cap voltage ratings: peak or ave V?

[Al (astral highway)]

Are (mains electrolytic) caps rated for peak or average voltage?

[Ed_Dinning]

Hi Al, no electrolytics on the mains please, they go BANG!
If you mean the types that are used for motor start and are in tubular cans, these are normally rated for system voltage (RMS).
The smaller class X types are rated for more than 250V RMS, there are also 440V types for 3 phase work, the Y types are similar.
The input electrolytics used in SMPS etc are rated for the rail voltage expected, which, given their large values, usually covers up to peak mains (typically 400V) they normally also have a 10 sec rating of a higher voltage, say 500V.

Typicial line operated SMPS systems with boost mode PFC use a stabailsed DC line of 370 to 400V (caps of 400 or 450 VW). These will withstand 500 to 550 volts for say 10 secs.

Ed

[Herald1360]

Not sure what you mean by a mains electrolytic... AC rated capacitors usually state the AC voltage rating on the case- that would be RMS voltage- but they aren't normally electrolytic. I would presume that a reversible electrolytic voltage rating would be peak voltage.

If you mean the first capacitor after the rectifier in a PSU, old ones often had two ratings- working voltage and surge (or peak) voltage- which reflected the fact that they could stand a higher voltage for a short period (usually the time between the rectifier valve starting to conduct and the rest of the valves taking full operating current), which was a way of using a cheaper capacitor than one rated for the highest voltage all the time in mass produced kit.

Modern capacitors might well stand the same sort of overvoltage, but without confirmation from the datasheet it's probably best to pick one rated for the highest voltage the circuit could ever reach- on a one off replacement basis the odd penny or so isn't worth the risk.

HTH

Chris

Oops- beaten on the draw by Ed

[Al (astral highway)]

Hi Al, no electrolytics on the mains please, they go BANG!

I initally put 'HV+ caps' but then realised that was a relative term. 'Mains' was a very poor choice of terminology on my part. I meant smoothing and reservoir caps, and see from your post that I should call these 'input electrolytics.' I wondered just how close to its maximum rated value a reservoir cap should be allowed to run, and you've answered that question here:

The input electrolytics used in SMPS etc are rated for the rail voltage expected ...they normally also have a 10 sec rating of a higher voltage....QUOTE

Thanks Ed.

[Al (astral highway)]

If you mean the first capacitor after the rectifier in a PSU...

Hi Chris, yes, sensibly guessed from my misleading terminology

It's comforting to see that a working voltage of say 400 really does mean that. I wanted to know more about the abilities of the first cap after the rectifier in particular, as it sees 1/2 the peak-to-peak voltage across the xformer.

I've put in a cap rated 400 that because of anti-surge measures I've put in, now only ever sees 334V.

But earlier, as I was testing progressive stages of a power supply rebuild, I had wondered just how close you could get to the maximum rating. (Especially when I was testing the circuit after I'd rebuilt the pi-filter and was doing a quick zero load test run...)

[Herald1360]

I believe that you can safely use a capacitor at the manufacturer's voltage rating provided it can never be exceeded long term, but a modicum of derating is good for lifetime, in terms of the voltage, temperature and ripple current that the capacitor is subjected to. Your 344 volts on a 400V cap sounds pretty good if the ripple current is also within ratings. For valve kit, probably 105C temp rated components are best, though if ventilation is good and you know the temperature is less than maybe 50C, 85C parts should give a reasonable life.

I think it's been mentioned elsewhere that too low an operating vs rated voltage isn't a good idea for electrolytics, but that's not likely with any sort of HT supply!

Thinking about your pi-filter.... beware of any resonances in the LC filter that are anywhere near mains frequency or its harmonics (or indeed anywhere in the operating frequency range of whatever it's powering if the current draw varies with signal)- in fact any undamped resonance can be dodgy- any sudden supply voltage change may provoke ringing when you can potentially get twice or more the normal voltage on peaks! (Sounds horrible, but in practice most inductors have enough series resistance to damp things down OK.)

