the 1M resistor in this box. 2-wire measurements are made via the 'Sen Hi' and 'Sen Lo' terminals.

This project will be part the next round-robin Aussie Cal Club Ref swap. Currently there are 5 members sharing calibration data for a range of instruments 6.5d to 8.5d DMMs.
enut11

[branadic]

Box is hermetically sealed plastic 115x90x55mm from Jaycar (Australia).

Well, honestly this is just a sealed plastic box, far away from hermetic.
You might want to exchange the box with a one made out of aluminum and spend some temperature sensor inside the box. Just as an idea for improvement.

[Andreas]

The resistors are meant to be measured 4-wire although this is really only needed for the lower 3 values.

Sorry but I see only a 3 wire connection for the resistors above 10 R.

with best regards

Andreas

[enut11]

Thanks for the feedback branadic. The box has a rubber seal between lid and base. Is this not good enough?
enut11

[enut11]

The resistors are meant to be measured 4-wire although this is really only needed for the lower 3 values.

Sorry but I see only a 3 wire connection for the resistors above 10 R.

with best regards

Andreas

Thanks andreas. I copied this design from Vishay. The lowest value resistor on this design was 1R.
enut11

[Dr. Frank]

Thanks andreas. I copied this design from Vishay. The lowest value resistor on this design was 1R.
enut11

So you have copied Vishays errors.

In your design, you always have the wire resistances inside the measuring circuit.

Re-wire it like in my sketch, this  gives a true Kelvin connection only.

Also, change the colors of the jacks, your box is totally confusing.
The two common ones ('Hi') should be red, the pairs on the other side of each resistor ('Lo') should be black (10EA required)

Frank

[manganin]

Good advise from Dr. Frank.

To avoid mechanical stress through the resistor wires, one or two turns around a drill bit would be a good idea. Which also allows to use a heat sink during soldering. For example a tight fit aluminium rod, but there are also special clamps available.

The cheap Chinese binding posts may not be the best alternative. A few years ago a speaker builder friend asked my opinion about the samples ordered from Ebay. All advertised as pure copper of course. I expected brass, but no. Most single binding posts were made of steel and the double version some zink alloy.

Even the gold was not always gold but some poorly conductive decorative plating of almost same color.

[Andreas]

thanks Frank,

you were faster, that would have been also my suggestion.

Besides this:
Edwins resistors are usually measured  0.375" from the resistor body.
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg463643/#msg463643

With AWG22 (0.6 mm) (if this applies) wire diameter you have 55 mOhm/m or 0.055 mOhm/mm change with wire length.

For the 100 Ohms resistor you have to correct Edwins measurement values by 1.8 ppm for each mm if you do not connect the sense lines at the same distance.

With best regards

Andreas

[David Hess]

Box is hermetically sealed plastic 115x90x55mm from Jaycar (Australia).

Hermetically sealed plastic?  With all of those holes cut into it?

Thanks for the feedback branadic. The box has a rubber seal between lid and base. Is this not good enough?

It could be good enough but hermetically sealed it is not.

[enut11]

Hi All
Thanks for all the valuable feedback. This was an interesting project in so far as I had to imagine that very tiny resistances matter.

Thank you Dr Frank (and Andreas) for alerting me to the wrong 4-wire wiring. Looks like even Vishay got it wrong in their design!

I am also now alerted to what constitutes a hermetically sealed box, which of course mine is not! Thanks branadic and David Hess.

The mechanical stress relief on the resistor wires is probably a good idea and one for the future. Thanks manganin.

I am actually not sure what the binding posts are made of. I know that they are non magnetic but when I cut one it showed a silver colour metal so it was not brass. The 'gold coating' seems to work as the posts were easy to solder unless the Chinese have worked out how to get around having to use real gold.

Back to the workbench.
enut11

[feedback.loop]

Re-wire it like in my sketch, this  gives a true Kelvin connection only.

Frank

I don't see what is wrong with the wiring, and why proposed re-wiring is better. Could you elaborate please? Thank you.

