Kelvin Varley divider Gen Res DV-4107D

[VintageNut]

I opened up this beast for the first time and was surprised to find that this divider appears to have an adjustable first decade. I have traded emails with the present-day factory that built this divider and they do not have any documentation for adjusting this divider.

I am opening up the floor for suggestions about how to go about a performance verification. I am paying around to see if I can get some stable readings with a Fluke 731B and a DMM7510.

Attached are pictures.

[VintageNut]

more pictures

[Vgkid]

Thanks for the teardown, im trying to figure this one out. Will be easier on my laptop.

[VintageNut]

If you look carefully at the resistors, you can see that decade 2 has 2.2k ohm resistors. Decades 3 through 10 have 480 ohm resistors.

As a working assumption, I believe that it is necessary to accurately measure output resistance of dial setting 099999X and adjust the lowest first decade so that the the resistance of dial setting 1000000 has exactly the same resistance as dial setting 099999X. Then each step of the first decade must be adjusted to this resistance.

The other method would be to shunt the second decade so that dial setting 099999X becomes exactly 10k ohms.

These are just educated guesses at this point. 

Measuring the divider with a DMM7510, 100K is high by 35ppm. The DMM7510 100K total uncertainty is 33ppm so the DMM7510 is not good enough to know exactly what the 100K resistance actually is. I need either a better DMM or a 100K resistor with very tight uncertainty to make a divider/bridge for accurately determining what the resistance is.

Thinking ahead, it would be good to have an ESI SR1010 that has 10K steps and have that calibrated to a very tight uncertainty.

[Kleinstein]

There is no need for the absolute resistor value to match exactly. For the divider only the ratio has by right.

There are only 6 adjustment pots - which implies that the first decade is not a normal classical KV stage, but uses more switches to reduce the number of resistors: e.g. have the chain in steps of two and switch one resistor in parallel so an other to get half the value. At one of the ends.

The small board with just one resistor and the pot seams to be there a a kint of shunt for the second decade, to adjust for the total resistor (to make adjustments easier).

Adjustment / verification procedures should be similar to other similar build bridges. There should be a few plans around from good old times when such instruments came with full schematics and repair instructions.

[Vgkid]

You have 11 adjustable resistor sections, with the 1 adjustable shunt resistor.  I thought the 2(19.96k) resistors are in parallel with the 100ohm in parallel with the 350Ohm+series trimmer. Though while stable, this doesnt give to 10K that was expected. It is a few ohms off.

[VintageNut]

There are only 6 adjustment pots - which implies that the first decade is not a normal classical KV stage, but uses more switches to reduce the number of resistors: e.g. have the chain in steps of two and switch one resistor in parallel so an other to get half the value. At one of the ends.

There are 12 pcbs total with 12 adjustment pots. 5 pcbs inside and 7 pcbs outside.

[VintageNut]

You have 11 adjustable resistor sections, with the 1 adjustable shunt resistor.  I thought the 2(19.96k) resistors are in parallel with the 100ohm in parallel with the 350Ohm+series trimmer. Though while stable, this doesnt give to 10K that was expected. It is a few ohms off.

I believe this is wired the same way Fluke does it. One 19.96k is in series with the 350 + 100 + pot. The pot allows the higher resistance string to be over 20k enough to make it in parallel with 19.96k to be exactly 10k.

The output resistance at dial setting 1000000 is only a few ppm over 10k.

[Assafl]

There are only 6 adjustment pots - which implies that the first decade is not a normal classical KV stage, but uses more switches to reduce the number of resistors: e.g. have the chain in steps of two and switch one resistor in parallel so an other to get half the value. At one of the ends.

There are 12 pcbs total with 12 adjustment pots. 5 pcbs inside and 7 pcbs outside.

12 PCBs but the one on the outside right (in the picture) is different. It is probably a shunt resistor for stage 2 (so its input R will match the stage 1 2xR).

[VintageNut]

There are only 6 adjustment pots - which implies that the first decade is not a normal classical KV stage, but uses more switches to reduce the number of resistors: e.g. have the chain in steps of two and switch one resistor in parallel so an other to get half the value. At one of the ends.

There are 12 pcbs total with 12 adjustment pots. 5 pcbs inside and 7 pcbs outside.

12 PCBs but the one on the outside right (in the picture) is different. It is probably a shunt resistor for stage 2 (so its input R will match the stage 1 2xR).

Yes agreed. The second stage is comprised of 2.2k resistors. 10 x 2.2k is 22k. The 12th pcb has a 220k resistor. 220K // 22k is exactly 20k. The pot allows the 220k to be larger than 220k for tweaking. It is a simple yet beautiful design and implementation.