[volt-nuts] Back to voltage

[Dick Moore]

I'll stick the oar of my opinions in here -- I've owned a couple of Fluke 732As, both of which needed some serious TLC from me, and so I got *very* familiar with them. Fluke used a Motorola monolithic device with a transistor and a buried zener as references in the 732A, the 8505/6 meters, the 5440A and B DC standards, the 510A AC standards, the 515, and a host of other products. In the critical units, like the 732A and the 5440, the reference parts and the critical resistors all cohabited inside a fairly massive assembly, surrounded by resistance heaters controlled by a thermistor placed inside the assembly near the reference. 

For the purposes of even a very good home lab or shop, the 732A is overkill. Fluke did things we don't have to do, since we have complete (or near enough) control over the way and place that a reference will be used. I have one of Geller's AD587 10V references, and if placed in a simple oven, it's good. But it isn't as good as an LTZ1000 or 1000A running in a closed cabinet at room temp, +/- a couple of deg C. This, anyway, is what my HP 3458 tells me, and that's as good as I'm going to be able to have as a means of verifying a reference's performance.

An LTZ1000A, a 723 regulator, a few decent 20ppm wirewound resistors and a wirewound trimmer, an IC chopper amp like the LTC1150, plus a box roomy enough to hold it all, will make a very good 10V reference with sufficiently low output Z to be useful. All that's missing is a way to measure the trimming of the output. If any of you make such a simple device, I will happily set the output for you using my 3458, if you pay the shipping of your box both ways. I can reliably measure the ambient to about 0.2 deg C with a Pt100 probe and provide that info as well. I think you can reasonably expect an overall stability of 1ppm over several months to a year, providing the unit is run for a month or so continuously before adjustment. 

This isn't magic, so the only real question is: how good is good enough? I sold my 732As and my 5440B. If I want to know what something measures, I turn on the 3458. I'm one of the lucky few who decided it was worth it and could at the time, anyway, afford it. Other things come close at more reasonable prices, such as the Datron 1081/2. I'll know better by October how well the 3458 did over a year, since I will have it re-cal'ed in September. It's about a 20 year-old dog that has some new parts, but its LTZ1000 is very well aged.

And Bill is right -- the real bears are AC standards, which are much flakier in every way, no matter how good they are. To that end, I'm laying out a board for Bob Cordell's state-variable oscillator, which has somewhat better distortion performance than the Fluke 510A and which, thanks to a very tight AGC loop, has good amplitude stability, providing the DC supply is adequately stable over temp and metal film resistors and polypropylene caps are used. Again, a simple oven would be helpful to hold the circuits to within a couple degrees C. I don't know yet how good this can be -- much fun lies ahead.

Dick Moore