I used a pair of 10uF polar ones in series to get about 5uF. I tied the (-) ends together and this works fine. This replaced a 4.7uF nonpolar (also called bipolar) cap in a DC nulling servo used in a Yamaha C-70 preamp. The cap served as the feedback cap in the integrator circuit. Old cap leaked and the integrator would not zero out the DC offset. Offset is now okay, at about 0.1mVdc. A film cap would be best in this position, but I could not fit one in there because of space limitations. It's important that leakage current be similar in both caps so that DC voltage divides evenly between them when sizable DC is present.
I'm no electrolytic capacitor expert, but I can't think of any reason why it wouldn't. I mean, when the formed electrode in a normal electrolytic starts to leak excessively when there is no voltage across it for a very long time, why wouldn't the same happen with the two formed electrodes in a bipolar electrolytic when there is no voltage across them for a very long time?
As far as I know, an electrode that has not been formed can still handle about 1.5 V to 2 V, so maybe you can get away with it when the voltage across the capacitor always stays below 1.5 V in normal use.
As far as I know, an electrode that has not been formed can still handle about 1.5 V to 2 V, so maybe you can get away with it when the voltage across the capacitor always stays below 1.5 V in normal use.
Do a search for Cyril Bateman Capacitors Sound: he's done extensive and definitive tests on this topic.
The short version, in his conclusion:
"Having measured a considerable number of aluminium electrolytics using test voltages from 0.1 volt to 3 volt, with and without
bias, a single Bi-polar type produced lower distortion than larger, more expensive, specialist polar capacitors..
Much better results were obtained by connecting two double capacitance value Bi-polar electrolytics in series. Using 1 volt or
smaller test voltages and no bias, distortions for a double Bi-polar and the metallised PET assembly were similar.
With increasing bias or with increasing test voltage, the metallised PET assembly produced less distortion than any electrolytic
I tested."
The short version, in his conclusion:
"Having measured a considerable number of aluminium electrolytics using test voltages from 0.1 volt to 3 volt, with and without
bias, a single Bi-polar type produced lower distortion than larger, more expensive, specialist polar capacitors..
Much better results were obtained by connecting two double capacitance value Bi-polar electrolytics in series. Using 1 volt or
smaller test voltages and no bias, distortions for a double Bi-polar and the metallised PET assembly were similar.
With increasing bias or with increasing test voltage, the metallised PET assembly produced less distortion than any electrolytic
I tested."
Each capacitor of a bipolar capacitor (I am talking about the two main capacitors, not the additional parasitic ones in series with those) behaves as an electrolytic diode junction when its voltage rating is exceeded.
The forward voltage is very low, like 1 ~2V, meaning that all the rest will be applied to the reverse-connected one, helping reforming it.
Not a very clean process, meaning a non-linear current, at least when the new voltage is first applied, which partly explains the non-linear behavior of these components.
The forward voltage is very low, like 1 ~2V, meaning that all the rest will be applied to the reverse-connected one, helping reforming it.
Not a very clean process, meaning a non-linear current, at least when the new voltage is first applied, which partly explains the non-linear behavior of these components.
I tried two 5600 uf caps back to back on the speaker output of a .02% HD PA amp. At about 1/4 watt it sounded funny. Direct transistor outputs sound better, however dangerous that is to your speaker. Test track was top octave solo grand piano, Peter Nero Young & Warm & Wonderful.
Single supply amp with a polar electrolytic biased to 1/2 rail voltage at the output does not sound funny at the same wattage.
Last edited:
There is a absence of information on what happens to bipolars in use or in storage. I cannot see why bipolars should behave differently. I am going to try to measure this, with some old polars that have been in a drawer for 20 years. Connect back to back and apply a large AC signal to simulate what happens in a speaker crossover.
Back to backs have four "junctions" compared with two in a true bipolar part. This may have some importance.
I tried two 33uF/25V parts that had been unused for 25 years, horribly tarnished and hard to solder leads. Both were almost exactly on value.
I applied 20V peak to peak at 10Hz via 300R triangle wave from a function generator. No visible current waveform distortion. After 5 minutes the capacitance had reduced 1%, something I often see on re-forming. Leakage was less than 1uA.
The centre tap measured about 1v offset with a 10M scope probe, to be expected with a leakage mismatch.
I tried two 33uF/25V parts that had been unused for 25 years, horribly tarnished and hard to solder leads. Both were almost exactly on value.
I applied 20V peak to peak at 10Hz via 300R triangle wave from a function generator. No visible current waveform distortion. After 5 minutes the capacitance had reduced 1%, something I often see on re-forming. Leakage was less than 1uA.
