Repair worklog : Old rusty HP 3458A

[Dr. Frank]

The other question is where the 5 V is actually needed for.
 My guess is, it might be used for Ohms  - but the diagram will tell us. Depending on this the 5 V may not even be critical at all.

The +5V reference provides the 'minor slope' currents +S1, -S4, +S16 and -S64, i.e. the residual A/D conversions. See hpj 4/89, page 12.
These are created a bit differently in practice in the 3458A, i.e. by the network R150-153, the 37k5 and 16k7 resistors inside the ASIC and R185 (3M)
 

The OHM circuitry creates its own +10V reference directly from the LTZ1000A output, by U400A, RP400 and then to 13.0V, 10.3V by U305 and RP300A/B/C/R310.

Frank

[Kleinstein]

Just for the minor slopes, the 5 V is less (e.g. by a factor of 100 or more) critical than the +-12 V. So these two resistors don't have to be the absolute best.

I am not sure the VPG series is the right choice as they are made for long term stability - something that is not needed here. The only critical parameter is TC matching.

[TiN]

-10ppm over 16 hours, -0.625ppm/hour

+12VREF,-12VREF stay around 1ppm stable. Perhaps time for 50K ASIC replacement?



I'm lucky to have few 50K mica fluke wirewounds suitable to task, I believe.

Also "jumps" are truly seem to come from LTZ (pink plot on graph, ZR_HI2-ZR_LO2 measurement). Perhaps ebay's 3458a A9 refs aren't all that good after all?

[Vgkid]

Couldn't you just use a single 50k resistor instead of the (4) 50k resistors in series parallel annangement. Unless the 4 resistors are for a course adjust(currently i'm on my phone, clips is on my computer).

[TiN]

How many Fluke resistors should I put into 3458 to have it beat my K2001? 



+1 more today (50K instead of ASIC network).
Issue with all these external resistors is thermal constant is getting very large. Meaning that meter takes lot of time to get into stable state when all resistors get their working temperature.
Little internal ASIC network get hot really quick, but these big-ass 3W wirewounds are slow as snails. I'm still seeing 2-3ppm/minute warming up after 30 minutes.
Such DMMs need multiple hours of warmup time, so not a big deal, but definately making use/transport less "easy".

[TiN]

Removed all wires to scan card (no monitoring), replaced integrating 330pF cap with new polypropylene one, shorted power supply for -12V opamp.
Nope, still doing it's pesky drift (this time up, +3.88ppm/hr).

Let's go extreme, make own resistor, as all those I have now are all over the place in terms of tempco or tempco matching (Fluke's 5ppm/K, Vishay 0.3ppm/K, RNC 25ppm/K), which is no good, given what HPJ says about resistor network:



I took one of my bad Fluke 156448 250K 0.05% 3ppm/K and started adding taps. It's not a job for faint-hearted shaky-handy people, I can tell you that.  And forget of doing so without microscope. Resistor wire (manganin?) is not solderable, so I'm using single strand from copper AWG30 cable to hop around resistor body and fix it under tension with blobs of solder in strategic spots.
Trimming done by adjusting wire position on the edge, where it touches resistive wire. I hope this will make same tempco or at least close to that, as it is same resistive wire, on same mica holder.

If that works (which I don't quite believe, it's really tricky and touchy), I'll make hermetic nitrogen-filled can for it, to make it useful.

So far resistor sections archieved:

10.162K, 24.860K, 10.451K, 10.902K, 6.087K, 4.187K. I still have about half of resistor body left, so I'll make pair of 50K one on it as well.

[Vgkid]

Looking forward to this impressive work. Just thinking about modifying those resistors 
Still mad at myself for not bidding on those resistors...

[Macbeth]

Wow! I was just reading Edwin Pettis' history of wirewounds and the nightmare of trying to weld wires to them. Interesting to see that the Minuteman and Trident nukes, not to forget the Apollo spacecraft all had dodgy resistor leads 

Good luck with those wire taps TiN! 

[TiN]

Test number "I lost the count already"



I failed with 10K -12VREF network, but found my VPG Z202's 10K's instead  80.6K PTF is just 25ppm to bring 5VREF back to ~5.3V (with 22K/10.14K it's 5.8V, bit high to my like).

With this contraption meter can't calibrate to 10V (10V applied reads 9.58V, I cal'd it to 9.75V readout, it failed, but constants get updated to get 9.75ish).
Value does not matter, all I need is to see how drift goes. If it's better by some miracle, then I can try more effort.

[TheSteve]

Is the dream dead or alive?

[TiN]

Dead-live. Did not do anything much since last week, just put in 10k+10k VPG resistor network on negative -12v reference, it still drifts.

[TiN]

I thought worth to give a little update:

Our member, ManateeMafia was kind to hook up his 3458A for stability/datalog test, so we spent some time to get it all working in realtime via Raspberry Pi GPIB + FTP + D3.js.
Data already being accumulated for over a week for first meter (red graph), a week for second (green graph) and bit over a day for third one (blue graph).

All three measuring same Fluke 732A 10V, and both ppm/dev and HP's temperature being logged.



As of my meter, it's getting downhill. I did not do anything for it over last 2 weeks, just sitting running 24/7 on shelf, and it began to throw ERR114 - multislope rundown convergence errors now (once every minute or two). Busy with other nuttery projects for now, will get back to it next year, when I get more time to waste. 

[TiN]

Alright , transplanted ASIC back to it's original second A3 PCB, started 10V 3245A test once again.
Also replaced AD848 and used new film 330pF C119 cap.

