Repair worklog : Old rusty HP 3458A

[Jacek_Paw]

Option 1 works with DS1230's after jumper JM600 relocation.
Another question, people who have real opt002, does it report OPT 0,2 for OPT? query? Yea, I know it's pointless, but still 

I do have 3458A with option 002 and no, after OPT? query it displays: "OPT0,0"

[TiN]

Repair of such gear is always a risky business. There is no proof that some complex fault still present, which cannot be detected by basic self-test diagnostics.
And just to prove my words, here's one. I left unit running , reading 1Meg ohms since afternoon, and when I got back just now, there is error: 114 : "SYSTEM ERROR -- multislope rundown convergence", and measurements halted.
If I press any button, it start working again and after few seconds get stuck with same error again. Any range or function.

Agilent had service note 07B for units till 2823A08000 (approximate) that says to replace A3 A/D board (and nothing else), but we don't go that road...

The delay time compensation on the A3 assembly (03458-66503 board) may cause an occasional
“Multislope Rundown Conversion” error. Design changes on this assembly have addressed this issue.
Because of the infrequent random nature of this error it is best to replace the A3 assembly and be
certain that the problem is corrected.

My unit S/N is past 08000, but who knows... I would like to know what parts are those "delay time compensation" if it's external logic..

[EEVblog]

I am the same, but then I think of all of Dave's projects on the long finger and feel better

I'm glad to be of assistance 

[TiN]

Cleaned board, will see tomorrow after it dry up.

Few mention this issue on web before - here and here, but noone posted solution 

[ManateeMafia]

@Dr Frank,

Thanks for pointing that out. Looks like smd and TO-92 versions are available.

@TiN,

Good luck, hopefully a spit and polish will clean it up. I know you have seen this, but it is worth linking to this thread. https://www.eevblog.com/forum/testgear/3458a-'error-209-harware-failure-timeout-unable-to-read-ad'/msg681957/#msg681957

At least someone had luck getting some discontinued parts for this meter.

[TiN]

Nop, cleaning magic did not happen. Still doing same thing.
It does not happening if readout below 0.004, but happen for anything higher.

Checked voltages (as well as tried little higher and lower voltage, with drive from SMU) and references -12, +12, +5, they all seem to be ok and stable.
Hotair therapy with hairdryer did not change things either.

Perhaps change comparators, need find replacement part for those EL2018's..

[Macbeth]

The hairdryer trick - you did also use freeze spray too? 

[TiN]

No, i dont have that.
Hm, i old cal NVRAM in, just to try, same result.

[Macbeth]

Worth getting a can of freeze spray. When a fault shows up spray the suspects until it goes away. Hairdryer to trip the fault again (It's otherwise used to bring up a fault condition that normally doesn't show for a long time). I've fixed countless pieces of equipment with that and the "percussive maintenance" methods of fault finding intermittent stuff. 

[TiN]

Well, it shows up within seconds, so it's not like come and go.
Also I was wrong about all functions, Freq, period and AC voltage, AC current have no problem. Perhaps they not using A/D in same way?

Decided to take a break and do something else first, or I would kill something..



Will have 3D STEP files so folks with 3D printer could print some caps for LTZ ref's .

[Macbeth]

But it's otherwise working for a few seconds then the fault shows up and stays? That is the number one fault mode calling for freeze spray. The hairdryer just makes the fault worse, as most semiconductor failures seem to show up when they warm up after power on. I was a bench service engineer a couple of decades back and countless faults were isolated with a can of "Freeze-it". They didn't teach me that in college so as an apprentice I would laboriously go through schematics fault finding the proper way until the old guys taught me the tricks of the trade.

Honestly just get a tin. I was just watching Daves HP1740A repair vid with loads of fault finding all the while thinking "just get the bloody freeze spray!". Sure enough he eventually tracked it to a bridge rectifier that the freeze trick would have picked up instantly. In that case the hairdryer would be used to trigger the fault first as waiting for it is boring. In your case the fault shows quickly?

[Zucca]

Infrared Camera?

[TiN]

Caps done STEP and PDF. I'll have also Solidworks source in article.

top and bottom caps drawing with PDF

STEP model, bottom cap
STEP model, top cap

[lukier]

Great restoration. Your story was like a good drama

Sometimes I think I should trade my SR850 for 3458A for the sake of having 3458A , but then I realize that is not everyday multimeter, I would rarely use its full capabilities and all the other time I would curse  the front panel UI on this thing

Now, if you are seriously volt-nut, you need to carefully choose the material for your 3D print  PLA will absorb moisture, deform due to heat and biodegrade. ABS might be better, but I suppose a real volt-nut would like LTZ1000 cover made of teflon or Valox like LM399 housing

[ManateeMafia]

Great job on the covers.

If you change the nplc settings and NDIG to something low like .01 and 4, do you still see the error? AC rundown might be more relaxed or it is input through a separate part of the A3 board for AC. Will have to look at the clip again.

