Favorite filter design for generic switchmode output

[Psi]

Hey, what are some good filter designs for a typical switchmode supply output?
Obviously there's multiple factors to consider like the switchmode frequency, but are there any good starting points?

I need to design a 5V@3A buck and 12V@3A sepic switchmode supply for powering video transmitters which tend to get faint lines over the image unless the power is smooth. Load will be around 1.5A

I've built many switchmode powersupplies in the past, usually by adapting the reference design in the datasheet to my needs, but it's always been for powering things that don't care about typical switchmode noise.

LC filter seems like a good starting point, perhaps two stages of 22uH and 1000uF,  (1khz cutoff)

Any tips?

[T3sl4co1l]

Sure, sounds good.  Don't forget the shielding, and use lots of ceramic caps and ferrite beads too (making sure the beads are rated for several times more amps than required -- they aren't rated the same way regular chokes are).  The challenge in RF terms is getting a good filter at asymptotic frequencies, but ESR and ESL (and the respective characteristics of chokes) poke your transfer function full of zeroes, making dips and peaks at various high frequencies.

Also mind common mode noise, which can be more important than differential (e.g., if the equipment has filtering on internal sections anyway).  Assume the switcher acts like a noise source between the input and output cables, and go from there.  Example, if both sides of the supply are firmly grounded -- at RF, not just galvanically -- it should be fine; but if the equipment is not grounded, that noise can be superimposed over the whole thing, and find its way along input and output cables, introducing ground noise and weird stuff like that.

Tim

[HackedFridgeMagnet]

Have you considered Cuk for a buckboost instead of the sepic.
I have never made one but I have read that they produce less noise than the Sepic.
Disadvantages are, higher FET Vds required and importantly the output voltage is reversed relative to the input. In some uses this doesn't matter.

The other obvious one is using a linear regulator, I have read if you do the filtering right you shouldn't need a linear regulator to reduce ripple. In my experience nothing works better than a linear regulator.
Obviously power, heat and pass/rejection frequencies would have a big influence on any descision but I would appreciate some discussion as to when to use passive LPF vs a linear regulator which I guess is basically some sort of active filter.

[Psi]

Yeah, the module has to be small so a linear reg is out due to heat issues, unless i use a tracking pre-regulator.

I've done a few test on it with quick and dirty RC filters and i can get rid of the lines on the video with enough filtering. I was just wondering what sort of filter values/design was usual for removing switchmode noise.

RC filter isn't practical due to the vdrop vs current, so LC will be the next attempt.

The output needs to be the same polarity, so CUK is out.

[nctnico]

A common mode filter works even better than just an inductor.

[T3sl4co1l]

Assuming it's common mode noise that needs filtering, of course.