[SI-LIST] Re: Voltage rating of a Ceramic capacitor

to contribute to the email flood you're getting...
I had run across some cap vendor specs that showed a strong deviation of 
capacitance from the nominal depending on the DC bias and dielectric 
used.  Since most of the caps are used in a DC environment, but tested with 
an AC signal, the stated capacitance isn't there.
- Bill

At 05:34 PM 06/21/2001 -0400, you wrote:
>Since I have received a flood of emails both through the list and directly, I
>will try to respond to all of you in a single email to the list concerning
>your questions on my post.  If I miss anything please email me directly.
>
>First, yes, NPO has a higher failure rate than X7R.  This is because there
>are two principle failure modes in X7R and only one in NPO.  Yes, I said this
>correctly.  The failure mode is due to complex processes but in X7R they are
>shared so the overall stress is less while all of the stress in one mode
>causes early failure in NPO.  The current leakage rate increases in NPO by a
>power of 2.46 while X7R increases by a power of only 1.42 when aged under
>heated bias.  This leads to failure rates several orders of magnitude higher
>for NPO caps at low voltages.  The failure rate is even higher than NPO for
>Z5U or Y5V caps since one of the failure modes is now again under principle
>stress.  In HALT testing, X7R caps will reach a 30 percent failure rate in
>about 170 hours and Z5U will reach the same failure rate in only 45 hours.
>So for highest reliability use X7R followed by NPO and lowest reliability
>option would be the Z5U or Y5V caps.  I will send a pdf version of a
>technical paper on the HALT screening test I developed for zero failure high
>reliability designs to anyone who requests it.  (This will probably elicit
>another flood of emails so be patient if it does not arrive tomorrow.)
>
>Second, why I don't recommend Z5U caps.  Beyond the reliability discussion
>above, the principle reason that should concern EMI and SI designers is the
>very nonlinear voltage to capacitance response at low voltage.  I have seen
>some transients and switching effects become magnified because of this effect
>and lead to a closure of eye-diagrams.  The ferromagnetic aging effect is
>another problem.  Manufacturers will sometimes reduce the aging effect by
>heating Z5U and Y5V caps above their Curie temperature and then cool them
>under bias to reduce the aging effect while they sit on the shelf.  But the
>effect returns when you expose them again to unbiased heat above the Curie
>temperature (soldering, reflow, etc.).
>
>The out of circuit HALT tester I designed uses a machined Teflon block bolted
>to an aluminum base.  The Teflon block was machined with channels about 75
>mils wide and 150 mils deep.  At one end of and common to all the channels
>was a nickel/gold plated copper metal strip which was the circuit ground.  At
>the other end of the channel was a spring loaded push pin which was also
>nickel/gold plated.  It is a simple operation to load caps by retracting the
>push pin, dropping the cap into the channel and releasing the push pin
>against the cap to hold it in and make electrical contact during heated bias.
>  Please be aware that a failing cap will usually short and explode during
>HALT testing.  If you protect each cap with a fuse then keep it outside the
>oven -- high temp fuses are expensive.  You can eliminate the fuse if you use
>a high voltage-high current power supply.  Please be careful with this test!
>
>Next, You will almost always achieve your best results at high frequency by
>paralleling several X7R/NPO caps and using gnd/pwr plane capacitance to reach
>the capacitance, impedance and ESR characteristics you need.  There are
>papers already written on these procedures so I will not go further with this.
>
>I should mention, however, that I have seen people suggesting that you should
>use Z5U and Y5V because they have a lower inductance.  Yes, it is often true
>that X7R caps will have lower inductance.  The higher dielectric constant of
>Z5U and Y5V caps allows the manufacturer to get the same capacitance with
>fewer plates in the same size package which will lead to lower inductance.
>But the other problems remain!  The solution to this is to use low voltage
>X7R ceramic caps that are designed to be thinner with wider/lower resistance
>plates.  The thinner the cap the lower (in general) the inductance.  This
>should make perfect sense to those of us who are microwave designers using
>single layer caps.  There are a whole series of (and for some reason more
>expensive) low inductance MLC's on the market that use this principle.
>
>I have seen very few books on these subjects (maybe I should write one).  The
>best information comes from technical papers written for passive symposiums.
>
>Lastly, my original post was not meant to give those designers who use Z5U
>and Y5V caps indigestion.  If you find that these high dielectric value caps
>work in your design, then congratulations!  If you get a large number of them
>back in a year or two after the caps have aged or failed in or next to a hot
>processor then you have been forewarned.
>
>Harold L. Snyder, Jr.
>Scientist & Consultant
>
>
>
>
>
>
>
>------------------------------------------------------------------
>To unsubscribe from si-list:
>si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>For help:
>si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>
>List archives are viewable at:     http://www.freelists.org/archives/si-list
>Old list archives are viewable at: http://www.qsl.net/wb6tpu
>

----------------------------
Bill Owsley,   owsley@xxxxxxxxx
919) 392-8341

Compliance Engineer
Cisco Systems
7025 Kit Creek Road
POB 14987
RTP. NC. 27709


------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List archives are viewable at:     http://www.freelists.org/archives/si-list 
Old list archives are viewable at: http://www.qsl.net/wb6tpu

Other related posts: