How to check for bad smd tantalum capacitors ?

[kripton2035]

Hi All,
I recently came in front of bad notebook computer main board, which main failure was due to a bad smd tantalum capacitor.
It serves as output of the graphic card power supply, is around 220-330µF 2V-2.5V
you measure the voltage on it, and get 0.3V and you should get around 1V
you change the capacitor and the computer works again.


I removed these capacitors, and measured them. they all have ESR and capacity that are in spec or so.
How do I check these capacitor for bad/good ?
I read that these capacitors are known for lasting no more than 5 years.
is it a leakage problem ? (I don"t have -yet- build myself a capacitor leakage meter)


thanks for any advise.

[tautech]

Hi All,
I recently came in front of bad notebook computer main board, which main failure was due to a bad smd tantalum capacitor.
It serves as output of the graphic card power supply, is around 220-330µF 2V-2.5V
you measure the voltage on it, and get 0.3V and you should get around 1V
you change the capacitor and the computer works again.


I removed these capacitors, and measured them. they all have ESR and capacity that are in spec or so.
How do I check these capacitor for bad/good ?
I read that these capacitors are known for lasting no more than 5 years.
is it a leakage problem ? (I don"t have -yet- build myself a capacitor leakage meter)


thanks for any advise.

There is a Tant failure mode called 'sputtering'......sort of self explanatory.
Can be found with use of a scope.

[wraper]

I removed these capacitors, and measured them. they all have ESR and capacity that are in spec or so.
How do I check these capacitor for bad/good ?

Attach multimeter  in mA mode in series to it and apply rated voltage through current limiting resistor (to not damage multimeter, just in case). Sometimes they appear fine at low voltage but start passing high current when voltage is increased.

I read that these capacitors are known for lasting no more than 5 years.

Not true, they are very reliable given they are used/derated properly. http://www.kemet.com/Lists/FileStore/KO-CAPUpd.pdf Usual tantalum capacitors shouldn't be used above 35% of their voltage rating for high reliability applications, tantalum polymer types need much less voltage derating. This happens because insulation gets damaged during reflow and may cause high current leakage resulting in catastrophic failure. If voltage applied is low enough it will self heal eventually. Current surges (like during power on) through them must be limited as well.

[kripton2035]

I measured them with a multimeter in mA and µA mode, applying 2V to the capacitor.
I get a 2 to 10 µA current on the tantalums, and 0.5-1µA on the non-tantalum.
but the current varies down for a long time. which value is correct ?
Imust have a good tantalum, and don't see any notable difference in current with the bad one (sure they are bad on the board)

what kind of test can I do to differenciate the good from the bad ?

sometimes I can check the voltage on the live board, but sometimes it's impossible.
that's why I want to have a method to test them out of the board.


the only thing I can see is the "good" one has a 350µF capacity (rated 330µF)
and the bad ones have a 400 to 450µF.
ESR are 0.2 to 0.6 Ohms

thanks.

[Farid]

There is a Tant failure mode called 'sputtering'......sort of self explanatory.
Can be found with use of a scope.

Hi, Can you please explain a little bit more in detail?
Thanks.

[Shock]

In circuit can be hard as other components (e.g. caps, resistors, semiconductors) can cause the inaccurate readings.
The other problem is the different types of failures that can occur with caps.

From what you are saying in your post, at a working voltage in circuit the capacitor appears to draw a lot of current and act like a short to ground bringing the voltage down. So we can assume it has a high series resistance or acting like a resistor at working voltage.

An easy method you can use to measure leakage is a DMM/Ammeter in series with the capacitor on a current limited supply (e.g. bench supply) running at a suitable working voltage. But you have to consider the resolution of the DMM and burden voltage. Current limiting is mainly to prevent your equipment/fuses and smoke.

If you don't have a suitable ammeter you can also use the mV range on your DMM in series. With 10M impedance you will get 1mV per 100pA of current drawn, this works really well. With the multimeters high impedance on voltage, the capacitor will take a while to charge (depending on the voltage) so once you have it set up you might want to short across the multimeter so the full voltage is applied to the capacitor to speed up the process.

If you have a logging multimeter (min/max with a reset may work perhaps) you can leave the capacitor on constant test, allow it to warm up or even apply heat/cold if required in order to observe a failure. From there you could probably use a programmable supply to perform advanced tests.

This is a good method for "profiling" electrolytic capacitors as well, monitoring the leakage will show a capacitor reforming.

[Shock]

To find that particular failure in circuit you could use a current probe/tracer, or use the voltage drop along the trace method to observe what is pulling a lot of the current. You may be able to determine a failure by hot and cold testing. Inserting a signal may not work if the cap requires a certain voltage, temp or signal condition met prior to failing. Similarly testing the DC resistance across the cap may not be useful if the cap has already tested fine for capacity and ESR.

[tautech]

There is a Tant failure mode called 'sputtering'......sort of self explanatory.
Can be found with use of a scope.

Hi, Can you please explain a little bit more in detail?
Thanks.

Not personally observed but told to me by a Dr of EE that's seen it.

Breakdown/short>self heal>short>self heal>short>self heal. (sputtering)
Think of voltage across cap like this: v^v^v^.........

