Is it worth buying huge value power filter caps?

[Austin]

Would you only need one cap and no resistors for your power supply if you bought a big enough value? I have found some caps 3900 uf for 15 bucks @ 400v. Would one of these be enough filter for a power supply or would you still need a resistor and another cap?

I am trying to buy some filter caps but in trying to figure out the best thing to buy I saw some huge ones and was wondering about this. Capacitors at The Electrostore.com - Electronic Surplus Parts & Equipment


Really I am just looking for about about ten 100 or 200uf 400v or more as inexpensive and as small size as I can find. Any suggestions? Thanks!

[Merlinb]

Would you only need one cap and no resistors for your power supply if you bought a big enough value?

In theory yes, but it would probably be a false economy because of the greatly increased ripple current. You would end up needing a power transformer that is rated for several times more current (power) than you actually need for the amp. The rectifier might also need to be more heavy-duty than usual. Any fuses would also have to be highly rated to withstand the inrush at switch on, which would make them of little use as protective elements.

Plus, a 3900uF cap charged to 400V would be a scary shock hazard!

[Austin]

Yeah I am thinking it would vaporize the flesh in between the two electrodes if you laid your arm on it or something...



Could the added stress to the rectifier and power supply transformer be relieved by adding a resistor before the cap?

[R.G.]

In theory, theory is the same as practice. In practice, it isn't.

In addition to the rectifier stress (which is a non-issue with solid state diodes in most cases), there are several things that the R-C filters do for you that a single big filter cap can't.

One is ripple reduction. Yes, an infinitely big cap would reduce ripple infinitely. In practice, you reduce ripple much faster and more cheaply by putting in an R-C section. It's only with insanely big caps that you get to here. Or with a regulated power supply. It's pretty easy to get 20db (i.e. drop of 10:1) in ripple reduction per R-C stage. That makes for a 60-80db (1000 or 10,000 to one reduction!) in ripple from the first filter cap to the final R-C stage. Ripple goes down linearly with capacitance (if you can ignore ESR, which is a happy accident which happens sometimes), so you need not just a bigger cap, but a 1000 to 10,000 times bigger cap to get the same ripple reduction. A 30uF first filter cap becomes 30,000uF for a 60db reduction in only one cap, and a 390,000uF cap for an 80 db reduction. That isn't just flesh-eating - it could set off a fusion reaction or power a railgun.

Also notice that the commonest linear regulators lower ripple about 40-50db. You might have to do a double-regulator stage to keep the ripple out of a high-gain input even if you linear regulated.

But the biggie is that there is some impedance in the power supply, even with BIG caps. The wire resistance, ESR of the cap, ESL of the cap, etc. form a way for currents used in the high power section to make a voltage appear reflecting that current into the drain and bias sections of earlier stages. These can be small, but then the gains of amplifiers can be and frequently are BIG too. And the classical triode or pentode gain stages have no power supply rejection whatsoever, so whatever noise is on the B+ coming to their plate resistors goes right through to the next stage grid. And wiring resistances are outside the cap entirely.

This has two ugly effects. One is motorboating, where the gain and phase add up to a low frequency oscillation caused by the phase shifts in each stage. The other is high frequency oscillation. In less severe cases where there is not enough loop gain to actually make the thing oscillate, it may have odd resonances at the almost-oscillate frequencies.

So... R-C decoupling is a quick, easy way to get where you want to go, given that you can afford to lose the volts in the voltage drop in the resistors. This is nearly always possible in audio gain stages with triodes and pentodes, so it's very heavily practical to do so.

[trobbins]

From a practical wiring sense - it forces you to single star distribution points directly at that big cap's terminals for all grounds and rails. From a design perspective, it forces all stages to operate at the same rail voltage.

Series resistance or inductance in the rectifier side circuit will lower stress on that side, and reduce the ESR/ESL induced voltage across the bulk capacitor that is also seen by the signal side.

[woodyc]

I worked on an amp that redplated for a while every time you turned it on. It turned out somebody had "upgraded" the caps in the bias circuit. It was one of those amps where the bias is tapped off one side of HV secondary thru a large resistance. The charging time was several minutes. I think it was a marshall head, but I don't actually remember.

[printer2]

100mA load driven by a 230V transformer with 33 ohm secondary through a bridge rectifier.

2200uF = 428V 1.5V ripple
220uF = 427V 3.5V ripple
22uF - 220 ohm - 22uF = 400V with 12V ripple
220uF - 220 ohm - 22uF = 405V with 1V ripple

More or less.


Brute force. 770VA

[woodyc]

Another thing to consider is that the stored energy in your amp is mainly the sum of .5*C*V*v for the the supply caps. Some guitar amps have about the same stored energy as a portable defibrillator. You could probably save a life with an SVT.

[Steve Conner]

I've used 3300uF, 400V capacitors in a pulsed power application where I really needed the stored energy. That stored energy can get released when things go wrong: I had one of my prototype circuits blow out like a shotgun blast, scattering pieces of itself in a 10 foot radius.

So, I don't recommend using any more stored energy than necessary, and as RG shows, a multistage RC or LC filter will give the same ripple attenuation with much less energy storage.

[Steve Conner]

I've used 3300uF, 400V capacitors in a pulsed power application where I really needed the stored energy. That stored energy can get released when things go wrong: I had one of my prototype circuits blow out like a shotgun blast, scattering pieces of itself in a 10 foot radius.

So, I don't recommend using any more stored energy than necessary, and as RG shows, a multistage RC or LC filter will give the same ripple attenuation with much less energy storage.

[J M Fahey]

WoW !!! 312 Joule capacitors!!
I NEED 30 of them for my new magnetizer !!
Pity passing them through Customs will be murder.
Plus they show: "out of stock"

[Chuck H]

WoW !!! 312 Joule capacitors!!
I NEED 30 of them for my new magnetizer !!

Of course you do

Awesome

[Chuck H]

But the biggie is that there is some impedance in the power supply, even with BIG caps. The wire resistance, ESR of the cap, ESL of the cap, etc. form a way for currents used in the high power section to make a voltage appear reflecting that current into the drain and bias sections of earlier stages. These can be small, but then the gains of amplifiers can be and frequently are BIG too. And the classical triode or pentode gain stages have no power supply rejection whatsoever, so whatever noise is on the B+ coming to their plate resistors goes right through to the next stage grid. And wiring resistances are outside the cap entirely.

This has two ugly effects. One is motorboating, where the gain and phase add up to a low frequency oscillation caused by the phase shifts in each stage. The other is high frequency oscillation. In less severe cases where there is not enough loop gain to actually make the thing oscillate, it may have odd resonances at the almost-oscillate frequencies.

This was my first consideration. Especially as it relates to something like a guitar amp where the input signal is amplified into clipping following stages and then the clipped signal is amplified into clipping stages following that!!! With the current of such a power supply it wouldn't take much resistance at the filter node (the chassis might be enough on it's own) to impress on the signal path across several stages. Whomp-whomp-whomp.

[Austin]

Thanks for the replies, so it is better to use more stages with smaller caps to get more filtering rather than larger value caps. Makes sense.

[mooreamps]

Thanks for the replies, so it is better to use more stages with smaller caps to get more filtering rather than larger value caps. Makes sense.

Yes, it does...and as a small side note, if your power stage is push-pull, you don't need the big caps at all....They don't buy you anything due to the common-mode rejection...

-g