I did a search on converting polarized electrolytic capacitors to non-polar, but get conflicting info.
The DIY FAQ states:
"You can simulate a non-polarized by using 2 electrolytic caps. Connect them together negative to negative and use the positive leads as the component leads; which is probably why someone notated it as +-||--||-+ Each of the capacitors should be double the value that you need for the circuit because of the series capacitance formula:Ctotal= 1 / (1/C1)+(1/C2)"
R.G. sez in a post:
"You can make a quick and dirty NP cap by tying together the negatives of two equal-sized polar caps.
In the series-NP connection, the capacitance value is funny. Normally caps in series are a smaller capacitance than either cap by itself. If you had two 3.3uF polyester caps, then the expected value for two of them in series is 1.65uF. However, electrolytic caps actually conduct in the reverse direction, so two 3.3uF polarized aluminum electrolytic caps act like they each have a diode in parallel with them that conducts when the voltage is backwards for that one cap. So two 3.3uF caps hooked up as series non polar (i.e. negative to negative) look like a single 3.3uF NP cap.
... except for tiny region near zero volts where they withstand a tiny reverse voltage, so they look like 1.65uF there..."
So will it cut the value in half putting 'em in series (negs connected) or remain the same value of each cap?
I am correct.
An aluminum electro cap conducts in the reverse direction, something like a diode, but with bad diode characteristics. The resistance of the side that is properly polarized is much much more than the reverse biased half, so the charge inside apportions itself across where it's blocked - the properly biased side. So only the properly biased side is acting as a cap, and there's only one cap's worth of capacitance there.
There is a small region around 0V where both are acting like caps but it's best avoided.
Even better, use my rule of thumb: never, ever rely on an aluminum or tantalum electrolytic cap to have a specific value. Use them only where you can make a case for "OK, if I just get a big enough capacitor..." and never where the exact capacitance matters. This is because even if the capacitance is correct today, it will change as the cap ages. And the tolerances on electros is generally bad. As bad as -20% +80%.
If you designed your circuit well, the electro caps are bigger than is absolutely needed, so make each one of the polarized caps be the value of the NP cap you want, and the circuit will work.
If you relied on the value of an NP cap for specific values, like a rolloff frequency - well, may Mother Nature have mercy on you.
Thank you. I have updated the FAQ
Wow R.G., a helluva big thanks for that quick response!!! Please don't be insulted or offended, but have you ever actually measured the results with a meter or is your statement purely theory? I just realized after posting I have a "CAP" mode on my tester and I'm getting the half reading with two 4.7uf tantalums..... or maybe my multi-meter tests at the "near zero" voltage you mentioned?
Like you said the tolerances are rather loose and it'll probably be fine either way, just trying to get to the bottom of it all.
Tantalum will withstand a reverse voltage of a few volts before the leakage comes up on the reverse biased side. So with tantalum the "funny area in the middle" is much bigger, maybe as much as 4-5V depending on the cap. Also, most DMM testers test with a small voltage across the cap so they give better answers when the caps are tested in-circuit. You're probably getting both effects.
In your conditions, I would expect just the results you're getting.
Thanks again. I'm getting the same results with electrolytics, but could just because I'm testing the caps out of circuit, as you mentioned. What I'll do is go bigger in value and then test again when it's powered up in the circuit.
I'm trying to get a non-polar 2.2uf that will be used to connect a 4 diode bridge arrangement(?) to vbias. It's for the Moosapotamus A/DA Flanger clone.
For those cases I need an NP electro, I usually use a pair, reverse, with diodes across each to pass the reversed voltage.
usually a 1n4007 ( well you can use a lower number as long as the voltage isn't exceeded, but the 1n4007 is CHEAP and
that way I only carry one item in my already overflowing junque box ;).
chuck
If all you need is 2.2uF NP then why not simple use a 2.2uF Mylar or metal Film or simular Cap?? or even two 1uF caps in Paralell??
Cheers
Quote from: Minion on March 09, 2007, 01:40:22 PM
If all you need is 2.2uF NP then why not simple use a 2.2uF Mylar or metal Film or simular Cap?? or even two 1uF caps in Parallel??
In theory, you are quite right, but there is a point at which a pair of back-to-back polarized caps will be both much smaller and much cheaper than their non-polarized functional equivalent. For instance, I can recall a Boss schematic where the LFO has a pair of 33uf caps back to back. The odds of ANY nonpolarized component/s taking up anything less than 3-4 times the equivalent space is small and the odds of it being anything less than 3-4 times the price of two electrolytics is similarly small.
