HP 3458A A/D timeouts on ACV test.

[Jwalling]

This 3458A came to me from a customer for repair. I don't normally work on these; in fact this is just the second one I've worked on (The other was just an NVRAM replacement job.)

It had two symptoms

1. On ACV and ACI, the readings were way out with the inputs shorted. On ACV, the meter would read 3955VAC. It was not pegged, though. The lesser digits would go up and down a bit.

2. When I ran the full self test, and it got to the ACV part of the test, it would error out with "209 HARDWARE FAILURE -- TIMEOUT: UNABLE TO READ A/D: 198"

My hope was that the two problems were actually related - fix the pegged ACV problem and it would be repaired and could be calibrated.

The unit had a defective Elantec EL2039 on the AC board which I replaced with a part from Keysight. According to the datasheet, it is a "very High Slew Rate Wideband Operational Amplifier" HP part # 1826-1859 OR 5063-1448.

Now ACV is no longer pegged. It has about 800uV or so offset with the inputs shorted. Not unexpected, as the unit will now need a formal calibration.

However, the full self test still fails when it gets to the ACV part of the test.
I can manually go through all the ranges and I never see any timeout errors. It keeps sampling at about once per second or so. It reads about 1VAC with 1VAC applied to the inputs.

I guess the question is: Could the large offset from replacing the chip cause this failure? Would a formal calibration "fix" it?

[coromonadalix]

If you could find an good / stable enough 10vdc reference you could try a calibration to see if you can get rid of this offset ?   since you say it will eventually need a cal later

[Kleinstein]

There are different AC modes: one with the old analog classic RMS chip and one with the new fast digital conversion and numerical RMS. Those two modes are quite different and it may be only one is failing, e.g. the ADC at very high speed.

The ADC timeout would be more like a separate problem, more with the  A3 board, maybe the fiber optics. So the first would be back to basics.

It could also be a good idea to do a test on the possible ADC drift problem. So check for drift in ADC-gain  (CAL72) - at some point his would cause ADC problems, possibly also too long a time.

[Jwalling]

There are different AC modes: one with the old analog classic RMS chip and one with the new fast digital conversion and numerical RMS. Those two modes are quite different and it may be only one is failing, e.g. the ADC at very high speed.

The ADC timeout would be more like a separate problem, more with the  A3 board, maybe the fiber optics. So the first would be back to basics.

It could also be a good idea to do a test on the possible ADC drift problem. So check for drift in ADC-gain  (CAL72) - at some point his would cause ADC problems, possibly also too long a time.

Thanks.

When I SETACV SYNC the display stops updating.
Then I SETACV ANA and the display starts updating again.

Am I on the right track, that this is the problem manifesting itself, or am I sounding completely ignorant?

Not sure at all what CAL72 is, I couldn't find it in the manual. Very much a newbie with this DMM.

[Kleinstein]

There is a service note (AFAIR SN18) about drifting ADCs.  Some of the ADC boards seem to be failing in the mode describe. The essence is to repeat ACAL and not the CAL72 value over a few days  maybe a week of so at a constant temperature. If the values drifts too  (AFAIR some 0.4 ppm/day) much this would indicate an failure of the special ASIC and need to replace the A3 board.  So it before doing much other test on the A3 board it may be a good idea to do (at least start) this test. There is a small chance the failure could also explain the problem / not working fast mode, though there are normally other faults first.

AFAIK one could also test a super fast conversion (e.g. 10 µs integration) in "DC" mode. This would test much of the same part as the AC mode, but not the trigger part.

Is the analog AC mode giving a correct reading ? This part should share much of the analog front end with the digital AC.

Without looking at the instructions / diagrams: not working in sync mode could be a problem with the line synchronization or trigger. This could be one of the more simpler faults.

[Jwalling]

There is a service note (AFAIR SN18) about drifting ADCs.  Some of the ADC boards seem to be failing in the mode describe. The essence is to repeat ACAL and not the CAL72 value over a few days  maybe a week of so at a constant temperature. If the values drifts too  (AFAIR some 0.4 ppm/day) much this would indicate an failure of the special ASIC and need to replace the A3 board.  So it before doing much other test on the A3 board it may be a good idea to do (at least start) this test. There is a small chance the failure could also explain the problem / not working fast mode, though there are normally other faults first.

AFAIK one could also test a super fast conversion (e.g. 10 µs integration) in "DC" mode. This would test much of the same part as the AC mode, but not the trigger part.

Is the analog AC mode giving a correct reading ? This part should share much of the analog front end with the digital AC.

