Ceramic capacitor failure

I am seeing some failures of ceramic capacitors in a product I designed. These are 1206 22uf 10V ceramic caps, we do not specify a manufacturer. The caps are on the output of a 3.3V 1amp 2MHz dc to dc converter (5 amp inductor 2.2uH), there are two in parallel. They become resistive, around 200 ohms. The ripple current is around 0.4 amps total, so I cant see that it is excessive ripple current, they are not near mounting holes so I cant see mechanical issues, and they initially pass test but fail in the field. Heating the cap does not have much effect on the resistance.
Any pointers would be appreciated. 
Thanks

steve <steve.jones@scannex.co.uk> wrote:
> I am seeing some failures of ceramic capacitors in a product I
designed. These are 1206 22uf 10V ceramic caps, we do not specify a
manufacturer. The caps are on the output of a 3.3V 1amp 2MHz dc to dc
converter (5 amp inductor 2.2uH), there are two in parallel. They
become resistive, around 200 ohms. The ripple current is around 0.4
amps total, so I cant see that it is excessive ripple current, they
are not near mounting holes so I cant see mechanical issues, and they
initially pass test but fail in the field. Heating the cap does not
have much effect on the resistance. 
> Any pointers would be appreciated. 

Do the capacitors get warm? 
-- 
Uwe Bonnes                bon@elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. 06151 1623569 ------- Fax. 06151 1623305 ---------

On Thursday, February 21, 2019 at 11:09:23 AM UTC, Uwe Bonnes wrote:
> > I am seeing some failures of ceramic capacitors in a product I
> designed. These are 1206 22uf 10V ceramic caps, we do not specify a
> manufacturer. The caps are on the output of a 3.3V 1amp 2MHz dc to dc
> converter (5 amp inductor 2.2uH), there are two in parallel. They
> become resistive, around 200 ohms. The ripple current is around 0.4
> amps total, so I cant see that it is excessive ripple current, they
> are not near mounting holes so I cant see mechanical issues, and they
> initially pass test but fail in the field. Heating the cap does not
> have much effect on the resistance. 
> > Any pointers would be appreciated. 
> 
> Do the capacitors get warm? 
>
I don;t know, it doesnt in the lab.

On 21/02/19 10:07, steve wrote:
> I am seeing some failures of ceramic capacitors in a product I designed.
> These are 1206 22uf 10V ceramic caps, we do not specify a manufacturer. The
> caps are on the output of a 3.3V 1amp 2MHz dc to dc converter (5 amp inductor
> 2.2uH), there are two in parallel. They become resistive, around 200 ohms.
> The ripple current is around 0.4 amps total, so I cant see that it is
> excessive ripple current, they are not near mounting holes so I cant see
> mechanical issues, and they initially pass test but fail in the field.
> Heating the cap does not have much effect on the resistance. Any pointers
> would be appreciated. Thanks

SMPSs can hammer components; the devil is in the details.
Ceramic caps can have very variable specifications and
construction, e.g. notoriously w.r.t. the capacitance as a
function of DC voltage.

The obvious question is whether you are using the capacitors
within the datasheet and/or application note limits.

The former will be difficult to ascertain if the manufacturer
is not specified.

On Thursday, February 21, 2019 at 12:27:42 PM UTC, Tom Gardner wrote:
> On 21/02/19 10:07, steve wrote:
> > I am seeing some failures of ceramic capacitors in a product I designed.
> > These are 1206 22uf 10V ceramic caps, we do not specify a manufacturer. The
> > caps are on the output of a 3.3V 1amp 2MHz dc to dc converter (5 amp inductor
> > 2.2uH), there are two in parallel. They become resistive, around 200 ohms.
> > The ripple current is around 0.4 amps total, so I cant see that it is
> > excessive ripple current, they are not near mounting holes so I cant see
> > mechanical issues, and they initially pass test but fail in the field.
> > Heating the cap does not have much effect on the resistance. Any pointers
> > would be appreciated. Thanks
> 
> SMPSs can hammer components; the devil is in the details.
> Ceramic caps can have very variable specifications and
> construction, e.g. notoriously w.r.t. the capacitance as a
> function of DC voltage.
> 
> The obvious question is whether you are using the capacitors
> within the datasheet and/or application note limits.
> 
> The former will be difficult to ascertain if the manufacturer
> is not specified.

