Leaky SMT Ceramic Caps

[James T. White]

I was troubleshooting some boards manufactured by a contract assembler
recently and ran into a strange problem. Some of the 1206 SMT ceramic
caps conduct slightly at supply voltage but show open with an ohmmeter.

In one case, a 0.1uF cap pulled to +5 via a 4.7K resistor was
sufficiently leaky that the processor on the board wouldn't come out of
reset. Upon removal, the cap was tested and showed the correct
capacitance value. When tested with an ohmmeter it showed infinite
resistance but with 5V applied across it, the cap it would conduct >2ma
of current. Replacing the cap solved the problem. I've also seen the
problem in another part of the circuit where a 0.1uF cap is pulled to +5
by pulses from the processor or pulled down by an external load. Same
symptoms and fix.

I suspect the contract assembler used the cheapest parts he could find
but was a bit surprised by the high leakage at just 5V. Has anyone else
seen anything like this?

Thanks.

 

[Lostgallifreyan]

I was troubleshooting some boards manufactured by a contract assembler
recently and ran into a strange problem. Some of the 1206 SMT ceramic
caps conduct slightly at supply voltage but show open with an ohmmeter.

In one case, a 0.1uF cap pulled to +5 via a 4.7K resistor was
sufficiently leaky that the processor on the board wouldn't come out of
reset. Upon removal, the cap was tested and showed the correct
capacitance value. When tested with an ohmmeter it showed infinite
resistance but with 5V applied across it, the cap it would conduct >2ma
of current. Replacing the cap solved the problem. I've also seen the
problem in another part of the circuit where a 0.1uF cap is pulled to +5
by pulses from the processor or pulled down by an external load. Same
symptoms and fix.

I suspect the contract assembler used the cheapest parts he could find
but was a bit surprised by the high leakage at just 5V. Has anyone else
seen anything like this?

Thanks.

If the contractor used a 'no-clean' active flux it can leave residue that
might be slightly conductive if it gets humid. If you replaced it, using a
rosin based flux, you'd probably dry and seal any residue while it was hot,
so it's possibly that. If you can rule that out, great, but check it out
carefully if you can't.

 

[Eeyore]

I suspect the contract assembler used the cheapest parts he could find

Pretty likely.

Graham

 

[Eeyore]

I suspect the contract assembler used the cheapest parts he could find

Do you issue component specs ? I don't mean in ultra detail but I tend to use
the term 'major vendor commercial quality generic part' to indicate to the
subbie that I have some expectation that they don't stuff the board with junk.

I will sometimes talk to them specificially about certain parts that concern me
most.

Graham

 

[PeteS]

I was troubleshooting some boards manufactured by a contract assembler
recently and ran into a strange problem. Some of the 1206 SMT ceramic
caps conduct slightly at supply voltage but show open with an ohmmeter.

In one case, a 0.1uF cap pulled to +5 via a 4.7K resistor was
sufficiently leaky that the processor on the board wouldn't come out of
reset. Upon removal, the cap was tested and showed the correct
capacitance value. When tested with an ohmmeter it showed infinite
resistance but with 5V applied across it, the cap it would conduct >2ma
of current. Replacing the cap solved the problem. I've also seen the
problem in another part of the circuit where a 0.1uF cap is pulled to +5
by pulses from the processor or pulled down by an external load. Same
symptoms and fix.

I suspect the contract assembler used the cheapest parts he could find
but was a bit surprised by the high leakage at just 5V. Has anyone else
seen anything like this?

Thanks.

Apart from what others have said, the cap appears to be defective after
removal. I've had that problem and it was simply that the reflow
thermal profile was creating hotspots on the PCB, and beyond the limit
of the devices.

If it was a hand assembler, check on the equipment he used to put the
parts down quite apart from what parts were used.

Cheers

PeteS

 

[Spehro Pefhany]

I was troubleshooting some boards manufactured by a contract assembler
recently and ran into a strange problem. Some of the 1206 SMT ceramic
caps conduct slightly at supply voltage but show open with an ohmmeter.

