Cap Voltage - How Near the Limit is Safe?

[thermionic]

Hi,

I need a 24v PSU.  There is an excess of 25v caps in the workshop. Providing I accurately set the voltage to 24, where's the problem as rail bypass caps? Usually I'd have more of a margin to be safe and use 35v caps. Is this too close for comfort in your mind?

TIA.

Justin

 

[JohnRoberts]

If the power supply is regulated to 24V before it hits that capacitor, then you should have a 1V margin. If an unregulated supply you must add some fraction of the Vp-p ripple voltage on the cap, on top of the average DC voltage, then the expected mains high line voltage, then... etc. Cap voltage ratings are for peak voltage not average. 

In general prudent capacitor manufacturer's design in some small amount of headroom above their rated tolerance. Prudent engineers also design in some safety margin on top of that. Less prudent engineers (I've known at least one), use the safety margin designed in by others to save a few cents on a BOM (capacitor voltage can be significant marginal cost factor in power amp designs). I have seen products with poor reliability (from one brand of capacitor) because of such corner cutting. Apparently the engineer at the cap company had a similar philosophy of using the other engineer's safety margin.  : I would like to think that engineer learned his lesson, but knowing him he probably just avoided that cap manufacturer.

JR

 

[thermionic]

Thanks, John.

If a 1v margin is considered ok by an experienced engineer such as yourself, I'm sure it'll be fine. Usually I would use a 35v part, but I have space concerns, as well as several hundred Panasonic FC rated @ 25v.

As far as the unregulated portion is concerned, that's also pushing it slightly - the secondaries are rated @ 25v, and I've earmarked 35v caps... 25 x 1.4 = 35... Let's hope the mains voltage doesn't fluctuate too much where this device is used! The circuit being powered is only a 2-opamp EQ, so it won't be loading the PSU much (each rail on an opamp pulls around 30mA).

Justin

 

[PRR]

If you need it to work for a year, near a soldering iron, do it.

I would.

I have "35V" caps holding 36V for decades of few-hour-per-week work.

It's not a hard line: perpetual at 24.9V and BOOM at 25.1V. They are somewhat less sensitive than incandescent lamps, and you would not fret to use a 25V lamp on 24V. (But you usually socket an incandescent and plan to replace it several times.)

Life declines, and sudden death rises, above rating; improves below rating. And actual rated life is shockingly low, like 1,000 hours or 5 weeks 24/7 (however at full temperature and voltage). If you need to be trouble-free for many years, the wiser path is to go 35V and maybe 50V on a 24V rail. You can't, as some mil-spec reliability calculators say, use 450V at 2V and pretend very-very-very long life (in fact older hi-volt caps could rot fast at low voltage). But rating 1.5 to 2 times applied voltage is reasonable for professional audio in DIY economics. 1.04 over-rating is fine for self-maintained, a bit cheap when others will use the gear.

 

[bcarso]

Who is the manufacturer?

I think I related my war story in here already, when some dirt-floor vendor's caps got into a powered speaker that I'd designed based on my experience with good Rubycon parts.  The junk caps were rated on the sleeve for 16V.  I was running at about 17, which was regulated in a manner of speaking.  But it was risky with good caps and in a fair number of cases catastrophic with the junk caps, which were appropriate for about 10-12V if that.  The fact that some of the sleeves were even on backwards (those exploded promptly at least) should have been a clue.

And yet---once the junk caps got through the first half-hour of operation, they lasted indefinitely.  I have a pair of the speakers on my desk even now as I type, and they have been working fine for several years, on continuously.

During the investigation in the midst of the disaster, experiments with a ~100k series resistor and a ~50V supply showed that the good -con caps (Rubycon, United Chemicon, Nichicon) "zenered" at about 30-35V and standard room temperature when their rating was 16V.

 

[Curtis]

Just this week I had to replace the main filter cap on my house alarm system board - keypad wouldn't always register, pressing a "1" key would show up as a "3" on the screen, backup battery wasn't charging properly.  Measured ripple at the output of the main regulator and had 7VAC on a 10VDC supply (should've been 13.8VDC).  The main filter cap looked OK, but when I started to desolder it, it let out a hiss and promptly spilled its guts all over the board.  It was a 25V 2200u 85degC part.  The main power is fed from a 17VAC wallwart, which works out to be over 24VDC once rectified, worse if the mains voltage is a bit high.  Couple the marginal voltage rating with a few years up in the roof cavity during some hot summer months and the stock cap had cooked itself to death.

Replaced it with a 35V 105degC part and it all works perfectly, and will probably last a lot longer than the original did.

 

[mobyd]

One type that is extremely sensitive to any overvoltage is the "bead" style tantalum. Sometimes with older stock even running them at rated voltage can lead to smoke and unpleasant smells.
M