Hello Dr. Frank
I have a few questions about your DIY divider.
1. Do you have part numbers for the switches that you used in your divider?
2. Do you have pictures of the inside and outside layout of the switches for the divider?
3. What voltage do you apply to the input of your divider for transferring range calibrations?
The 100K resistors in the box posted above are rated at 2.3mA maximum so they are about 500mW max.
What power would you think will be safe to avoid self-heating errors? I will have a look at the 752A error calculation discussion in the manual.
One of the things that I want to do is to transfer the 10V setting on my Fluke 731B up to the 1000V source setting on my Keithley 237 and on my Fluke 332A. It looks like I need to join the discussion that you and xlymex are having about working voltage division.
Thank you again for answering in this thread.
1. It's a switch from ELMA, 04-3264. 6 positions (6-to-1), on 3 planes, 3µm gold plated contacts. It can withstand 500V per specification, and is tested at 1000Veff between adjacent contacts, so careful distribution of the HV lines is necessary. It costs about 83€ in 2004.
http://www.elma.com/en-eu/products/rotary-switches/rotary-switching-products/product-pages/type-04-detail/2. A picture can be found here, though it's not useful to reconstruct the schematic, I think:
https://www.eevblog.com/forum/testgear/hp34401-measurement-of-linearity/msg358713/#msg358713These cables are PTFE, and you see the 104 EA / 25kOhm resistors, and several trimming thin film resistors, for each of the 7 different partitions of the divider, i.e. 25k, 3 x 75k, 3 x 750k, and the Wheatstone bridge, 2 x 50k.
The switch circuit is also balanced for contact resistance.
My design is a mix of features from the Datron 4902 (100 MBF resistors) and the Fluke 752A (Hamon type auto-adjustment).
Like the Datron 4902, it falls into the category of 'equal temperature rise' of its divider legs.
The typical error at 1kV due to self heating for either leg of the divider can be estimated as follows:
(power dissipation per resistor) x (heating coefficient) x T.C.
4mW x 0.12°C/mW x 1 ppm/°C = 0.5ppm
If both legs have same polarity of T.C., which is more probable, the overall error will further decrease.
If they have inverse T.C. of each other, it might add up to 1ppm.
1ppm/°C is the typical T.C. for one resistor, which I have measured, but this value might also alternate between ~ 0.2ppm/°C as best case, and 2.5ppm/°C maximum, as specified.
Determination of T.C. of the seven different partitions of this divider, and selection of single resistors (requires maybe additional 20 spare resistors), would allow trimming of the overall power error, like it's done for the Fluke 752A.
Instead, I made a dynamical 1kV divider test, i.e. monitoring of the change of the 10V output voltage after loading the divider with 1kV.
This gave about 1ppm initial drift, as estimated for the typical case.
3. Like in the 752A, I use 20V for bridge excitation.
I had the very same intention, as you. At that time, I owned a 6 1/2 DMM (34401A), a 10V reference (DIY LTZ1000), a 332B/AF, and I wanted to calibrate 10V, 100V and 1000V ranges of the 332B by transfer from 10V. The 34401s DCV ranges of course could be calibrated also, see that other picture, below the reference divider.
If you dive into ppm uncertainty, especially at high voltages, a 'safe' power dissipation value can not be defined, that's the wrong approach.
In the 752A, there are no details given for the T.C. and power dissipation matching, I have to admit.
Therefore, you have to calculate the thermal drift of the ratio, by analyzing the type of divider you have, and the T.C.s and matching techniques of the resistors, like briefly presented above for my divider.
zlymex already has listed many of these HV dividers, and on his bbs.38hot.net almanac, he has nice overview and pictures and analysis of the 752A, already.
http://bbs.38hot.net/thread-137-1-1.html (needs google translator, maybe)
For the different categories of dividers, by applying similar calculations, or from experiments, I'd like to supplement their estimated 1kV errors.
Critic welcome!
Frank