HP34401A start-up problem - replace one electrolytic or all four?

[Lorenzo_1]

My 34401A came up with the start-up issue described in earlier post "HP/Agilent 34401A DMM startup problem" - display self-tests, then one beep then blanks out. Only Shift key works toggling display segment on/off.  After checking the numerous posts on this, I checked the power supply caps and find one secondary electrolytic 25V 2200uF has failed.  The three others - 25V 4500uF & 2 x 50V 1000uF all test OK.  Interestingly, the faulty one has red marker pen dot on top - perhaps it was dicey when this unit was 'refurbished'.  Still it lasted a few years so no complaints.  Question is do I leave well alone and just replace the bad cap or should I replace all four while I'm in there.  I lean towards just doing the one rather than mess about with the PCB unnecessarily.  Advice appreciated.

[coromonadalix]

normally  i recapped all the psu sections in the dmm

you remove the big caps   and remove the tantalums  too   with hot air  and protect the other parts with kapton tape

and you should be okay,   you need  good experience with this kind of job, you could damage the pcb

[Lorenzo_1]

Thanks. I think I can manage it with care. Will look at the tants as well. Why do u recommend hot air? I have a good Hakko air gun now but wary about local overheating. Does soldering iron risk lifting pads? If I use hot air should I preheat whole board? I have an iPad heating pad.

[Miti]

If one electrolytic was bad, chances are the others aren't too far behind. I wouldn't touch the tantalum though.

[Lorenzo_1]

Thanks. Took anor look last night and decided to replace 4 electrolytics and leave tants alone unless they fail. I need more experience with surface mount soldering. Have done some but not enough - better to try on lower grade boards. This is hobby stuff for me.  Mostly work on keeping older gear working rather than send it to the dumpster.

[factory]

If one electrolytic was bad, chances are the others aren't too far behind. I wouldn't touch the tantalum though.

I've got an older 34401A where the all the AC & frequency ranges stopped working, cause was one of the tantalum capacitors on the +/15V supplies, it took a Zener with it, thankfully this was found in time, as they can burn up the PCB if not noticed, I haven't repaired it yet due to lack of surface mount repair equipment here, we did have a nice rework station at work but it disappeared.  Don't want cheap soldering kit, been there done that, got fed up of having to repair the cheap soldering kit constantly.

Much better write up & pictures here from someone else, who had one fail after another. https://www.edn.com/repair-hp-34401a-capacitor-failure/

David

[Lorenzo_1]

That's sobering... I think I'll replace the electrolytics, confirm it comes back to life (hopefully) and then decide on preemptive replacement of tants.

[davidwenrich]

I would recommend hot tweezers over hot air for the tantalum replacement.

[robert.rozee]

for me, two soldering irons also work well, one in each hand. it may sound daunting at first, but it is no different to using a knife and fork to eat dinner!


cheers,
rob   :-)

[factory]

I've no idea how common tant failure is in the 34401A, usually people recap without mentioning if they had obvious problems.
If you do replace the tants, it's usually good to uprate the voltage ratings for the new caps, to prevent repeat failures.

David

[HighVoltage]

I never had a bad electrolytic capacitor in a 34401A
But the tantalum's like to to break.

[Wallace Gasiewicz]

You could at least check the tants for shorts, I think this is the worst failure mode
Check the resistance across the surface mount tants while they are in place and if one is less resistance than the others, it is probably bad.

[Lorenzo_1]

Trying to send an update but can't get a 670kB or 1.2MB jpg/jpeg/png image to load. Will be hard to understand without this image, Any ideas what might be causing this? Done it plenty times before with no trouble, Tried every which way and it just stalls on upload with Cancel option but never completes.

Here's my diagnostic results without the thermal image - will try and load separately tomorrow. 

Replaced C556 - 2200uF electrolytic on 5V VCC regulated line.  No change and removed cap tested fine after removal (it was reading 1pF in circuit before & after replacement).  Checked all the tants in situ using LCR meter - 10 out of 12 test fine (all the 22 & 47uF and 2 of the 3 1uF caps).  Couldn't get LCR tweezers onto 1uF tant C552 without removing anor electrolytic (which I’m trying to avoid) and couldn't get any reading with U1252B multimeter cap test (it read the other two 1uF caps OK).  Also C411 4.7uF tant reads 14.5uF at maximum 100kHz LCR setting. Not sure what's going on there,

Also tested all the power rails. 5V Unreg = +5.040V. -18V Reg = -17.965V. +18V Reg = +19.010V. 5V Reg = +4.953V. Confirmed with logic analyser that the 5V regulated output from U553 (an LM2925T) is stable and is not being reset when display blanks out and meter beeps once. So that seems to exclude the possibility of control unit sending reset b/c rear GPIB/serial ports not passing startup tests (as suggested in some posts). The 19V signal on the 18V line looks high but is in spec (+18 to +20V). So it looks like I have correct voltages throughout the main PCB rails after startup. Display is clearly functional as it flashes up all segments before going blank.

