Kelvin Varley Divider [and Precision Voltage Source]

[HLA-27b]

0,1mm is easy to achieve with steady hand, with aditional gentle filing it should be possible to improve precision by another order of magnitue. As you said strain would be a major concern. The initial strain from the wining process could be eliminated by annealing the assembly after winding.

On the other hand since glass has much lower tce than the metal wire, we could expect the wire to slacken when temperature is hot and to tighten (and get strained) when cold. If the wire is wound in figure of eight fashion there would be enough slack to prevent this from happening in most cases.

To prevent corrosion either immersion in oil or enclosing the resistors in inert gas should do the trick. It seems like a big deal but after the trouble undergone to achieve 1 ppm it should seem trivial.

[amspire]

0,1mm is easy to achieve with steady hand, with aditional gentle filing it should be possible to improve precision by another order of magnitue. As you said strain would be a major concern. The initial strain from the wining process could be eliminated by annealing the assembly after winding.

You cannot really anneal it. The wires are coated in varnish, so the heat would destroy the varnish. When making the Thomas Standard one Ohm resistors, they anneal the wire at something like 500 degC from memory. Not sure if it is possible to do that with a large value resistor on a former.

On the other hand since glass has much lower tce than the metal wire, we could expect the wire to slacken when temperature is hot and to tighten (and get strained) when cold. If the wire is wound in figure of eight fashion there would be enough slack to prevent this from happening in most cases.

To get it right is not easy. The former and the wire probably has different expansion coefficients, but also the glass has a greater thermal mass. So the wire can be hotter then the glass or cooler. It is hard to stop the wire being stressed at some point.

I remember reading about the Hewlett Packard refinement of the Rosa Standard resistor (the secondary standard resistors in the 40mm diameter cans with the two bent conductors out the top). I think the Rosa resistor uses a thin-walled brass tube as the former, and HP used something like a thin thickness polyester tube as the former that gets immersed in the oil. The idea is the polyester is not strong enough to stress the wire.

To prevent corrosion either immersion in oil or enclosing the resistors in inert gas should do the trick. It seems like a big deal but after the trouble undergone to achieve 1 ppm it should seem trivial.

[HLA-27b]

You cannot really anneal it. The wires are coated in varnish, so the heat would destroy the varnish. When making the Thomson Standard one Ohm resistors, they anneal the wire at something like 500 degC from memory. Not sure if it is possible to do that with a large value resistor on a former.

Uncoated wire perhaps. Strain should not be a problem if wound with a little slack. As long as windings do not short it should work.
I should seriously try to get my gands on some low tempco wire then...

[fmaimon]

The other was that soldering was very risky business and I always needed to have a few extra matched resistors to fine tune a decade after it was all assembled. This was true for the wire wounds as well, and I did use heat sinks.

I'm thinking on soldering the resistors with their bodies immersed in deionized water, with only the tips of the leads outside the water. This may help preventing the heat changing the value of the resistors.

A 1K pot is way to big. You will regret it.  If you have resistors that match between decades to at least 0.05%, you don't want the pot doing much more then that. I would have the 24.3K (25ppm) and a 820 ohm (1% 100ppm) in series and put a 10K pot across the 820 ohm resistor. I prefer this circuit as only a fraction of the current is going through the pot wiper.

Ok, I'll do that.

Thank you,
Felipe Maimon

[Conrad Hoffman]

Having built a couple high precision resistance standards, I can tell you that something as soft as a spray coating of Krylon will completely destroy the stability of the standard. The ones I built used a bite of wire, held at one end with a piece of thread, and loosely wound on a brass tube. I think I threaded the OD of the tube so the wire would have a nice predictable place to sit. IMO, you could do a great job with the KVD using wire, but for one problem. To get any sort of higher values using wire, you have to go to a really small gage. That still means a huge number of turns and working with the stuff is no picnic. Don't over estimate the inductance cancellation of the various winding schemes, as they work less well than you might think. Flat windings on mica cards are easiest to build. I've used other things as well, from pieces of PC board, to metal, to plastic. I saw a new (to me) form in diyaudio.com that I want to try. It was essentially two small star shaped pieces, with a stand off in the middle. The wire was wound top to bottom around the outside, with a phase shift between the legs of the top, vs the bottom. This gave a winding very similar to an RF inductor, where the wires are never close and parallel, but cross each other at an angle. It looked like it would give far less inductance than the various old fashioned ways you find in the electrical measurement texts like Stout, Laws, Farmer, Harris, Terman et al. FWIW, I don't think the high temp bake is as critical as for under 1 ppm standards like the Thomas resistors. Also, for whatever reason, people seem to use manganin for low resistance standards and Evanohm for higher ones. There are some really good "800 series" wires from Kanthal (formerly H.P. Reid in Palm Coast, Florida, USA) that have both low TC and low thermal emf to copper. So who's ready to machine up some tellurium copper binding posts?