In one case, a 0.1uF cap pulled to +5 via a 4.7K resistor was
sufficiently leaky that the processor on the board wouldn't come out of
reset. Upon removal, the cap was tested and showed the correct
capacitance value. When tested with an ohmmeter it showed infinite
resistance but with 5V applied across it, the cap it would conduct >2ma
of current. Replacing the cap solved the problem. I've also seen the
problem in another part of the circuit where a 0.1uF cap is pulled to +5
by pulses from the processor or pulled down by an external load. Same
symptoms and fix.

I suspect the contract assembler used the cheapest parts he could find
but was a bit surprised by the high leakage at just 5V. Has anyone else
seen anything like this?

Thanks.

IIRC, that's a known MLCC failure mode-- something like microcracking
caused by thermal shock. Possibly a combination of dubious parts and a
suboptimal process. Was it a lead-free process?

http://www.calce.umd.edu/whats_new/2001/Ceramic.pdf


Best regards,
Spehro Pefhany

 

[JW]

I was troubleshooting some boards manufactured by a contract assembler
recently and ran into a strange problem. Some of the 1206 SMT ceramic
caps conduct slightly at supply voltage but show open with an ohmmeter.

In one case, a 0.1uF cap pulled to +5 via a 4.7K resistor was
sufficiently leaky that the processor on the board wouldn't come out of
reset. Upon removal, the cap was tested and showed the correct
capacitance value. When tested with an ohmmeter it showed infinite
resistance but with 5V applied across it, the cap it would conduct >2ma
of current. Replacing the cap solved the problem. I've also seen the
problem in another part of the circuit where a 0.1uF cap is pulled to +5
by pulses from the processor or pulled down by an external load. Same
symptoms and fix.

I suspect the contract assembler used the cheapest parts he could find
but was a bit surprised by the high leakage at just 5V. Has anyone else
seen anything like this?

Thanks.

I've seen it happen to SMD caps that were placed under mechanical stress,
as when a PCB is flexed too far. Apparently, tiny fractures can occur in
the substrate, causing the failure mode you are seeing.

 

[James T. White]

Do you issue component specs ? I don't mean in ultra detail but I
tend to use the term 'major vendor commercial quality generic part'
to indicate to the subbie that I have some expectation that they
don't stuff the board with junk.

I will sometimes talk to them specificially about certain parts that
concern me most.

Graham,

Thanks for the input.

As we didn't get them built, I don't really know what component specs
the manufacturer were given but I suspect that they were likely
inadequate. The product and board design were done by a "design
service", the OEM then has the boards fabricated and assembled and now
that they are having problems, the came to us to try and figure out what
is going wrong. I know that the "design service" provided a data book
of component specs for the board but how much of that made it to the
manufacturer is something I'll have to investigate.

 

[James T. White]

On Mon, 18 Dec 2006 00:44:23 -0600, the renowned "James T. White"

IIRC, that's a known MLCC failure mode-- something like microcracking
caused by thermal shock. Possibly a combination of dubious parts and a
suboptimal process. Was it a lead-free process?

http://www.calce.umd.edu/whats_new/2001/Ceramic.pdf

Spehro

Thanks for the URL. I hadn't thought of trying methanol on them to see
if I could change the conductivity which would pretty well indicate
micro-cracking is the problem.

 

[Joerg]

I was troubleshooting some boards manufactured by a contract assembler
recently and ran into a strange problem. Some of the 1206 SMT ceramic
caps conduct slightly at supply voltage but show open with an ohmmeter.

In one case, a 0.1uF cap pulled to +5 via a 4.7K resistor was
sufficiently leaky that the processor on the board wouldn't come out of
reset. Upon removal, the cap was tested and showed the correct
capacitance value. When tested with an ohmmeter it showed infinite
resistance but with 5V applied across it, the cap it would conduct >2ma
of current. Replacing the cap solved the problem. I've also seen the
problem in another part of the circuit where a 0.1uF cap is pulled to +5
by pulses from the processor or pulled down by an external load. Same
symptoms and fix.

I suspect the contract assembler used the cheapest parts he could find
but was a bit surprised by the high leakage at just 5V. Has anyone else
seen anything like this?

