Resistors for high voltage applications

[John_BS]

I thought it might be useful to share some research and practical measurements relating to resistors for high-voltage probes and similar.

You can find resistors advertised with quite impressive-looking specifications (voltage rating several kV, temperature coefficient <50 or 25ppm/K etc.), but what is rarely quoted is the voltage coefficient, i.e. the relationship between the applied DC voltage and the resistance defined at that voltage.

I was unable to find a resistor with a quoted voltage coefficient less than 2ppm/V, and this was only achieved up to a certain resistance value: as the resistance approached the specification limit (for a particular resistor size) , the coefficient rose to (typically) 10ppm/V.

At first glance 10ppm sounds like a small ratio, but if you're wanting to build a fairly precise probe for a DVM, 10ppm/V with 5kV applied amounts to a change in resistance of 5%!! So all that money you spent on a 1% part with a stable temperature dependence was wasted.......

I think the lesson here (for critical applications) is
1) to ensure you have this parameter specified in the first place, and
2) try to use a number of lower-value resistors in series to obtain the desired resistance. I've just purchased some 33M VR37's, which were 5% tolerance but measured closer to 2%.
Vishay quote <10ppm/V, but as these are nominally 3.5kV capable, running them below 1kV resulted in much less variation with voltage (c 1.5ppm/V)

John

[GrimJosef]

I used to build very, very high voltage probes (hundreds of kV) at work. We put lots of identical relatively low-value resistors in series and then measured across the bottom one using a much higher resistance monitoring circuit. This approach compensated for any voltage nonlinearities. If we needed good temporal performance we also tried to locate the resistor chain within the equipment so as to minimise transient effects due to stray capacitances. As with so many things in the field of pulsed power we learnt all this from Charlie Martin's group at Aldermaston http://www.independent.co.uk/arts-en...n-1085802.html.

Cheers,

GJ

[John_BS]

As with so many things in the field of pulsed power we learnt all this from Charlie Martin's group

GJ:

That's a really interesting obit.

It did occur to me that a chain of identical resistors would be self-correcting, but of course, for a DVM with 10M input, the R's would need to be well under 1M to avoid the loading, and the probe becomes a bit impractical!

J

[GrimJosef]

Charlie was a really, really interesting person. Much of what he knew he couldn't tell us of course. But the few things he could tell us were fascinating and hair-raising in roughly equal proportion. His work is very well summarised in this book http://www.amazon.co.uk/Martin-Pulse.../dp/0306453029 (sadly too expensive for most people to own, but may be accessible through libraries) and a history of the field from a US perspective is given here http://lcweb2.loc.gov/pnp/habshaer/m...md1450data.pdf.

As far as monitor design goes, it depends on the source impedance of course and the duration of the pulse. The source impedances I was working with were typically of the order of ohms, delivering tens or hundreds of kilojoules in a few hundred nanoseconds. So I could easily live with a probe resistance of just a few tens of kilohms made up of, say a hundred 470R resistors in series. The bottom resistor would still see a few kV (hence the need for nonlinearity correction). So the monitor across it could be designed for a further 200:1 attenuation, say a further twenty 470R's feeding a 50 ohm co-ax out to a fast scope. This second attenuator would sometimes have to be built into a screening can as noise pick-up inside these machines could be 'an issue' to say the least .

Cheers,

GJ

[ppppenguin]

My homebuilt EHT probe is shown here: http://vintagetvandradio.myfreeforum...hp?f=15&t=1152

I built it originally in 1974/5 when I got my first ex-rental colour TV. I was working at BBC Research Labs at the time. 33M VR37 resistors were readily available there