In an application with low impedance supply rail it might quickly lead to Tant self destruct when it would be easy to find.
A Tant I found shorted in a BWD scope on a LV rail, the scope would power on (but not work) but the regulating componentry and current limiting (resistor) on that rail got warm and that was the only clue to a shorted rail.

Where Tants are used for remote bulk capacitance and decoupling (more common in days past)(and dipped bead types), they were across DC rails and sometimes used too close to their V rating.
In this application one leg of the rail is often at mains or DUT ground when it's a simple matter to monitor with a scope with say falling edge and single trigger.

[kripton2035]

I don't see that on a scope, but rather a lower voltage than expected (20 to 80% less) but the voltage is very stable.
when it comes too low, the graphic display shuts off.

[montemcguire]

In your original post, you note that the bad capacitor prevents the power supply from rising past 0.3V. If you had measured this bad cap, you would notice that it no longer works like a cap, but like some sort of resistor of relatively low value. That is how you check for a bad capacitor - by checking that it is not shorted or otherwise has enormous 'leakage current'.

Checking low level leakage current is troublesome, since a Ta capacitor's leakage current will change over time after the capacitor is installed in a circuit. Once the capacitor is driven to its operating voltage, any micro-shorts that can get cleared at that voltage will get cleared and the oxide layer at the short will be re-grown. As long as the voltage stays at this level, the leakage will settle to an extremely low value. Increase the operating voltage, and more sites that could short out at this new higher voltage will then short out, until all of those are cleared. In this mode, the leakage current is the current required to clear the shorts, and once the short has cleared, no further current will flow. Because of this, once a Ta capacitor is installed in a stable circuit, leakage current will become extremely low and its reliability will increase to extremely high levels.

So, the simplest answer is that if you can power the cap up to your desired working voltage and the 'leakage' current decays toward zero, the cap is good. The leakage current of a cap that is not good at a certain voltage will increase limitlessly when biased to that voltage. Catastrophically failed Ta caps will have such a large short that they will conduct current even at very low voltages, like those from an ohm-meter.

[kripton2035]

so, as these capacitors are rated at 2 or 2.5V, what testing voltage do you recommend ?
also these capacitors should have 1V or 1.5V at max on their pins, in a normal running situation.
I really don't understand that these capacitors don't work in the circuit (a power supply)
and are tested good for C, esr and leakage out of the circuit ?

[Jeroen3]

You find bad capacitors or shorted chips with a thermal imaging camera.

[montemcguire]

so, as these capacitors are rated at 2 or 2.5V, what testing voltage do you recommend ?

The operating voltage in the circuit. This is the only voltage that matters - if the cap works at that bias, it works in your circuit.

also these capacitors should have 1V or 1.5V at max on their pins, in a normal running situation.
I really don't understand that these capacitors don't work in the circuit (a power supply)
and are tested good for C, esr and leakage out of the circuit ?

I'm not used to such low voltage capacitors, but I guess if you need a super tiny part at a huge capacitance, this is the only way to go. If you could find a mechanically compatible part with a higher rated voltage, you'll probably be happier. For 1.5V, you really want a 3V cap or more - 2 or 2.5V is on the edge. In general, if you want to put X volts across a solid Ta, you need to get it rated for at least 2X. Also, the part cannot be subjected to limitless current pulses, such as accidental shorts or very fast voltage edges from low impedance sources. But, if you do all of that, solid Ta can be pretty reliable. Personally, I try to avoid them for new designs as I need to avoid starting fires, but again, if you can over-rate the part generously, and it will be exposed only to a predictable range of currents, then they can be very reliable, will last forever, and will have extremely low leakage once they settle in.

[Shock]

so, as these capacitors are rated at 2 or 2.5V, what testing voltage do you recommend ?
also these capacitors should have 1V or 1.5V at max on their pins, in a normal running situation.
I really don't understand that these capacitors don't work in the circuit (a power supply)
and are tested good for C, esr and leakage out of the circuit ?

Operating voltage would be as per the datasheet taking into account derating, 50% is common. If it does pass a leakage test out of circuit, then it could be a self healing problem as suggested and the only way you are going to categorically prove the failure is by reproducing the circuit plus all stresses.

[kripton2035]

the capacitor is filtering the output of a dc to dc converter, based on a ISL6xxx chip.
what kind of stress can they have, most of the job is made by the controller ?

[Shock]

Any type of stress including environment. With a laptop I would guess heat helped play a part, but it's really just a guess. You would have to check the datasheet and brand if it was up to the job, and analyze the circuit if it was a design issue. Hardly worth it unless it's a repeating failure or you can find others with the same issue. I'd derate it even further (as it's a known tantalum trait) and replace with a quality brand then wipe my hands. Here is some interesting reading.

http://www.kemet.com/Lists/TechnicalArticles/Attachments/199/2014%20EDFA%20Tantalum%20Cap%20Failure%20Analysis%20Review%20by%20Javaid%20Qazi.pdf

https://escies.org/download/webDocumentFile?id=62199

The video Kemet did a while back showed they were working hard to solve tantalum cap problems but your laptops vendor could have thrown any old shit in, if you know what I mean.

Couldn't find that video but I found this one which is just as interesting, he says here infant mortality can occur years after the tantalum goes into service. Anyway I always tend to do failure analysis if it's not obvious as part of a good fix, but the most important part is finding the defect easily not dwelling over it's death.