Finding the leaky MLCC's on the board

[mzzj]

Got some battery powered equipment to repair, battery flat sooner than expected

Board shows ~30uA current consumption, expected is probably below 10uA

6pcs 100uF 6.3v trough-hole! MLCC's parallel with the supply. 
Does someone have bright ideas how to find out which ones are leaky without desoldering them from the board?

I desoldered all of the similar caps from one board, 3 out of 6 are suspicious.

Tested at 5v and 25cel room temperature:

3uA
4uA
16uA
<0.1uA
<0.1uA
<0.1uA

[jaromir]

Is it possible to heat up the capacitors one by one? Not much, perhaps 10-15 degC above room temperature should be enough. The leakage is expected to rise with temperature, so you can find the ones with highest contribution to total leakage.

[Fraser]

If some are failing in the same role, I would advise replacing them all as the ones that pass any tests now, could be failing in the near future. For their cost, just fit new parts and sleep soundly knowing the new parts meet spec and have decent life left in them.

If the MLC parts are failing prematurely, you may wish to look for reasons why.

Sadly not helpful in this particular case due to the low currents, but I use thermal imaging to look for unexpected dissipation in components. The energy has to go somewhere and it us often thermal dissipation.

Fraser

[mzzj]

If some are failing in the same role, I would advise replacing them all as the ones that pass any tests now, could be failing in the near future. For their cost, just fit new parts and sleep soundly knowing the new parts meet spec and have decent life left in them.

If the MLC parts are failing prematurely, you may wish to look for reasons why.

Fraser

I was thinking same about replacing them all. Working with the boards is just bit annoying as they are dipped in some sort of silicone compound. 10-20 boards with 6 caps per board 

Failures are not under my control, these are RH+T dataloggers and some of them are used close to MLCC accrelerated aging test conditions. (hot and humid)

Not too many component alternatives either when you need 100uF trough-hole ceramic caps

[T3sl4co1l]

Would tantalum or electrolytic not be preferable?

Tim

[mzzj]

Would tantalum or electrolytic not be preferable?

Tim

I was wondering about the same thing.
Manufacturer of these usually seem to know what they are doing with more thought given to environmental conditions than many others but still.

+85C maximum specified operating temperature and battery power could be one  reason. Tantals are more leaky than non-faulty ceramic.
https://www.avx.com/docs/techinfo/Low_Leakage_Current_Aspect_Designing_Tantalum_Niobium_Oxide_Capacitors.pdf

 10uA at rated voltage, maybbe about 3uA when derated to 30% voltage. x6 and we have 18uA at room temp.
OK, probably no need for 6x100uF as tantals actually have some capacitance left under bias voltage unlike MLCC's.
3x 100uf would be almost acceptable at room temp =  9uA.

[mzzj]

Is it possible to heat up the capacitors one by one? Not much, perhaps 10-15 degC above room temperature should be enough. The leakage is expected to rise with temperature, so you can find the ones with highest contribution to total leakage.

This crossed my mind also and it seems like possible option at least on theory.

16uA leaky MLCC rises to 100uA @ 50cel
3uA leaky MLCC rises to 30uA @50c
Good one with 0.1uA leakage at room temp showed 1uA leakage at 50c.

I would be still inclined to replace them all like Frazer suggested, maybe with different type or make. Except the only (easily)available TH MLCC in this size is the exact one these boards already use.
https://www.digikey.fi/product-detail/en/tdk-corporation/FK22X5R0J107M/445-2902-ND/1008928

[mzzj]

3 out of 11 boards measured what I considered good, IE less than 10uA (around 4uA actually)
One board at 10uA, Rest of them were between 20-40uA. 

Cap specification is actually quite crappy, 1 Mohm insulation resistance for 100uF parts. So for 7x parallel caps on board   26uA would be still "ok" at 3.7v

Bucketload of new caps from digikey, all measure sub 1uA at 10v after 10 second "soak" so >10Mohm.  Tested one for longer soak time and leakage dropped to sub 40nA. So far so good. 

[Psi]

Does someone have bright ideas how to find out which ones are leaky without desoldering them from the board?

If you hit them all with freeze spray and then watched them with a good thermal camera you might be able to spot the ones that return to ambient temp faster than the others. Maybe.

16uA is tiny amount of current and im guessing the voltage is low too, so the heating will be tiny, but if you get the cap down to a really low temp that heating may become more obvious.

[Fraser]

Mzzj,

Excellent news. At least you now know what the likely cause of the issue is ..... the capacitors specification is just inadequate for the application. Good detective work on your part and at least you can have a good level of confidence that the new, better specification, capacitors will solve the problem in the long term 

Fraser

[mzzj]

Mzzj,

Excellent news. At least you now know what the likely cause of the issue is ..... the capacitors specification is just inadequate for the application. Good detective work on your part and at least you can have a god level of confidence that the new, better specification, capacitors will solve the problem in the long term 

Fraser

Actually the specs of the new caps are just as bad as the old ones. They just measure at least 10 times better than specification. Few ones that I tested for longer time leaked in tens nanoamps after sufficiently long soak so actually like 200x better than specification.

Who knows what they measure after few dozen cycles between -25 and +55 with 80% humidity but at least we have known good start. 

These were good reads
https://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20160014331.pdf
https://nepp.nasa.gov/files/24612/MLCCs%20Vabs%20DMR%20Teverovsky%202C_n181.pdf