LM2678: LM2678SD-12 Capacitor Failures

Hello Todd,
From your description it appears that some kind of over stress is causing capacitor failures.

As you know tantalum capacitors are susceptible to damage due to either excessive voltage, current or temperature. Any of these can
impact their long-term reliability. Reducing any of the mentioned stresses will have a positive effect on improved reliability. The most common
method for reducing tantalum dielectric stress to improve reliability and guard against catastrophic failure is to reduce the
applied voltage so that it is less than the capacitor’s rated voltage (normally a 60 % to 50 % voltage derating is recommended
for solid tantalum capacitors).
For 100uF capacitor, from datasheet table, Dc voltage rating is 16V. So maybe some over voltage or current may be causing damage.
Also 33uF capacitor is rated for 25V, while your input voltage is 24V.
So there is very little headroom on both capacitors.

We generally recommend using capacitors rated at least 50% above expected operating conditions.
Regards
Brani

Brani,

Let me provide some additional details. First, the temperatures are ambient so nowhere near the rated temperature min/max for these capacitors. We actually had the failure on the LM2678SD-12 AND LM2678SD-5, one on each board. Both input voltages are 24V. The 12 volt regulator failure was on the output capacitor, a 100uF 24V Sprague cap. This is the one that caught on fire. The most recent failure was on the 24V side of the 5V regulator, a 33uF 35V Sprague cap. We have been measuring the input voltage on this cap and it doesn't appear to ever exceed 25V, much less approach 35V. So, I don't believe this issue is related to over voltage or temperature. Can you explain what you mean by over current? Could that perhaps explain the failure? We are running current up near the top limit of the regulator (4A is typical).

Todd

Todd,
Thank you for detailed explanation.
Unfortunately, this is very unusual.
Is there any way you could provide schematic and layout of your design?
That should help us with troubleshooting.
Regards
Brani

One issue might be the RMS current rating of the capacitors.

You might calculate the expected RMS currents for both the input and output and

check that against the manufacturers ratings.

Here is the schematic:

The most recent capacitor that failed on the input side (24V) has a max ESR of 0.2Ohm and a max ripple (100kHz) Irms of 1.12A.

The original capacitor that caught fire was on the output side (12V) has a max ESR of 0.09Ohm and a max ripple (100kHz) Irms of 1.67A.

What is the expected ripple current from the LM2678? My output current is variable, but max is around 4A.

Hi Todd,
Schematics look OK.
Since voltage rating seem to be sufficient, i believe at this point current ratings for capacitor you are using should be examined.
It is possible that ripple current is too high for capacitors you are using, degrading their performance and ultimately causing them to fail.
Please check component datasheet and verify that ripple current ratings are sufficient to handle input/output current in your application.
To calculate RMS current through input capacitor you can use following equation
Irmsin=Iout*square root of (D*(1-D)), where D is duty cycle=Vout/Vin

And RMS current through output capacitor
Irmsout=((Vin-Vout)*(Vout/Vin))/(2*L*fsw*1.73)

An experiment to try would be to change the mix between type of capacitors used, so for example replace C4/C5 with 22uF/35V ceramic capacitors.
Also in other case 24vin/12Vout, replacing one of the output capacitors with combination of ceramic capacitors might be helpful.

My understanding is that the ripple current should be divided between the parallel capacitors, thus they should not be exceeding the rated ripple current of these capacitors.

I found similar ceramic capacitors that meet the capacitance and voltage ratings of the tantalum. If I replace all the tantalum capacitors in the circuit with ceramics, will that fix my problem?

It is correct that ripple current will be split between capacitors. However the capacitors are not exactly identical, so there is potential that one of them will just experience higher ripple current, and potentially fail.

It is not possible for me to say if replacing all tantalum capacitors with ceramic type would solve the issue you are seeing. You would need to verify that new capacitor configuration helps and solves observed issue. In terms of overall performance, having mix of ceramic capacitors placed close to the regulator and some tantalum capacitors placed as bulk, is often recommended.