Why steel? Unless you need magnetic shielding (simple steel won't work), Aluminum is usually easier to work with. Copper or copper-clad FR4 is sometimes easier for a one-off, because it can be soldered easily. Copper tape is also very useful for prototypes.
Why? This is almost always a *bad* plan. It gains nothing and is a royal PITA. Internal planes work (one ground, one power) and still allow access to the signals.
It's pretty straight forward. Anything over 1/20 wavelength is too big. Are you looking for a commercial solution?
Do you have problems with the SW when the RF generator is on, or do you just see things on the supply with your oscilloscope? The oscilloscope probes may show RF on the supply while it may not be there.
If it is there, the root cause can be direct electric or inductive coupling and/or coupling via common mode current (that enters via the cables).
To measure whether there is RF on the supply, you must measure directly on the points where you expect it. Don't use the ground lead of your probe, but use the circular ground around the probe tip and make contact to the suspected point directly (no wire in between the probe ground and the circuit ground). You may also add some ferrite cores around the probe cable. When you see difference between with or without cores, you have a measurement problem.
At these frequencies, it are many times the cables (or cable terminations) that cause the trouble. Best is to enter all cables via one side and decouple them with ceramic capacitors with short traces/ leads to avoid parasitic inductance. For data lines, decouple them with RC or RCL circuit to keep out the 27 MHz. You might experiment with ferrite cores around the cables, experiment with the number of turns through the center of the ferrite core (clip-on types are very practical in this case). When you notice difference, you can be sure that you have a so-called common mode problem. For the power supply, a
10nF MLCC with short traces between the supply lines will kill all possible differentially induced interference, but will not change the common mode interference that may cause a problem elsewhere at the PCB.
When it is direct electric/magnetic induction, you might experiment with copper foil or tape above and under the traces where you expect the RF. Ground the material directly at the ground nearest to the foil (no long wires between foil and ground).
It is recommend buying a practical book on EMC (for example EMC for product designers, Tim Williams, Newnes). The difficulty with EMC is, when you block all but one interference path, you may notice negligible difference (or even increase of interference effects). The effect may be visible just after you block the last interference path.
If the power supply secondary is floating (no connection to case etc.) use a sufficiently thick twisted pair and run it through a ferrite (both wires) close to the micro board ?
Does the power supply suffer from shutdown or erratic output voltages when the RF is present ? The RF is then entering through the DC output terminals and into the voltage sense feedback path, driving the differential amplifier into saturation etc. Some ferrites on the power supply side of the DC cable will usually help and some chip capacitor across the DC terminals as well. Of course the obvious place to filter would be in the differential amplifier inputs, but this may cause loop stability problems, unless you know what you are doing.
In the presence of a strong altering magnetic field, you should avoid any large wire loops, since the induced current is proportional to the loop area. If you have a row of buttons on the front panel, do not run a common ground wire to all buttons and separate wires to the button hot side, since in the worst case, there will be a large loop, when the button is pressed at the opposite side to the common ground wire. Instead, run separate twisted pair wires to each button separately. A chip capacitor at the exact point where the twisted pair enters the PCB will also help.
In general, use differential signaling (RS-422/485 or 20 mA current loop) when possible (instead of RS-232) etc. to simplify filtering. Optoisolation also helps in avoiding loops and antenna effects (capacitive coupling).
More important is avoiding loops with large areas and reserving sufficient space for filtering components on the perimeter of the PCB to act as the first line of defense.
The frequency is quite low (wavelength 11 m), so this should not be too hard to keep it out, but of course, you have to be careful about selecting a suitable ferrite material, in order to achieve a sufficiently large inductive reactance at that frequency.
When you place the coil to the material to be heated, is there a large metallic object in the vicinity. It "smells" like a common mode problem in that case. Even a good balanced inductive antenna loses balance when in an inconvenient orientation with respect to another large/long metallic structure.
You observed resetting of the CPU, is this because of the supply or via other inputs to the CPU/PCB. I think if you want better help via the forum, you should provide more info (picture or drawing showing all cables, PCB, cable routing, etc).
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.