In the other hand, very often you CAN get a couple .27uf NPs to stick in parallel (or a .27 and .33) for only a wee bit more than the space and cost of a series pair of 1uf polarized. It all depends on which value in particular. There are certain minimum usable sizes when it comes to through-hole parts. Some components aren't much bigger than that, and some need to be a lot bigger.
Quote from: R.G. on January 26, 2007, 07:40:35 PM
I am correct.
This should be in the header at the top of the page:
"I am correct." - R.G. Keen
This one's always an interesting topic. It all depends if you have enough bias to cause the "diodes" RG speaks about to conduct.
From Cornell Dubilier,
http://wiki.ece.rose-hulman.edu/herniter/images/0/03/Aluminum-Electrolytic_Capacitor_Application_Guide.pdf
First page,
"If two, same-value, aluminum electrolytic capacitors
are connected in series, back-to-back with the positive
terminals or the negative terminals connected, the
resulting single capacitor is a non-polar capacitor
equal in capacitance to either of the original pair. The
two capacitors rectify the applied voltage and act as if
they had been bypassed by diodes. When voltage is
applied, the correct-polarity capacitor gets the full
voltage. However, on a capacitor meter with no bias
voltage the two capacitors measure half capacitance as
you expect from capacitors in series. In non-polar alu-
minum electrolytic capacitors and motor-start alu-
minum electrolytic capacitors a second anode foil sub-
stitutes for the cathode foil to achieve a non-polar
capacitor in a single case."
FWIW, even that's not 100% right as some capacitance meters (not LCR meters) will have a permanent bias.
Wow I have never considered this before! I shall have to do some measurements to satisfy my curiosity.
QuoteI am correct.
Extraordinary claims call for something something!
I've just run some frequency response tests and so far I have not been able to produce the effect RG or Cornell describe. Test circuit shown below.
Whether in series or shunt, I always get a consistent cut-off frequency. Small signal (0.08Vrms input) or large signal (8Vrms input), whatever the bias voltage (0V to 30Vdc), the effective value of the back-to-back caps is always half the value of one. This was using a pair of 22uF 100V caps, a pair 33uF 40V, and a pair of 1uF 63V.
As I see it, there is always one cap that is correctly biased and therefore blocks the DC current to a negligible leakage value. So we always have a pair of identical caps and a pair of identical parasitic diodes in anti-series, so the cap-combo is always symmetrical for
AC , and the total signal voltage is always shared equally across both cap impedances. I get the same result if I put actual diodes across both caps, too.
(https://i.postimg.cc/14nqvQpf/dfb.jpg) (https://postimg.cc/14nqvQpf)
Probably an even better way to look at this is to not use electrolytic caps where a 2:1 change in value would make a difference in your circuit.
All electro caps change value as they age, and have wild tolerances if not pre-selected for a given value. So it's best to avoid using electros, polarized or not, in filters or where cutoff frequencies really matter.
Historically, there was a time when a circuit might have an impedance level that required 1uF to 10uF caps to set a rolloff, and you just had to use an electro. However, with today's box film caps available in uf range values in small sizes and affordable prices, why not just sidestep the issue?
QuoteI've just run some frequency response tests and so far I have not been able to produce the effect RG or Cornell describe. Test circuit shown below.
That's for putting up your results. Very interesting. There's probably more to this than just one answer or the other.
Cyril Bateman had some stuff about the diodes.
https://linearaudio.nl/cyril-batemans-capacitor-sound-articles
I don't think there's any doubt the diodes can kick in. I guess the question of when they do and when they don't. More the point when they do and don't with back to back caps. I assumed they would kick in when the voltage across the *reversed* cap got high enough. Not unlike exceeding the forward voltage of a diode.
You gotta love this stuff sometimes.
Quote from: R.G. on April 14, 2021, 09:58:32 AM
why not just sidestep the issue?
Because the OP deserves the right answer to his question.
I'm sure everyone has seen coupling caps/DC blocking cap in cases like C1 and C3 of power amps,
https://sound-au.com/project3a.htm
For the most part we are talking zero bias on power amp side and on the preamp side due to bipolar power supplies.
I've always worked of the assumption this practice fine provided the voltage across the caps was kept small. That's usually the case for a power amp since the cut-off frequency set by the caps is very low. In later years I have seen some amps using back to back diodes across C3 presumably to prevent large reverse bias.
The point is there are cases where an electrolytic will handle AC. I don't recall every seeing any any rectifying behaviour. Now is that because the cap is so large we just don't notice any changes in behaviour.
I've done some work on electrolytes and passing current through salt solutions. The changes take time so you may need some persistent DC to expose any issues.
------------------------
EDIT:
See page 6. Fig 2.2
https://www.nmr.mgh.harvard.edu/~reese/electrolytics/tec2.pdf
At -4V the leakage is quite small. It's not what you would call a strongly conducting diode. At -250V I should imagine a different picture!
