Poor Man gets his candies (6X weston cells)

[zhtoor]

-

[tggzzz]

Don't do it, any more than you would with any other type of battery/cell.

If you want to know the average voltage, measure each voltage on its own and calculate the average.

Apart from any other considerations, that will enable you to discount the effects of an individual failing cell.

I'm sure you already know that you must ensure there is zero current drain, and that you must be careful with the mercury and cadmium compounds that have leaked.

[tggzzz]

http://conradhoffman.com/stdcell.htm

Easily found using google.

[HighVoltage]

We had a long discussion about this already in the thread by Dr.Frank:

1000: My (hi)story of the Weston cell, of the Volt, and of being a volt-nuts
https://www.eevblog.com/forum/metrology/1000-my-(hi)story-of-the-weston-cell-of-the-volt-and-of-being-a-volt-nuts/

Your cells seem all ok.
Make sure to set your voltmeter to High-Z impedance.
A typical 10 MOhm load is too high for these cells.

[HighVoltage]

these cells are 1-tube design, i haven't seen anything like it in the literature. (might have missed a lot)

regards.

Can you take one of the cans apart and show us the inside?
I have also not seen a one tube Weston Cell before.
 

[Gyro]

these cells are 1-tube design, i haven't seen anything like it in the literature. (might have missed a lot)

regards.

Can you take one of the cans apart and show us the inside?
I have also not seen a one tube Weston Cell before.

The Muirhead K-375-C is a single tube design (saturated)...

https://www.eevblog.com/forum/metrology/muirhead-std-and-ref-cells-all-catalogue-info-uploaded-!-(ayce)/

[HighVoltage]

The Muirhead K-375-C is a single tube design (saturated)...

Thank you, I had not realized that!

[MisterDiodes]

Read your reply #2 above.

This is a big issue with these cells - do NOT draw current during a measure; you want that discharge rate as slow as possible - and watch the temperature.  Now you remembered why you wanted that KVD and null meter too....

[tggzzz]

There are many things you could do, but that doesn't mean you should do them.

In this case you should work out what benefit you expect from your idea, and whether there are better ways of achieving those benefits.

I strongly suggest you learn standard metrology processes and procedures, and why they are that way.

[VintageNut]

Read your reply #2 above.

This is a big issue with these cells - do NOT draw current during a measure; you want that discharge rate as slow as possible - and watch the temperature.  Now you remembered why you wanted that KVD and null meter too....

thanks MisterDiodes,

my mouth still waters at the mention of KVD+Nullmeter combo! 
i came across the following researching weston cells.

please comment.

As to the quote of your article....
If the cells were in parallel, they would not be for "comparisons" it would only be one comparison.

My guess is that the cells were connected to a switching multiplixer to take intercomparison measurements.

The primary lab that I visit once a year to have my 731Bs calibrated has a bank of references that are connected to a switch multiplixer. There are a couple of open spots in the wiring of the multiplexer for references to be measured.

The bank of references are compared to each other once every week. This data is saved over many years and used when calibrating an unknown reference.

[MisterDiodes]

Read. Your. Reply. Number. 2. 

Whatever moldy tomb of info your getting that from doesn't mention how long that arrangement lasted and at what discharge rate, either.  The parallel connection discharges some cells faster and leaves you with a mess of haz-mat gloop basic equivalent one cell (for a short while) that's going completely fail when any one of the group fails.  You lose all the advantages of having multiple cells to average and compare - and be able to spot the bad cells and remove those from the group as required.

I just realized I repeated your reply #2 but it will sink in at some point.

By the way, watch the bias current of your DMM also.  When everyone is telling you "draw zero current" from these cells, we aren't kidding.  This is a very big issue with these cells.

Null meter time, or really verify your meter's input Z and bias current.

[VintageNut]

Read your reply #2 above.

This is a big issue with these cells - do NOT draw current during a measure; you want that discharge rate as slow as possible - and watch the temperature.  Now you remembered why you wanted that KVD and null meter too....

thanks MisterDiodes,

my mouth still waters at the mention of KVD+Nullmeter combo! 
i came across the following researching weston cells.

please comment.

As to the quote of your article....
If the cells were in parallel, they would not be for "comparisons" it would only be one comparison.

My guess is that the cells were connected to a switching multiplixer to take intercomparison measurements.

The primary lab that I visit once a year to have my 731Bs calibrated has a bank of references that are connected to a switch multiplixer. There are a couple of open spots in the wiring of the multiplexer for references to be measured.

