Need to dejam an ATX power supply

[Franki]

hi folks

I need to inprive the dejamming of an ATX power supply, which is a PWM power supply with active PFC, from whirring harmonics of the device itself and different kinds of ripple current generated by devices hanging on the molex connectors.

note: the improved dejamming needs to be on the molex/ATX-connector side.

My question is: does anyone know whether a bunch of film capacitors and ferrite cores usually suffice or do I need to do more work on it?

The ATX power supply almost completely looks like this one:

full size image:
http://upload.wikimedia.org/wikipedia/commons/c/cf/ATX_power_supply_interior.jpg

THX

[Simon]

what exactly do you mean by dejam ?

[Time]

I've never heard of that either.

[Russel]

A quick Google search came up with this. I know that can't be it!

[apex]

I think he refers to noise on the power rails.

And here are my ideas to his problem:
1. BIG decoupling caps ONTO the devices that hang on the molex connectors
2. Depending on what kind of noise you get:
    - Inductors in the individual power rails
    - Ferrite Cores (and please real ring cores) onto the individual power rails

apex

[Franki]

what exactly do you mean by dejam ?

I'm not used to terms of electronical engineering in the English language, since I mostly do that in German. And believe it or not, 3 different dictionaries all proposed my to translate the German word "entstören" with "to dejam", which just felt awkward from the beginning. Who wants to know which ones are to blame?

I mean I need to suppress ripple frequencies on the power line both generated by the ATX power supply and by other devices attached to molex connectors.

Of course the power supply has big supposedly low-ESR/ESL electrolyte caps, but there are no film capacitors which would really have almost no low-ESR/ESL. There are some ceramic caps, but still there are many switching components that don't have any caps at all.

Because of that, the DC outputs of the power supply really suck. And it's not the only ATX power supply that sucks, they are all bad in this respect, unless you spent a fortune. It really seems that it's more reliable and cost-effective to just use a regular power supply and improve it yourself.

[Franki]

And here are my ideas to his problem:
1. BIG decoupling caps ONTO the devices that hang on the molex connectors

Big electrolytes would indeed suppress spikes and ripples (we are talking of HF noise modulated with audible NF), but only as good as low as their ESRs are, and they aren't low enough. And since I'm dealing with rather HF ripples and not big DC-fluctuations, film caps would be a better go. I'm not sure that this is only an ESR-specific issue, might be due to dielectric absorptions and other non-linear characteristics.

2. Depending on what kind of noise you get:
    - Inductors in the individual power rails
    - Ferrite Cores (and please real ring cores) onto the individual power rails

That's part of the plan. I would have to unsolder and resolder 36 rather thick several-Amps wires. Other option would be to buy ferrite cores big enough to put molex connectors through, which is a complete waste of space technically and the wallet would start itching. Other options would be to cut the cables and resolder them or to try to pull the pins out of the molex connectors and put them back afterwards.

So this would give me a second-order LC low-pass for each attached device if I'm correct.

[hacklordsniper]

or to try to pull the pins out of the molex connectors and put them back afterwards.

That is very easy and fast solution

[Franki]

or to try to pull the pins out of the molex connectors and put them back afterwards.

That is very easy and fast solution

Yeah, but sometimes, the pins won't come out of the molex or they break in the process