Quote from: R.G. on April 14, 2021, 09:58:32 AM... it's best to avoid using electros, polarized or not, in filters or where cutoff frequencies really matter....
True. But we had a case yesterday where someone had a common LFO triangle generator and wanted to replace the stock two back-to-back 15uFd polars with a single bipolar to save space. What value?
Obviously in many cases you can, if needed, raise the associated resistances to get back on the mark. But this already had a large value pot and higher could be awkward.
Worth noting the different uses. A coupling cap normally has very-very-small voltage at audio frequencies. OTOH an LFO timing cap has very large voltages, half the rail or so. The LFO is not as sensitive to distortion as aucio coupling can be, but a glitchy LFO waveform stands-out.
Quote from: bluesdevil on January 26, 2007, 08:30:40 PM
I'm trying to get a non-polar 2.2uf that will be used to connect a 4 diode bridge arrangement(?) to vbias. It's for the Moosapotamus A/DA Flanger clone.
Not sure if the OP is still paying attention, but another option here is to connect the cap to ground instead of vbias. I don't have a copy of the schematic at hand, but assuming the signal is swinging about Vbias, so a polarised cap should be fine in this case (with - to ground).
Just tried the following on my Electric Mistress clone. It has the usual triangle wave LFO, with 2 back-to-back 10uF caps.
I don't know what that capacitance is equivalent to, either 10uF or 5uF, right? Let's call it X.
I measured the LFO period and got a time of 0.38 seconds.
I happen to have a 2.2uF NP electrolytic handy from an old build, so I added that in parallel with 2x10uF that's already in the circuit. The LFO period increased to 0.55 seconds.
If the ratio of periods is proportional to the ratio of capacitances then we get
( X + 2.2uF ) / X = 0.55 / 0.38 = 1.447
That gives X = 4.91 uF.
So to me it seems those back-to-back 10uF caps act like about 5uF in the LFO application, not 10uF.
Would be interested to see what anyone else finds if they can repeat this sort of test.
EDIT: I should add that the NP was not a fresh unused cap, and I am just going by the 2.2uF value written on it. I don't know if its age (a few years) could have caused its capacitance to increase by a factor of 2, in which case X could be closer to 10uF.
EDIT2: The caps in the Electric Mistress are several years old too, so maybe they lost capacitance over time.
Would like to try this with some fresh caps but that old 2.2uF NP is all I have.
QuoteWould be interested to see what anyone else finds if they can repeat this sort of test.
The LFO on the Boss BF2 Flanger uses the back to back cap arrangement . The timing works out
that the two back to back 33uF caps acts like the two caps are in series ie C = 33uF/2 = 16.5uF.
That's consistent with your results and Merlins.
What I was getting at in my last post is electrolytics in reverse still work like a cap, and look like the marked value. The difference is they leak more as the voltage is increases (see graph). I suspect if you put a *single* electrolytic cap in the LFO integrator it would work, at least at low voltages. The leakage would cause the timing to be extended when the cap is in the reverse direction. When the leakage is the same order of magnitude as the circuit current it will stop working as the cap never charges.
When we put two electrolytics caps in back to back series the second cap blocks the leakage so the circuit doesn't misbehave.
From what I can see the point between the caps should develop a DC voltage due to the leakages (the diodes in the model). A DC voltage which biases each cap with the correct polarity.
Here's a crude sim which shows that for steady state conditions the back to back series Electrolytics
has a value of C/2.
Under steady state conduction a DC bias forms at the junction point of the two caps which correctly biases
the cap and this makes the combined series cap look like a cap of value C/2.
However, as you can see at the start of the (blue) waveform under transient conditions the waveform does not
rise correctly. In this example the rise is more sluggish because we are charging one cap of value C though
the diode. It's only when the leakage is reduced to a small value do we see the rise of the waveform match the
reference circuit with cap value C/2.
On a real cap if you keep the voltage across the cap low the leakage will be low. That follows why we get away
with single electrolytic caps with AC waveforms (see earlier power amp post).
Here's the waveforms and test set-up. For RL=10k there is a DC voltage on the tap point. The 10MEG case will also develop a DC bias but it will take a much longer time. When it does the combined cap will still look like C/2.
The diode and RL resistors are not added they simulate the Electrolytic cap. The resistor RL sets how much leakage you get when you reverse the polarity of the electrolytic cap. See link and comments at the bottom of Reply #16 of this thread.
[when the page opens click the image again to get a better resolution pic]
(https://i.postimg.cc/Rq6KRtqQ/Back-to-Back-Electrolytic-Capacitors-V10.png) (https://postimg.cc/Rq6KRtqQ)