[nukie]

I don't mind turning a couple but where do I find that raw material?

[amspire]

There are some really good "800 series" wires from Kanthal (formerly H.P. Reid in Palm Coast, Florida, USA) that have both low TC and low thermal emf to copper. So who's ready to machine up some tellurium copper binding posts?

I can't seem to find any mention of 800 series resistance wire. The best wire from Kanthal current products seems to be +/- 15ppm/C coefficient wire. 

http://www.kanthal.com/products/materials-in-wire-and-strip-form/wire/resistance-heating-wire-and-resistance-wire/list-of-alloys/

Richard

[amspire]

There are some really good "800 series" wires from Kanthal (formerly H.P. Reid in Palm Coast, Florida, USA) that have both low TC and low thermal emf to copper. So who's ready to machine up some tellurium copper binding posts?

I can't seem to find any mention of 800 series resistance wire. The best wire from Kanthal current products seems to be +/- 15ppm/C coefficient wire. 

http://www.kanthal.com/products/materials-in-wire-and-strip-form/wire/resistance-heating-wire-and-resistance-wire/list-of-alloys/

I have found  Jelliff Alloy 800 that could be like the "Series 800" wire.  It is rated at 5ppm over a -55 to 150 C range. They don't specify the coefficient near room temperature. It has good solderability and low EMF with copper.

http://jelliff.thomasnet.com/item/resistance-wire/jelliff-alloy-800-electrical-resistance-alloy/w08000056?

[Conrad Hoffman]

Most of what I know is old and obsolete! I think the old name was Rediohm 800. That Jelliff looks like the stuff. Usually it has to be spot welded because soft solder won't wet it. There's also a Kanthal wire down the list a bit that's available in 5 and 10 ppm/C- the Nikrothal LX. These are more modern and higher performance than manganin, but manganin has more history on its behavior since its been studied so much. I know the Julie KVDs used tellurium copper binding posts and you can get the copper rod from speciallty copper/brass suppliers, but no idea the cost. Then we have to talk about the low thermal emf solder used by Leeds & Northrup. It was loaded with cadmimum, so might be a big no no today. I believe they put a dab of paint (green?) on any joints that used it in their equipment and standards.

[amspire]

I found out a bit about making resistors.

After winding, you want to age them - cycling them between room temperature and a maximum temperature. So if you are making resistors or buying, it would make a lot of sense to set up a rig to cycle current on and off for a few days before testing the resistors. Cycling between room temperature and 100 deg C sounds reasonable.

If you have bare resistance wire, you never want to let the wire get about 100deg C or it will oxidize and you need to pickle it to remove the oxide.

With enamel coated wire, you have to keep the maximum temperature below 140 degrees so you do not damage the varnish.

One of the main problems with soldering is the high temperature coefficient of tin - 4600ppm/degC. That is why it is best to spot weld wire rather then solder.  However, for a 10K resistor, all you need is for the solder resistance to be less than 10mOhms and you are fine. Amazing the small things that make a difference when you are working to 1ppm.

Another thing that could be tried is if you have resistors - wirewound or metal film - and they have a slight negative coefficient, you could wind some extra turns of copper wire on the resistor and put it in series. Copper has a positive coefficient of something like 4000ppm.  So if you measure a 10K at different temperatures and decide it has a -10ppm coefficient near room temperatures, you could add 25 ohms of thin copper wire in series and you would then get a zero temperature coefficient. It would probably be easier to use lower resistances for the divider like 3k3 and put 3 in series for the first decade. That way, there will be less copper wire needed.

But don't tell anyone - this idea is patented in 1976 by a Japanese guy. Since when was combining different temp coefficients together a new idea - what kind of idiots are in the patent offices?

