My first KVD - the "Decade Voltage Divider" by James G. Biddle Co

[Alex Nikitin]

Just couldn't resist. It is an ancient thing, almost (but not quite) as old as me... which is to say pretty old  . If we take "11/71" as its DOB, it should turn 50 soon. It took a while to arrive from the US, and when it eventually landed on my doorstep and I've connected it to check what condition it is in, I was quickly disappointed, as one of the steps on the second decade was obviously shorted. Opened it, admired the view and quickly found the fault  . One of the quickest repairs I've ever done. Now it operates almost in spec, below 10ppm full scale errors across the range. Switches are smooth and consistently accurate. Below some photos. Yes, it came with a full manual too  .

Cheers

Alex

[YetAnotherTechie]

Can't see any picture   

[Alex Nikitin]

I can see the pictures, and see no reason why you shouldn't  .

Cheers

Alex

[Alex Nikitin]

Ok, now the photos are added as attachments.

Cheers

Alex

[TimFox]

In the manual pages you posted, there are good specifications to linearity, etc., but no thermal specifications.  Are any temperature effects quoted elsewhere in the manual?  A beautiful piece of equipment, especially the switches.

[Alex Nikitin]

I will scan the complete manual properly at some point, it is about 20 pages. It mentions further on that the wire used has tempco <10ppm/C . Here is the second page.

Cheers

Alex

[TimFox]

One can infer a tempco spec for the voltage ratio from the statement that the advertised accuracy is good over 25 +/- 5 deg C range. 
Assuming they used the same wire for the critical resistors (more significant digits), the ratio tempco should be quite a bit better than the wire tempco.
The "sensitivity" keys are very important for the original application, where balance between (the reference voltage divided by the ratio) and (the voltage being measured) was checked with a very sensitive analog ammeter (galvanometer).

[Alex Nikitin]

I've played a bit more with the divider as it settled down in the lab, and it appears to be fully in spec  . I cannot make it to produce an error above ~ 5ppm (of the input value). At the moment I only have the Keithley 2015 to measure the output voltage, however I use the Fluke 731B as the source and it is pretty stable. I can dial 123456 and get 1.23456V on the 2015. On the other hand if I dial 590000 I get 5.90004V. The input resistance is about 100.002K, my 2015 is not that accurate on resistance, unfortunately. However I can see that the input resistance does not change more than 10ppm over the full range of first two dials.

I am pretty happy about this purchase. $500 well spent. 

Cheers

Alex

[tggzzz]

I've played a bit more with the divider as it settled down in the lab, and it appears to be fully in spec  . I cannot make it to produce an error above ~ 5ppm (of the input value). At the moment I only have the Keithley 2015 to measure the output voltage, however I use the Fluke 731B as the source and it is pretty stable. I can dial 123456 and get 1.23456V on the 2015. On the other hand if I dial 590000 I get 5.90004V. The input resistance is about 100.002K, my 2015 is not that accurate on resistance, unfortunately. However I can see that the input resistance does not change more than 10ppm over the full range of first two dials.

I am pretty happy about this purchase. $500 well spent. 

A KVD has a constant input resistance but a varying output resistance. Hence the load should be infinite. That is, by definition, the case when a KVD is used in conjunction with a null meter to measure voltage.

Hence when working out the division ratio, the 731's output resistance shouldn't matter. But it is necessary to ensure the voltmeter is configured to be high impedance, not merely 10Mohm.

[Alex Nikitin]

On 10V range and below the Keithley 2015 has high input impedance.

Cheers

Alex