The bank of references are compared to each other once every week. This data is saved over many years and used when calibrating an unknown reference.

the lab that you visit, does it have weston cells on a switch matrix?

No. There more than 10 total of a mixture of 732B, 732A, 7001, .....  They are all very well aged and the 732Bs are calibrated at Fluke every year.

Nobody would do serious work today with Weston cells today as far as I know. The temperature control needed is a serious bath and control equipment.

Weston cells are very highly temperature sensitive compared to a modern compact voltage standard such as a Fluke 732B.

[MisterDiodes]

...no matter, we all know how to do it better now.

Again - when you connect all 6 in parallel, it takes on cell to fail to bring the whole group down - and now you're drinking from the cup of sorrow.  You're also the proud owner of 6 dead cells when you could have had 5 remaining good cells - if you had only listened to your nutter-brothers trying to help you.

In 1929 you could buy more cells too...and that was before we knew what haz-mat meant.

One lower-risk way:
You can also measure each cell individually, average them, and then pick -one- cell that you will use for measures.  On that one cell you use it's value offset from the average.  Only use that cell to measure (measure + offset to relate that back to average), but check that cell  against the newest recalculated average value occasionally.  Or do the same technique using all the cells round-robbin for measures so each cell gets about the same amount of energy withdrawn at each measure.

There are other ways as well but in Metrology it's all about checking across as many -separate- references as possible, with as many different techniques as practical, and looking for bogus measures and eliminating those from the measuring system.

You're going to keep your measures relatively quick, and not very often (compared to a Zener Diode Vref that will run all the time).  Remember that you only have a limited amount of energy in each cell.  You're going to keep your measure current at zero, or at least as low as possible, say a few hundred fA or less during a measure.  The more you discharge those cells the faster you'll be out of good Vrefs, so plan accordingly.

You might look at some of the various older app notes etc. that describe very low input current / low noise buffers etc. for these cells.  HINT: Fet or tube buffer followers might be more help than op-amps here but you'll have to research that on you're own.  Everyone here will have a passionate opinion on a good buffer circuit I'm sure.  Even with a buffer amp you'll keep your measures relatively short, and keep the cells disconnected and at a stable temp  as much as possible.

The main thing you'll probably learn quickly is what an absolute pain in the ass Weston cells are to use and maintain - but Have Fun!

[Edwin G. Pettis]

Naw, CK5886 tubes aren't that hard to find, I've got them.  They are made for low voltage operation and input current is on the order of 25fA.

If you want to shell out the bucks, you can get a Weston cell measured to .04 PPM but that comes with restrictions.  They are interesting devices, probably fun to play with but they have been obsolete for quite a few years, solid-state Vrefs are much more robust, less temperature sensitive and possibly even more stable with time.  Plus, in many places, because of their chemistry, they have been classified hazardous, harder to buy and even harder to get rid of 'legally'.

Properly cared for, properly used and kept under constant temperature, they are very good voltage references and also low noise to boot, but today, it just isn't worth the hassle of keeping and handling Weston cells when much better technology is available.  But have fun experimenting with them, just be careful of how you handle them.

[MisterDiodes]

That transistor in the op-amp note is replaced by just disconnecting standard cell when you turn off the amp.  Maybe think about a set of switch contacts on your On-Off switch as a replacement for that transistor switch.  I wouldn't leave the cell connected to any complete circuit while not in use.

Remember that's a slightly higher current design, looks like it's pulling some 10's of pA from the cell. which (I think) the Eppley's were a little more forgiving than what you have, maybe.  Somebody might have more info on that.  I might think about a lower input current design on the amp...   You can probably use just a unity gain buffer, you don't need the complexity of the resistor-feedback gain necessarily.

Edwin suggested a tube and that's really the bee-knees for low current.  Makes your project even more fun!

There is another standard cell buffer sample circuit floating around either at Linear / AD or TI/National.  I think it was around a pA current on the cell but I might be wrong on that.  But have a look around.  The older app notes will have other ideas - if they are still online.  But you should be able to find an appropriate op-amp if you want to do it that way.  Use the parametric search tables.

I'll have to let you consider those thoughts for now.

[VintageNut]

I have exactly one good Weston cell. It is the kind that degrades normally without any help of drawing current.

The way I measure it is a >10G DMM to measure the different of the cell vs. a 731B divided down via a KVD.

