Re: [SI-LIST] : Some questions about capacitor reliability

Chris Heard/US/3Com (Chris_Heard/US/3Com%3COM@SMTP1.3com-ne.com)
12 Dec 97 9:55:21 EDT

I have lived through tantalum capacitor reliability hell also. My rules are
the following:

1) Voltage rating must be 2.5x the operating voltage.
2) ESR and frequency of operation must be looked at. Measure the voltage
excursions on the board and make sure the voltage amplitudes are within
tolerable limits of the cap.
3) Absolutely minimize the inductance of the connection from each side of the
cap to vcc and gnd planes, use big vias and at least 0.100 mil wide etch, and
keep the etch incredibly short.
4) Only use screened, fused Tantalum capacitors. Count on them eventually
failing, shorting, and the fuse will open. Test and characterize the fuse
clearing characteristics and make sure system behavior is acceptable.
5) Make the vendor screen the parts by applying an appropriate test like 1.3x
the rated voltage, at 85C, for 10 hours.

Chris Heard
3Com Corporation
Southboro, M.A.
508-490-5616

----- Previous Message ----------------------------------------------------

To: perlman @acuson.com @ SMTP1
cc: si-list @silab.Eng.Sun.COM @ SMTP1
From: fred@apsimtech.com (Fred Balistreri) @ SMTP1
Date: Thursday December 11, 1997 05:54 PM
Subject: Re: [SI-LIST] : Some questions about capacitor reliability
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Bob Perlman wrote:
>
> Hi -
>
> One recurring issue among designers and signal integrity engineers is
> how to properly bypass component supply voltages. I bring this up not
> because I want to reopen the issue of just how to do bypassing; I'm
> firmly convinced that bypassing in the form of discrete capacitors (as
> opposed to just relying on plane-to-plane capacitance, or power supply
> capacitance, or faith) is a requirement for high-speed designs.
> However, this does bring up a related, troubling issue: capacitor
> reliability.
>
> It's been my experience that capacitor reliability is a serious
> problem. Tantalum capacitors have a distressing tendency to fail short
> and burn up, and even multi-layer ceramic capacitors sometimes short,
> albeit with less spectacular results. Adding to the severity of the
> problem is the fact that these parts are used in great numbers,
> increasing the probability of failure. I'm familiar with some rules of
> thumb for selecting and using capacitors: don't use the largest value in
> a given package size; derate tantalum voltages by 2 to 1; follow the
> manufacturer's temperature profile recommendations when assembling the
> devices onto boards, etc. However, while following these rules lessens
> the chance of mishap, failures still occur at rates that dwarf those of
> other components (your mileage may vary, of course).
>
> So, my questions are these:
>
> 1) What rules do you follow when selecting and assembling bypass
> capacitors?
>
> 2) Do you know of any "capacitors-are-my-life" consultants who can give
> me better insight into capacitor failure mechanisms and how to avoid
> them?
>
> Thanks,
> Bob Perlman
> Acuson Corporation
I may be wrong but I'm sure that I have seen MTBF data on components
such as capacitors. Generally speaking the more expensive a capacitor
the better the MTBF. However that's not the case with tantalum. Needless
to say that capacitor voltage ratings should be taken seriously. Bypass
capacitors are expecially tricky becuase of voltage spikes than can
occur during turn on/off if the supply is not designed properly. Again
generally speaking the cheaper the supply......well you get the point.
Years ago I had the same problem. Except for the case of tantalum the
problems stopped once we stopped the practice in indiscriminately buying
by the bulk without getting engineers involved.

If the manufacturer does not publish the MTBF data ask for it! If they
can't come up with it find a manufacturer that will. At least one would
know what to expect.

Best regards,

-- 
Fred Balistreri
fred@apsimtech.com

http://www.apsimtech.com