Repair : HP 3245A precision DC/AC source

[sdk.61]

Hp  3457A mult?meter ?

[Vgkid]

Why 34401? This beast does not even have ADC 

meant 3457, wrong decade.

[TiN]

Fun with ARB function :

[PTR_1275]

That is really awesome.

[dom0]

Now that's just arbitrary.

[Vgkid]

Impressive.

[crispy_tofu]

+1, fantastic! 

[TiN]

Both brothers assembled and ready for testing and initial DIY calibration 

[babysitter]

Gee your 3458A has that nasty display bug showing gibberish at the last digits! See! Those Frankennumbers.
I will gladfully do the waste disposal for you.

[deadlylover]

Grats mate, I don't dare to ask how much damage the 3458A repair did to your wallet. 

Next time you have a fiddle around, can you please check what the residual reading is with inputs shorted for the ACV range on the 3458A? 10mV/100mV range on analog mode, just ballpark readings will do, cheers.

[TiN]

ACV.. Hmm, that's interesting, I'll run noise script modified for ACV, just like we did on DCV before.
What AC mode should I use? Not familiar with them yet.

[TiN]

I think better late than never. Here's repair video and some testing. Hope you will not be bored.



Also took a closer look on source board... Guess what, DAC was replaced before. Somebody did not bother cleaning flux 



Little lumped inductor made me smile:



Don't you love it?

Voltage reference upgrade



Good bye, good old HP-marked LM399...



Added pair of coax wires instead. What was heater power now becomes LTZ module power. What was zener diode, now is Kelvin-connected LTZ1000ACH output (different module used, not the exact one from photo). Took photo with older module since it's the only one without insulation wrap around..



Removed resistors R400 and R401, as these not needed anymore. Unit is powered and hooked to DMM to log output. It's out of cal ofcourse, but no big deal, as we recalibration is easy with just 3458A.



Here's warmup graph, 40 minutes span. It's slow and ramping...

[Vgkid]

Impressive

[ManateeMafia]

Wait until you see the next project. Fluke 87V with LTZ1000! Don't ask about battery life.

[TheSteve]

Wait until you see the next project. Fluke 87V with LTZ1000! Don't ask about battery life.

The best 4.5 digits of your life!

[crispy_tofu]

  Can't wait!! 

[nidlaX]

Wait until you see the next project. Fluke 87V with LTZ1000! Don't ask about battery life.

The best 4.5 digits of your life!

Don't give Fluke the wrong idea. We don't need voltage standard wars between multimeter manufacturers like they did resolution wars for digital camera sensors.

[TiN]

Hmm, no good.



Perhaps I should hook more DMMs to log LTZ itself (7.1V) and +11.6V output REF, to see if it's output stage drifting, or issue with reference circuitry.

Also I'll separate heater drive power, as I suspect that one..

[TiN]

Aha.

Separated heater supply from opamp supply, now it's playing nice. Trap for young players, pay attention of proper signal and power current paths on your precision designs, or it can easily cause utter failure like in post above.

Also rigged some measurement, source running open cover with ref module hanging on side.
LTZ itself stays ~0.4ppm window over last hour (measured via 2002)
3245A output set to 10V on DMM (actual output setting is +9.71203V) stays within 0.35ppm (meas with 3458A) in last hour.
Ambient temps +28.3°C (no AC in room, to prevent pesky airflow).



For those who still feeling dizzy grasping all these PPM scales, here's same 1ppm division graph:



Day and night, eh?
Now time to pack unit back in with covers, run it few hours to get thermal equilibrium and perform calibration.
I tried calram backup tools from 3458A to capture 3245A calibration data, but both linux and windows verisons failed. Seems 3245A using different address for CAL data. Not big deal anyway.

P.S. horisontal time frame of latter graphs - ~1 hour.

[Dr. Frank]

I tried calram backup tools from 3458A to capture 3245A calibration data, but both linux and windows verisons failed. Seems 3245A using different address for CAL data. Not big deal anyway.

As the control PCB is exactly the same as in the 3458A, all memory chips must be on the same physical location!

If you look onto the 3245A board, it does not have the 2kx8 nvRAM, its socket is empty !

It only contains the two 32kx8 nvRAMs (and also the four 32kx8 SRAMs).

Therefore, the calibration data probably will also be inside these, but without this elaborate access-lock.

Another question is, if the 3245A firmware also contains the hidden MREAD command.


