HP 3478 lost calibration

[cncjerry]

I have three 3478a meters that I plan to sell and was checking the calibration of one against my 3457a using an Agilent USB GPIB adapter.  I was plugging the adapter between the 3457a and 3478a as I ran a series of resistance checks using an ESI deca box.  Everything was coming out near perfect, many times they matched to the lower resolution of the 3478a.  Then all of a sudden the 3478a meter calibration went off the tracks.  I thought it was the cable at first, checked it all, then turned it off and on.  At that point the meter went 'uncalibrated'.  I had the calibration active at the time I was checking the meter.  Thinking back, I know it isn't a good idea to swap the GPIB cable while the meter is on but I've done it a lot more than not. 

I checked the battery and it is still at 2.95V though I realize it doesn't have much of a load on it.  My current plan is to swap the battery and recalibrate the meter to see if it holds calibration.  I've had this one a year, it is in great shape, never had a problem, etc.  Wondering what happened? USB cable static?  Battery?  Failing NVRAM (assuming it has one or equivalent).  Should I swap the NVRAM (if it has one) before calibration again?  I haven't dug out the sheets on it yet to see how the calibration is stored.  I would hate to sell it to someone and have it happen to them down the road.  What other preventative maintenance can I do to it to limit the exposure of it dumping calibration again?

Thanks

[MarkL]

I don't see anything you did wrong that would have caused the loss of calibration data.  Are you getting a "CAL RAM FAIL"?  Or just bogus readings?

The NVRAM is probably fine.  Make sure it's being powered properly.  With the meter on, measure pin 22 on U512.  That's the NVRAM Vcc.  It should be 4.4V or more.  Keeping the probe on pin 22 turn the meter off.  It should drop to around 2.5V or more.  2.0V is the minimum required for retention.

Always leave the CAL switch off unless you're doing a calibration.  It disables the gate that's connected to the NVRAM R/W pin.  The processor can't write to the NVRAM even if it went crazy.

If you have a signature analyzer, you can perform an NVRAM test as described in the service manual section 7-D-23.  The first part is a read test and the second is a destructive read/write test.

[cncjerry]

MarkL,  I have a sig meter and will dig out the pages.  Something happened in that process.  I measured the voltage with the Meter off and it was 2.95.  I'm not getting CAL RAM FAIL, just uncalibrated.  I only had it in cal mode incase I had to make an adjustment and going forward I will turn the switch, calibrate, then turn it back.  Sort of sucks because this meter was really performing.  I have three of them and since I picked up two 3457s I thought I would just test them and sell out.  Now I have to go thru all the calibration, which isn't a big deal, just wasn't expected.

[MarkL]

I wouldn't jump into a calibration unless you know the constants are really gone.

The calibration constants are checksum protected, and if they're corrupt you would see "CAL RAM FAIL".  Unless the unit somehow wrote some bogus calibration constants, which seems unlikely, I think you're problem lies elsewhere.

A common issue on these meters is the front/back switch.

Are all the ranges and functions out of cal?

[cncjerry]

The calibration looks fine but the meter reports Uncalibrated when I turn it on.  From what I remember you have to run thru and calibrate all the ranges to get rid of that error.  Right now the plan is to replace the battery, recalibrate it and see what happens.

[MarkL]

Found it in section 4-56(a).  You're right - the meter reports UNCALIBRATED if the checksum is bad.

I'm not sure then exactly what causes the CAL RAM FAIL.  Perhaps there's some sanity data that's also being checked.  The manual is not very specific.

At any rate, good luck with the re-cal.  I would still perform the NVRAM check if you have a signature analyzer so you're not wasting your time.