Chris

[Al (astral highway)]

...any undamped resonance can be dodgy- any sudden supply voltage change may provoke ringing when you can potentially get twice or more the normal voltage on peaks!

Hmmm. Thanks for the warning Chris. I'll really have to watch out for that. There is going to be a lot of stray inductance in the pi filter because it exploits part of a physically massive current-limiting resistor chain used for the voltage regulator. Also, the last smoothing cap is mounted a long way from its earthy end.

Is there a good, systematic way to check for ringing in power supplies, with a 'scope?

[Herald1360]

Sounds like no problem if your pi is C-R-C- I was assuming you had C-L-C. A few stray micro- or even milli- Henries won't matter. You only get the problem I was raising with Henries in a high quality smoothing choke.

I only ever ran foul of this once, with a large low voltage supply where the filter resonance was around 70Hz- not obviously (to a younger me anyway) a problem with 300Hz three phase ripple. Unfortunately it fed an amplifier which was designed for dc-a few hundred Hz- the voltage excursions when the test sig gen was wound through filter resonance were impressive!

To look for ringing... perhaps try switching a resistive load on and off the HT rail whilst looking at HT volts with a 'scope?

Chris

[Al (astral highway)]

Sounds like no problem if your pi is C-R-C- I was assuming you had C-L-C.

Actually, my pi is LCR!

10H, 10Uf, 1K

It's a long story but basically, the old xformer was O/C and I got hold of an NOS one that wasn't ideal but had most of the right windings. They're expensive, as you know, and I didn't want to start all over with yet another new one. So mine is 350-0-350 and yet we needed to end up with just enough to give the voltage regulator enough to produce a regulated 149V.

I tried various things but the fastest, neatest and most efficient way I could shed around 100V was to use a choke input filter instead of the usual capacitor input. Of course, this way, there's no initial C to charge right up to the peak voltage of each rectifier half cycle. It works out just as expected. The R is 1K of the current limiting resistor (total 3K6) for the voltage regulator.

There's an additional 4.7 uf of smoothing after the 1K, also.

I'm hoping run a simulation for this but I've only newly acquired the software and haven't had time to try it for ripple yet

[Herald1360]

Hi, ran a sim on what I guessed was your cct, with various resistances added for likely strays (are you using valve or solid state rectifier?).

Red trace is volts on 10u, green is volts on 4.7u.

Blue is result of damping the circuit with (100u in series with 150R) in parallel with 10u reservoir. It's a dodge you can do with cheap(ish) modern capacitors which are available in values 50's designers could only dream of. The 150R gives somewhere near critical damping to the resonant circuit, the 100u just keeps DC off the resistor so it doesn't overload the PSU / cook itself.

Resonant frequency of LC bit is around 15Hz. Switch on overshoot goes to about 400V; if you're using a valve rectifier, the soft turn on would probably stop this.

You can do this sort of run with the demo version of Simetrix- I think the circuit is within its size limits (I used my old complete one, though).

Chris

[Al (astral highway)]

Hi, ran a sim on what I guessed was your cct,

Thanks Chris! I appreciate the considerable effort you've gone to there.

...are you using valve or solid state rectifier?)

It's a valve: an 80. In fact there are a couple of tiny further small differences between your sim cct. and the 'as built'. There is a 370 R resistor in series with each anode. In my case, the choke has a DC resistance of 396R. And between the cathode of the valve voltage regulator (not zener, as show) and xformer CT, is a current limiting R of 52. Apart from the anode Rs, there is an NTC thermistor in the heater.

Apart from preserving the life of the elderly rectifier valve, the soft-start means that the voltage regulator sees a safe strike voltage: it gently builds up to 180V, instead of starting above its maximum strike voltage.

I am very interested in your damping idea but when I looked at the output on the 'scope last night, I found the major ripple component to have a period of 4 milliseconds at 40 mV peak-to-peak. (This was through a x 10 probe).

This is a much higher frequency than I expected. Any thoughts on that?

It made me think of taking your idea but perhaps with a 'small' (about 1 millihenry) choke (instead of the R) and taking a wild guess, something like the size of your cap in series. I haven't done the calculation yet.

Interested to hear what you think.