I did this every month or two for over a year and made a spreadsheet.

The best way to use the Weston cell is also with a KVD.

If I recall correctly, the 731B has a procedure to calibrate it vs a Weston cell of known voltage. That was back in the day when Weston cells were not obsolete.

Since I have had my 731Bs calibrated, I have lost interest in the Weston cell as it requires regular calibrations to know what the voltage is as it degrades.

[MisterDiodes]

You can get a price quote at NIST to measure your cells against their JVS down to around no less than .04ppm uncertainty, runs around ~$4200 last time I knew as a special quote (might be more or less today), but you pay extra fee on top of that for hazmat shipping and handling, and it is sent in a thermal isolated box, last I knew.  I think you have to have something like an insurance bond in place in case one of the cells break and they have to clean the lab, I know at one time that was the case.  Just getting the cells to and from NIST is not easy nor cheap, and you do that every year or 6 months or so to find out the unique discharge rate of each of your cells.

That cost is for -each- cell by the way, not for a set of cells.  There might be some discounted price for a cell set.

I know Boeing Cal lab used to do this service also, but they no longer accept hazmat cells for testing.  That will be very common at most labs these days, and that's why nobody except you is trying to use them today.   That's another big reason these are such a P.I.T.A.

Yes I was going to suggest the technique shown at LT app note 9 (same as your 1052) - in this use the cells are stacked in series which will give you a higher unknown voltage (and also means you can't tell which cell is bad), and inherently averaged into one unknown cell voltage:

http://cds.linear.com/docs/en/application-note/an09f.pdf

The trouble with 1052 (and similar chopper IC styles)  is the very high current spikes so you want to isolate the amp from the cells via RC filter at least. You want to bypass the amp input to give those current spikes a place to go someplace besides the standard cells - the makers of zero-drift chopper amps always leave that little part out.   Those short switching spikes (at chopper freq) can be up to several hundred times higher than what you see listed on the datasheet for the low freq current noise. 

Pay attention to your op-amp offset current AND bias current, no matter what.

The problem with something like LT1097, LT1008 etc. is the offset voltage and drift those introduce (that's why those example circuits include gain added with a variable pot - now input Voffset is not important).  The trouble with IC amps in general is you tend to trade off low input current for higher voltage noise.  You can get various electrometer IC op-amps but look at the voltage noise and offset.

That's exactly why Edwin and I were suggesting something more along the lines of a Tube or Fet type buffer as a discrete design.

Really:
In your case I'd just stick to using a DMM with 10G input impedance, and I would at least try to find out which way around your input bias current runs on that meter so that at least you're not trying to "charge" those cells.  They are not batteries.

Remember: You are trying hard to not discharge nor charge these cells. At. All.  These devices work best in an age where mirrored galvos, KVD's and null detectors meters were the norm. Normally you'd put a safety limiting resistor in series with your cell while it's connected to your bridge balance measure with your glavo or null meter, and then as you get thing close to in balance on the KVD, then you remove the safety resistor and do the final adjustments on the KVD to bring the bridge into balance.  Then you record the measure and remove the cell from the circuit.  The idea is to keep the current flow through the cell as close to nothing as possible.

Why not just use a few LM399's?  For what you have to work with you'll still have an Vref voltage source, steady but unknown.  Or for what you have to work with for measuring, why not just use an LT6655? You don't have much budget to work with, and at least something like a '6655 will get you in the ballpark of a Vref at a close known value.  Or use several of those in parallel.  That might be a better starting out place for your lab's local definition of a "Volt" reference.

If you just need an unknown but quiet voltage source - say for instance you need to do a a ratiometric measure - then a NiCd battery or sealed lead acid battery can work as a very quiet voltage source.  Rechargeable and relatively cheap too!

For a step up:

As VinatgeNut points out - you can find 731b's around for relatively cheap, and they WORK GREAT for what you're doing - you don't need a 732 at your level.  731b's will replace every function of a standard cell with ease, and much more reliable and easier to use than anything you're doing with a standard wet cell.  You can even get them cal'd at most any lab for around $100 or so.  Easy Peasy Cheesy.

[Gyro]

I don't think anyone has mentioned the technique of comparing cells connected back to back.

Obviously it doesn't give you the absolute reference voltage, but it does allow you to compare the voltage deviations of the cells more frequently, with minimum current draw and maximum resolution. You can then share the burden of measuring the absolute reference voltage across all of the cells.

