Electrolytic Capacitor Voltage Rating

I remember reading a technical paper regarding applied voltage for an electrolytic, but can't seem to find it now. Essentially the claim was that it was detrimental to use too large of an overrated capacitor voltage.

Meaning for example you would target 1.5-2x the necessary voltage.  For a 15V supply a 25V cap is better than a 35V or 50V.

I'm wondering if this was perhaps an old vs new technology thing  or if this is still true today.  Certainly there after cost,  size,  ripple current implications,  but what about filter performance?

john12ax7 said:

I remember reading a technical paper regarding applied voltage for an electrolytic, but can't seem to find it now. Essentially the claim was that it was detrimental to use too large of an overrated capacitor voltage.

Click to expand...

it is really detrimental to use not enough rated voltage.

Meaning for example you would target 1.5-2x the necessary voltage.  For a 15V supply a 25V cap is better than a 35V or 50V.

Click to expand...

not that important... it cost more to use more than needed but in large scale production you generally standardize on a handful of values.

I'm wondering if this was perhaps an old vs new technology thing  or if this is still true today.  Certainly there after cost,  size,  ripple current implications,  but what about filter performance?

Click to expand...

Generally don't use electrolytic for audio filters.

JR

I was meaning filtering in terms of power supply.

But that does bring up another point,  when an electrolytic is necessary to block DC in the signal path. I tend to overate these to account for worst case scenarios,  while others use  low voltage caps.

I can think of two places where audio signal goes through an electrolytic capacitor. One is in power amplifiers, the output going to the negative input (transistor base) of the differential pair - it's a voltage divider, say 10k from output to the negative input, then 470 from there that WOULD go to ground, but that would mean both the signal and offset would be amplified by 20. There's usually an electrolytic capacitor in series with the 470, so that at sub-audio and DC the gain is 1 and so the output offset is much lower. This electrolytic often goes bad because there is very little or no DC on it. The solution to having this electrolytic in the circuit is to use a servo (basically an op-amp sub-audio-frequency lowpass filter) to zero out the output offset.

The other is in a Class  A power tube output such as 6L6 or 50C5 with the common cathode bias resistor. Putting an electrolytic capacitor across this resistor brings the tube up to "full" gain. If the output sounds tinny and lacking bass, it's usually because this capacitor has given up and lots most of its capacitance. The only solution I know is replace it with a new electrolytic capacitor ...

Current leakage in electrolytics is higher with higher rated voltage. Look at the CV value in a datasheet which is capacitance in uF * rated voltage * some CV factor which is usually like 0.01-0.02. And V is the rated voltage, not applied. So leakage will be greater if you use a larger rated part regardless of applied voltage.

For example, leakage for a 16V part would be something like:

100uF * 16V * 0.01 = 16uA

whereas a 100V part it would be:

100uF * 100V * 0.01 = 100uA

So does 100uA vs 16uA leakage matter? Maybe. It might necessitate a smaller drain resistor on an output or cause noise with a pot or switch or shift the bias in a high gain circuit. But for most things, no it won't matter.

john12ax7 said:

I was meaning filtering in terms of power supply.

But that does bring up another point,  when an electrolytic is necessary to block DC in the signal path. I tend to overate these to account for worst case scenarios,  while others use  low voltage caps.

Click to expand...

