RE: Capacitor Leakage Current over Temperature

shabaz — Wed, 26 Jun 2019 18:36:47 GMT

Hi Michael,

Very interesting results!

Recently I took a look at a couple of computer motherboards, to see if I could spot any trends. One was an old motherboard (2010), and the other was about 5 years more recent in manufacture. They were from different manufacturers, both major brands, and both intended for server use, not consumer.

The older motherboard had (as expected) lots of MLCC, and the remainder were a mix of aluminium polymer, some tantalum (clustered in one part of the design only, near some connector), and some aluminium electrolytic. There were also some I could not determine (rectangular case like tantalum, but all black in colour, they may have been tantalum polymer - I could not tell from the code on top).

The newer motherboard had the MLCC as expected, and then _all_others were alu polymer, save one, which was aluminium electrolytic. There were zero tantalum. The lone alu electrolytic (Nichicon HM series) was at a corner of the board, perhaps it was decided it would run cool and have a long life there. It was part of the output of a DC-DC converter, in parallel with some MLCC. The DC-DC converter chip was the same as that used in a few other places of the board, but the other places used alu polymer in parallel with some MLCC. (EDIT: it was 1000uF 16V, so perhaps that DC-DC converter has a higher voltage output, for some ancillary purpose not as power-intensive as the lower-voltage output DC-DC converters on the rest of the motherboard).

It was interesting seeing the complete absence of tantalum, at least for that server motherboard, and near-absence 10 years ago. For a different market (with ultra high reliability requirement) out of uni, the capacitors were almost entirely tantalum (+ MLCC), no alu electrolytics.. but then there there were no alu polymer available at the time.

RE: Capacitor Leakage Current over Temperature

neuromodulator — Wed, 26 Jun 2019 16:54:29 GMT

Very nice report Michael, the chamber and the parallel recording through a DAQ surely help a lot in making the testing much more rigorous!

I'm currently working on designing a chamber to do what you just did, test different arbitrary temperature curves...

To reduce noise I set the number of power line cycles to 2, and averaged 100 measurements, this took approximately 3 s.

In the plots, which is current, the red or the blue? What is the other color?

As there appears to be hysteresis, it would be interesting to see an i vs T plot.

To see the effect of the dT/dt it would also be interesting to test different values of dT/dt.

RE: Capacitor Leakage Current over Temperature

genebren — Wed, 26 Jun 2019 15:18:15 GMT

Very interesting tests and data. I look forward to your further tests and conclusions.

Gene