# Computing stuff tied to the physical world

## TK – Voltage accuracy

In Hardware on May 10, 2012 at 00:01

Welcome to the Thursday Toolkit series, about tools for building Physical Computing projects.

(this is a bit of a side excursion, about checking the quality of a measuring instrument)

“Ah, but is it any good?” – that’s the inevitable question to ask when getting a precise instrument, right?

I’m referring to the 6.5 digit 34401A HP (now Agilent) multimeter I got my hands on, recently. This translates to: better than 1 ppm (part per million), i.e. 10,000 times more accurate than one percent!

This is the sort of thing the members of the volt-nuts mailing list ponder about, I would imagine.

In my case, with now over half a dozen ways to measure voltage here (numerous hand-held multimeters, mostly), I just wanted to know which one to trust most and to what extent.

The solution comes in the form of a transfer voltage standard – an item you can order, gets shipped to you, and which then gives a certain level of confidence that it will provide a fixed voltage reference. As it turns out, Geller Labs offer just such a thing at low cost – it’s called the SVR 2.0:

Put 15V on its input (left), wait 30 min, and the output pins (right) will produce exactly 10.00000 Volt – magic!

Each board is “burned in” (kept turned on) for 200 hours and calibrated at the temperature you specify (I asked for 21°C). You even get the measured temperature coefficient at that spot (mine is 1.7 ppm/°C), so you can in fact predict the voltage it will generate at a slightly different temperature. Now that’s serious calibration!

My bench-top multimeter will indeed go down to 1 ppm in 6-digit mode, i.e. steps of 10 µV when measuring 10 V:

And guess what – after a 30-minute warm-up (both the 34401A and the SVR), it’s spot on.

No last-digit jitter, nothing. A constant 10.000,00 readout. The current room temperature is 21.1°C, heh.

Think about it for a second: as hobbyist, you can order a precision second-hand instrument from eBay, Google around a bit to find a little voltage standard, have ’em shipped from different parts of the planet, get them here within two weeks, hook up some wires, wait 30 minutes, and they match to 0.000,1 % precision.

Given that this instrument is from the 90’s, I’m massively impressed. This 34401A HP thing rocks!

Voltage? Current? Resistance? Game over – for me, this is more than enough precision for serious use.

1. An interesting reference.

I have a question. Having just read the spec for the SVR 2.0, it says that a 15.0v supply is required for accurate operation, which sounds like a bit of a catch 22! How much effect does supplying the module with 14.9v or 15.1v have on its accuracy?

• Good question – the AD587 specs put line regulation at max ±100 µV/V, or 0.01%. So to stay within 1 ppm, the input voltage has to be within ±0.1 V of 15V (at which it was calibrated). Easily checked with a multimeter of say 0.5% accuracy.

• So a bit of a trade up… If you have a 15v supply accurate to 5000ppm you can use it to get a supply of 10v at 1ppm.

Sounds like a good deal.

Just a pity they chose 15v for the supply. What an odd voltage.

2. Really impressive performance! Actually +/- 15V supplies used to be very common for opamps and analog work in general.

3. Even if your multimeter was way off in calibration so you didn’t know if you had 15.0 V, you could “bootstrap” the calibration- so long as it was linear. Or if you had , say three 10k resistors for a voltage divider to give you 10V = 2/3 of 15 V, and a null meter, that needs no calibration at all. You could confirm your resistors were matched by rotating their places in the divider and comparing the null value.