Richard

[robrenz]

tellurium copper rod   http://www.onlinemetals.com/merchant.cfm?pid=15266&step=4&showunits=inches&id=1112&top_cat=0

The tellurium is for machinabilty not electrical properties.  Pure copper is nasty to machine.

[nukie]

The tellurium is for machinabilty not electrical properties.  Pure copper is nasty to machine.

Tellurium by itself might not be good for electrical conductor but it's Tellurium Copper which makes it ideal for low thermal emf connection.

http://www.tequipment.net/Pomona3770.asp

Very nice it makes copper softer and easier to machine. I have machined 101 oxygen free copper with carbide and coolant. Need to add a bit of patient and slow feed rate.

[amspire]

A thing I have never understood.

Why is gold plated tellurium copper alloy better for EMF then gold plated copper or gold plated brass terminals? I cannot see how the metal under the plating makes any difference.

Now if you are talking about unplated terminals , then it does make sense to me.

Richard

[robrenz]

A thing I have never understood.

Why is gold plated tellurium copper alloy better for EMF then gold plated copper or gold plated brass terminals? I cannot see how the metal under the plating makes any difference.

Now if you are talking about unplated terminals , then it does make sense to me.

Richard

  Doesnt the base metal and the plating form a another thermal emf junction.  that would explain why they are different even though they all have gold plating. Any two disimilar metals in contact whether its just physical contact or wetted metals(solder) or plating are a thermocouple arent they?  My understanding is all these need a temperature gradient to generate a voltage. There is no thermal EMF if there is no temperature gradient.

[amspire]

As far as I can see, if the connector is gold plated, and you are not touching the underlying base metals at all, then any gold to base metal EMF voltages are irrelevant.  All that matters is the test lead to gold EMF voltage.

The only thing I can think of is if you are working at a high current, then there can be local heating at the contact points with the leads.  Perhaps the gold to base metal EMF voltage differentials caused by this heat are enough to cause a voltage gradient on the surface of the gold plating. The must be a reason like that since gold plated brass does appear to be a no-no for low EMF work.

Richard

[robrenz]

As far as I can see, if the connector is gold plated, and you are not touching the underlying base metals at all, then any gold to base metal EMF voltages are irrelevant.  All that matters is the test lead to gold EMF voltage.

Richard

I am not an authority, I am just thinking out loud.  IMO you cant pick and choose what different metal to metal joints you are going to account for. Every single connection in a circuit that is a dissimilar metal counts as a themocouple. and the connection of a plating to the base metal counts just as much as a copper wire crimped to the gold plating.


As far as I kinow the EMF of a thermocouple is not generated at the junction of the two metals. It is generated in the thermal gradient of each wire.  So how does this apply when you have few microns of plating?

[amspire]

As far as I can see, if the connector is gold plated, and you are not touching the underlying base metals at all, then any gold to base metal EMF voltages are irrelevant.  All that matters is the test lead to gold EMF voltage.

Richard

I am not an authority, I am just thinking out loud.  IMO you cant pick and choose what different metal to metal joints you are going to account for. Every single connection in a circuit that is a dissimilar metal counts as a themocouple. and the connection of a plating to the base metal counts just as much as a copper wire crimped to the gold plating.


As far as I kinow the EMF of a thermocouple is not generated at the junction of the two metals. It is generated in the thermal gradient of each wire.  So how does this apply when you have few microns of plating?

Here is what I have found out so far. The best connections are copper to copper  - less then 0.3 uV/C.  The problem is that copper oxidizes easily, and copper to copper oxide has 1000uV/C EMF. Obviously any copper oxide is a disaster. So if you were after the absolute best possible EMF performance, you would use a solid copper wire with a copper post, and you would lightly sand both before using to remove of all traces of copper oxide.

Copper to gold or silver is 0.5uV/C.  So perhaps they flash the copper with enough gold to restrict corrosion, but for EMF reasons, they don't put the usual nickel plating first to seal the copper. This means the copper can still oxidize through the pores in the gold. Tellurium-Copper has a similar EMF properties to pure copper, but has a high corrosion resistance, so you don't get much copper oxide leaching through the pores in the gold plating. Tellurium bonds tightly enough to copper to prevent reaction of the copper with oxygen. I guess if the Gold plating is not very thick, the Tellurium-Copper somehow still dominates in the EMF voltages.

