Kelvin Varley Divider [and Precision Voltage Source]

[lowimpedance]

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

Ouch, I had better treat it well  then
This unit plus a 1 k version were rescued from dumpster doom at work (some old custom bit of gear that no doubt cost a few arms and legs in its day!!)
By the way there is a Fluke 720a collecting dust in the instrument store at work, if anyone is interested in some internal pictures let me know.
John

[BravoV]

By the way there is a Fluke 720a collecting dust in the instrument store at work, if anyone is interested in some internal pictures let me know.

John, yes please, really love to see how is it look inside.

[amspire]

This unit plus a 1 k version were rescued from dumpster doom at work (some old custom bit of gear that no doubt cost a few arms and legs in its day!!)

EIT only mention the price of the 3 decade version, but that one is probably worth well over $6000 to buy one new - possibly creeping up towards $10K.  A pretty good find in a dumpster.

[lowimpedance]

EIT only mention the price of the 3 decade version, but that one is probably worth well over $6000 to buy one new - possibly creeping up towards $10K.  A pretty good find in a dumpster.
[/quote]
 
Well that is rather scary for what is in it (of course precision costs and a little, and a bit of profiteering on top too!! .)

By the way there is a Fluke 720a collecting dust in the instrument store at work, if anyone is interested in some internal pictures let me know.

John, yes please, really love to see how is it look inside.

Ok as requested attached is a collection of pic's of the Fluke 720a with covers off.
The first thing that hits you are the trailer load of trimmers (a good number also hidden from view in the pic's.)
The green block at the rear is a 'can of precision resistors in oil!.
The green switches are special low thermal types (Leeds and Northrup)
The reference used in the 'lash up test' is a ltz1000 around 12 years young'
Again the 34401 is set for an input r of >10g ohm.
I hope that satisfies your curiosity
cheers
John

[amspire]

EIT only mention the price of the 3 decade version, but that one is probably worth well over $6000 to buy one new - possibly creeping up towards $10K.  A pretty good find in a dumpster.

Well that is rather scary for what is in it (of course precision costs and a little, and a bit of profiteering on top too!! .)

Not necessarily profiteering. It looks like the resistors are more or less hand made. They are probably made, aged and then someone has to adjust the value of every resistor until the whole set of resistors match. With things like this, it is easy to match the resistors, construct everything, a night or two in the thermal chamber to age, test again, and then find it is a little out. Then you pull it apart again, adjust, assemble, another night in the thermal chamber,  and check again. This is repeated till it is right.

It looks like there are no adjustments in it, so everything has to be accurate.

I don't think they can be made very quickly at all, and at that price, they will be made to custom order.

The green block at the rear is a 'can of precision resistors in oil!.

The can of resistors for the first decade is the huge silver looking box taking up most of the rear with the "CAUTION" sticker on. It didn't look green to me, but the colour may not show up in the photo. It should contain 12 sets of 3 resistors, so 36 large Fluke precision wirewound resistors in total. I do see large green switches.

Thanks for the photos. I haven't seen inside a 720a before. So that is all you need to make a 1 part in 10^-7 divider! 

Richard

[lowimpedance]

Richard
 When looking at it from that perspective of hand made it can easily mount up in price.
Did a search and found the price for the DP1311 to be mid $5000 , must say I'm still a little surprised.
May be it has something to do with being around this sort of gear in my work that I dont give it much thought as to the costs.
(I dont do the purchasing of this type of gear! , more just general test equipment).
Sorry about the color rendition in the photo, but the box is green.
The manual says if something goes wrong inside the box, send it back ( read betwwen the lines; you are screwed!).

John

[amspire]

The manual says if something goes wrong inside the box, send it back ( read betwwen the lines; you are screwed!).

I think the current price for the IET KVD-700 (the same as the Fluke 720a) has a "Starting Price" of $28,895, and that box could be worth at least $20,000 of that.

Richard

[BravoV]

Ok as requested attached is a collection of pic's of the Fluke 720a with covers off.
The first thing that hits you are the trailer load of trimmers (a good number also hidden from view in the pic's.)
The green block at the rear is a 'can of precision resistors in oil!.
The green switches are special low thermal types (Leeds and Northrup)
The reference used in the 'lash up test' is a ltz1000 around 12 years young'
Again the 34401 is set for an input r of >10g ohm.
I hope that satisfies your curiosity
cheers
John

Thanks John, great pics.

About those truck load of trimmers, just curious, I assume those golden trimmers are not like your average trimmer out there right ?

It didn't look green to me,

May be all your gadgets/instruments, monitor included are maintained and fully calibrated properly, you just forget to calibrate your eyes Richard.    ...j/k ...

