View Single Post
Old 31st Jan 2016, 8:00 pm   #2
mhennessy
Dekatron
 
Join Date: Jul 2007
Location: Evesham, Worcestershire, UK.
Posts: 4,241
Default Re: A precision voltage calibrator

The LT1021C and LT1236 both seem to be very impressive - a headline spec of 0.05% basic accuracy isn't bad at all. Of perhaps more importance is stability over time - 15ppm (0.0015%) is very respectable (some manufacturers will "burn-in" their references before fitting them). Tempco is also good.

For context, the best-case DC accuracy of the top-line handheld Fluke meters is 0.025% - a more commonly found meter (e.g. Fluke 75) is 0.4%.

So such a thing seems like a worthwhile project to me, though I'd suggest that for the intended use, a cheaper reference would probably suffice. As he mentions the LM4040, I'll add that I have used it many times, and it's pretty foolproof - and the price is highly attractive too (64p plus VAT from Farnell). My experience suggests that the 0.1% specification is pretty conservative. If absolute accuracy really matters to you, I'm sure someone could measure it for you - indeed, most professional references come with a label stating the exact voltage as measured for reference.

He mentioned the multiple resistor trick - it works well in my experience, and is usually cheaper than buying precision resistors.

Given the cost of the LT ICs, I'd add as much protection as I can (inverse-parallel diodes, etc). I'd also study the datasheets extremely carefully to make sure I was using them correctly - e.g. is there a risk of the potential divider loading down the output voltage if you choose lower resistor values? Linear Technology have some of the best datasheets and appnotes out there.

BTW, I'd be tempted to split the 2k into 1k8 and 200 ohms to get 100mV. Most Fluke meters have a separate "mV" range, so it's nice to test that separately.

When using analogue meters on the 1V output, "loading effect" will cause errors as Zout of the calibrator will be 1.8k with the values suggested in the project. Into the 1V range of an analogue meter with a sensitivity of 20k/V, the reading will be 0.917V, an error of 8%. The error reduces to 3% on the 3V range of an AVO 8 - IIRC, the AVO is 1% at FSD?
mhennessy is offline