12V smps EMI-filtering

[zilog]

This smps is not intended for commercial applications, but I still like the idea of constructing things that are follow general guidelines.

I have a maximum average input current of ~70A, the mostpart of the ripple on top of this is hopefully caught by the 5x parallell 1000uF electrolyte capacitors feeding the push-pull stage. The switching frequency is 2x53kHz, the transformer is thus fed by a square wave pulse-train with 106kHz frequency.

What I can see from searching the forum, and reading literature, the most common way of filtering the input of such a smps is through a CLC-filter. What frequency range must I suppress, and how much? Furthermore, most inductors of this size have quite low self-resonance frequency and therefore the filter will stop filtering above a few MHz - is this enough to keep conducted/radiated EMI below regulations? Common practice seems to be to parallell the bulk capacitors at the output of the filter with a low-ESR/ESL ceramic capacitor. Should I have or not have this stage (i.e. are my 5x1000uF caps enough?) I believe in having some inductance at high frequencies in the switching loops since it makes switching transitions less harsh, and in my mind it creates less EMI.

I have an aluminum metal casing in mind for this project (the local shop sells nice heat-finned ones). To get a better PCB layout, I want to place the to220-mosfets and rectifier diodes directly beneath the PCB, screwed onto the aluminium bottom plate. This will cause capacitively coupled EMI from the transistor tabs to be coupled into the casing, at a distance from the point where I connect the input ground for the differential mode emi filter. Will these currents give raise to any measurable EMI outside the casing? I dont have a single millimeter height left in the casing that allows me to add a aluminum plate that is isolated from the bottom, so I must mount the semiconductors directly to the bottom plate, or route them to the side of the PCB where I can mount them to an insulated heat-transfer. What would be the best route given that I have access to 2-layer PCBs?

When it comes to the CLC-style input filter, it is said that this filter should be directly where the power cables enter the casing to avoid radiated EMI to be picked up by lengths of these wires. I would very much like to have the filter inductor mounted to the PCB, will this cause any degradation in EMI performance, as long as I have the ceramic input capacitor directly across the chassis mounted terminals?

What kind of inductor should I use? I lean towards using a -2 material toroid because of the high DC current bias, but rod inductors should also be able to carry significant current because of the large air gap. What worries me is that the fluctuating magnetic field from a rod inductor should couple into surrounding components, what is the case with this?

[Electrone]

I commend your attention to details concerning EMI. For the CLC filter, I can see merit in making the outermost capacitance have especially low impedance at high frequencies.