I've been noticing a trend while calibrating 3458A's lately. They are failing the zero tests and/or the 100mV ranges. If you exercise the front/rear switch several times it will change the reading. We ordered a bunch of these 03458-66510 boards and started replacing them with good results. Has anyone else run into this before?
Yes. Two primary culprits. The first one is the molex connections. If you are lucky, that ends up being the problem. In some instruments with front/rear switch issues, I find loose connections to the many molex pins. I gently tighten and re-seat, and that fixes the problem. When that doesn't fix it, unfortunately it ends up being a new switch needed. I have tried unsuccessfully cleaning the switch. They are such low thermal modules that I don't very much trust cleaning or exercising.
The quickie test I do is with the copper wire short attached to both front and back, I test 2W and 4W ohms by switching back and forth. I monitor the stability and repeatability of both of those. If either is flaky (even if still in tolerance), or if the value significantly changes between iterations of front, rear, I consider the switch to be bad.
I have a very weird 3458A issue I might as well bring up here in case someone else has seen it.
I had a 3458A connected with very good low thermal cables to my SR104 10 kOhm resistance standard (in 4W). I ran AUTOCAL ALL, and got about a +30 ppm error, and very stable and consistent. Let it sit over night, ran AUTOCAL ALL again, and this time got about a -30 ppm error, again, stable and consistent.
I even exercised the front/rear switch, and no changes. I ended up taking it out of service for the moment, as I don't have time for it right now (doing my annual 10 VDC hot cal with standard from Fluke this week - one of my favorites).
Any way, just in case anyone has seen the weird 3458A symptom...
Would not worry about it, this is normal on 3458's (HP and Agilent....Keysight I don't know), normal is more than double this.