CLC pi filter - voltage drop Q.

[Diabolical Artificer]

I have a power supply that incorporates a pi filter, HT is 430/437v which is ok for the first cap in the filter, however the other cap is only rated at 400v, therefore I need to drop 30v ish so thought about using a dropping resistor, but where do I put it?

There's three choices as far as can see, 1) put the resistor before the first cap, 2) put the resistor in parallel with the choke, 3) put the resistor after the choke before the 2nd cap, see att.

Any ideas? Andy.

[GrimJosef]

2) won't lower the voltage on the second capacitor - putting a resistor in parallel with the choke will only lower the impedance of the combination, and lower impedances drop fewer volts.

1) will work by lowering the voltage on both capacitors. But the resistor will see not a continuous current but a sequence of current pulses - current will only flow when the voltage out of the rectifiers exceeds the voltage on the first capacitor. This means that the resistor gets hotter than it would if it was between the two capacitors. You might also find that you need a different resistance value to deliver 400V on the second capacitor from what you'd calculate using just the average current.

3) might be simplest. Obviously it doesn't matter (significantly) which order the choke and resistor come in. The one thing you do get with 1) that you don't with 3) is lower stress on the transformer and rectifier since a resistor before the first capacitor limits the peak current in those components.

If you're interested in trying different component options then I'd strongly recommend the free power supply modelling software psud2 available from the Duncan Amps website. It allows you to try things out in cyberspace and see the voltages and currents at different points in the circuit.

Cheers,

GJ

[Argus25]

Your 400V capacitor is not suited to the task in any of those circuits.

It could possibly be if there was a shunt divider to drop the voltage for a low current stage somewhere downstream.

The reason is you cannot rely on the current via a choke or resistor or other series element to drop the voltage down below a capacitor's rated voltage because that current might not be there if the stages it is feeding have not fully warmed up or are disconnected.

Also you should allow at least 15 to 20% headroom for fluctuating input voltages. So the capacitor/s should not be over-voltaged even with zero load on the supply and a higher than expected peak input voltage, if you want to avoid failure.

The capacitor should be 500V rated absolute minimum in my view, to be reliable.

Personally, for a supply that delivered over 400V, I would use a 600V rated electrolytic.

[G6Tanuki]

Yes, as Argus25 says, the 'second capacitor' needs to be rated to handle the maximum voltage which it will see in the absence of any current-draw through the choke/dropper-resistor. This sort of 'worst case' scenario is often overlooked - particularly when people replace a valve rectifier with a semiconductor one. A semiconductor-rectifier will provide full HT-voltage within a few tens of milliseconds of switch-on - well before the valves have warmed-up and started drawing current. So *all* the smoothing/decoupling capacitors see this full HT-voltage for quite a few seconds at switch-on.

In your case I'd consider using either 600V-rated ones (expensive/not-so-available) or alternatively for each of your existing electrolytics use a pair of series-connected 400V-rated ones (with suitable voltage-sharing resistors wired in parallel). I've got a HF linear-amplifier here that runs with 1250V HT under load but at switch-on - before the valve starts drawing current - the HT goes up to 1600V. I'm using five 470uF 400V-rated electrolytics in series, just to be sure.

[trobbins]

Modern electrolytics have to pass a short duration +10% voltage application in a reliable manner. There is also built in margin if the capacitor is cold. That may help for turn-on events where B+ is somewhat higher than normal operating conditions, and the amp hasn't been on for hours.

But of course that all may not provide any margin for higher mains voltage. So it is worthwhile trying to identify how variable your mains AC is, as some can experience much more than 10% variation, so making a spot measurement only is risky.

So characterising what '430VDC' actually varies between is a good start.

[Diabolical Artificer]

Thanks all, I may have resolved this, in the end I went for option 4) moving said cap in parallel shunt regulators. I was concerned about doing this as I know VR valves can oscillate, but all seems ok. Left it on for a few hours, no probs.

I realised I had a problem when I heard several loud cracks and bangs and the music went off, on investigation the cap was hot and had vented.

I agree using 500v+ rated caps would be a good idea but as usual trying to use what I have and save dosh. 600v high capacitance (over 100u) caps would be silly money. Space constraints rule out series connection.