Btw, it looks green here.

[lowimpedance]

Ok as requested attached is a collection of pic's of the Fluke 720a with covers off.
The first thing that hits you are the trailer load of trimmers (a good number also hidden from view in the pic's.)
The green block at the rear is a 'can of precision resistors in oil!.
The green switches are special low thermal types (Leeds and Northrup)
The reference used in the 'lash up test' is a ltz1000 around 12 years young'
Again the 34401 is set for an input r of >10g ohm.
I hope that satisfies your curiosity
cheers
John

Thanks John, great pics.

About those truck load of trimmers, just curious, I assume those golden trimmers are not like your average trimmer out there right ?

It didn't look green to me,

May be all your gadgets/instruments, monitor included are maintained and fully calibrated properly, you just forget to calibrate your eyes Richard.    ...j/k ...

Btw, it looks green here.

Actually they are just single turn trimmers all 25ohm ! but I suspect a little bit better quality than some
Also there are a truck load of multiturn trimmers under the PCB too (access from the front, shown with the 'flap' up).

[amspire]

It didn't look green to me,

May be all your gadgets/instruments, monitor included are maintained and fully calibrated properly, you just forget to calibrate your eyes Richard.    ...j/k ...

Btw, it looks green here.

I thought that was a board sitting on top of the box, but now I have done the eye recalibration, I see it is the top of the resistor box.

[HLA-27b]

Thank you very much for the pictures John,

One thing that I could not see from the pictures - What is the winding type of the resistors? Is it normal helical winding or is it bifilar?

[lowimpedance]

Thank you very much for the pictures John,

One thing that I could not see from the pictures - What is the winding type of the resistors? Is it normal helical winding or is it bifilar?

Hang tight until Monday and I will check when I get to work.
John

[Conrad Hoffman]

Cool, thanks for the photos; I'd never seen what's under her skirt either. IMO once you have a procedure down for building the resistors, the rest is just hard work. Remember that the values on the upper decades are high, so it involves a huge amount of very fine and difficult to work with wire. Creating a bight (I misspelled that some time back) is a PITA. Julie wound their resistors on plastic bobbins with separators. So many turns one way, then move to the next section and wind the other way. Loose is the rule to avoid stress. Then they added a short length of resistance wire externally on the switch terminal to trim to final value. Sensible solution, IMO, and they're said to remain stable for many years. The Fluke trimmers look just like the trashy ones I find in old hi-fi amps for bias. They may be wire wound. The low value resistors are card wound and very likely are a simple unidirectional winding. All the Fluke resistors I've seen are. For a while the Fluke KVDs could be had cheap, while everybody was going out of business in the USA, but the test equipment market seems to solidified at a high price now.

[lowimpedance]

Just had a look at the ESI dekapot and the Fluke 720a and the card resistors are as wound as Conrad suggested. The others ...well cannot see inside the can!.
The single turn pots that ideed appear somewhat cheep on the pictures are wire wound and were made by CTS.
John

[amspire]

Just had a look at the ESI dekapot and the Fluke 720a and the card resistors are as wound as Conrad suggested. The others ...well cannot see inside the can!.
The single turn pots that ideed appear somewhat cheep on the pictures are wire wound and were made by CTS.
John

They will be totally adequate. In the critical first decade, the way they are connected, they are adjusting the 10K divider stages by +/- 0.2 ohms. So they pick a Select-On_test resistor to get each divider section in the first decade matching to within 10ppm, and the pot only has to be stable to 1% to get a final 0.1ppm accuracy.

The pots may look cheap, but Fluke have probably made sure that the same type of resistance wire is used in both the pots and the divider resistors, so all the temperature drifts will match. They probably do a better job in the 720a then a modern 10 turn trimpot would do.

Richard

[amspire]

I am starting to pull the elements together for a KVM, so I think I will set a target.

I want to make a 6 digit KVM with a budget (excluding case) of $60. I could spend more money on better resistors, but I just don't want to. I thought about winding resistors, but the wire is not cheap at all, I don't know where to get good formers from, and when you are talking resistors in the 1000's of ohms, there is a lot of winding to be done with very fine wire. So scrap that.

Here is the plan:

Switches: $4 each  (ebay item 300562748033). I will review these switches in my next post.

Resistors: 3k3 3W 1% 50ppm metal film. $10 per 100 resistors. A total of 103 resistors will be needed, but I can use cheap  1% 1/w watt on the lower ranges. To calibrate, I will use 100 10 ohm 1% resistors.

