5
\$\begingroup\$

I am working on a waveform generation board.

The board contains +-200 V supplies and +-24 V supplies. The board takes a low voltage (+-15 V) waveform from a DAC on separate board and outputs +-200 V waveform. This output powers piezoelectric tubes (mostly capacitive.)

The +-200 V supplies are controllable by an MCU and are switched off when the output is not in use. The DAC waveforms are linear ramp with 5 ms rise and 1 ms fall, 100 Hz. The worst case load on the op-amp is 2.4 μF (all piezoelectric tubes together). The +-24 V supply has a sufficient capacitor bank to take the dump current from the piezo capacitance.

The opamp has a 100 nF, 450 V capacitor (C3216X7T2W104M160AA) on its supply pin (attached section of schematics.)

We have seen failure of this 100 nF on two boards after significant hours of operation. The capacitors fail short and have burn marks on their leads.

I have done many design checks to find out possible causes for this failure, but could not find any possible reason. I have done analysis for current through the capacitor on turn on/off of power supply, current through capacitor on load turn on and temperature rise, checked for spurious voltage on +-200 V - this op-amp is only load on 200 V supply.

Am I missing any other checks (analysis/tests) I should be doing?

Another important point is that these boards were manufactured/stress tested and commissioned ~2 years ago and were in powered off state for a really long period after that. The failure was observed soon after they were powered back on.

These boards are geographically at different locations but have seen similar operation cycling before being powered on and having the failure.

In a nutshell, completely tested working board, powered off for long time and fails in capacitor short when powered up.

The place where the boards are stored is an office environment, so storage temperature and humidity wouldn't be something this part can not handle.

What could be reason of this capacitor failing after a power off time period?

Any pointers to understand the issue would be greatly appreciated.

enter image description here Failed Capacitors pictures enter image description here enter image description here

\$\endgroup\$
6
  • \$\begingroup\$ Where is the -200V supply? The capacitor is listed as not recommended for new designs, so maybe they had some reliability problems? If the supply voltage is applied suddenly, perhaps there is a high surge current. \$\endgroup\$
    – PStechPaul
    Commented Aug 10, 2022 at 6:04
  • \$\begingroup\$ When boards (with some electrolytic capacitors) are powered off for a long time (for example some old TV for 15 days off at vacancies ...), some capacitors should be (re)charged with a slowly rising voltage for a little time ... for "reforming" electrolyte. \$\endgroup\$
    – Antonio51
    Commented Aug 10, 2022 at 7:49
  • \$\begingroup\$ These boards are geographically at different locations where and what environment etc.. \$\endgroup\$
    – Andy aka
    Commented Aug 10, 2022 at 8:05
  • \$\begingroup\$ -200V is for second amplifier which generates negative ramp. The supplies have soft start feature and soft start time is such that the current into cap is way lower than it can handle. Australia and Germany, but both are being used in office environment. Controlled temperature and humidity. @Antonio The part failure we have seen is on ceramic cap. We do have some electrolytic caps on board, which are working fine. \$\endgroup\$
    – Nirmala
    Commented Aug 11, 2022 at 0:56
  • \$\begingroup\$ OK. Perhaps this : Conductive anodic filaments (CAF)? mclpcb.com/blog/guide-pcb-caf-issues Do you have a picture of this? \$\endgroup\$
    – Antonio51
    Commented Aug 11, 2022 at 8:04

2 Answers 2

Reset to default
9
\$\begingroup\$

Micro cracking and delamination (typically due to handling such as board flexing or manual soldering) is a major cause of failures in MLCC capacitors.

The failure is typically not immediate (nor even detectable) but the latent defect allows moisture to enter the capacitor and failure (to short) occurs later. See this NASA report, for example.

As well as design and handling considerations, there are capacitors made with "soft" terminations that are less susceptible to such damage.

\$\endgroup\$
3
  • \$\begingroup\$ Thank you for the explanation and link. What baffles me is that the failure happened immediately after significant power down duration. They worked fine for more than a year of continuous operation. \$\endgroup\$
    – Nirmala
    Commented Aug 11, 2022 at 3:22
  • 1
    \$\begingroup\$ I wonder if the warmth of continuous operation kept moisture from getting in. \$\endgroup\$
    – Spehro Pefhany
    Commented Aug 11, 2022 at 3:50
  • 1
    \$\begingroup\$ I did suspect that. I searched for information on whether ceramic caps are affected by moisture in off state, but it turns out ceramic are better than other counterparts at not absorbing moisture (ieeexplore.ieee.org/document/5138959), plus the boards are in office environment where humidity is controlled. There is possibility of some mechanical stress on these parts due to mounting of boards. These parts are not very close to mounting hole but close enough to be in strain path. I guess I will close my analysis by replacing with soft termination part & see what happens. \$\endgroup\$
    – Nirmala
    Commented Aug 11, 2022 at 6:05
1
\$\begingroup\$

From this 2013.

See the discussion:

V. Discussion
Contrary to our initial expectations that ECM in cross-sectioned capacitors is limited to the metals used in internal electrodes of the capacitors (Ag/Pd for PME and Ni for BME), experiments show that all metals, including those used in the termination metallization system, and even metals used in the solder, play an important role in the dendrite growth and failures of MLCCs.

And as @Spehro Pefhany pointed out:

  1. Due to the ease of electromigration of termination and eutectic Sn/Pb solder metals, typically used for hi-rel parts, cracks that extend through the terminals, rather than the body, are the most dangerous and are much more likely to cause catastrophic or intermittent failures in MLCCs.
\$\endgroup\$

Not the answer you're looking for? Browse other questions tagged or ask your own question.