Frank

[TiN]

No, it's not empty, there is ribbon cable running to source board. And X2816BP U18 ROM + few RAM and logic on inguard section chips (not battery backed) there as well.
It's not socketed, and only DMM needed to do full calibration, so I did not bother socketing RAM.

I'll take few more photos with everything in place, to clarify this.

[TiN]

Few more photos, as promised

Top side:



Bottom side:



Modification on source PCB:



Remove R400, R401, LM399 reference.
Solder coax wire in place of LM399's diode location. This connected to Kelvin output (direct connection at LTZ's zener pins).
Solder coax wire to +15V and GND to power LTZ's opamps and control circuitry
Solder coax wire to +18.5V and power GND to provide separate current path for LTZ's heater, so it would not upset reference output.

Now, find a best spot for module location. I think it's best on bottom side, between front panel and HV Opt.002 board. It's far from mains transformer, far from pesky fan.
To have it isolated from airflow I added big piece of airlocked foam used in packing electronics.



It tied to metal chassis via dual-sided 3M tape, for better contact.



Now soldered piece of copper foil for sake of it. It's soldered to ground on power coax wire in single point. Idea to have it for even temperature distribution and bit of shielding. It's likely doing nothing at all, like audiophoolery, but hey, it's shiny and I had it without use anyway. Why not?



Piece of 4mm rubber pad is holding everything in place securely, screwed down to existing two metal stubs on chassis. Coax wires not touching anything and running at least 1cm from output wires/front panel.

Now, let's calibrate it.



Password is standard, but calibration have to be done by GPIB.



Temperature in room around +26°C. After entering calibration mode, it's just showing "Calibrating..." and output have voltage or current, which need to be measured by DMM and written back by command "CAL VALUE x.xxxxxxxx". After running all values from cal manual, it automatically saves CALRAM. Optionally you can leave a string with CALSTR "TEXT" later.

3458A was ACAL'd and initial calibration was just quick sanity check, to make sure I can get output 10.000V. I did not have too much effort on it to do.



It worked well, so I calibrated again, keeping temperature within 1°C span and measuring all at NPLC200



Result after calibration



Now I'll leave it overnight without AC, to see how good is tempco.

I checked 3458A+EDC MV106 earlier today, outputting 10VDC. That setup had +1.23 ppm/°C tempco.  I hope 3245A with LTZ + 3458A can be better.

[Dr. Frank]

Hi TiN,

on the main controller PCB, the address space for  the original DALLAS 2kx8 nvRAM is used for the steering of the analogue board.
That's what I meant, that the CALRAM is not installed..
Therefore the CAL dump program for the 3458A cannot work.

On the analogue board, I found the XICOR X2816, as you already described.
This is an EEPROM, 2kx8, therefore should not loose its content.

Congrats to the final repair and to this new, nice instrument

Frank

[Zucca]

Congrats TiN,

one day with my (was yours) LTZ module I will do something similar!

Now soldered piece of copper foil for sake of it. It's soldered to ground on power coax wire in single point. Idea to have it for even temperature distribution and bit of shielding. It's likely doing nothing at all, like audiophoolery, but hey, it's shiny and I had it without use anyway. Why not?

Mmm why not place there some silica packs?

https://www.eevblog.com/forum/projects/ultra-precision-reference-ltz1000/msg181096/#msg181096

Just because you mentioned audiophoolery....  well sometime if it costs nothing I am a proud audiophool! 

Take Care!

Z

[alanambrose]

OK I'll just add onto this thread if it's OK with you TiN...

Received my 3245A - end '92 vintage, same 2843 firmware, only ever one factory cal. This is the one with 2 x 10V output. Everything seemed OK for an hour and I was  doing some basic cal checks when snap, crackle, pop ... thunder and lightning etc ... it smoked, smelt and flashed a bit and then flipped the mains RCD - despite me turning it off pretty sharpish. I assumed it was the transformer but actually it turned out to be the xxs£$%ss:{@~}@xx^&*xx Schaffner mains filter. Since the filter is outboard of the fuse etc, it has no protection and it smokes 'til it goes pop and relies on a trip somewhere else to protect it - so it's a bit criminal to have a fault like this on it. This was even more galling as I already had a new filter and fan sitting on my desk waiting to be installed. A mess like used sticky engine oil - gory pictures for your amusement. Several hours later, IPA clean up / drilling out the damned rivets and it's operational again. Thinking about it, the switch to 220V from its previous many gears of running at 110V was probably the last straw. Now I'm a bit concerned what else of my HP gear has the same dodgy inlet.

BTW TiN - did you write some 3245 / 3458 cal software?

Alan