UNI-T UT61D failure, possible repair [solved!]

[newbrain]

My cheapish UT61D (Euro version, with MOVs and PTCs) yesterday suddenly started to consistently give low readings in DC and AC V.
Resistance and current were spot on, though.

Tried a new battery, no change.
Took out the PCB, gave it a good IPA (no, not the beer) bath and scrub, still no dice.

Well, let's have a look at the precision resistors...gotcha!

The 10M resistor for the input divider in the datasheet of the FS9922-DMM4:


Its physical realization:


The value of R24C, supposedly 2.5M  :

Yes, my other multimeter is a Fluke

The other three are exact to the last digit.

First a question: is this a common failure mode for SMD high value resistors?
I've seen inductors and capacitor dying, and cooked some resistor myself, but never a sudden rise in value.
Note that the DMM has never been stressed in terms of applied I/V, shocks, or humidity and temperature.

Then, the next step.
Looks like an easy repair? Not so much!
It seems that there's no such beast as a 1206, 0.1% 2.5M resistor, I've checked Digikey, Mouser, RS and Elfa.
Even AliExpress.

Instead, I can easily find 0.1% 1M  resistors in 0805: for the price, and no shipping cost (shop pick up) I have ordered some already.

Considering that the meter is out of warranty, and that it does not go anywhere near high voltages, I was thinking of this mega botch (long live MS paint):

where the smaller resistors would actually be 1M, and would exploit the ample pad areas (I'll need to cut the small trace between R24C and R24B).

Apart being laughed out of the forum, are there other considerations the would advise against it?

The first to come to mind is safety: the 0805 are rated for 100V, rather than the usual 200V or the 1206.
Even if having a series of three would look like an improvement, the PCB mod will probably worsen the overall rating...

Precision should be OK (and I would not need to recalibrate), durability...whatever, probably not worse than the original!

Opinions and alternative suggestions welcome!
Light chiding and mocking accepted.

[bd139]

I wouldn't worry. If you've got a Fluke only stick that near the mains

You could pull the faulty 2.5M out and replace it with 4x 10M stacked on top of each other (yes I know that's just as ugly) but it'll maintain the voltage rating and give you the same 0.1% precision? You'd have to buy 100 or so of the 10M and hand select as you can't get 0.1% 10M's by the looks.

I've only seen SMD resistors change in value from crappy soldering really. They might be crappy resistors too however.

Edit: they might only be 1% in the UT61D

[crazyguy]

replace all the voltage divider/reference resistors with Caddock 1776-C48

http://www.caddock.com/Online_catalog/Mrktg_Lit/Type1776.pdf

[newbrain]

Thanks for the answers!
Specifically:

replace all the voltage divider/reference resistors with Caddock 1776-C48

That's very interesting (my 30+ yo Hung Chang uses one of these networks, a Norma I'm trying to repair two), but, apart from price considerations, it would mean extensive modification to the PCB and flying wires, as the 4x2.5M are quite far away from the other resistors in the divider (close to the main chip). This would probably defeat any remaining voltage safety.

I wouldn't worry. If you've got a Fluke only stick that near the mains

Yes, and I don't usually need to.

You could pull the faulty 2.5M out and replace it with 4x 10M stacked on top of each other (yes I know that's just as ugly) but it'll maintain the voltage rating and give you the same 0.1% precision? You'd have to buy 100 or so of the 10M and hand select as you can't get 0.1% 10M's by the looks.
Edit: they might only be 1% in the UT61D

This is a good mechanical solution, but yes, I'll have to select the right 4.
The stated 61D accuracy in DC V is 0.5% in most ranges and I must say it was in very good accordance with the Fluke, before the malfunction. The other 3 are precisely 2.500 M on the Fluke.

I've only seen SMD resistors change in value from crappy soldering really. They might be crappy resistors too however.

I'm also puzzled by this. I could expect a resistor to go open, or change value if cooked, but to go up abruptly about 33%?
The soldering is not so bad, as you can see.

[alm]

Note that the DMM has never been stressed in terms of applied I/V, shocks, or humidity and temperature.

Could it be mechanical stress that degraded the end cap attachment due to shock or flexing of the PCB in that area?

It seems that there's no such beast as a 1206, 0.1% 2.5M resistor, I've checked Digikey, Mouser, RS and Elfa.
Even AliExpress.

250 (249) is an E48 value. You would likely have to special order that from the factory (MOQ >> 1). Or find someone with surplus. How about two 1206 5 MOhm resistors in parallel? Looks it might just fit with less hackery than your proposed solution.

It seems that there's no such beast as a 1206, 0.1%
Apart being laughed out of the forum, are there other considerations the would advise against it?

The first to come to mind is safety: the 0805 are rated for 100V, rather than the usual 200V or the 1206.
Even if having a series of three would look like an improvement, the PCB mod will probably worsen the overall rating...

It's a UT-61D. What rating? 

The obvious issue to me is that the clearance between pads is severely reduced (the distance between the pads where you will cut the trace and parallel two resistors looks like only 0.4mm or so to start with). This means that in my opinion you should remove all CAT II/III/IV ratings (sharpie, label, whatever) and keep it away from high energy sources. If exposed to a high voltage (transient), I could see arcing between those pads, possibly leading to a cascading failure of the other resistors.

The second issue is the attachment between the left two resistors. There is no pad there. so the resistors will be unsupported. Some glue might help against this. Or even create your own pad with a tiny piece of copper sheet/tape.

For accuracy I see no issue as long as the resistors have the same performance (tempco, stability) as the original ones. It is not like contact resistance or current handling capability is much of an issue.

[newbrain]

Could it be mechanical stress that degraded the end cap attachment due to shock or flexing of the PCB in that area?

The DMM has been bought new, from a shop. It was never dropped, and internally, there's no real stress on that PCB side...

It's a UT-61D. What rating? 

But mine is the GS tested euro version!
It sports 2 PTCs, 3 MOVs, real HRC fuses and slots galore!
(Yes, I agree though).

So in the end I went for the 4x10 Mega, arranged 2x2.
It's the only one that's reasonably safe and mechanically sound:


Well, the resistor I got are 100ppm/K, but I think I can live with it.
The result is even slightly better than before the fault (in the background, my 3083 based PSU, ugly, but accurate enough):


Thanks to all for the worthy suggestions!
 

Bonus picture, my first DMM, Hung Chang HC-5010T, with a friend's Norma D1010: