A supercap dilemma

I was assembling some boards, and I noticed that one of my brand-new 10F, 2.7V supercaps was assembled wrong.
Normally the negative lead is shorter than the positive one. On this cap, the positive lead (as indicated by the case wrap) is the short one. So, which is it - is the case wrap correct, and the negative lead was cut long, or are the leads correct, and the case wrap is installed rotated 180° ?

Is there a simple test I can do that won’t destroy the cap?

No test I know of, but I’d go with the markings.

But, hey, if it’s wrong, you’re out, what, $2 or $3?

Yeah, but that’s 2oz of Starbucks…

I wonder if the negative lead is connected to the case.

I just checked - it’s not.

If you have a similar cap of known polarity, you could connect the known cap in series with the uncertain cap according to the markings and apply a small test voltage. Measure the voltage across each cap, if the markings are correct, you sould see the same voltage over each.

Even at a low voltage I would think the leakage current would be much lower when the capacitor is “correctly” connected.

Apply a small voltage to the cap, e.g. 100 mV. (To do this you can use a 1.5 V cell and resistor divider. Just keep in mind there will be a long RC time constant. Keep the resistance values low to keep RC low.) Measure current after it comes to a steady state. Discharge the cap and then repeat the test with the polarity reversed. If there is a big difference in the current between the two tests, the polarity with the lowest current is the correct polarity.

Another test - perhaps quicker:

  1. Discharge cap.
  2. Set DMM to measure resistance.
  3. Select one of the lower ranges. (Do not use auto ranging.)
  4. Connect DMM to cap. Monitor reading over the next few seconds.
  5. Discharge cap.
  6. Reverse polarity and repeat test.

The polarity with the highest resistance reading is probably the correct polarity.

Doing anything neat with the supercaps?

These are for a power controller circuit for a Pi Compute Module, for an embedded controller system I designed. There are (6) 10F caps in series, connected through diodes to 12v, and a micocontroller that sequences the power to the Compute Module. The µC enables 3.3v when the caps are charged, and then sends a power-fail signal to the CM when input power falls. Once the CM shuts down, the µC turns off 3.3v, and waits for power to be restored (there are actually a few more states, but this is the basics). This solves the “SD card corruption” problem that many Raspberry Pi controllers have.

One of the factoids I discovered when designing this circuit is that electrolytic capacitor lifetime is strongly dependent on operating voltage. So, even though 5 caps in series (5 x 2.7v = 13.5v) would have been sufficient, I used 6, because it more than trippled the expected lifetime of the caps. Instead of some complicated balancing circuit, I used 20K precision resistors in a ladder to equalize the voltage on each cap.

Nice. I am working on a similar project, albeit commercial. We have exactly the same problem. Instead of supercaps, which are expensive and won’t fit, we are using a more expensive kind of memory chip and a power safe file system. I am concerned this may not work…

Well, this is a commercial product, too…
I looked at using YAFFS, but stuck with EXT-4, since the shutdown will be controlled. What memory are you using?

ext4 with journaling on and none of the volumes are mounted except for one. A Sandisk emmc with an onboard controller.

So far there haven’t been any failures from this, at small (around 100 units) volumes. Product is planned for far larger volumes. One concern is the engineering staff asked for either supercaps or a shutdown signal from the host, we haven’t gotten either. (supercaps raise the per unit cost and there is not enough space allocated for them to even fit).

In theory since none of the volumes that are writable have executable files, and the state model of the application doesn’t depend on any data or preferences stored in the device, this will work.

What do the writeable volumes have on them? If the file system gets whacked, what will that do to your application?

Logs. We hope nothing.

I guess if the logs are expendable, than that system should work. But, I would test how the logging fails if that volume is corrupt.

I ended up contacting the manufacturer, and received some information that might be of interest to all the capacitor aficionados in doper-land: