Reforming electrolytic capacitors (was Re: Sony CPD-1302 monitor -- IIGS cable?)

[Eric Smith <eric@brouhaha.com>]

"Michael J. Mahon" <mjmahon@aol.com> writes:
> The typical failure mode of electrolytics is progressive loss of
> capacitance, so there is no catastrophic effect of this failure,
> just a progressive loss of performance.

The loss of capacitance is just the beginning.  The oxide layer
(dielectric) breaks down, reducing the effective voltage rating
of the capacitor as well.  Eventually it gets to the point that a
voltage well below the rating will punch through the remaining
dielectric, making it (briefly) act like a dead short.

This basically happens when the electrolytic sits unused for a long
time.  The time varies depending on the exact construction and materials
of the capacitor, but usually takes longer than five years of non-use.

Or if they are used, but at less than the rated voltage, the
oxide thickness drops but they remain useful at the lower voltage.

The standard way to fix electrolytics that have degraded (but not
shorted or burst) is to put them on a current-limited DC supply
to "reform" the oxide layer.  This is basically a repeat of part
of the process that was done to initially create the oxide layer
at the factory.

On the PDP-1 Restoration Project at the Computer History Museum,
one of our first tasks was to reform all of the electrolytic
capacitors in the many power supplies of the PDP-1.  We used a
very overengineered approach which used a programmable power supply.
The software set an initial current limit, then set the voltage
to 0.5V, and waited until the supply reached 0.5V.  If the capacitor
is good, that happens quickly, but otherwise it takes a while with
the power supply in current-limiting before it charges the 0.5V
and the power supply enters constant voltage operation.

Then we set a new current limit and set the voltage to 1.0V, and
again wait for the cap to charge to 1.0V.  This process was repeated
until the rated voltage of the capacitor was reached.

The current limit for each voltage setting was determined based
on the maximum power dissipation that we expected the capacitor
to handle.  So at higher voltages, the current limit was lower.

We only had a few capacitors that could not be reformed (after about
eight years of storage).  Many were however degraded, and the
reformation process was in fact necessary to bring them back to
spec.

If we had just turned on the PDP-1 without reforming the capacitors,
some of them probably would have been destroyed.

The software I wrote to control the reformation process is open
source called "WREC" (for "WREC Reforms Electrolytic Capacitors"),
and is available here:

	http://www.brouhaha.com/~eric/software/wrec/

I am not aware of anyone other than the PDP-1 Restoration Team
actually using it though.  The normal approach is just to use
a current-limited DC supply at a single current setting, and
monitor the voltage.  None of the step-and-wait stuff is
actually necessary.

Eric