Knobs: 6mm shaft cheap comfortable knobs with a printed mini-dvd disk stuck/screwed to the bottom with the 0 - 9 printed on it.

Terminals: Cheap gold plated brass 4mm terminals.  Thermal voltages are only a problem when temperatures differ, so I will make sure they don't.

Adjustment Pots: I will first test the resistors, then work out how much adjusting I will allow. Bulk packs of Chinese multiturn pots are easy to get.

The first decade will use three 3k3 resistors in series per divider. The other decades will use a single 3k3 resistor.

I will age all the resistors first. This will be done by connecting all the resistors in parallel and cycling between 80V and 0V for a day or two. To do the cycling, I will use a mains timer with alternate setting tabs up or down.

I will then work out the temperature coefficient of the resistors, and assuming it is a bit negative, I will wind some extra copper wire on the resistor to cancel out the temperature coefficient. This will be done by putting a bunch of resistors between two pieces of metal with just the leads sticking out, add some thermocouple probes, and heat to a bit over 100 deg in an oven. When it comes out, I will wrap in some insulation, and let the temperature stabilize. Then as the resistors slowly cool, I will measure the resistance of each. I only need the first decade done really well.  The each lower decade only has to be 1/10th as good as the decade above it. The copper wire has a positive coefficient of about 4000ppm/C, so if a 3k3 has a -20ppm coefficient near room temperature, I would need to add 33 ohms of copper wire in series with the resistor to get a zero coefficient. 40 gauge copper wire is 1 ohm per foot, so that would be 33 feet or 10M wound around the resistor. The 3W resistors are large, so there is plenty or room to add the copper wire.

Errors between resistors in a divider set are cancelled out by adding selected 10 ohm resistors in series. If the resistors turn out to be stable in the short term, I will aim for better then a  100ppm match in the first decade plus use a pot to adjust down to 1ppm. Realistically, if I can get 10ppm reliably from the KVM, I have done well.

Richard

[lowimpedance]

Richard
 Lood forward to see the result when your done. The old 'sperry', couldnt bring yourself to use the switches from that (or are they not suitable?).
I like the idea of the mini dvd disk dial to get the numbers.

Just to note in the 720a oil filled resistor can, each section in the divider is made up of 4 resistors of 2.474k in series then externally in series with those is a combo of
 factory selected R + (8.45k+5k trimmer//100R), getting the temp co. right here must have been real fun. This divider uses the low thermal (green) switch.

cheers
John

[amspire]

I received my 1 way 2 wafer switches.

https://www.eevblog.com/forum/index.php?action=dlattach;topic=5442.0;attach=16882

The ceramic wafers are 25mm diameter, and the shaft is a 6mm metric shaft.

It looks like the contacts are silver plated brass.

So are they good enough?

Quality: Contacts are very firmly riveted onto the ceramic wafer and do not move. Alignment is poor, so that some contacts are not centered properly. The contacts seem to work though. The batch I received had some tarnishing on the silver. Initial contact resistance was high - sometimes over 0.1 ohms, but after turning the switch back and forward 10 times, most contacts were 12 milliohms. I pulled some switches apart and detarnished the contacts and resistance reduces down to 9 milliohms on the good contacts. Some contacts can be bad, so get extra switches and pick the best one for the first decade. I think they are good enough to get started.  I compared it with some 50 year old super quality switches - the 50 year old ones had a contact resistance of 0.002 ohms with a variation of 0.0001 milliohms. To at anything like that today costs hundreds of dollars per switch.

Construction: Can be taken apart. Good quality ceramic wafers which are the  best for low leakage. Ceramic wafers typically have resistances between contacts of over 900GOhms. Heavy duty beryllium copper spring for the detents is not going to wear out or break.

Feel: Nice positive detents. Satisfied.

Thermal EMF voltages. In a practical test with soldered copper wires, and heating up one contact with my finger, it looks like there will be less then 0.3uV thermal voltage in total, so I am not going to worry there. If nothing is heating the pins significantly, it will not be measurable. The first decade resistors will not be mounted on the switch, and they are the most likely resistors to produce heat.

This switch is about $4 in one of quantities, and you can get much cheaper if you buy in quantity. I can get as low as about $1.60 if I buy hundreds. If you try and buy something similar at Digikey, Mouser, etc, you will probably pay 5 to 10 times the price.

So I will give them a go.  If I later get my hands on better switches, i can replace the higher decades.

Richard

[amspire]

Richard
 Lood forward to see the result when your done. The old 'sperry', couldnt bring yourself to use the switches from that (or are they not suitable?).
I like the idea of the mini dvd disk dial to get the numbers.

