Terry suggested that I use Manhattan style construction to protect my IGBT drive circuit against EMI that may happen inside a welding machine.
I am reading about Manhattan style construction and have to wonder about something. The boards inside my welding machine, including the ones mounted next to the high frequency arc starter, are regular printed boards. I will open my welder and look at them again, but they do not look different from any other boards that I encountered in various equipment.
If that is so (I will verify that tonight or tomorrow), then is my circuit special in some way (say, due to higher switching speeds) and different from the rest of the boards inside my welder?
Also, note that there are two compartments inside my welder. The compartment with the high frequency arc unit, reactor, and commutator, and the controller unit with all boards, and the compartment with the primary transformer and fan and primary supply power contactor. These compartments are separated. My circuit will be installed in the transformer compartment, mainly because that's where there is more space and ventilation.
The main thing I would suggest is to minimise the area of the loop around which a current flows, since a larger loop area produces larger inductance, and also more magnetic field which can lead to interference. I would also suggest that where there are small signals, logic signals, etc. that you always remember what the reference of 'ground' of that logic signal is. Think about the loop formed by the signal source, the signal wire, the signal load and the ground connecting the load circuit back to the source circuit. If you keep the area of that loop small, and keep that loop well away from any conductors carrying high currents or especially rapidly changing currents, then you will have more success.
If a terminal carries a large current to one place but is also the reference for a low level circuit in another place, then try to minimise the common impedance by spiltting the signals apart and running separate wires. (e.g. the emitter of an IGBT is connected to the power source which is a high current path, and also connected to the gate driver circuit where it is the reference for the gate voltage which is a relatively smaller more sensitive signal. Therefore you need two separate wires for the emitter of the IGBT, and they only join where they get to the IGBT.)
For the logic circuits that you are making you could probably use ordinary vero-board (I think you guys call it strip board) which has a lot of holes and copper strips that connect rows of holes together, and which you can break up with a 1/8" drill rotated between thumb and forefinger. As long as you mount any logic boards etc. inside a closed metal box with capacitors bypassing any supplies right where they pass through the box, and take care of any inputs and outputs by either slowing them down with capacitors or by using coaxial cable, you should have no problems, even inside a welder. The metal box will not allow much EMI inside, as long as you protect the wires coming in and out of the box.
For very high frequency low level circuits, you can get good results by finding a piece of fibreglass-epoxy panel which has one side totally covered with copper foil, which serves as a ground plane, in other words a fairly low inductance reference voltage for many components. Note that by saying ground, I don't mean that it is connected to the safety ground that goes to a stake in the earth, just that it is a reference against which other signals are compared in the circuit. You can solder the ground pins of the components directly to the copper, and solder the components together up in the air, or where necessary for robustness make solder junctions out of small (1/8") squares of epoxy - fibreglass glued to the big sheet. I actually use small squares of double sided copper clad epoxy fibreglass and I solder one side of the little square to the big ground plane, and solder my components to the other side. You can support the power supply pins of ICs by connecting them directly to a ceramic capacitor to the ground plane, thereby providing good decoupling as well as some mechanical support. This is a good way of prototyping but not as robust as a proper printed circuit board.
As a rule of thumb, remember that 1mm of wire has an inductance of about a nanohenry (10^-9), or 1 inch is about 25 nanohenries, and that you will get an induced voltage proportional to the rate of change of current times the inductance, e.g. if you switch 100Amps in 100ns then the rate of change of current is 1e9 amps per second, and you will get 25 Volts induced in a 1 inch length of wire, even if it is very fat wire. You will also get nearly as much voltage induced in a wire that is right next to the one carrying the current. It is a real voltage which can really blow things up or stop them working. You can measure this with a scope probe where you clip the ground clip onto the tip of the probe and hold it near a wire carrying large currents, but wrap it up in insulation tape before putting it near your welder in case you touch something!