[pelule]

You may better understand  if you make e schematic where you replace the wires with resistors (that they are) and connect a ohm measuring DMM.
You will see, the voltage drops between the binding posts and the precision resistor adds a failure. The recommended Kelvin connection measures the resistor (in fact the voltage created by the reference current) without that.
/PeLuLe

[VK5RC]

Hi enut11,
Spent the last few days testing most of the Aus Cal Clubs round 1 references (and my set-up)!
I suspect a lot of the devil is in the detail for most of this area, unfortunately there are lots of areas to try to get close to right.
Below is my version of a Vishay 10k resistor. Paper insert to allow better photo contrast only.
Below is my thoughts on what I did.
-I tried to get a Kelvin type connection by having the connecting wires separate and connecting on either side of the resistor lead.
-used relatively small diameter wire to get more rapid thermal equalisation- dissimilar metals (connector - lead - solder - resistor)
-The resistor is not touching anything - less heat transfer to it.
-Metal shielded box with earth connection.
-Pomona connectors - low EMF
Possible problems
- metal box conducts heat too well - thermal changes
- mechanical stressors on resistor 'floating'
- measure temp inside box
- resistance of wire itself
I am still getting modest interference from some test / gear inside my lab - I get the most repeatable results with just a linear PSU and the 3458 switched on.
Things I have observed to alter readings by a few (up to 10 or so) uV  but it is difficult to collect serious data without include
- switching on a laptop/computer nearby
- switching on laptop power supply
- connecting a USB cable +/- GPIB-USB device
- lights (esp SMPSU LEDs)
- temperature changes in the room
An interesting area - lots of challenges ahead!
Robert VK5RC

[feedback.loop]

You may better understand  if you make e schematic where you replace the wires with resistors (that they are) and connect a ohm measuring DMM.
You will see, the voltage drops between the binding posts and the precision resistor adds a failure. The recommended Kelvin connection measures the resistor (in fact the voltage created by the reference current) without that.
/PeLuLe

I am afraid I am not following your explanation. What do you mean by "precision resistor adds a failure"?

[not1xor1]

You may better understand  if you make e schematic where you replace the wires with resistors (that they are) and connect a ohm measuring DMM.
You will see, the voltage drops between the binding posts and the precision resistor adds a failure. The recommended Kelvin connection measures the resistor (in fact the voltage created by the reference current) without that.
/PeLuLe

I am afraid I am not following your explanation. What do you mean by "precision resistor adds a failure"?

My English (besides my EE knowledge) is not that good but I guess he means that the voltage drop adds a measurement error since the supply current and the sensing signal share the same path

[Gyro]

I suspect a lot of the devil is in the detail for most of this area, unfortunately there are lots of areas to try to get close to right.
Below is my version of a Vishay 10k resistor. Paper insert to allow better photo contrast only.
Below is my thoughts on what I did.
-I tried to get a Kelvin type connection by having the connecting wires separate and connecting on either side of the resistor lead.

Wouldn't you still get a more predictable connection it you connected the force wires near the ends of the resistor leads and the sense wires a little closer to the resistor - keeping lengths and spacings matched? If you try to connect both at the same point as shown ('loose' mechanical connection between the three wires) surely there's a danger that at least some of the force current could influence the sense wire by passing through the solder rather than the resistor lead. Thermocouple effects would still be minimal.

[cellularmitosis]

I was thinking about resistor boxes, kelvin connections, and binding posts recently.

My question is: are the requirements for the sense and force binding posts different?  There has been some great progress in finding affordable unplated pure copper binding posts in another thread, which got me thing about this topic: https://www.eevblog.com/forum/metrology/diy-low-emf-cable-and-connectors/

For the sense leads, both leakage currents and thermal EMF voltage offsets are important to consider.  But for the force leads, is it only leakage that you need to worry about?  If the force leads are driven by a constant current source, then any voltage errors shouldn't really matter, I would think -- the only error would come from current leakage or injection, is that right?

In other words, for the force binding posts, it would be a good idea to mount them on PTFE, but you probably wouldn't care if they were made from nickel plated steel, rather than unplanted tellurium copper.  Is that right?

Would love to hear anyone's input on this thought experiment!

[ap]

First of, both the force and sense binding post leakages contribute, sinply because they are in parallel. But not only the plastic material is your enemy, but also humidity/condensation. This caused e.g. a few ppm decrease after an ESI SR104 had shipped and trapped some humidity. On a side note, the ESI SR104 does neither have PTFE insulated posts nor cable. Reason being, it is not relevant at that resistance level. Different at say 10M and more. In a real world, if you measure precision resistors with a precision meter, they compensate EMF voltages (3458A, K2002). Other less accurate meters that do not support this will not get you to the required accuracy anyway. Also watch out for dielectric absorption, Frank wrote about it extensivele with relation to the 3458A (use DELAY). Also, I would be very carefull with pure copper posts, because copper oxide is the worst EMF contributor of all (thermoelements aside ), and it is hard to clean posts. So unless you are working in a primary lab, and unless you are working at nV levels, I would always prefer gold plated telurium copper over pure copper. Generelly, in a good setup like this, I should say EMF is one of your least problems doing measurements.