Non polarised electrolytics

[jimmc101]

That made me stop and think, I've always assumed half capacity so I thought I better find out...

I grabbed two 220u 25V aluminium electrolytics (NOS from RS) and measured them on a Wayne Kerr B424 Component meter.
The meter measures capacity at 100Hz with less that 6.5mV AC and allows DC bias of up to 50V.
The results are shown below (Sorry I omitted to measure the ESR of the individual  capacitors.)
Just to clarify the caps were connected +ve to +ve, C1 normal / reversed refers to the polarity applied to the pair in series.
The B424 only does positive bias, the capacitors were swapped over to reverse the polarity.

   

Allowing for experimental error, it looks as if the combined capacity does not vary with bias and is half the individual value.
I would love to investigate further but don't have the time at the moment.


Jim

[Terry]

Jeffrey, could you resolve the problem empirically?

Finish the restoration but leave those unpolarised electrolytics alone - unless there is a very good reason for not doing so.

When you've got the set working, see how good or bad the convergence is, then change one of those caps at random with either 2 of the same value or double the value - your choice - then see what effect it has on the convergence. This should tell you if what you've just done is right or wrong.

Then carry on with the others in the light of what you've just learned - and let us know, of course!

[ppppenguin]

Jim, many thanks for doing the measurements. The results are very good to see. They support the theory of using a pair of double value caps.

Terry, I may well do just that. The caps on the convergence panel aren't in positions that will give me huge grief if faulty.

The only other bipolar cap in the set is the input luminance coupler on the video PCB. Here the value isn't critical, just needs to be big enough to have minimal signal across it over the video band. Simply replaced with a pair of double value caps in inverse series.

[Kalee20]

Interesting study from Jim!

What I'd have been intrigued to measure, is with the 50V applied DC bias, what were the individual voltages across the capacitors? It would need a really high resistance DVM, but I'd still guess that one would be seeing 25V in reverse, perhaps quickly diminishing to zero, while the other rises from 25V to 50V.

[jimmc101]

Some further thoughts.

It appears that the capacity is sensibly independent of voltage, including a small amount of reverse bias.

With the two capacitors in back to back in series, in the steady state,  the reverse biased capacitor only has to supply the leakage current of the correctly biased capacitor.
(Of course it also has to supply the initial charging current as well.)

Even if the reversed capacitor behaved as a good semiconductor diode (in parallel with a capacitor) it's slope resistance would be several hundred ohms assuming that the leakage current is low.
e.g. 50uA gives >500 ohm (26/I)
As this is much greater than the reactance (100uF @ 100Hz is 16ohm) it has little effect.

Extending this to the case where the back to back capacitors are subjected to significant AC, each capacitor will charge to the corresponding peak of the AC and only leakage current will flow thereafter.

I've simulated this using 'perfect' capacitors, schottky diodes to minimise reverse voltage and resistors  to crudely model leakage.
As can be seen both the capacitors are correctly biased over almost the whole of the cycle, polarity reversal occurring only at the peaks of the waveform.
It also shows the initial reverse bias of C1 as C2 charges.


   

My conclusion is that the overall capacity is always the simple series value.

I am concerned as to whether repeatedly passing the initial charging current through the reversed biased capacitor will lead to degradation in the long term. I think I would be tempted to use a diode in parallel with each capacitor to avoid this possibility. This may well increase the 'hardness' of any distortion at the peaks but feels safer to me.

Jim

[ppppenguin]

Jim, thanks for doing all the analysis.