G'day all,
After recently acquiring a couple of nice 3478a to compliment my 3457a I've begun to do some comparative checks.
One of them is for all intents and purposes almost bang on compared to the in-cal 3457a, and one reads consistently high (~50ppm). I'd like to be able to track/log drift for a while to see if it requires adjustment or is a deeper issue.
I've been playing around with trying to get a vaguely stable DCV source for comparison and before I started to bodge up some Manhattan style madness I thought I might ask for advice.
I tried using my basic Dick Smith 30V 2.5A linear psu through a large RC filter. That wanders all over the place. I reckon it moves 10 millivolts if I fart near it.
My best result thus far has been to use a Tek PG506 on the 50V output into a divider comprised of 5 220k resistors in a 3/5 configuration. That allows me to put the PG into vari mode and get close enough to 30V. It's all over the place (relatively) with local condition variation and the weather here in Perth has been up and down over the last week, so getting any relatively stable comparison has been "fun". When I say all over the place, it has dropped 2.5mV over the time it has taken me to compose this post. That is related to a change in airflow in the room which has probably pulled half a degree C out of the PG in the last 20 minutes. Some of that is probably the divider also as it's swinging in the breeze, but it's part of the problem.
As an aside, I have a 3 meters hooked up in parallel on DCV, and I plonked a 330uF cap across the input to see if it helped. With the impedance of the PG output and the extra half a meg of the divider it took hours to charge up and get stable, but I've been working on my GPIB interface again and found that sending a device reset to one of the 3478a caused a significant drop in the voltage across the cap (like half). Like something in the meter seems to be presenting a relatively low impedance during the reset cycle. Having been through the manual, I can't quite figure out what might be happening, but don't want to risk damage to the meter so I've removed the cap for now. Anyone ever seen something like that?
Anyway, back on topic my plan was/is to just build a linear regulator based around a vaguely stable reference (~2.5-5v), chopper stabilized op-amp, some reasonably precision divider resistors and a PNP series pass transistor capable of delivering stable 300V & 30V outputs. I figured if I built a rudimentary oven around the ref/amp/divider it'd likely give me a more stable voltage to monitor/compare.
Bear in mind it doesn't have to supply much current and absolute accuracy or long term drift is not a parameter. Being able to stay stable to a couple of hundred uV over an hour would be really nice, but I might be making that sound easier that it really is. I won't really know until I try I suppose.
Is there a better approach? I've looked longingly at the Power Designs bench power supplies, but due to the weight postage is just too expensive to justify over here.