I learned something recently that might interest the assembled multitude: several thin conducting layers do a much better job of low-frequency magnetic shielding than one thicker layer.
A couple of years back we did a very pretty diode laser controller that runs off +5V. On one credit-card-sized, 8-layer board, it has four switching regulators, an ultralow noise laser diode driver, a Class H Peltier driver and temperature controller, and an MCU for control and communications.
The switchers and MCU are on one side and the analogue stuff on the other. We put the laser driver under the MCU and the TEC stuff under the switchers, because the TEC leads can be filtered easily since they're so slow anyway. The switchers are underneath a tinned steel shield, but there's no iron underneath, just the PCB. The PCB was an eight-layer job with blind vias, which improves shielding a lot since the ground plane is featureless.
Despite switching a few amps at 2.15 MHz, there's no measurable switching junk in the laser supply, even with a 10-Hz IF bandwidth on the spectrum analyzer. (The switchers are locked to the MCU clock, so this isn't a silly statement.)
A similar circuit built on a 4-layer board was much worse. It turns out that having ground pours on six layers below the supply was the golden ticket. (We did that for belt-and-suspenders since we didn't need those layers in that part of the board.)
I took a bunch of data on Friday, suspending a B field loop above one of those cute Murata power modules with the embedded toroid that JL likes so well. The module is soldered to a my usual prototyping setup (a piece of 1-oz two-sided copperclad held inside the lid of a cast-aluminum stomp box by bulkhead BNC connectors). It has a 1.5" x1", two-piece Leader Tech tinned steel can soldered over it.
The B field loop is a piece of RG-174 with a 3/8"-diameter loop in one end; the end of the shield is insulated but the centre conductor is soldered to the shield about an inch back from the end. It's a DC short, but at AC it's a fully-shielded magnetic dipole pickup. Works great, especially for the price. ;)
The module runs at 90 kHz, and the B-field that gets out of the can is basically solenoidal. (A progressively-wound toroid effectively has a1-turn, air-core solenoid in series with it.) You can null out the field completely (50 dB or more) by rotating the loop by 90 degrees. The null angle depends on how far the loop is from the axis of the toroid--you have to tip the loop a bit as the direction of the B field changes. So far so hoopy.
I put the proto in a rubber-jaw Panavise and taped the loop on top of a piece of clear PET plastic sheet, fixed to the tops of the jaws so that there was about a 6-mm gap between the shield can and the plastic. That made it convenient to put in various bits of shielding and see what they did.
The spectrum analyzer displayed a forest of spurs going up to at least10 MHz. (I was using an HP 89441A FFT analyzer without the RF section, and that's as far as it goes.)
The first thing that helped a lot was to solder the lid of the can to the frame. That was good for 15 dB or so.
I have a roll of 3-mil Kapton with 0.5-oz copper on one side, so I cut2" x 5" strips of it and insulated the copper with packing tape so it wouldn't short out the input and output pi-network filters. I did the same with some 10-mil nickel plated steel from a 70-mm film can and some 1/16" 1100-T0 aluminum.
(Parenthetically, when I was looking at the kickout from the supply input, I noticed that it got much _worse_ when I closed up the stomp box! Apparently the eddy currents in the box material were coupling the stray field from the toroid into the input circuit.)
Getting back to the multilayer thing: each 0.5-oz (17 micron) copper layer knocked the measured B field down by about 11 dB, at least until other factors took over at around -40 dB. The 1.6-mm aluminum sheet was about equal to five layers of copper, despite being almost 20 times thicker overall.
The multiple layers seem to work best when they're insulated from each other, as you'd expect.
I thought I was pretty good a grounding and shielding, but this was an effect I'd never come across.