Have looked at a lot of data sheets very few give ripple current specifications

On Thursday, February 21, 2019 at 12:46:25 PM UTC, steve wrote:
> On Thursday, February 21, 2019 at 12:27:42 PM UTC, Tom Gardner wrote:
> > On 21/02/19 10:07, steve wrote:
> > > I am seeing some failures of ceramic capacitors in a product I designed.
> > > These are 1206 22uf 10V ceramic caps, we do not specify a manufacturer. The
> > > caps are on the output of a 3.3V 1amp 2MHz dc to dc converter (5 amp inductor
> > > 2.2uH), there are two in parallel. They become resistive, around 200 ohms.
> > > The ripple current is around 0.4 amps total, so I cant see that it is
> > > excessive ripple current, they are not near mounting holes so I cant see
> > > mechanical issues, and they initially pass test but fail in the field.
> > > Heating the cap does not have much effect on the resistance. Any pointers
> > > would be appreciated. Thanks
> > 
> > SMPSs can hammer components; the devil is in the details.
> > Ceramic caps can have very variable specifications and
> > construction, e.g. notoriously w.r.t. the capacitance as a
> > function of DC voltage.
> > 
> > The obvious question is whether you are using the capacitors
> > within the datasheet and/or application note limits.
> > 
> > The former will be difficult to ascertain if the manufacturer
> > is not specified.
> 
> Have looked at a lot of data sheets very few give ripple current specifications

Also have life tested with 1amp load, no noticeable temperature rise, most of the rise appears to be from the inductor warming up (3 degress rise)

On 21/02/19 12:46, steve wrote:
> On Thursday, February 21, 2019 at 12:27:42 PM UTC, Tom Gardner wrote:
>> On 21/02/19 10:07, steve wrote:
>>> I am seeing some failures of ceramic capacitors in a product I designed.
>>> These are 1206 22uf 10V ceramic caps, we do not specify a manufacturer. The
>>> caps are on the output of a 3.3V 1amp 2MHz dc to dc converter (5 amp inductor
>>> 2.2uH), there are two in parallel. They become resistive, around 200 ohms.
>>> The ripple current is around 0.4 amps total, so I cant see that it is
>>> excessive ripple current, they are not near mounting holes so I cant see
>>> mechanical issues, and they initially pass test but fail in the field.
>>> Heating the cap does not have much effect on the resistance. Any pointers
>>> would be appreciated. Thanks
>>
>> SMPSs can hammer components; the devil is in the details.
>> Ceramic caps can have very variable specifications and
>> construction, e.g. notoriously w.r.t. the capacitance as a
>> function of DC voltage.
>>
>> The obvious question is whether you are using the capacitors
>> within the datasheet and/or application note limits.
>>
>> The former will be difficult to ascertain if the manufacturer
>> is not specified.
> 
> Have looked at a lot of data sheets very few give ripple current specifications

In which case I presume you are using those capacitors, and
are operating well within the limits.

If not, especially if you are using something like Farnell's
"multicomp", then you will have to qualify the caps yourself.

On Thursday, February 21, 2019 at 1:32:50 PM UTC, Tom Gardner wrote:
> On 21/02/19 12:46, steve wrote:
> > On Thursday, February 21, 2019 at 12:27:42 PM UTC, Tom Gardner wrote:
> >> On 21/02/19 10:07, steve wrote:
> >>> I am seeing some failures of ceramic capacitors in a product I designed.
> >>> These are 1206 22uf 10V ceramic caps, we do not specify a manufacturer. The
> >>> caps are on the output of a 3.3V 1amp 2MHz dc to dc converter (5 amp inductor
> >>> 2.2uH), there are two in parallel. They become resistive, around 200 ohms.
> >>> The ripple current is around 0.4 amps total, so I cant see that it is
> >>> excessive ripple current, they are not near mounting holes so I cant see
> >>> mechanical issues, and they initially pass test but fail in the field.
> >>> Heating the cap does not have much effect on the resistance. Any pointers
> >>> would be appreciated. Thanks
> >>
> >> SMPSs can hammer components; the devil is in the details.
> >> Ceramic caps can have very variable specifications and
> >> construction, e.g. notoriously w.r.t. the capacitance as a
> >> function of DC voltage.
> >>
> >> The obvious question is whether you are using the capacitors
> >> within the datasheet and/or application note limits.
> >>
> >> The former will be difficult to ascertain if the manufacturer
> >> is not specified.
> > 
> > Have looked at a lot of data sheets very few give ripple current specifications
> 
> In which case I presume you are using those capacitors, and
> are operating well within the limits.
> 
> If not, especially if you are using something like Farnell's
> "multicomp", then you will have to qualify the caps yourself.

