Dc blocking cap voltage

[bmaughan]

This is probably a stupid question so I apologize…
but what voltage rating would be best for a dc blocking cap coming off an opamp running 24v bipolar? (I’m grabbing a feed for a transformer direct out.)
I have some 63v caps around I was thinking about using.. I’d there any trade off for using too high of a voltage rated cap?

 

[JohnRoberts]

cost?

JR

 

[Newmarket]

All other parameters being equal then no disadvantage. But bear temp parameter in mind also. 105 Vs 85 Deg C also makes a big difference.
Check out datasheets that offer this info.

 

[JohnRoberts]

size?

JR

 

[scott2000]

Some Nichicon KZ with their big fat leads

 

[cyrano]

I remember reading once that electrolytics that are way overspecced in voltage range don't last as long. The reason would be the dielectricum needs appropriate voltage to rebuild after not being used for longer periods.

The same theory is used when reviving old analog gear that's been in the attic for decades. Powering up slowly at lesser voltage. That's why you use a variac. Except maybe for digital gear?

So don't use a 250V elco as a DC blocking cap if you can avoid it.

I'm not at all sure this is science, though. Dielectricum forming is not my specialty at all. What do you guys think?

 

[john12ax7]

Consider what the worst case voltage will be across the cap, then add a 1.5x to 2x margin.

 

[Monte McGuire]

Nichicon UES series usually work well for this application since they're bipolar. Digikey is marking them not recommended for new designs, but the Nichicon UES datasheet says that only those caps with 11 mm height cans are being discontinued.

 

[john12ax7]

Would be a shame if the UES line got discontinued. Maybe it's being replaced with an SMD version?

 

[Monte McGuire]

I'm not certain what the real story is, but the Nichicon datasheet is only NRND-ing the 11mm height parts, which in my mind are less desirable anyway. Now that you mention it, the smaller cans would be the ones you'd use for an SMD product though, so you could be right. I've always found UES to work well, and even though they are only rated for 1000 hours, they seem to be reliable and long lasting.

 

[scott2000]

Wonder what's up with those polymer electros?

 

[user 37518]

I remember reading once that electrolytics that are way overspecced in voltage range don't last as long. The reason would be the dielectricum needs appropriate voltage to rebuild after not being used for longer periods.

The same theory is used when reviving old analog gear that's been in the attic for decades. Powering up slowly at lesser voltage. That's why you use a variac. Except maybe for digital gear?

So don't use a 250V elco as a DC blocking cap if you can avoid it.

I'm not at all sure this is science, though. Dielectricum forming is not my specialty at all. What do you guys think?

Dielectricum sounds like latin, but yes when powering old caps it is best to start slowly, the dielectric is not the one being reformed but a thin oxide layer is restored, which is a dielectric. However, Ive never believed on cap reforming, its just best to replace them if they are too old. But definitely power up with a VariAC to avoid something else getting burned.

 

[cyrano]

I'm sorry, dielectricum is Dutch. I figured it would be the same in English...

the dielectric is not the one being reformed but a thin oxide layer is restored, which is a dielectric.

You've lost me here. If the oxide layer is a dielectric, then why is it not a dielectric? Are there several layers?

As I said, I'm not familiar with the chemistry involved in electrolytic caps.

 

[JohnRoberts]

Capacitors are funny when it comes to forming in... I recall having to deal with some old capacitor inventory while working at Peavey (Hartley didn't like to write down inventory). This was a bunch of large relatively high voltage caps that had been sitting for years. I decided the specialized handling to reform them in made trashing them a good option (I won that time)..

In a perhaps unrelated incident I had one new series of electrolytic capacitors that crashed and burned in production due to excessive leakage current. In phantom DC blocking application there was audible noise in the mic preamp inputs. I didn't bother to dig into the failure mechanism because there was a long list of alternative cap manufacturers begging for the buasiness.

With years of hindsight I can speculate that maybe the caps were not properly formed in by the manufacturer (manufacturers apply a small DC current to grow the oxide layer). Note: this is pure speculation on my part, that this was responsible for the input current/noise.

JR

 

[bmaughan]

Consider what the worst case voltage will be across the cap, then add a 1.5x to 2x margin.

Im a serious novice..so...how does one figure the worst case dc offset?

Nichicon UES series usually work well for this application since they're bipolar. Digikey is marking them not recommended for new designs, but the Nichicon UES datasheet says that only those caps with 11 mm height cans are being discontinued.

the ones I have are the UVP series. Would you advise against? What's to be gained from the "muse" UES series?

thanks for all the help!

 

[Monte McGuire]

Im a serious novice..so...how does one figure the worst case dc offset?

I think a more important consideration is that higher rated voltage capacitors will show fewer losses than those with a lower voltage rating. This is shown on the datasheets as "tan-theta" or the tangent of the loss angle, basically the real component of loss. This isn't quite ESR, but the combination of all dielectric losses. So, while the cap may never see more than a few volts of offset, a capacitor rated for 25V, 35V, or even 50V will probably perform a lot better regarding tan-theta, and thus should be more linear overall.

the ones I have are the UVP series. Would you advise against? What's to be gained from the "muse" UES series?

thanks for all the help!

