[EE]: Interesting observations testing old capacitors from ~1973 Rotel RA-211 Amp

An interesting observation.

An old cap with these markning:-

    35V 1000uf CEO4W SP 33 85DegC

when charged from 32.4V via 1k resistor will quickly rise to 28v and almost
*stop* rising. Measured 29.85V to 30.32V in 1 minute.

Its measured parameters are 1112uF 1.2ohm ESR and 2%Vloss (whatever that is)

I am guessing, if I measured the rail voltage in the amp I would get about
28V.

Read on for rambly details.

I use an old Rotel RA-211 Amp in my kitchen and every few years a channel
will play up.

It is old so parts are easy and I like the challenge of surgically finding
the culprit and not changing parts unless they test faulty.  On this
occasion a 10uf between pre-amp and amp had failed.  From the outside it
looked fine..

However, this time, out of character, I thought I would remove and test
some caps that are clearly leaking brown stuff and down the rabbit hole I
went.

To my surprise they all test better than the nominal value.  It does not
matter which way I slice or dice them, they test better than new components
with the same specs. All the new ones I have measured both in my collection
and at Altronics store measure less than the nominal value (I think they
think I am nuts as I  randomly test their caps on the shelves)  . It seems
many manufacturers are using that tolerance thing to their advantage where
they test just within the range.

I use one of those cheap component testers that I had begun to distrust
because all new caps as mentioned measure less and the old ones measure
more, so I made a cap testing circuit with a PSU, CRO and
precision resistors.

I measure the time it takes to go from Vo to Vo/e which should be approx RC
and again the old caps all test better that the nominal value and the new
ones test less than  nominal value.

i.e for the 1000uf caps above I get a calculated RC of 10s but a measured
time of 14.3sec for both.

for a 100uF discharged thru 10k should take 1s to go from 48.4v to 17.65v
but these two take 1.2s and 1.5s.

Justin