[TiN]

Polypropylene would work well too, I'd believe. But since I got original caps from fellow member here, I will not go there. Heck, I don't even have 3D printer anyway

Issue is not present on any of AC functions (ACV/ACDCV/ACI both fine, any range), as well as freq/period. Also if I set NPLC let's say 1 or 100 on AC functions, it seem not changing, and speed is still same. Is there any other settings need to be done for AC NPLC to work?

I'll order some comparators and perhaps make that agilent-bodge from your picture in other thread.
Seems finding modern comparator with supply >18VDC and <40ns is a real next to impossible, according to my search on Linear,Digikey,MAXIM,TI :/

If I use 0.001 NPLC or faster, no error. If use 0.01 or slower, errors. Let me make a video.
I think it's because HP's ADC using only integration runup method to resolve on those speeds, and rundown used only for more precise measurements.

[Dr. Frank]

Issue is not present on any of AC functions (ACV/ACDCV/ACI both fine, any range), as well as freq/period. Also if I set NPLC let's say 1 or 100 on AC functions, it seem not changing, and speed is still same. Is there any other settings need to be done for AC NPLC to work?

I'll order some comparators and perhaps make that agilent-bodge from your picture in other thread.
Seems finding modern comparator with supply >18VDC and <40ns is a real next to impossible, according to my search on Linear,Digikey,MAXIM,TI :/

If I use 0.001 NPLC or faster, no error. If use 0.01 or slower, errors. Let me make a video.
I think it's because HP's ADC using only integration runup method to resolve on those speeds, and rundown used only for more precise measurements.

Hi TiN,

NPLC works differently for ACV, and it also depends on the ACV mode (3 different!) used. Resolution is limited to 6.5 digits, in any case.
See manual, command reference for NPLC, p. 6-67, Remarks.

You may choose the different modes for ACV and see, if that changes something.
ACV also uses a different A/D input path (AC_AD vs. DC_AD, see CLIP).

The conversion anyhow, should or may be using the same algorithm, so maybe the problem is not with the comparators.

For faster conversions than 100us, there is a difference in the conversion algorithm, see hp journal 4/1989, p.13.
Maybe you fiddle around with the aperture parameter, and determine, when the error occurs.

Do you use ACAL frequently?
This is the most important function of this instrument, it will help, if you engage that function after your recent calibration, or after big temperature changes.

Frank

[TiN]

Thanks, I'll look into all that.
It cannot run any ACAL or selftest now, as soon as that error happens, it fails operation until i press any key on front panel. I did CAL and ACAL before when everything was working fine, and everything was booting and working without a single error or message. And ACAL is kinda pointless, as there are no covers/shields on unit anyway.

Ordered EL2018's from UTSource, as well as making order for all rest parts on DK. I ordered few ISL55143's comparators as well, can make adapter board to bodge them it. And yes, they are obsolete too, but so far available on DK.

I'll test with different mains frequency too, just too see if it would be less frequent due slower conversion speeds.

Resolution is limited to 6.5 digits, in any case.

Huh? Unexpected
K2002 does show all 8.5 digits. If they are meaningful, that's another story, but it does show them 

[ManateeMafia]

Without trying to get anyone turned around I was thinking about U150 on the A3 board. It is listed as "Minor Slope Drivers". I have not tested it yet, but would it be possible that the minor slopes are the last few slope run downs the converter performs and that possibly the DC ranges are using it but not the AC?

I will stick a LA on the four inverter inputs and check the different ranges and modes. Hopefully, someone has been there already, but if not, I will collect some data and look for a pattern.

[TiN]

That one is ok, checked it already.

Few captures on integrator's comparator.

Left to right, NPLC 0.0001, NPLC 0.001 and NPLC 1



Need a scope to see anything, LA not suitable for such debug...

[Dr. Frank]

Resolution is limited to 6.5 digits, in any case.

Huh? Unexpected
K2002 does show all 8.5 digits. If they are meaningful, that's another story, but it does show them 

On the internal display, hp adapted the useful number of digits for each mode.

That is:
DCV: 8 1/2, DCI: 7 1/2, ACV, ACI: 6 1/2, freq/per: 7

Criterion is stability of references and of range components, noise.

Over the GPIB, up to 9 1/2 digits will be transferred.

Other 8 1/2 digits instruments like Fluke 8508A always display 8 1/2 digits, but even the specification is not appropriate for that.

Frank

[TiN]

I understand the meaning and difference behind resolution and accuracy, just interesting how different products aproached this. Low ranges on DCV also provide less digits on screen.

[BFX]

TiN: excellent work
Thank you

What logic analyzer you using?

[TiN]

I listed all used gear in article. And it's TLA714 + TLA7AA4 with P6810 probe.

[TiN]

Schematics draft WIP of adapter board for EL2252CM to pop on A3.