Last thing I did was run the thermal camera over the board on startup (see pic),  U301 (listed as LM356M dual op-amp in service manual) appears to be getting hot. It was reading a bit over 90 degrees C on one startup and about 70C on a second run-up.  The other thermally white components pictured were at 35-45C. U403 (the 7.1V voltage reference) and U153 & U105 (both AD706JR dual op amps) are warm but not hot.  Can anyone comment on whether the U301 temperature looks excessive? U301 is part of the AC measurement circuit and it’s unclear to me why it should be this hot with no inputs.

Looking at the service manual I think next steps (in no particular order) are to (a) remove PCB and test C552 and install a replacement if needed (b) check for ripple on the voltage rails (c) reseat the voltage reference and (d) try and run diagnostics on GPIB.  Grateful other suggestions/ideas in case I’m missing something. I can post some extracts of circuits etc from service manual if that's useful.
 

[coromonadalix]

pls check if the ribbon cable is properly seated and as not damaged wires

they are stiff and easy to break after few manipulations, in one of my meter the wires got kinda stuck / glued on the connector pins, it was difficult to remove it slowly and properly
some people in the past reported some zenner and or capacitor  on the front panel assembly

the front pcb was having issues ??

thats why i mentioned, in my case all tants where changed before putting new big alu  caps back .... yeay  it's not fun  but i did not wanted to take chances .. but the meter was fully functional before doing so

the habit of automatically changing caps is not always the good one, sometime you put in parallel new caps and see  if it help    my 2 cents

[Lorenzo_1]

Finally got thermal image referred to in my last message to upload.  No idea why it failed over two days. Interested in whether the temperature of U301 is excessive - running at 90 degrees C.

[Lorenzo_1]

After more diagnostics, I think the LM2925T 5V regulator may be faulty. Would appreciate any comments before I look at replacing it.

Double-checked power rails after replacing C556 - main filter cap on the 5V rail (floating rail).  All look stable with low/very low AC ripple (~15mV on +/-18V and ~9mV on +5V floating supply and 3mV on +5V earthed rail).

Checked GPIB section by connecting to controller. Device is recognized on port 22 but not identified as 34401A and won’t respond to commands – just times out.  However, U504 & U704 optocouplers appear to be passing traffic.

Resoldered display cable header W601/J500 on main PCB b/c looked to have dry solder joint on one pin and flashing on/off when cable manipulated.  Checked inputs to display module at connector. Much of it OK – 5VAC across the display filament (pins 5 & 11) and power rails look good.  However, I think there should be  +5V at pins 4, 6 and 10 but there is 0V on Pin 10 – which comes from Pin 19 on the main PCB microcontroller.

Pin   Nominal   VDC      VAC
1   A GND   
2   IGFPDG   0.467   0.007
3   +18      +18.99   0.026
4   IGFPSCK   +4.94   0.041
5   VFIL1   -11.88   2.525
6   IGFP01   +4.94   0.096
7   -18V      -17.97   0.016
8   IGFPRES   +0.13   0.209
9   A GND   0.00      0.003
10   IGFPINT   0.019   0.029   
11   VFIL2   -11.88   2.532
12   +2.5V   +2.49   0.002

Looked more closely at the main microcontroller reset circuit that comes from U553 (the LM2925T 5V reg). Input was stable at ~11V. Output stable at 5V. Delay line (pin 4) goes to +5V at ~300ms after startup – consistent with datasheet specs for 0.1 uF C558 on pin 4 C558 tests OK). Datassheet says Reset pin should be held low for duration of delay, but it shows regular charge/discharge cycles immediately after power on, which suggests delay is inoperative and it must be resetting the main uC every few 100us (see photos of Reset & Delay lines at startup). 

Pin 5 leads through R503 & C504 to Reset Pin 16 on main uC and Pin 13 on the ASIC but nowhere else.  R503 & C504 both test OK. I can’t see how the faulty signal on U553 Pin 5 can be generated externally given the inputs all look OK. So I’m thinking the 5V reg must be faulty. Grateful suggestions whether I should replace U553 or do some additional tests.







 

[Swainster]

As is often the case, the MCU can pull its own reset pin down to reset other devices on the same reset circuit - have you checked that this is not the case here? A possible reason for the MCU resetting itself is bad RAM or ROM (U503/U502).

[Lorenzo_1]

Thanks for the suggestion.  Had a look at MCU manual for reset causes.  Confirmed 12MHz on xtal1 and 6MHz on CLKOUT so that looks OK. Waveform was a bit untidy but pretty sure only due to differential probe setup - only had sharp tips on long leads. Not sure how to check RAM/ROM so will have to research that. Can't get GPIB or RS232 comms, so can't get Error report.