I have commented on this at length over the decades. Yes, it is often most practical to use electrolytic capacitors in DC blocking applications, where these capacitors form a single pole HPF. My strategy after thinking about it for years was to set these HPF poles octaves below the actual passband , then use one film capacitor HPF to establish the dominant LF cut off. Most non-ideal capacitor phenomenon are related to changing terminal voltage (voltage coefficient, etc) so if the terminal voltage doesn't change inside the audio band , the cap should not cause audible shenanigans.
=====
Back in the 70s I used a capacitor in series with a modest value resistor to establish the LF cut off for a typical phono preamp gain stage. After inspecting this closely on my bench I determined that the electrolytic cap in that circuit node while executing a HPF at roughly 20 Hz, it was also contributing tens of degrees of phase shift at 20 kHz (due to series inductance). I will not veer off into thoughts about the audibility of 20 kHz phase shift, only to state that it is not the ideal transfer function of that circuit with those values.  (BTW modern electrolytic caps have lower ESL and ESR so would likely measure differently on my circa 70's test bench).
===
Another modern audio myth is that servos magically eliminate capacitors from the audio path. BZZT they don't, but they allow higher quality film capacitors (in combination with op amp gain) to replace larger lower quality electrolytic caps. 
===
In decades of manufacturing I have only once experienced one leakage problem from new (good?) electrolytic capacitors. These were used as phantom voltage DC blocking caps in a mixer mic preamp front end. The leakage was so noisy and apparent the line QA person notified my design engineers. We isolated the problem to a new series of capacitor that the cap manufacturer was proud of but did not cut it for real world use. At the time I just gonged the cap and disapproved it for use in our system without digging deeper. In hindsight decades later, it may have been an issue with not being properly formed in by the capacitor's original manufacturer.

Leakage can be empirically measured with modern VOM, I have not personally measured anything on the bench close to worst case data sheet numbers, but we must design to still work with worst case components.

JR 

this could be a good subject for a master thesis at some graduate school somewhere, if it has not been written up already,  you need an obscure subject for a thesis so as to impress the teacher with your original idea and not raise any questions about originality,
you might even get a patent out of it to help pay off citibank,

i would start my thesis with the following abstract:

"Dielectric relaxation is the momentary delay (or lag) in the dielectric constant of a material. This is usually caused by the delay in molecular polarization with respect to a changing electric field in a dielectric medium (e.g., inside capacitors or between two large conducting surfaces). Dielectric relaxation in changing electric fields could be considered analogous to hysteresis in changing magnetic fields (e.g., in inductor or transformer cores). Relaxation in general is a delay or lag in the response of a linear system, and therefore dielectric relaxation is measured relative to the expected linear steady state (equilibrium) dielectric values. The time lag between electrical field and polarization implies an irreversible degradation of Gibbs free energy."

CJ said:

this could be a good subject for a master thesis at some graduate school somewhere, if it has not been written up already,  you need an obscure subject for a thesis so as to impress the teacher with your original idea and not raise any questions about originality,
you might even get a patent out of it to help pay off citibank,

i would start my thesis with the following abstract:

"Dielectric relaxation is the momentary delay (or lag) in the dielectric constant of a material. This is usually caused by the delay in molecular polarization with respect to a changing electric field in a dielectric medium (e.g., inside capacitors or between two large conducting surfaces). Dielectric relaxation in changing electric fields could be considered analogous to hysteresis in changing magnetic fields (e.g., in inductor or transformer cores). Relaxation in general is a delay or lag in the response of a linear system, and therefore dielectric relaxation is measured relative to the expected linear steady state (equilibrium) dielectric values. The time lag between electrical field and polarization implies an irreversible degradation of Gibbs free energy."

Click to expand...

That sounds a little like DA (dielectric absorption). Kind of like magnetic domains in an inductor, capacitors have polarized molecules that don't immediately relax to a neutral (0V) state, when driving voltage is removed.

JR

john12ax7 said:

I remember reading a technical paper regarding applied voltage for an electrolytic, but can't seem to find it now. Essentially the claim was that it was detrimental to use too large of an overrated capacitor voltage.

Meaning for example you would target 1.5-2x the necessary voltage.  For a 15V supply a 25V cap is better than a 35V or 50V.

I'm wondering if this was perhaps an old vs new technology thing  or if this is still true today.  Certainly there after cost,  size,  ripple current implications,  but what about filter performance?

Click to expand...

It seems that, for a given capacitance value, there is a variation of ESR according to voltage, with an optimum between 25-50V for low-voltage caps (<100V).  However the variation is about 1:1.5, which should not be significant in a well designed circuit. There is certainly as much or even higher difference between brands and lines.