Good thing - AC issues are fixed now, it was A3 failing for SELFTEST ACDCV 10mV/100mV/1V.
Selftest pass now, CALd to 10V.

Will the evil machine drift? I'm ready to bet bottom dollar that it will....

[TiN]

I wonder who got this one... (that's not me, I promise!   )
eBay auction: #381500231334

[nidlaX]

I wonder who got this one... (that's not me, I promise!   )
eBay auction: #381500231334

I saw three "cheapos" come and go from express_test, but they are described as "incomplete" with no pictures of the actual units!

[TheSteve]

I wonder who got this one... (that's not me, I promise!   )
eBay auction: #381500231334

I saw that one and was tempted, then I recalled this thread and came to my senses.

[TiN]

Drifting bonanza continues...

Here's what done so far these days:

* Removed drifty U180 ASIC from original bodged up A3 PCB.
* Cleaned second A3 PCB with ASIC, replaced integrating cap to DK 399-7719-ND, installed new AD848
* Cleaned ASIC between legs with cotton swap and IPA
* Soldered in ASIC to A3

Tested stuff, ACDCV issues are gone now, meter pass selftest and calibrate okay.
1ppm jumps in readout were still there. As we tested this before, it was reference voltage jumping, so eyes to A9 PCBA (which I got from eBay, as some may remember).

* Added copper-foil hat around A9, thinking stray air-currents around A9 cause jumps.



* Proven useless, copper hats removed.
* Removed LTZ1000ACH
* Installed LTZ1000CH (1990) instead.
* Removed 95K resistor to get temperature setpoint back to +80°C

Since we talking sub-ppm now, I changed source at meters from HP 3245A 10V output to one of my LTZ1000-references 7V output. Both K2001 and 3458 measuring same output. LTZ powered from K2400 12V ~30mA.



Now, no ppm jumps, only drift, at much slower rate ~0.2ppm/hour.

Funny part - LTZ1000CH non-A warms up much slower than original -A, this is visible by having 10V on meter and power it on from cold. Voltage creeps 3-4mV/second till it reference voltage stabilize after ~2 minutes.
If one try to run selftest or CAL 10V during this period - you will be greeted by HARDWARE FAULT 204: REFERENCE CONTROL LOOP FAILURE or something like this Once it warmed up, selftest and CAL works ok.

Could it be just leakage on ceramic surface of ASIC causing drift, all this time? 

[Kleinstein]

Leakage on the ceramics is a possibliliy. As the level of leakage is likely not stable this my look like bad resistors and cause quite some drift. However it takes quite some leakage to disturb the 10K/10K voltage inverter.

Even when resonable clean, it may need a little for the ASIC (together with board) to warm up - In the warm state surface water layers evaporate and reduce leakage. It might a good idea to have critical parts run slightly warm, e.g. 10 degrees above normal ambinet, just to reduce leakage. So it might not be the LTZ1000 that needs the longer warmup, but the ASIC.  In the initial phase there can also be larger thermal gradients on the board. I would not trust the ACAL very much thats done so fast after turn on.

For the ceramics is might be a good idea to wash it with clean water too - some of the dirt (e.g. salts) might be better soluble on water than IPA.

[acbern]

Or the non-A LTZ you soldered in is drifty!?

[TiN]

However it takes quite some leakage to disturb the 10K/10K voltage inverter.

Even with inverter voltages stable (via external resistors) it was still drifting about same rate or worse. There are still other both analog and digital signals around ASIC involved.

So it might not be the LTZ1000 that needs the longer warmup, but the ASIC.  In the initial phase there can also be larger thermal gradients on the board.

Did day worth test before with original A9, did not see this warmup effects so long before.

I would not trust the ACAL very much thats done so fast after turn on.

It was just sanity test, to check if other issues raised in process, not by any means meaningful calibrations.

For the ceramics is might be a good idea to wash it with clean water too - some of the dirt (e.g. salts) might be better soluble on water than IPA.

Will try it. I noticed purple residue (same color of ceramic) on swab after cleaning. I thought that should not be there.

Or the non-A LTZ you soldered in is drifty!?

Let me check it.

[TiN]

Hooked my LTZ module output to reference input on 3458A instead, so we can see both results.



K2001 reading this reference voltage
3458 reading 3245's 10V output

I need more meters around 

[branadic]

Could it be just leakage on ceramic surface of ASIC causing drift, all this time?

Not necessarily leakage. I was told by a representative from IST AG, that ceramic is also somewhat conducting and with increasing temperature the conductivity rises, thus limiting the maximum temperature for platinum sensors. For higher temperature application an additional insulating layer is required.

[plesa]

Could it be just leakage on ceramic surface of ASIC causing drift, all this time?

Not necessarily leakage. I was told by a representative from IST AG, that ceramic is also somewhat conducting and with increasing temperature the conductivity rises, thus limiting the maximum temperature for platinum sensors. For higher temperature application an additional insulating layer is required.

Ceramic is conduction at much higher temperatures than temperatures below 400K.

[branadic]

Ceramic is conduction at much higher temperatures than temperatures below 400K.

No, there is conductivity right below 400K, but it becomes more critical as temperature increases.

[TiN]

Hm, it drift down over today logging, at same rate ~0.2ppm/hr. Is it because meter was upside down (as i cleaned A3 again) and electrons fall off LTZ1000 chip?

Turned it back with face up, lol.