The centre tap measured about 1v offset with a 10M scope probe, to be expected with a leakage mismatch.
Hi, a newbie and apologies for waking this old thread:
there seems to be some differences of opinion on this topic, does this sum it up?
... and do the ??? circuits work? (sorry about the character pics):
--|(-)|-- OK 10+10 => 5
--)|-|(-- OK 10+10 => 5
--|(-)|-|(-- ??? 10+10+10 => 3.333 ???
. !--|(--!
--| |-- OK 10+10 => 10
. !--)|--!
. !--|(--! !--|(--! !--|(--!
--| |----| |----| |-- OR:
. !--)|--! !--)|--! !--)|--!
. !--|(--! !--)|--! !--|(--!
--| |----| |----| |-- 10+10+10 => 3.333 ???
. !--)|--! !--|(--! !--)|--!
thanks - a.f.d
there seems to be some differences of opinion on this topic, does this sum it up?
... and do the ??? circuits work? (sorry about the character pics):
--|(-)|-- OK 10+10 => 5
--)|-|(-- OK 10+10 => 5
--|(-)|-|(-- ??? 10+10+10 => 3.333 ???
. !--|(--!
--| |-- OK 10+10 => 10
. !--)|--!
. !--|(--! !--|(--! !--|(--!
--| |----| |----| |-- OR:
. !--)|--! !--)|--! !--)|--!
. !--|(--! !--)|--! !--|(--!
--| |----| |----| |-- 10+10+10 => 3.333 ???
. !--)|--! !--|(--! !--)|--!
thanks - a.f.d
When placing caps in series, it's important that any voltage across the combination splits up evenly between the two caps. If one cap hogs the voltage due to unequal leakage currents in the caps, then an over-voltage condition can occur in the cap with the lower leakage of the two caps. Placing equal, usually large-valued resistors across both caps to swamp the leakage and force equal DC voltage sharing is the usual way to accomplish that.
Many thanks for the replies:
@MarcelvdG - Yes 20uF is what I would have expected but I read a couple of posts on a different forum that suggested that each cap was only 'working' half the time. I'm don't understand your point about the maximum voltage - most posts I've seen about this suggest either the same or double the voltage of a single cap?
Anyway I tried it out as far as I could (these are 33uF 35v caps):
I am also keen to know whether the polarity 'solution' works for 3 caps in series (my picture3):
I ask this because I have a speaker crossover that uses series 2 x 100UF => 50uF and I want to reduce it to 33uF.
As regards voltage, if I put a multimeter across the output of my little LC Meter it fluctuates 5.8 - 7.1v on AC and 2.5 - 4v on DC.
I haven't tried higher voltage yet and don't really want to waste caps <g>.
thanks - a.f.d
@MarcelvdG - Yes 20uF is what I would have expected but I read a couple of posts on a different forum that suggested that each cap was only 'working' half the time. I'm don't understand your point about the maximum voltage - most posts I've seen about this suggest either the same or double the voltage of a single cap?
Anyway I tried it out as far as I could (these are 33uF 35v caps):
I am also keen to know whether the polarity 'solution' works for 3 caps in series (my picture3):
I ask this because I have a speaker crossover that uses series 2 x 100UF => 50uF and I want to reduce it to 33uF.
As regards voltage, if I put a multimeter across the output of my little LC Meter it fluctuates 5.8 - 7.1v on AC and 2.5 - 4v on DC.
I haven't tried higher voltage yet and don't really want to waste caps <g>.
thanks - a.f.d
For me all circuits haven't a good design as long big capacitors in non polar version are necessary. An exception is this DC Filter under
https://www.scintilla-buizenverster...erkers/zelfbouw/zelfbouw sites/DC-Blocker.htm
where each elcap have it's own diode.
A normal (polarized) electrolytic capacitor also works as a (poor) kind of diode: when you apply more than 1.5 V ... 2 V in the wrong direction, it starts to conduct DC. When you connect two of them in antiseries with their minus poles tied together, the diode behaviour will ensure that the minus poles will (eventually) end up at or close to the most negative potential. (This means that each of the capacitors is subjected to voltages up to the peak value of the applied AC voltage across the pair.) From that moment onward, neither of the electrolytic capacitors is driven into conduction, so they then behave as a normal pair of series-connected capacitors.
Connect two electrolytic capacitors in antiparallel, and there is always one going into conduction when the voltage exceeds 2 V or so. Hence the poor voltage handling.
I think it should work, as long as the peak AC voltage is less than the DC voltage rating of each single capacitor. It is better to use a proper non-polarized capacitor meant for loudspeaker crossovers, though (usually more accurate, relatively small ESR, and no rectification effects, not even at the start of a loud musical passage).
Last edited: