Keysights reaction on modified 3458a (lower ltz temp, FRAM,etc) at calibration?

[aronake]

Hi,

I am thinking of sending one of my 3458a for service (classic swap of A3 board), calibration and adjustment.

I currently have it running with 100K modification for lower LTZ1000 temp and use another A3 board.

Ideally I would want to keep the 100K modification, but could see a small risk that Keysight have some reaction to it.

Anyone with experience sending in a somehow modified 3458a for service, calibration or adjustment?

[aronake]

My own guess is that they would not react or notice 100K LTZ temp modification, if it isn't that the LTZ board is faulty. For other more visible modification like FRAM memories they may have some comments. And probably replaced with fresh Dallas memories.

[maat]

I had an HP3458A with the FRAM modification a few years back. It went to calibration a few times and they didn't mind back then. They also replaced the front panel switch at one time and did not complain. The DMM finally died with a

114,"SYSTEM ERROR -- balanced rundown convergence" error and a
202,"HARDWARE FAILURE -- "SLAVE TEST: CONVERGENCE" self test error,

the famous ADC drift bug.

At that time, the DMM was under extended warranty. This sudden death came about a few months after they had released the new black edition and they decided to gut the old device and replace all PCBs except the power supply and the reference during repair. So this DMM is now essentially a black edition inside. At any rate, they never complained about any modification I made to the device.

[aronake]

I had an HP3458A with the FRAM modification a few years back. It went to calibration a few times and they didn't mind back then. They also replaced the front panel switch at one time and did not complain. The DMM finally died with a

114,"SYSTEM ERROR -- balanced rundown convergence" error and a
202,"HARDWARE FAILURE -- "SLAVE TEST: CONVERGENCE" self test error,

the famous ADC drift bug.

At that time, the DMM was under extended warranty. This sudden death came about a few months after they had released the new black edition and they decided to gut the old device and replace all PCBs except the power supply and the reference during repair. So this DMM is now essentially a black edition inside. At any rate, they never complained about any modification I made to the device.

Thanks! Thats a pretty generous approach on all areas! Maybe they feel they need to compensate for being extremely ungenerous by stop selling bare A3 cards and charging 3500 USD for a repair...

Did you ever do the A3 drift test in the beginning of your ownership or any time later? If so how did it come out? My suspicion is that most or all 3458a that eventually get the "HARDWARE FAILURE -- "SLAVE TEST: CONVERGENCE" error are fualty and have A3 drift already from shipment.

[maat]

I guess you are talking about Service Note 18 (https://xdevs.com/doc/HP_Agilent_Keysight/3458A/service/3458A-18A.pdf). IMHO, this service note was probably written for people unable or unwilling to write up a quick script to continuously poll that CAL 72 constant and plot/fit it. Its meaningfulness is therefore limited. Using only two points to do the drift calculation needs to account for other environmental factors that affect the CAL 72 constant (like temperature). Therefore the 0.43 ppm/d figure given in the SN is extremely high. My 3458A drifted -0.027 ppm/d before being called to its creator 4 months later.

I recommend continuously collecting the CAL 72 constant along with the internal temperature of the device over a few days, then doing a linear regression against both temperature and time. If you see any noticeable drift over time in either direction and a fairly large tempco of the CAL constant your device is likely gonna need a new A3 board in the not too distant future.

I have attached my fits for both the tempco and the drift as an example. The plot shows the 10 V range. The 1 V range drifted in the opposite direction at +0.014 ppm/d. The drift plot also explains the necessity of the huge error margin HP factored in. The data fairy noisy. So giving a number like "worse than 0.01 ppm/d" would trigger a lot of false positives. To measure the temperature I stuffed a 10k thermistor into the case.

[iMo]

If I were KS I would easily accept the FRAM mod as it has nothing to do with the DMM performance, imho.
The 100k mod with the LTZ temperature is a different story, of course..

[DavidKo]

Why? They can use computer solution with CMOS and CR2032 battery or even with bigger battery (like in Keithley 23x). My first computer also have had this Dallas 1287 chip (battery, BIOS config and RTC) and today we do not see such a solution in computers, some even have CR2032 soldered to the mainboard to save the money. Keysight as the producer can even go further and use EEPROM for CAL values and volatile memory for the user values.

[EC8010]

Some even have CR2032 soldered to the mainboard to save the money.