You could even have a 'sacrificial' cell, which is used for providing absolute voltage measurement, and which is only referenced to the other cells by back to back comparison.

[MisterDiodes]

...Yes that can be done to compare cells.  The only problem is if you only have a DMM to use for measure (with bias current) then you're always trying to charge one cell and discharge the other.  Discharge is bad enough, but trying to charge a cell can really throw off it's discharge rate - which we hope will be at least kinda sorta constant over time.  At least when the cell is newer during its usable period.  As the cells age the real discharge rate gets a bit wobbly anyway, so it might not even matter in this case.

[fcb]

I've used a silicone conformal coating to seal epoxy packaged references and the edges of the (AN82) style PCB relief, but I do an elevated temperature run first for a few days.

[fcb]

If you can't get hold of MG (what we use) silicones in your country, then you can forget getting parylene.

I've had some success mounting epoxy voltage references in little glass bottles, but the seal is awkward and you have to really use solid state potentiometers or a DAC to perform adjustments. It does yield a stable humidty and isolation from air-pressure changes (if you get the seal right!).  There are some cheap 42ml jars on amazon.

Might be cheaper to use references that are already sealed though.

[Kleinstein]

For the humidity problem I would consider having a sealed metal container for the whole circuit, not just the reference part itself. With sufficient size and some silica or similar buffer one could get away with epoxy feed through.

The weston cells are not really practical for every day use. More like use them for a monthly or maybe 6 month comparison with you next best reference (e.g. LM399 set). They might be an alternative for a long time stable reference if checked against each other and a few checks again a precision calibrated reference a few years apart.

Shipping those refs is best avoided - more like keep in you lab only, preferably a cool place in a basement where they can stay the next years. Shipping or an earthquake might cause changes. That is one reason the got obsolete.

For a not so frequent use, the loading with a high Z DMM may not be such a big problem. Just check the bias of the DMM before use. Depending on the quality of the DMM / amplifier it would be more like each cell separate, maybe 5 in series to compare to a 7 V reference. Keep no. 6 separate, just inc case something bad happens.

[Conrad Hoffman]

I've never seen a single tube cell and without reading every word here twice, is it certain that these are saturated or unsaturated?

[MisterDiodes]

Be careful blaming an observed drift on humidity - that is an effect that usually doesn't appear for several days or weeks after humidity changes.  Sometimes it happens faster but that effect can have the fastest observed effect in the DMM itself.  Another sneaky effect is if you have a rapid temp change in humid conditions where you get an almost invisible film of condensation on your board that mixes with any contaminates left on the board - so always make sure your boards are 100% clean and flux free, and keep them in clean & dry enclosures.  Distilled water (condensate) is non-conductive, but -any- ion contaminates in the water raise conductivity VERY rapidly.

Most practical: An LM399 + PWW resistors and PWW adjust pot will be far more humidity resistant than anything else you have - including the DMM.   An 8 pin plastic dip amp won't see a real humidity change effect for several days at least, SMT version is much more vulnerable of course.

Coatings don't have to be "electronic spray on" plastic.  Coatings from a old school wax dip (keep away from bugs and it attracts dust, and make sure the board stays cool) can work, glass bottles, canning jars or paint can enclosures filled with CO2 / LN2 or whatever the local welder can get you, house paint can slow down absorption (test first to make sure it is non-conductive enough!!), clear wood varnish, linseed oil, run the circuit under mineral oil, a storage cabinet with a desiccant packet or saturated salt solution stored inside - there's a million ways to control humidity, some are better than others.  Test away!  Don't put anything on your board until you know what it does, and I'd wait 6months to a year to make sure your coating under test is really plastic-safe and non-conductive. 

I would also look at setting up your lab space with better temp / humidity control if your delving into the low ppm world.  Get creative.  For instance a humidifier can be made from a cheap box fan pulling air over or thru moist furnace air filter media, and a dehumidifier can be rigged from an air conditioner and a heater.  There are other ways also.  For what you're doing even if you kept humidity controlled to with some ballpark of 30% change range would probably be fine, and even that would be hard to see on a 3468a.  Generally you want to stay over 50% humidity to avoid static problems - the more you go below that the more every surface becomes a static generator.




 

[Gyro]

I've never seen a single tube cell and without reading every word here twice, is it certain that these are saturated or unsaturated?

I can't speak for the OP's cells directly but the Muirhead k-375-C single tube cell is definitely saturated (so it is certainly possible that the OP's are).