Just to note in the 720a oil filled resistor can, each section in the divider is made up of 4 resistors of 2.474k in series then externally in series with those is a combo of
 factory selected R + (8.45k+5k trimmer//100R), getting the temp co. right here must have been real fun. This divider uses the low thermal (green) switch.

cheers
John

I will do a seperate post on that Sperry today, because it is is very interesting.

I bet is would be really hard finding anyone who could make a switch today like that green 720a switch. Very consistant low resistance, high voltage as it has up to 1.1kv on it. Very low current leakage. If someone does make an equivalent, the cost would be incredible.

[alm]

Metal film should have a positive tempco if I remember correctly. Apart from this, sounds like a well thought-out project. Will make sure to follow your progress. Thermal EMF will be an issue if you go down to the uV/nV levels. Copper to gold is about 0.5 uV/K, so with 10mV across the divider, the resolution is 10nV, requiring the temperature of the two binding posts to be within less than 0.01 or so K. If you stick to reasonable voltages, it becomes less critical. With a resolution of 1uV, you only need to be within 1 K or so. Not unreasonable, but not in the 'you need a blowtorch to see any difference' territory either.

[amspire]

Metal film should have a positive tempco if I remember correctly. Apart from this, sounds like a well thought-out project. Will make sure to follow your progress. Thermal EMF will be an issue if you go down to the uV/nV levels. Copper to gold is about 0.5 uV/K, so with 10mV across the divider, the resolution is 10nV, requiring the temperature of the two binding posts to be within less than 0.01 or so K. If you stick to reasonable voltages, it becomes less critical. With a resolution of 1uV, you only need to be within 1 K or so. Not unreasonable, but not in the 'you need a blowtorch to see any difference' territory either.

The thermal voltages are only an issue if something is generating a temperature differential. My testing showed that with the switch carrying no current,there is just no measureable thermal emf (with a 0.1uV resolution meter).

I will not mount the first decade resistors on the switch, so the only place where a significant heat can be generated to the switches is in the second decade resistors. These will be large 3W resistors and even if I apply 1000 volts to the KVM, their heat will be 3mW per resistor which could possibly cause a rise of one degree in the wire to the switch. With 10V applied to the KVM, the power will be 3uW per resistor which is nothing. So if I have a problem, it will only be at extreme voltages, and at 1000v, I can afford to have emf's up to 1mV without any change in accuracy. I think I would only trust the switches up to about 250V, and I do not have any supplies that go above 60Volts.

I will probably need to somehow shield the heat from the first decade resistors from the rest of the system, but I wont go as far as putting them in a can of oil.

As far as the terminals are concerned, I can put a foam insulator over them inside, so they are not getting any significant radiated heat. Thermal EMF's generated through handling the front panel terminals with fingers will change as they cool again, so if the emf's are relevant, I will wait till the drift stops. I am just not going to spend $20 per terminal to get the low EMF copper alloy ones.

Richard.

[amspire]

Metal film should have a positive tempco if I remember correctly

The 1/4W one I have tested here have a negative tempco, but I will just have to wait and see what the 3W ones do when I get them.

[BravoV]

Richard, looking forward to see your build in progress in the new thread, lots of pics please ! 

[fmaimon]

Richard, where did you find those 3W 50 ppm resistors? Digikey nor Mouser seem to have them...

[amspire]

Richard, where did you find those 3W 50 ppm resistors? Digikey nor Mouser seem to have them...

Only the best for me.

Brand is totally unknown, so I will have to see if they are good enough to equal the world famous "One Hung Low" resistors. 

http://www.uxcell.com/wire-leads-33k-ohm-metal-film-resistors-100-pcs-p-161865.html

I will test them first, and if after aging, I think I can get enough short term stability, I will press on. I ordered them today, so in a week or two, I should know.

I did look at Digikey and Mouser, but it seems as soon as you get above 1/4 watt, the resistor prices skyrocket.

I just have a feeling that if I can't afford to buy the Vishay 1ppm foil resistors, then I will get better results from nice big chunky resistors of lesser quality then expensive brand name 25ppm resistors in a small package. I think it is the duty of the hobbyist to try and do things incredibly cheaply that normally are done by spending big bucks. Well, it is fun anyway.

Now between the decades made of 3k3 resistors, you need to add a 8k2 resistor + some extra calibrating parts ( a pot + small resistors) that goes across the next decade. So I ordered 8k2 3W resistors as well.

Between the  first decade with 10K (three 3k3 resistors) per division and the second stage with 3k3 per division, I will need a resistor across the second decade of 50K. I haven't decided what I will use for it yet.

Richard.