[Swainster]

You should be able to determine if the reset is from the MCU by probing either side of R503 with your 'scope, or if that fails, removing the resistor altogther and doing a 'manual' i.e. tweezer aided reset by shorting across C504. You may need to wire in an external pull up resistor if you remove R503

[Lorenzo_1]

Sorry for the delay responding - other commitments... Tested as Swainster suggested - signal was identical both sides of R503 so I removed it carefully. Reset pulsing remained same but now isolated to MCU side.  5V supply RESET is stable HI. So certainly looks like MCU is pulling RESET pin 16 low.  This useful post https://www.vintage-radio.net/forum/showthread.php?t=179008 describes symptoms very similar to mine, including GPIB symptoms, which all turned out to be due to a faulty SRAM. Rather than mess around trying to diagnose that, figured easier to replace U503 and see if that fixes it.  Will have to await parts arrival.

Meantime would appreciate some advice re scope probing on floating side.  So far I've used differential probe but would be helpful to use multiple channels which I can only do with standard probes earthed to scope ground.   When I test the voltage differential between scope probe ground and the floating ground on the 34401A I only get +0.4VDC difference. The result's much the same whether I have the DUT earthed or floating through an isolation transformer.  Given there's not much potential difference between the scope ground and the floating ground, can I use standard grounded scope probes on the DUT without causing problems?  These earthed/floating systems still cause me a few headaches...

[Swainster]

In this particular case you can ground the floating side using your scope probe without any worries (so long as you are not doing anything daft like using the DMM to measure mains voltage at the same time). If you want to probe the meter whilst taking measurement for troubleshooting purposes, then the (safe) signal source should ideally be floating, or at the very least you need to ensure that the common mode voltage is observed i.e. both signal source and DMM commons are at the same voltage.

Just for completeness, consider what happens if you get this wrong - i.e. you try to supply say 10V to the common of the DMM. If both are referenced to the same point (earth ground for a typical 'scope) then the 10V will be shorted out by the common path through the DMM and scope. Whether this is a problem depends of the current sourcing ability of the shorted supply - if you are using a voltage reference or signal generator then typically nothing bad will happen. If you are using a high current supply then likely something will go pop. Most probably the fusible trace on the 34401A. That's the short thin section of PCB trace without solder mask that comes from the common input - at least, that's what my 34401A has. If this burns out then at least the damage should be localised and can be jumpered over (though doing so may add to the thermal coefficient of the meter). That said, 99% of signal sources for troubleshooting are going to be either floating or current limited. If you really want to cover that remaining 1% then you could always put a current limiting resistor between the signal source common and the DMM common.

One thing the be careful of in the case of the 34401A is that the display 'common' is not the same as the floating logic common, but is also not floating with respect to the logic common. The schematic makes this clear but easy to forget when looking for somewhere to clip and 'scope ground. Probably not an factor for this particular fix though.

[Lorenzo_1]

Thanks for that useful advice.  On fault-finding, did some logic traces on U503 RAM and it is cycling at exactly the same 213us intervals (4.7kHz) as the RESET line.  Can't be sure about whether it's the cause or just another symptom but I am seeing 0FFH signals on the data lines that the MCU manual says will cause a RESET.  Fingers crossed it's the RAM causing the problem.

[Miti]

Ok, I come back to this thread to retract my statement, I did not think that such a fine instrument can have such stupid design flaw. I bought two 34401A a couple of months ago and somehow I missed that one had a blown tantalum at C312. The series Zener was short so instead of -15V, U308 saw -18V, and all was ok...apparently... functionally.
Looking at the schematic I can't believe HP engineers selected 20V caps on a 15V supply voltage. I doubt the 2X+ derating rule for tantalum caps wasn't know then.
Anyway, there are 5 capacitors that are more likely to burn C311, C312, C322, C323 and to some extent C559, although it sees 12V only and it is rated at 20V so please REPLACE these tantalum caps ASAP.
All the others are pretty much ok, 5V with 20v or 10V caps, 18v with 35V caps. I replaced all those mentioned before with caps rated at 35V and I'll be ordering a bunch of 3.3V Zener diodes.
I'm curious if anybody saw any burnt tantalum caps in 34401A, other than the five that I mentioned before.

Cheers,
Miti

[Lorenzo_1]

RAM finally turned up. Installed and the 4.7kHz reset signal disappeared - but only to be replaced by a 1kHz reset with much the same signal levels display behavior remains the same photo attached).  Might be some time until I can devote adequate time to further debugging. I  am suspicious of the display connector as display does flash on briefly and a bit randomly when the ribbon cable is jiggled. Not the best termination and a bit hard to test b/c other end is hiding under the display panel.

[Swainster]

I'd be a bit worried that the new RAM was faulty - did you get it from a reputable source?