Andy.

[GrimJosef]

I opened up a large 800V electrolytic once. It had two smaller ones inside wired in series (and without balancing resistors too !). There are plenty of 500V ones available. There are also some 550V and a very few 600V but I'd be more nervous, and I suspect further out of pocket, using those than using two lower-voltage ones in series.

Cheers,

GJ

[turretslug]

I recall reading about those "800V" electrolytics (quite likely here!) and how they were two 450V types in series, no balancing resistors. I have heard an argument (and I'm prepared to give it at least some credence) that series connection does have a tendency to be self-compensating and correcting as regards leakage current and voltage-sharing without balancing resistors, but it would have to be identical new capacitors from the same batch before I'd entertain doing it. I have to admit that the few times I've connected HV electrolytics in series, I used balancing resistors.

There are a wide range of values of plastic film motor-run capacitors available nowadays, often 450VAC rated and data sheets also often give a DC rating. They are bigger and more expensive than HV electrolytics, but not madly so.

[trobbins]

I reckon the 'self-balancing' view would be due to the series caps reaching a voltage where each starts to independently exhibit internal micro-discharges. Up to that point, the leakage current would be common to each cap, but the voltage on each cap could be somewhat different. A spark would reduce that caps voltage a bit, and consequently raise the other caps voltage - with the likely outcome being an equilibrium of internal sparks at the common leakage current level.

Two related papers on the general topic of ecap behavior at voltages exceeding rated level.

https://pdfs.semanticscholar.org/58f...0d6f2ac4a6.pdf

https://pdfs.semanticscholar.org/8a7...218f114870.pdf

[Diabolical Artificer]

I have a few 500v rated caps G, F&T 16u 16u & 50u 50u as well as 4u 600v and 2.5u 600v. All either too big physically or too small capacitance. I am using an F&T 50/50u 500v as part of HT decoupling but the 820u is really needed to squash ripple if this PSU ends up powering a phono preamp, which is probable.

The solution as is, with the two VR150's clamping HT @ 310v across the 820u isn't ideal, but I've scoped the HT and there's no oscillation. I'll run the preamp today and see how it goes.

I can only assume the cracks and bangs I heard was arching and shorting inside the cap. The dialetric must have broken down. I'm surprised though as I've come across lots of caps run over their V max in amplifiers and other devices and the cap in question wasn't a cheap type being made by Aerovox.

I'll have a look at those trobbins thanks for linking them.

Andy.

[kalee20]

Must say I'd buy new capacitors.

Failing that, Option 1 would be my preferred one. The resistor is handling pulse currents, so will get hotter than simple Idc² x R power calculation would indicate, but the rectifiers will have an easier life as the current pulses will be broadened and flattened. Though, I'd do it as a temporary measure till my new capacitors arrived.

[Diabolical Artificer]

"Must say I'd buy new capacitors." I would but am cash strapped.

I'm using four 800v PIV rated diodes, two in series for each 330v wndg. Thinking about it I suppose I should put two resistors in parallel with each one, 220r - 470r maybe. I'm now using two 220u 385v rated caps in series for the first cap on the choke, with the resistors mentioned above in parallel, would a resistor still be a god idea before the two diodes? If so, what value? 100r?

The 820u 400v cap has to stay, there's no room or cash for another cap. I'm measuring 310v, slowly rising from about 150v on it, so well within spec.

Andy

[Argus25]

Originally Posted by Diabolical Artificer

"Must say I'd buy new capacitors." I would but am cash strapped.

I'm using four 800v PIV rated diodes, two in series for each 330v wndg. Thinking about it I suppose I should put two resistors in parallel with each one, 220r - 470r maybe.

Andy

There is no requirement to put resistors in parallel with series diodes, if they are of the same type. Creating artificial diode leakage with parallel resistors won't do anything useful. If you were fitting suppression capacitors across each diode, they should have equal value, but mostly, its hardly required.

If you want a single high piv rated diode, just go for a BY228:

https://www.vishay.com/docs/86003/by228.pdf

[Diabolical Artificer]

"There is no requirement to put resistors in parallel with series diodes" Yep, just realised after I'd typed I was talking utter rubbish.

Thanks, A.