Creating some high-precision resistors, for calibration purposes

[RandallMcRee]

I hope to create some precision resistors that I and other members can use for generally checking their 6.5, 7.5 and hopefully 8.5 digit meters. What I have available to me are Alpha Electronics MPP and MPQ types which are 0.2ppm, typical (see data sheet).  I plan to put 4-16 together to create a stable, robust value with good temperature characteristics.

Now, the real question: is it a good idea to have a switch in the resistor path? You see, I have this nice 4 deck, 12 position rotary switch. This would allow four wire measurements on 12 different resistors (or two different configurations of 6 resistors). Since a four wire measurement would be required that should nullify the contact resistance of the switch, right? It would be nice, obviously, to not have to pull one resistor out and pop another in, over and over. But if the repeatability is not there the switch would defeat the purpose. What are your thoughts?

Thanks,
Randall

[alm]

Given the accuracy levels you are targeting, I'd worry about thermal EMFs for any external connections or switches. So I would be concerned about the thermal EMF of the switch. Contact resistance is no big deal for four wire measurements that you would be using. However, low thermal EMF binding posts are not cheap either.

[RandallMcRee]

I do have four low-thermal binding posts that I purchased from TiN in the group-buy some time ago. I could use those.

The switch and resistors would all be in the same small box. That would negate thermal effects, presumably.

Thanks for the thoughts,
Randall

[WattsThat]

A low TCR does not a standard make. What’s the long term drift of a foil based epoxy encapsulated smd part? Do you wonder why it’s not on the datasheet? I do.

There’s a reason why 100ppm is tightest tolerance you can order. Add a possible soldering shift of 300 to 500 ppm and you don’t have much of a standard left.

And you’re worried about a switch in the path?

[alm]

A low TCR does not a standard make. What’s the long term drift of a foil based epoxy encapsulated smd part? Do you wonder why it’s not on the datasheet? I do.

Good point about the packaging. Humidity will get into the resistors and affect the value, and SMD makes them susceptible to stress as the PCB swells and contracts due to temperature and humidity. The performance probably won't rival hermetic through-hole foil resistors or precision wire-wound resistors.

There’s a reason why 100ppm is tightest tolerance you can order. Add a possible soldering shift of 300 to 500 ppm and you don’t have much of a standard left.

Or maybe 100 ppm is the lowest tolerance they can economically do production testing on? And what does soldering shift have to do with this? Surely you calibrate a resistance standard after assembly?

[WattsThat]

The much smaller die area of smd parts likely limits the classic Vishay calibration technique and .01% is the best they can do. I’m more concerned about the ratio of the connection tabs verses die size. There is no doubt a lot of stress imparted into the chip when soldering, given the delta R’s shown in the datasheet. Note the datasheet note about solder pad size. It matters. Oh, and soldering is after calibration so it shifts the absolute value. Your previous .01% part is now .05%.

I don’t think a small epoxy smd packaged part is a good idea as a standard, even if they’re free. But, hey that’s just my opinion. Besides, it’s the Internet.

[Conrad Hoffman]

Look at the size of commercial reference resistors and try to list the reasons they're not small. I've had the same ideas, but considered starting with a spool of manganin and winding them. There are still many pitfalls- search the various posts from @kleinstein and many others with a lot of experience in the area. You might look at the Fluke 5450A as an example of a 4-terminal relay-switched device. It's interesting in that the device doesn't get adjusted, being made with fixed resistors, but a new value is simply entered for the display to reflect the actual value of resistance.

[RoadDog]

Look at the size of commercial reference resistors and try to list the reasons they're not small. I've had the same ideas, but considered starting with a spool of manganin and winding them. There are still many pitfalls- search the various posts from @kleinstein and many others with a lot of experience.

Perhaps a group buy of manganin from Pelican wire? I don’t necessarily want to spend $500 or $750 on a spool. I’d toss some $ in though.

Then there’s the question of the heat treatment. Xdevs has an article from the 1940’s about it requiring hundreds of hours at a certain temp. I think more modern wire doesn’t require quite the same extensive process. It needs a couple temperature cycles to at least 150C for a while though. Edwin and others have posted a lot of good info over the years. I’d want something around 5-10 ohms per foot for the projects I’m looking for. Probably better off just buying something from VPG and waiting but it would be an interesting project and then send to TiN or someone else that can reliably test the TC and see how we do.

I’ve seen smaller rolls of wire online. Edwin has mentioned in the past to avoid cheap wire though as it won’t perform.

[Dr. Frank]

E.M.F. is no issue, as you will use 4W + Offset Compensation for precision measurements. Use sufficient delay, though. Therefore, a clever Kelvin Varley configuration over the switch will do no harm. AE resistors are fine, got a picture of those?
Frank

[jonpaul]

I was at Julie Research Laboratoires 1967, the 0.001...0.005% resistance were handwound, using special low TC wire, with a temp compensation series wire.