Tin/Lead solder connections to copper is pretty bad at 1 to 5uV/C, but if you have a copper-solder-copper joint, then to you need a temperature differential across the joint to generate an EMF. Definitely could happen at high currents, but probably not at low currents. Cadmium-tin solder is down at 0.3uV/C but is very toxic. Companies like Leeds and Northrup used it a lot, so be warned if you are working with old precision equipment.  10%tin/90% lead solder is sometimes used as a safe compromise.

Richard

[HLA-27b]

Do we know the thermal noise generated by a Manganin - Copper connection ?

I recall that insteard of copper, phosphor bronze was preferred as connection material in a paper about precise measurement of temperature.
If its thermal noise properties are suitable phosphor bronze is decidedly easier to keep free of oxidation.

[nukie]

Hold on, does oxygen free copper 101 turns into copper oxide? The HP34401a input connectors are all made of copper.

[amspire]

Hold on, does oxygen free copper 101 turns into copper oxide? The HP34401a input connectors are all made of copper.

No, they are gold plated tellurium-copper alloy connectors. The specifications describe them as "Copper Alloy", but they are tellurium-copper Alloy.

[robrenz]

What about the precision multi pulse capacitor discharge spot welders for making the connections to the resistance wire. Would the extremely quick but high temperature be worse for long term stability.

[HLA-27b]

What about the precision multi pulse capacitor discharge spot welders for making the connections to the resistance wire. Would the extremely quick but high temperature be worse for long term stability.

We don't really know if it would. On one hand we hope that even if it changes value at least the ratio would stay the same. On the other hand one would anneal the whole thing just to be on the safe side.

[lowimpedance]

Just thought that it would be interesting to show a few pictures of an old ESI dekapot DP1311 (1OK) Kelvin Varley divider for reference on how a commercial product was built.
note; the 34401 has been set for an input R of 10g ohm.
(Since this is not a real review, and not many would be interested anyway I thought this would be the best place to post, hope no one minds).
cheers
John

[amspire]

I have been contemplating the issue of building a KVD. I am not impressed with the quality of affordable resistors. Some people are luck enough to snag a 6/7 digit KVD at a bargain, but the sellers seem to now know that a good one is worth $1000+

So how else can you get great resistors + switches affordably?

I saw this and bought it for $16.50. Worth half a US family average income back in 1960. See the row of decade switches - could that be a Kelvin-Varley divider?



I have it now, and it is not a KVD, but what it does have is very interesting and as useful as a KVD. I am sure someone in the forum recognizes what the switches are. From the build dates and the last calibration date, I suspect it was used for calibrating avionics in Lockheed Martin Orion P-3C aircraft. I know for sure it was built for Lockheed Martin. Last calibrated  about 2000 and at this time, Australia was upgrading all its Orions to AP-3C's with all new avionics. I think I will do a teardown, before I "re-manufacture" it.

The next idea is that the Fluke Differential Voltmeters on ebay all contain Kelvin-Varley dividers, and you do see some on sale under $100.

Now there are some that have 3 decades + a 0-100 pot and some that have 4 decades + a 0-100 pot.  The 4 decade ones are the way to go as they have a fully calibratable first decade, and I think they use 0.01% resistors in the second decade. So using it as it is, you can get at least 0.001% out of the divider, and knowing Fluke, it may be much better. The first decade is usually 5K resistors in the divider.

The 3 decade ones are all fixed resistors with no calibration. The first decade is 100K resistors in the divider which is very high.  To have less then  0.0001% error from the following voltmeter or amplifier, it would need an input resistance of over 1000 GOhms.

I feel a bit guilty suggesting messing with these great old instruments, but if you turn something that is now hardly used into something useful, well it is not such a bad thing.

Richard

[amspire]

Just thought that it would be interesting to show a few pictures of an old ESI dekapot DP1311 (1OK) Kelvin Varley divider for reference on how a commercial product was built.
note; the 34401 has been set for an input R of 10g ohm.
(Since this is not a real review, and not many would be interested anyway I thought this would be the best place to post, hope no one minds).
cheers
John

Thanks John. I had been wondering what was in them.

I mentioned the price before  - I would love to have a home lab kitted out to the same dollar value as that one divider.