Just a thought: Was that assembly accidentally done on a RoHS line or
with a RoHS temp profile? Maybe the caps weren't rated for that and got
fried.

 

[Nermal]

I also suspect microcracking. The cracking may be very close or covered
by the metallization at the end. I would suspect failure was due to
mechanical stresses rather than thermal shock. Look to see if the
circuit board at the capacitors has been subjected to bending. One
thing I have learned over the years is that ceramic capacitors do not
take too well to bending stresses.

 

[reglarnavy]

I was troubleshooting some boards manufactured by a contract assembler
recently and ran into a strange problem. Some of the 1206 SMT ceramic
caps conduct slightly at supply voltage but show open with an ohmmeter.

Jim,

Leakage can be the result of either damage to the cap, or contamination
underneath / around it. Typically a cracked cap will read some nominal
ohmic value on a meter - these parts don't repair themselves once
cracked. Cracked caps can be caused by stressful dapanelization,
mechanical assembly at standoffs w/ screws, pressfit connector
installation, attachment or separation of daughterboards, overclamp ATE
fixtures, touch-up rework with a hot soldering iron, etc. There's lots
of things that can fracture mlcc's. Sometimes mapping out the failure
sites on a grid can lead you to the source of the problem. Do you used
these parts at multiple locations, and do the failures show up randomly
at all of them, or mostly at the same locatins? I think you can find a
nice tutorial on cracked caps on the AVX or Kemet website.

The reflow smt process, lead-free or not, is not likely to crack a
ceramic part. Those assembly profiles just don't have the ramp rate
(typically 2 degrees C / sec) or excursion (250C-260C) to fracture
these parts. Now, a bottomside wave part can easily be cracked. If
the preheat temperature is more than 100C below the wave temperature,
then the thermal shock they experience going over the wave can
certainly crack the larger parts. Once again though, those cracks,
which often are invisible underneath the end terminations, cause ohmic
changes in the parts which are usually detectable with a meter -
several hundred kohm vs. open.

If these are bottomside glue parts - more and more of a rarity in these
days of selective pallets, then you might be seeing something going on
with voiding in the adhesive, which entraps flux at wavesolder. This
has the potential to be a very serious problem, especially underneath
caps which are usually across Vcc-gnd. Certain adhesives are very prone
to voiding and subsequent leakage after wavesolder. The flat
cross-sections look like a slice of tomato with flux in the cavities
between the terminations.

What do you know about the board's assembly chemistry - "leave on" or
OA washable? In general the leave-ons are OK just as long as they're
properly reacted in the process. Left on the board as a liquid, say
for instance when used in a "post assembly" touch-up process where the
operators like to use a flux bottle, they can be trouble. If the board
is an OA process build, then cleanliness is a huge issue because the
chlorides and bromides in those fluxes must be washed off completely.
Look around for any signs of etching around the leads of other smt
parts (QFPs, underneath BGAs) that can be the result of improper
cleaning. Any "fern-like" things (dendrites) growing across these
caps, or smt resistors?

Do these boards undergo any ESS / humidity testing or use in a harsh,
polluted environment (near power plants, dense automobile traffic,
sulfur mines)? All can cause interesting problems with passive
components.

Something else to consider is the layout of the board. Do you have
vias underneath these caps? A tight layout that does this can get into
difficulty with solderballs between the end termination and the vias.
These could look OK on an unpowered board, but show up as you mention,
when it's powered up.

As for the parts themselves, you could have some bad material,
especially if it's coming from a broker. It's worth finding out where
they came from. As for the CM buying the cheapest part they can find
.... that's most likely true; however, CM's typically buy to the
customer's AVL, which calls out the generic component description and
the allowable vendors w/complete part number. I'd be very surprised if
your cap is just listed by value, voltage, tolerance and size, and the
CM is free to buy any vendor's part that fits that generic description.
Your AVL probably calls out a Kemet, AVX, or some other part number(s)
that the CM must purchase. True, they will hunt down the least
expensive part they can within your allowable selections, but anything
they buy will be a part the OEM has specified. No CM that I can think
of would want the liability for a non-working assembly that was the
result of having independently selected a component inappropriate for
the application.