ideal capacitance of ATX power supply in DC rails

[laser]

I want to know what's the ideal capacitance in the 3.3v,5v and 12v rails of any power supply

the more capacitance the better? the less ripple?

for example I have this Antec sp450 power supply



as you can see in the pic there is lot of place in the left near the fan to insert 3 caps to reinforce capacitance of 3.3v 5v and 12v rails
also will be ideal because caps there will be refreshed by the rear fan

can I put 3 big capacitors there ?
I'm thinking about something like 4700mf for each DC line

thanks

[mariushm]

Re: ideal capacitance of ATX power supply in DC rails

There's no ideal value.

It depends on how the power supply is designed and what other components are used together with the output capacitors to stabilize and regulate the output voltage.

Older power supplies need more capacitance, newer models may need less for the same wattage rating, so there's no "best" choice or ideal value.

For a 400-500w power supply, 2200-3300uF capacitors are very common, some even use 4700uF but it's arguable if it actually makes a difference.

No, you can't move the capacitors away from the board (well, without consequences). The capacitors on the secondary side are low esr and it's very important that they have VERY SHORT leads, ideally installed like you see them right on the pcb board.

In some cases, even half a centimeter of leads can mess up the quality of the output voltage because those wires act as inductors or extra resistance in the circuit.

You can move some minimum load resistors further away from the capacitors so they're not subjected to heat that much and that's about it.

[370forlife]

Re: ideal capacitance of ATX power supply in DC rails

To understand how ripple and voltage regulation is effected by capacitors, you must understand how the secondary side of a power supply functions. The transformer puts out (in most cases,) 12V and 5V AC power that is stepped down from the primary side voltage of 320-380V AC. This 12V and 5V AC must be converted to DC voltage, and is done so by use of diodes, typically in TO-220 or TO-247 package, Often rated for about 30V and 30A in most sub 300W power supplies.

Diodes are very simple, basically they only let voltage flow one way. AC voltage flows both ways at a certain frequency (Hertz) and DC voltage only flows one way. So when the AC voltage flows one way, the diode conducts and lets it flow through, but then when the AC voltage flows the other way, the diode does not let it flow through.

So what happens on the secondary side of the power supply you get a 12V DC voltage that quickly goes from 12V to 0V back to 12V and back to 0V at nearly the same frequency of the AC voltage that is fed to the diode pack. Now obviously computers don't like to have 0V fed to them, so you have to smooth that fluctuation out using inductors and capacitors. The capacitors build up a 12V charge so that when the voltage falls back down to 0V at the diodes, the capacitors discharge their 12V so that the voltage does not fall too far down. In fact, many times it only falls a couple dozen millivolts.

(Deep Breath...)

That was a very simplified explanation of what capacitors do on the secondary side of a power supply. There is many more complex things that go on to do with the ESR of the capacitor and the inductors used (specifically power supplies that employ CLC-filters on the secondary).

Now, Capacitance doesn't really matter once you reach a certain point. When it does matter is If you have a no-name cheap unit that only has a 680uf cap on the 12v rail, then what will happen is when the voltage at the diode falls down to 0V, the capacitor will discharge completely before the voltage comes back up from the diode. This causes huge amounts of ripple that can sometimes be measured in Volts.

When you have a power supply that already has two 2200uf capacitors on the 12v rail, then you are good. Adding more does not really have any effect. It may decrease ripple some, but it can also have a profound effect of increasing ripple. Like I said above, most power supplies use CLC filters, (capacitor inductor capacitor) that are carefully designed to decrease ripple as much as possible by using certain uH value inductors coupled with certain value uF and ESR capacitors. Your best bet would be to just recap the power supply by using caps that closely match the ESR and capacitance of the originals.

[laser]

Re: ideal capacitance of ATX power supply in DC rails

There's no ideal value.


No, you can't move the capacitors away from the board (well, without consequences). The capacitors on the secondary side are low esr and it's very important that they have VERY SHORT leads, ideally installed like you see them right on the pcb board.

In some cases, even half a centimeter of leads can mess up the quality of the output voltage because those wires act as inductors or extra resistance in the circuit.