Without looking at the instructions / diagrams: not working in sync mode could be a problem with the line synchronization or trigger. This could be one of the more simpler faults.

I could try the cal72 test, however, I am not working in a temp controlled environment, but a drafty old house built in 1850. Given this, would I be wasting time?
The serial range in that document indicates a serial range from  US28031400 to US28032927. The serial # on this one is 2823A06296. No "US" (but perhaps that's presumed?), and seemingly outside that number range.

If I set NPLC 0 then it works fine; no timeouts.

The analog AC mode seems OK, but is off somewhat from having replaced the wide band opamp which was defective.
ACAL ALL works fine till it reaches ACAL ACV RATIO 1 after about 11 minutes into the calibration when the timeout occurs.
https://www.youtube.com/watch?v=kTIMMwZSqpg&feature=youtu.be

If I run ACAL AC, it fails instantly.

[Kleinstein]

The ADC drift for the SN18 test is easier to measure if the temperature is stable. With changing temperature it would take a few more days to decide if variations in CAL72 are due to temperature changes or if there is a strong drifting background. I would not give too much importance to the serial numbers effected. It does not really hurt writing down the CAL72 numbers and check for a few days, just to confirm the ADC is OK and not drifting too much.

If fast (so that the AC path and 10 K resistor to the ADC is used) DC conversion works (not sure what setting PLC 0 gives), may suspicion would be with the trigger or line sync part (e.g. U581 and surrounding parts)

[TiN]

"UNABLE TO READ A/D: 198" means meter does not see value expected for calibration constant 198, which is 198 flatness DAC 10V. This directs you to A2 board issue on 10V range, before you start any witch-hunt on A3 ADC Check the other Elantec part on A2, it could be toasted too.

Normal external calibration using 10V/10kOhm standards differ from ACAL procedure only by additional steps to acquire actual value of internal 7V/40K references. Then those references are used to calibrate/adjust all of the other ranges/internal functions, including AC. Only exception is high-frequency flatness past 1 MHz, for which you use external TVC setup.

[Jwalling]

If fast (so that the AC path and 10 K resistor to the ADC is used) DC conversion works (not sure what setting PLC 0 gives), may suspicion would be with the trigger or line sync part (e.g. U581 and surrounding parts)

NPLC 0 sets the DMM to 4.5 digits and a much faster conversion. I was trying to do faster conversions that you mentioned when you said

AFAIK one could also test a super fast conversion (e.g. 10 µs integration) in "DC" mode.

[Jwalling]

"UNABLE TO READ A/D: 198" means meter does not see value expected for calibration constant 198, which is 198 flatness DAC 10V. This directs you to A2 board issue on 10V range, before you start any witch-hunt on A3 ADC Check the other Elantec part on A2, it could be toasted too.

The only other Elantec part is U501 which generates LEVEL TRIG. When I set the DMM for analog AC, it's a 60Hz square wave with 4 Volts amplitude derived from the line input. When I program the DMM with the command SETACV SYNC (and the display stops updating) it is 1KHz which is the input I'm applying from a Keithley 2015.

Maybe I need to send more than SETACV SYNC? Can someone try this GPIB command on their DMM?

[Jwalling]

So after giving up on this more than once, I spent a bit of time probing around the A/D board. The Elantec chip marked in red did not appear to be working correctly.
I also noted some ugly rework on the chip highlighted in blue.

I PM'd TiN to see if he ever saw a rework like this - he did not think this was a factory mod. (Thanks TiN!!!)

The Elantec chip is EOL and Keysight won't sell it, but I noted that there was a REV.C board on TiN's website that showed a replacement board that Agilent made for these two chips.
(No way was I going to source a chip from a Chinese seller on Ebay)
They are only $55 each so I ordered one and replaced both chips with the module. Note that this board (# 03458-66530) is "now available till gone", so there may not be many left...
Now fast conversions in AC mode is working, and the full self tests are passing! 

No doubt that this will need a formal calibration.

[TiN]

You see old Elantec, you shoot on sight. Ask questions later

[Jwalling]

You see old Elantec, you shoot on sight. Ask questions later

I've not seen much of their stuff in anything. Is that true for just the 3458A or anything?

[TiN]

Common for 3458, as these need to run hot and fast there. Multiple old 3458A's have issues due to failed Elantec's on either A2 or A3.
Agilent patched newer A3 with board that has dual elantec (which is available only in SMD package) like the pic above shows. It's hard to find ultra-fast comparator that would be direct replacement while +/-15V rails. I have even designed the patch before, but never ordered gerbers, since had to replace whole ADC anyway.