Cheapskates. Not in the same league as 3458A, but my Keithley 2400 failed a few years after I bought it new due to failed CR2032 soldered to PCB (the populated PCBs had been sitting around for a while before assembly into finished instrument). Fortunately, I was able to desolder battery and fit a CR2032 battery holder on flying leads enabling future replacement in a few minutes without having to remove a PCB and disturb lots of connectors. I chastised Tek for penny-pinching that seriously annoys customers.

[Oldtestgear]

I modified my 3458A  to reduce the LTZ1000 temperature, replaced the NVRAMs, plus a few capacitors before returning to Keysight for calibration. They just calibrated & adjusted it before returning it with a 12 month warranty. The warranty was part of a special offer they did about 5(?) years ago.

I sent another one to them in 2011 that I had re-capped the PSU, fitted new NVRAMs in sockets, new RIFA caps, plus new mains filter.  No complaints this time either although they made it really difficult to deal with them by insisting that I had to prove I was not a consumer.  This, of course, after they had sold me a bunch of new instruments a while before & were completely willing to give me a substantial trade in deal on a load of junk.

The last time I sent my 3458A for calibration in November 2023, they did not adjust it even though it was outside the limits. My 34470A was within limits but was adjusted automatically for optimum performance.

[aronake]

I guess you are talking about Service Note 18 (https://xdevs.com/doc/HP_Agilent_Keysight/3458A/service/3458A-18A.pdf). IMHO, this service note was probably written for people unable or unwilling to write up a quick script to continuously poll that CAL 72 constant and plot/fit it. Its meaningfulness is therefore limited. Using only two points to do the drift calculation needs to account for other environmental factors that affect the CAL 72 constant (like temperature). Therefore the 0.43 ppm/d figure given in the SN is extremely high. My 3458A drifted -0.027 ppm/d before being called to its creator 4 months later.

I recommend continuously collecting the CAL 72 constant along with the internal temperature of the device over a few days, then doing a linear regression against both temperature and time. If you see any noticeable drift over time in either direction and a fairly large tempco of the CAL constant your device is likely gonna need a new A3 board in the not too distant future.

I have attached my fits for both the tempco and the drift as an example. The plot shows the 10 V range. The 1 V range drifted in the opposite direction at +0.014 ppm/d. The drift plot also explains the necessity of the huge error margin HP factored in. The data fairy noisy. So giving a number like "worse than 0.01 ppm/d" would trigger a lot of false positives. To measure the temperature I stuffed a 10k thermistor into the case.

This is interesting. My suspicion was that drifty A3 was bad for long time, and in many cases from original sale of the 3458a. But here is a case where the A3 drift suddenly started to happen after long time of use.

[aronake]

I guess you are talking about Service Note 18 (https://xdevs.com/doc/HP_Agilent_Keysight/3458A/service/3458A-18A.pdf). IMHO, this service note was probably written for people unable or unwilling to write up a quick script to continuously poll that CAL 72 constant and plot/fit it. Its meaningfulness is therefore limited. Using only two points to do the drift calculation needs to account for other environmental factors that affect the CAL 72 constant (like temperature). Therefore the 0.43 ppm/d figure given in the SN is extremely high. My 3458A drifted -0.027 ppm/d before being called to its creator 4 months later.

I recommend continuously collecting the CAL 72 constant along with the internal temperature of the device over a few days, then doing a linear regression against both temperature and time. If you see any noticeable drift over time in either direction and a fairly large tempco of the CAL constant your device is likely gonna need a new A3 board in the not too distant future.

I have attached my fits for both the tempco and the drift as an example. The plot shows the 10 V range. The 1 V range drifted in the opposite direction at +0.014 ppm/d. The drift plot also explains the necessity of the huge error margin HP factored in. The data fairy noisy. So giving a number like "worse than 0.01 ppm/d" would trigger a lot of false positives. To measure the temperature I stuffed a 10k thermistor into the case.

This is interesting. My suspicion was that drifty A3 was bad for long time, and in many cases from original sale of the 3458a. But here is a case where the A3 drift suddenly started to happen after long time of use.

Among my 5 3458a, 3 show evidence of having had A3 swapped quite early after first sale, but are now very good. One was drifty when i got it, and now have reached convergence error. One have original A3 and very stable.