You can move some minimum load resistors further away from the capacitors so they're not subjected to heat that much and that's about it.

I'm not talking about moving the caps,I'm talking about adding more caps in parallel in the primary side to reinforce capacitance or just in case one of those in the secondary fail
Those good caps which I replaced could fail in the future because this unit works hot despite it have 2 fans..both are thermal controlled both works at low RPM... is a very silent unit

btw,
this Antec unit comes with buggy chinese fuhjyyu caps which some of them bulged then I replaced all by rubycon MBZ series except for a big panasonic 25v -3300 mf for the 12v rail

[tom66]

Re: ideal capacitance of ATX power supply in DC rails

MBZ is not a good idea for a PSU.

[laser]

Re: ideal capacitance of ATX power supply in DC rails

Your best bet would be to just recap the power supply by using caps that closely match the ESR and capacitance of the originals.

as I said in my previous post this Antec unit comes with lame fuhjyyu caps...I don't know if they were ESR or not ...I discarded all and replaced by Rubycon MBZ series which I consider the best ones

[laser]

Re: ideal capacitance of ATX power supply in DC rails

MBZ is not a good idea for a PSU.

why not?

[mariushm]

Re: ideal capacitance of ATX power supply in DC rails

It's not a good ideea to use Rubycon MBZ (or Nichicon HZ) in power supplies, they're ultra low esr and power supplies don't need ultra low esr capacitors.

You can use capacitors with slightly higher esr (but still considered very low esr) like Panasonic FR series which has 6000-10000 hours @ 105 rating compared to Rubycon MBZ which has 2000h if I'm not mistaken and you'll get the same performance out of the power supply.

The capacitors on the primary side are large and handle heat much better than the capacitors on the secondary side and generally they're not subjected to as much ripple and high currents and overall stress as the secondary side capacitors.
Increasing capacitance won't do anything to the secondary side, I don't understand what you think will happen to the secondary side if you "reinforce" the primary side.
Nothing's gonna happen, those primary capacitors and their value has no relation to what happens on the other side.

Bigger capacitance will only keep the power supply running for a longer time if there's a sudden drop in input power (atx standard says the psu should stay on for one cycle of AC power - 1000ms / 60 Hz = 16.6ms) and maybe improve other things on the primary side, but won't affect anything on secondary side.

Those caps on the secondary are squeezed there with cables covering them and they're near inductors and heatsinks all generating heat and blocking airflow. Using MBZ there? Not really a good choice.

[mariushm]

Re: ideal capacitance of ATX power supply in DC rails

The Fuhjyyu were probably TM series:



3300uF 6.3v : 1870mA ripple , 0.035 ohm esr

You've using Rubycon MBZ:



3300uf 6.3v 2800mA ripple, 0.012 ohm

Ripple OK, esr three times lower. Not wise to alter the esr that much.

[PeteS in CA]

Re: ideal capacitance of ATX power supply in DC rails

First, there are components in the regulator control circuit that are selected for the type of caps the manufacturer used, in prticular, for the ESR of those caps. If you use a part with significantly lower ESR, the regulator stability may be affected. Second, MBZ or HV are 2000 hour parts, and use an electrolyte that doesn't do well with heat. Both characteistics are notmgood for a P/S you want to run form really long time. If you must use low ESR parts with water-based electrolytes, consider Nichicon HE, Rubycon ZL or UCC KZE. Those are long life series and can even take some abuse.

[laser]

Re: ideal capacitance of ATX power supply in DC rails

The Fuhjyyu were probably TM series:

yep...check:

http://www.overclockers.com/forums/s...d.php?t=495380

[laser]

Re: ideal capacitance of ATX power supply in DC rails

you must use low ESR parts with water-based electrolytes, consider Nichicon HE, Rubycon ZL or UCC KZE. Those are long life series and can even take some abuse.

chemicon KZE are low ESR I saw tons of them on motherboards
are those long life...are you sure?

[PeteS in CA]

Re: ideal capacitance of ATX power supply in DC rails

Yes, laser. I've looked at their datasheets many times and tested many of them while working in Component Engineering at my previous employer. They are rated for 5000 or 10000 hours and handled my torture tests rather well.