High voltage reference standard design (that's one way to do it!)

[alexwhittemore]

I've been thinking a lot lately about voltage standards. I've picked up lately a few multimeters and a bunch of test gear in general on ebay and craigslist that I'm reluctant to trust (getting less so!), and I started out thinking to myself, "how can I cheaply get some kind of reference standard I'm confident in to measure these devices against?" My first thought was "buy a precision IC," but it turns out you can only get those precise to .05 or MAYBE .005%, and that's if you're willing to spend many dollars or tens of dollars per chip. That led me to ultra high stability references like the LTZ1000, whose initial accuracy isn't even really specified - you design around these things for stability, then calibrate them later with an outside standard and just keep track of the measured value. This "secondary standard" calibration got me looking into primary standards where I learned a bunch about Weston cells and the new kid on the block, the Josephson junction. The latter is what Sandia national lab uses as a .05ppm transfer primary standard from NIST. My understanding is that NIST uses the same for their primary standard, presumably good to the same precision, but I guess because they're NIST they get to say "our unit is the definition of the volt"? Is there any way to get a transfer standard for the Volt better than .05ppm?

Anyway, now that I know a bit more about precision references and secondary standards, I got to wondering "what about for high voltage stuff?" I can go down to my corner ebay or ConRes and get a fluke calibrator that gives me 300V ±10ppm, but how's that made? I suppose if you have a high precision and stability 1-odd or 10-odd volt reference, you could use a highly matched resistor network and opamp to multiply that, but then opamps are only good up to, say, 30V. Beyond that you could do some sort HV semiconductor circuitry with pass transistors in a follower config divided down, but then that's only good to a few hundred volts. Surely some HV guys somewhere need a good 5ppm reference for a kilovolt. Or 10kV. Or 100kV! I happened across this fun little app note from TI:

http://www.ti.com/lit/an/sbaa203/sbaa203.pdf

Gist: A couple HV guys needed these sorts of standards, and they figured out that if you wire a REF5010 appropriately, you could make a pretty-darn-good shunt regulator closely approximating an ideal zener diode. Then they multiplied that 10V reference by 10k to get a 100kV standard. Quote from the paper: "This design probably hits the record of simultaneous use of TI parts in a single device." Fun!

What other ideas (or examples) do you guys have? Say I have 10V ±5-odd ppm. How could I make from that a precision 300V? Most cheaply and easily? Most precisely? Of course I realize those are totally different goals, I'm just looking for ideas and open discussion.

[cellularmitosis]

I've been curious about this myself, for calibrating the 300v range on my old HP 3478a

[wiss]

1: Read the manual for Fluke 341A, 10 uV to 1 kV, I remember it to be a high-side driver thingie.

2: Set divider ratio at 10 V, then apply 100 V (so the divided voltage is the same as your 10 V)

3: Google "Hamon divider".

[alexwhittemore]

1: Read the manual for Fluke 341A, 10 uV to 1 kV, I remember it to be a high-side driver thingie.

2: Set divider ratio at 10 V, then apply 100 V (so the divided voltage is the same as your 10 V)

3: Google "Hamon divider".

So I suppose the basic "high value divider with an opamp and series pass transistor for regulation" is actually a reasonably solid plan, then. Thanks for pointing me to the Hamon divider, neat little circuit for basic switchable order-of-magnitude divisons!

they are they not because they are "NIST" or how many silicon they have. they are they because they know what they are doing, including the question you asked. and they are they because they collected enough funding to do what they know and made who they are.

I think your meaning may not have translated properly . I can't say I know what you're getting at.

[David Hess]

Oddly enough I was considering the same calibration issue with my HP 3478a.

Using a series reference as a shunt reference is nothing new and unlike the TI parts in that excellent and hilarious application note, some are specified to operate that way.

Linear Technology application note 6 shows a 0 to 100 volt reference which is adjustable in 1 millivolt steps.  It would be straightforward to extend it to higher voltages and make it operate at several fixed values.

Another option which I would consider is a suspended operational amplifier design but short circuit behavior could be tricky.