Shielding of power supply.

I've built a Leach low TIM amp and it works fine (I've just had it going in a temporary cabinet for the last few months). Now that I'm going to stick it into a permanent case I was going to have the cabinet partitioned into a "noisy" side and a "quiet" side by a sheet of metal.

On the noisy side I will have the transformers and obviously the circuit boards etc shielded from the transformers on the quiet side. What side should I put the bridge rectifier and the power supply smoothing caps on?
 
Harvardian, you are right. This is how I shielded transformer and made connections in my SOZ PS. The active circuit, additional caps and heat sinks go on top.
 

Attachments

  • shild.jpg
    shild.jpg
    38.1 KB · Views: 680
Aluminum isn't very good at shielding

The cage that a standard PC Power supply is enclosed in is pretty good. I built a switcher/single chip amp in one and it doesn't bother anyone. MuMetal is damned expensive, but if you find a dead oscilloscope, chances are that you have a nice little supply. If I recall correctly, MuMetal changes its characteristics if bent, however. If you really, really need to shield you should have a brass screen on the fan! (Decades ago I worked in a completely RFI free quiet room -- multiple brass screens, brass finger stock on the doors -- we even used lead acid batteries to power the stuff inside! All this to see how some protons changed their spin when monkeyed with). Oh, and btw again, National Semi has hints on shielding on their white paper dealing with EMI/RFI.
 
split the caps like harry says

i put all my caps on my amp's signal PCB (15,000uF per rail), right by the output devices, and i think it was a mistake. i think there was some noise radiation from the ripple currents in and around the caps which in turn caused some audible hum. better to put a good amount of storage capacitance right after the rectifier to take care of ripple, and keep that part of the supply isolated from the signal circuits. then add on some more bypassing right by the output devices to lower impedance, provide instantaneous reserves, and compensate for any lead inductance. in any case the power going to the signal PCB should already be as close to pure DC as possible to minimize noise. i would say put around 10,000uF per rail locally, and put the rest by the rectifiers. that's assumming you have at least 20,000uF total storage per rail. the amp should still be able to work w/o the local supply decoupling, the extra caps by the output stage are just the icing on the cake IMHO.

marc

p.s. mccormack did something like this a while back. they called it "Distributed Node Amplifier" or something. i don't think it was really much more than supply caps spread out over the circuit.