Termination by welding.

The lab standard resistance were air bath or oil bath température controlled.

No switches.

I doubt if any methods mentioned, either with stock SMD parts or DIY with resistance wire can be useable better than 0.1%.

We used resistance wire from Wilbert B. Driver and Driver Harris, in 1967, as I recall.

reccomend the OP check out the TC, accuracy and drift spec sheet of the resistance standard they try to achieve.

Fortunately I kept a few of the JRL résistance and later found some Genrad and L&N laboratory standards.

Checking with Keysight 34465a, all are still in cal.

just the ramblings of an old retired EE

Jon

[RandallMcRee]

I also have a few hermetically sealed AE resistors. Those should be suitable, I guess.

For the SMD ones I find that there are many conformal coatings and potting compounds that claim to hermetically seal circuits. Seems like that could work in this application? Does anyone know of actual research that shows SMD to be unsuitable?

Here is a photo of some AE SMD ones in parallel and the AE hermetics.

JonPaul--thanks for your comments, they are always thoughtful and interesting. With regards to switches not being in IET/Julie calibration standards: perhaps this is because they wished to support two-wire measurement as well?

Thanks again for all of the comments.
Randall

[jonpaul]

Rebonjoure Randall. , mille mercis pour votre gentil notes....

0/ No idea re te AE.

1/ Any potting or coating is chemically active and will degrade the contacts or seals.

2/ Julie Research and other hi precision stds never use switches , AFIK, all are 4 wire.

If any taps they use solid gold plated split bolt and plugs.

J

[WattsThat]

Any decent reference standard must be a hermetic device with some thermal mass (or isolation). The ESI SR10x’s are an example of what a reference standard should look like and realistically, that’s what would be needed for 7/8 digit device accuracies. How could you possibly check a 7.5 or 8.5 digit meter with a device that could drift 10-20 ppm or more between cal cycles?

Speaking of switches, Vishay made the 1300 series of decade boxes back in the 70’s. They had custom rotary switches which had thick gold over nickel plating and resistors that were all hand selected S102 hermetics. Even with selected resistors and expensive custom switches, the boxes were only good to about 100 ppm absolute. You could use them for tighter values but you had to confirm with a KVD or other bridge referenced to a standard resistor.

In todays world with HPAK 3458 class devices having replaced most other measuring methods, you’d be using the digital meter to confirm the resistor was in range for its intended use rather than the other way ‘round where you’re using the resistor to check the meter. n'est-ce pas?

[mendip_discovery]

As you can see the mob has mumbled about better kit that is outside the reach of most people. Unless you want to spend daft amounts of money for fun. But that is because you mentioned 7.5 and 8.5 digit meters.

So RandallMcRee I say to you. Have a go. See what you get. You might find that you do a fairly good job the first time around, if not you will learn and try something better. As long as you are honest to yourself on how good it is then you can work on that.

With a bit of thinking you can try to negate some of the problems that others are harping on about. If you calculate it for use in a 20°C±2°C which is fairly easy for anyone planning on using it, measure it a few times, each time recording the temp. See if it makes a difference. The resistor doesn't need to be precise, what it needs to be is stable and predictable.

Though I would agree, keep away from rotating selector switches unless they come from a high end decade box they are often a cause for plenty of error, even in the high end ones. I have toyed with the idea of using relays and building a poor mans version of the Fluke 5450A, it would have to be a relay with multiple connections to reduce contact errors and be a type that can switch and stay until a signal is sent to switch off so that the switching circuit can de-energise to reduce noise etc.

[Conrad Hoffman]

The difficulties at the 7.5 and 8.5 level are large and expensive. Limiting myself to 6.5 digits makes a lot of DIY stuff possible that might be questionable at the high end!

[RandallMcRee]

. . .

So RandallMcRee I say to you. Have a go. See what you get. You might find that you do a fairly good job the first time around, if not you will learn and try something better. As long as you are honest to yourself on how good it is then you can work on that.

. . .

I had a go! I did use a rotary switch, since perfection is the enemy of the good (or something). Here are some photos of the assembly, along with measured values using my existing (non-calibrated) equipment. The spreadsheet shows the resistor type, some are good, some not so good.

It would be great if someone with calibrated 7.5/8.5 meter could also check it, preferably on the west coast, USA. Send me a PM.

Thanks,
Randall

P.S. I recently purchased a Fluke 5450A, as well. That also needs calibration, however. So I will work on that after applying Dr. Franks 5450A medicine, that is, the relay replacement.

[RandallMcRee]

Assembly photos

[martinr33]

When it comes to resistance standards, they are generally approximations of their nominal value. You'll generally see standard resistors some way off of nominal.

Resistance calibrators usually store the measured value of the reistor in non-volatile memory - as your 5450A does. I'd also avoid messing with your 5450A unless you identify a problem.