A couple of months ago I managed to acquire an Ernest Turner Electrostatic voltmeter on ebay. The thing that attracted me to this one was that it has a 1kV fsd, with useful scale resolution down to less than 400V. I used to have a similar 10kV one as a kid, acquired in Lisle street, but like many treasures back then, I was too young to appreciate what I had. The intended use is for mesuring high impedance bias supplies, low current voltage multipliers, GM tube supplies etc.
A very hopeful sign from the leakage perspective was that it acquired a reading as I pulled it out of the bubble wrap which stayed until I shorted the terminals with my fingers, the light brown resin rear panel obviously has very low leakage (I have a small 2.5kV one in standard black bakelite which instantly drops to zero when disconnected).
I found the pointer slightly catching on the scale at a one point, so a good excuse to open it up and take a photo while correcting it. It has three fixed vanes and two very light foil moving ones, dimpled and ridged for rigidity.
Anyway, characterising. The two important parameters are voltage accuracy and loading (DC leakage and Capacitance).The first was simple, my Hunting Hivolt insulation tester is just able to drive the 10Meg input resistance of my DMM up to 1kV on its highest 100uA range. The meter scale indicates that it is a Class 1 instrument and I was able to confirm that the readings matched to within within a couple of percent of fsd, at least down to the point where scale graduations become too cramped.
For the loading parameters, my first idea was to simply use my picoammeter in series with the meter Earth terminal and feed the high potential terminal from the insulation tester to measure DC leakage. Unfortunately, while picoammeters are great when the test sample is nicely enclosed in a screened enclosure with a quiet bias supply, there was far to much AC hanging around any noise from the insulation tester to get any stable reading. The meter is relatively large and I didn't have a suitably large enclosure to hand. I obviously need clear visibility too. The Earth terminal is also indirectly connected to the black bakelite case, which has far too much leakage so I would need to put it on a PTFE platform which I didn't have.
The alternative method was to measure the discharge time of a known value capacitor across the meter. I happened to have recently acquired some 50nF 1.5kV Polystyrene capacitors (2.5kV test), so air wired one of these across the terminals. I though that it would at least swamp out the capacitance variation of the meter with reading (which, it turns out, I grossly overestimated). I went for using a 20V drop at 100V intervals, using I=C dV/dt. It turned out that 50nF gave me plenty of time for sleep and normal life and still being able to hit the readings, visiting it occasionally (did I mention that I got it a couple of months ago!). I started at 1kV and then after each 20V drop, discharged it to the next start value using a 10G resistor.
The attached plot shows the resulting leakage curve. It is reasonably linear up to 700V and then displays a slight upward curve rising from 2.2pA at 700V to 4.4pA at 1kV. I think [Edit: thought?] this is mostly down to surface leakage, for entertainment I've included the 'Leakage NOT ENOUGH CLEANING' plot, where current increases sharply to 19pA at 1kV. Luckily I spotted this fairly early and cleaned the rear panel several times with IPA with a long drying off period.
The other loading factor is capacitance. Again a change of plans! It was easy to measure the capacitance at 0V as approx 8.5pF. My original idea was, now knowing the leakage current at any voltage, I could time the same 20V drops at 100V intervals with the bare meter and calculate capacitance. In practice it wasn't nearly as straightforward - I kept missing the points with sufficient accuracy and the resulting plot made no sense. Also the needle movement was much slower than would be expected from the leakage current order of magnitude and zero reading capacitance. In the end, I took the pragmatic approach, removed the front, and with the meter, capacitance meter, and short lead positions undistubed, gently moved the needle to each voltage reading and noted the capacitance. The readings aren't perfect as it was a fiddly operation, but a second order polynomial best fit curve looks pretty close. Capacitance rises to around 16pF at 1kV reading, less than I thought - maybe there is a certain amount of parasitic fixed capacitance, to the fixed metalwork and scale plate, that masks some of the actual swing.
Looking again at the plots again today, I am still not happy with the leakage results. Clearly they are the combined leakage of the meter and polystyrene capacitor in parallel over a long period of time. I had assumed that the polystyrene capacitor leakage would be negligeable (I can't clean
that with IPA!) but the bare meter should drop to zero in less than a second when it actually takes several hours to drop the last few hunded volts. It is difficult to fault the capacitance readings as they are directly measured and repeatable. Clearly I still have more work to do, maybe using a chain of air wired capacitors in series (I have four) to better quantify their leakage contribution.
Home
Help
Search
Login
Register
