Reforming capacitors (technical description, not politics)

Eric Smith spacewar at gmail.com
Fri Jul 31 16:41:45 CDT 2015


Time to reform is to a first approximation proportional to the rated
capacitance, rated voltage, and the length of time that the capacitor has
been unpowered. However, there is still a lot of variation. I've had a few
40-year-old capacitors (unused for at least 30 years) that took less than
two hours tp reform to 135% of rated voltage, but others of the same
ratings that took more than 48 hours.

I usually use a lab power supply with adustable current limiting. I adjust
the voltage in increments of 1V, and set the current limit to a different
computed limit for each voltage increment so that the maimum power (PS
current limit setting * PS voltage setting) doesn't exceed a value believed
to have no risk of damage to the capacitor.

(This is basically the PDP-1 capacitor reformation procedure devised by Bob
Lash, except the voltage increments were 0.5V.)

At each voltage increment, the leakage current starts out higher than the
PS current limit, so the actual applied voltage doesn't rise to the PS
voltage setting immediately. As the oxide grows, the leakage current goes
down, so the voltage rises. Once it hits the PS voltage setting, the
voltage stays the same, but the leakage current continues to drop. I leave
it there until the leakage current stops declining (no further reforming
occurring), then go to the next voltage step.

With the few capacitors that wouldn't reform, at some voltage setting the
leakage current leveled off at a value higher than the maximum rated
leakage current. In some cases the leakage  current declined some but then
started to rise. When thag happens I conclude that thr capacitor has a
fault and cannot be reformed.

I usually use an lab supply that can be controlled over HP-IB or serial,
and use software to automate the process. The software sends a text message
to my phone when the process has completed or failed, and logs a fair bit
of information to a text file.


More information about the cctalk mailing list