Electrolytics back-to-back => non polar capacitor

[AB2EZ]

A post in another thread recommended the use of a pair of electrolytic capacitors... back-to-back, to produce a non-polar capacitor.

I had never seen this... but it sounded very interesting... and useful.

After doing a little research, on line, about the physics and chemistry of electrolytic capacitors... I decided to do a simulation to see how two electrolytic capacitors... placed back to back would behave.

To add additional safety, I included a diode across each capacitor, pointing in the direction that is opposite the direction of the correct polarity of applied voltage.

The simulations are shown below. V(n001) is the voltage across C1 (from left to right), and V(n002) is the voltage across C2 (from right to left). The location of the ground is not relevant, except to provide a reference point for producing the voltage displays. The applied voltage source is a 100V (amplitude) 60Hz sine wave.

As predicted, the combination of the back-to-back capacitors, and the diodes acts like a single capacitor whose value is 0.5 x the capacitance of either of the two back-to-back capacitors. More important: neither capacitor experiences a voltage in the wrong polarity.

I've never tried this in practice... but it looks like a great idea for creating a non polar capacitor out of polarized electrolytic capacitors.

Stu

[WD5JKO]

Stu,

  Glad you did the simulation. As I said in one of those posts, with big electrolytic's, and audio frequencies, the diodes proved unnecessary as the slightly back-biased (if there is a DC component in the AC waveform) capacitor gets leaky right away preventing much dissipation in the unit. The capacitance seen in the circuit is more closer to the value on either capacitor, and not 1/2 of either.

   We used this concept commercially with a bridged pair of big Crown amplifiers, the M600, making 2000 watts and passing a PWM 7.5 Khz bipolar square wave (like +/- 80v peak) to a high voltage tansformer and multiplier stack making 40KV @ 50ma. The DC path was interrupted with the "AC" capacitor. Back in the early 1980's off the shelf HV Glassman, or Spellman power supplies did not exist, so using the big Crowns was attractive. The PDM duty cycle was way below 20% at no load, and about 80% at full load, so the series "AC" capacitor saw a DC component. They ran this way for years and years without failure. I once tried the diodes, and found them unnecessary. I think we used dual 10,000 MFD @ 100v, big computer grade jobbies from that era (Sprague 36DX rings a bell).

Edit: Also at the same company, analog op-amp circuitry needing an AC capacitor in the feedback circuits used dual polarized Tantalum caps in series with plus or minus sides together. In a high impedance circuit, a back biased tantalum capacitor was a short circuit. Worked like a hose.

Jim
WD5JKO

[WA1GFZ]

You don't need the diodes Stu. The leakage  of the correct polarity dominates the current through reverse cap  breaking down so there is no damage.

[AB2EZ]

Jim
Frank

Thanks for the comments.

There is a behavior indicated by the simulation that may just be an artifact of how I am modelling the capacitors... but please refer to the three sets of simulation results below.

In the first attachment, I have set the resistors R1 and R2 to each be 10000 Ohms. In the second attachment, I have set R1 and R2 to each be 1000 Ohms. In the 3rd attachment, I have set R1 and R2 to each be 100 Ohms.

What is shown is the current flowing in each of the resistors.

My thinking is that an electrolytic capacitor would (without any added diode across it) act like it has, in parallel with it, a diode in series with it's reverse leakage resistance.

Note that the charge that flows through each capacitor (the area of the current v. time in each current spike) in the reverse polarity direction, each cycle, is much larger when R1 and R2 are larger.

It is this flow of charge in the reverse direction that (I think) damages the oxide layer of an electrolytic capacitor).

Therefore, it would seem that putting a real diode across each capacitor would protect the oxide layers.

Stu