The life testing was done on the current manufacturing batch that have the issue and I assume the same capacitors fitted, they look the same.

On 2/21/19 7:27 AM, Tom Gardner wrote:
> On 21/02/19 10:07, steve wrote:
>> I am seeing some failures of ceramic capacitors in a product I designed.
>> These are 1206 22uf 10V ceramic caps, we do not specify a 
>> manufacturer. The
>> caps are on the output of a 3.3V 1amp 2MHz dc to dc converter (5 amp 
>> inductor
>> 2.2uH), there are two in parallel. They become resistive, around 200 
>> ohms.
>> The ripple current is around 0.4 amps total, so I cant see that it is
>> excessive ripple current, they are not near mounting holes so I cant see
>> mechanical issues, and they initially pass test but fail in the field.
>> Heating the cap does not have much effect on the resistance. Any pointers
>> would be appreciated. Thanks
> 
> SMPSs can hammer components; the devil is in the details.
> Ceramic caps can have very variable specifications and
> construction, e.g. notoriously w.r.t. the capacitance as a
> function of DC voltage.
> 
> The obvious question is whether you are using the capacitors
> within the datasheet and/or application note limits.
> 
> The former will be difficult to ascertain if the manufacturer
> is not specified.

If the OP's ripple waveform has a lot of nasty steps in it, he could be 
exciting a piezoelectric resonance, leading to fatigue failure.

Cheers

Phil Hobbs

-- 
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

On Thursday, February 21, 2019 at 2:33:56 PM UTC, Phil Hobbs wrote:
> On 2/21/19 7:27 AM, Tom Gardner wrote:
> > On 21/02/19 10:07, steve wrote:
> >> I am seeing some failures of ceramic capacitors in a product I designed.
> >> These are 1206 22uf 10V ceramic caps, we do not specify a 
> >> manufacturer. The
> >> caps are on the output of a 3.3V 1amp 2MHz dc to dc converter (5 amp 
> >> inductor
> >> 2.2uH), there are two in parallel. They become resistive, around 200 
> >> ohms.
> >> The ripple current is around 0.4 amps total, so I cant see that it is
> >> excessive ripple current, they are not near mounting holes so I cant see
> >> mechanical issues, and they initially pass test but fail in the field.
> >> Heating the cap does not have much effect on the resistance. Any pointers
> >> would be appreciated. Thanks
> > 
> > SMPSs can hammer components; the devil is in the details.
> > Ceramic caps can have very variable specifications and
> > construction, e.g. notoriously w.r.t. the capacitance as a
> > function of DC voltage.
> > 
> > The obvious question is whether you are using the capacitors
> > within the datasheet and/or application note limits.
> > 
> > The former will be difficult to ascertain if the manufacturer
> > is not specified.
> 
> If the OP's ripple waveform has a lot of nasty steps in it, he could be 
> exciting a piezoelectric resonance, leading to fatigue failure.
> 
> Cheers
> 
> Phil Hobbs
> 
> -- 
> Dr Philip C D Hobbs
> Principal Consultant
> ElectroOptical Innovations LLC / Hobbs ElectroOptics
> Optics, Electro-optics, Photonics, Analog Electronics
> Briarcliff Manor NY 10510
> 
> http://electrooptical.net
> http://hobbs-eo.com

The unit has a sleep current of 40uA, a running current of 70mA with a 1 amp load whenever the GSM modem kicks in, this has a 4700uF cap on it and when turned on and is driven by a 1amp current limited switch which takes around 15ms to charge up, then current is around 350mA. Unit runs approx every 15mins and uses the modem once a day for a couple of minutes.