LC power supply capacitor value question.

[Chris Wilson]

I hopefully attach a schematic for a PSU based on a transformer and five 1 Henry chokes. I have bought the transformer and chokes cheaply, from a guy who bought several sets as it was cheaper to have them made in quantity and sell off the excess to others following the same project. The schematic shows mainly 150 uF 400V electrolytics used for smoothing, 2 in series, so 800V and 75 uF. My question is mainly academic. Are the chokes and caps "tuned" to one another? If I say used 450uF caps 450V rated, would the added filter capacitance "upset" the way the chokes operated? I believe because it is not a "resonant" supply more cap value won't do any harm and may improve smoothing further, but want to be sure. Thanks.

[Paul Price]

This is not a resonant supply and does not need to be tuned.  JYou would benefit by placing about a 1-meg resistor across each of the capacitors in series to equalize the voltage drop across each capacitor, because any two series capacitors might not match in leakage currents causing the voltage across one of the capacitors to exceed its voltage rating over time.

I would replace this whole mess with a single capacitor and a H.V. transistor or n-chan MOSFET to regulate the voltage or even to just to almost completely filter out any power supply ripple.

Using 450V or 500V individual capacitors would be better because they would  take up far less space and require fewer components.

BTW it seems you are missing one electrolytic on the output of the rectifier on the 150V winding (bottom).

[Chris Wilson]

You mean get rid of the chokes?    I thought they'd add to the ripple reduction very well? I do not intend to use the zener chain, I just bought the transformer and chokes as they were cheap and the right output voltages, (made for the application). I am using a complete build it yourself board to control the tetrode PA valve, it's for an RF linear amp.  http://www.ifwtech.co.uk/g3sek/boards/tetrode/tetrode-1.htm  This will do the regulating and protecting stuff. Thanks Paul. Will check out the missing cap on the schematic... Hmmmm.

[Paul Price]

Forget about the "zener chain", only one zener may be needed, if that, with a semiconductor regulating approach.

Just remember that chokes and caps were the only economically and practical, but now antiquated method of trying to regulate and filter power supplies before semiconductors and IC's became cheaply available to do this job.
MOSFET and/or regulator IC's are so much better, and with so much less cost and space used, there's such a big saving in the work to do all the associated wiring and mounting parts, etc.

[Chris Wilson]

OK, thanks Paul, I see a typo in my original post, the chokes are of course 1 Henry (300mA, 16 Ohms), not 1 uH. I have corrected it.

[SeanB]

If you are going to place caps in series make the bleed resistors 33K 1W units, so that they have enough voltage withstand capacity and are large enough to run cool. 1M will not give good voltage division as the current is about the same as the leakage current. The chokes are very good on a transmitter to reduce the RF fed back into the mains though, but you need them on both sides of the rail though. Depending on the insulation from choke to core you can do this, but even if doing this mount them on an insulated board to reduce voltage stress to the core, and connect the core to ground with a 100n 1kV ceramic cap in any case, and if you want place a 1k 5W resistor across the choke for damping. Bypass the smoothing caps with some 470n 1kV film caps as well in a transmitter, it reduces the RF heating of them. You will need some good film units near the transmit valve in any case, look at microwave capacitors, 1u2 will be a good value easy to find, plus they have built in bleeder resistors.

[Paul Price]

Sean8, I think you should check your calculations for the bleeder resistors.

The leakage current of a recently mfg'd electrolytic cap are much better than you think, check the spec sheets.

 2 series 33k bleeder resistors will have to dissipate over a watt each at the 400V supply voltage to use about 6mA of shunt current across each capacitor.

I may be modern, but that is a hell of a lot of bleeder current!

Since there are 6 electrolytic capacitors that means 6  33-k resistors and about 7.3 watts of heat wasted just to excessively balance leakage current, perhaps you are being just a little too conservative. I would expect from a spec sheet and my calculations that there is  not more than 1.3ma max leakage per cap to balance out, and you are using 6mA to shunt this amount.

According to one mfg, there is a max of .04CV plus 100 uA of leakage after 1 min at the rated voltage, but they are running at 1/2 their rated voltage, so there should be considerably less leakage current per cap.

Rubycon says max leakage would be 3 * (sqr root of(C*V)) where I= leakage in micro amps,  C is in uF and V in volts, this yields about .7mA of leakage at the rated voltage, but they would be used at 1/2 their rated voltage and the leakage current would again be much less.

[SeanB]

Conservative yes, though if the caps are guess name brand ones that did not explode at power on I would use them, rated at 5W. After all they are going to be fed into a class A amplifier, so an extra watt or ten makes no difference in the toaster oven inside the case. Name brand ones I would use 100k 1W resistors, and check the centre tap voltage after power up and after a 4 hour run.