A little while back, I made some APx 555 distortion measurements of a Nichicon UES1C221MPM1TD (220µF 16V) capacitor being driven at 2kHz with a +15dBV signal through a 100Ω resistor to the cap to ground, which is basically putting 56mA of 2kHz signal through the cap, and looking at the distortion voltage across the cap. In that test, there are no distortion voltage components greater than around -165dBV, the noise floor of the FFT used to measure the distortion residual. That's a lot of gobbledygook to digest, but the important point is that, compared with most other "normal" high quality electrolytics, this is an extremely low distortion level. Almost all other caps at the same drive level will show some distortion component, either 2nd or 3rd harmonic, but no harmonics were visible at all. For example, a 470µF Panasonic FR series (which is polarized, but driven with 22V DC bias) will show a tiny bit of 2nd harmonic and -140dBV of 3rd harmonic.

I have not measured a UVP cap under these conditions, but tests with dozens of other types of high quality caps almost always show some distortion component, either near residual or significantly higher. So, the point is that measurements seem to show that the UES series caps are measurably cleaner than most other types. I will make no claims of audibility, or whether this is significant, but if it's a simple choice and you have a part that measures better, I see no reason to not use it. Or, if you think this 'improvement' is insignificant, there's no reason to make a choice - use whatever you want and be happy

edit: re-reading, you say you already have some UVP, and those will probably work fairly well, so why not use them. A comparison will probably not show significant differences but if it comes time to buy more, it's worth trying the UES series. Leaning toward the higher voltage size and larger capacitance will also help, The drawback to a large capacitance cap will be increased leakage, which may be important for some circuits, so gross excess is probably not wise. But generally sticking to larger caps with larger voltage ratings will result in a cleaner cap. Setting any LF time constant to around 1Hz or so is "large enough" IMHO, and will help to reduce the signal voltage across the cap and make it misbehave less.

 

[abbey road d enfer]

Im a serious novice..so...how does one figure the worst case dc offset?

Worst case is the cap sees the full rail voltage. That usually happens when turning on or off the unit.
It may be very short, but the level of damage increases exponentially when exceeding rated voltage. And it's cumulative.
With current technology, I'm happy to use 35V-rated caps in my +/-17V rail circuits.
When it comes to distortion; the magic ingredient is using super large values. Distortion is proportional to the AC voltage across the capacitor, so doubling the nominal value decreases distortion by 6dB.
Typically I use 100uF caps with 10k loads, which give a LF -3dB point of 0.16Hz. To be compared with the old RoT of "calculate for -3dB, then multiply by 10", which results in about 9uF.
Thanks to modern technology, it doesn't result in huge bulky caps.

 

[abbey road d enfer]

I think a more important consideration is that higher rated voltage capacitors will show fewer losses than those with a lower voltage rating. This is shown on the datasheets as "tan-theta" or the tangent of the loss angle, basically the real component of loss.

I don't know how we could relate it to distortion, which is the main subject here. Since a significant part of losses is actual "real" resistance, I'm not sure it matters much.

A little while back, I made some APx 555 distortion measurements of a Nichicon UES1C221MPM1TD (220µF 16V) capacitor being driven at 2kHz with a +15dBV signal through a 100Ω resistor to the cap to ground, which is basically putting 56mA of 2kHz signal through the cap, and looking at the distortion voltage across the cap. In that test, there are no distortion voltage components greater than around -165dBV, the noise floor of the FFT used to measure the distortion residual. That's a lot of gobbledygook to digest, but the important point is that, compared with most other "normal" high quality electrolytics, this is an extremely low distortion level. Almost all other caps at the same drive level will show some distortion component, either 2nd or 3rd harmonic, but no harmonics were visible at all. For example, a 470µF Panasonic FR series (which is polarized, but driven with 22V DC bias) will show a tiny bit of 2nd harmonic and -140dBV of 3rd harmonic.

I think it would be interesting to compare caps of similar size. The Nichicon cap you have tested is quite big (dia 10mm H 20mm)compared to the 100uF 35V I favour, which measure dia 6.4mm H 12mm.
Actually height is not a concern, let's just concentrate on footprint.
Nichicon at same diameter is either 22uF 25V or 33uF 16V. For reasons I expressed before, I wouldn't go below in terms of rated voltage.
Or conversely compare the Nichicon with a 1000uF 25V cap of same size.
I don't have access to an AP anymore for this experiment, so maybe you would...? .

The drawback to a large capacitance cap will be increased leakage, which may be important for some circuits, so gross excess is probably not wise.

Agreed. I apply my oversized caps only in circuits that have very small DC.

But generally sticking to larger caps with larger voltage ratings will result in a cleaner cap.

+1

Setting any LF time constant to around 1Hz or so is "large enough" IMHO,

I support the idea that choosing a significantly lower cut-off (typically 0.1-0.2Hz) results in better performance whilst being economically viable.
That also helps deflating any criticism from purists of the dreaded VLF phase-shift.

and will help to reduce the signal voltage across the cap and make it misbehave less.

+1

 

[bmaughan]

this is all super helpful! Thank you for the detailed explanations.. definitely helping me get my head around it.

 

[john12ax7]

One thing to also consider is where the signal is coming and going. For example if you are using caps on an output you should consider the possibility that it may get plugged into a mic pre and may accidentally get phantom power applied. In that case you need to protect against +48V, so a 63V cap or higher might be appropriate.