Helpful Hints Updated March 2006
Take safety seriously. Be very afraid of getting an electric shock. Some of this stuff is one touch leathal. Beware of capacitors which remain charged even after the power is switched off. Put resistors across big electrolytics to discharge them safely.
If your new to this then your first design should be small and simple. Look at everybody elses designs first.
All my previous SSTC work used mosfets but increasing igbt's are being used or even the brick igbt seen in my OLTC. These are simply getting bigger and faster and cheaper. Mosfets still have an avantage at over 200kHz but the general rule now is to use igbt's and hold the frequency to under 100kHz.
To avoid complaints from the FET preservation society - Experiment in a sound proof basement so no one can hear them squeal.
Things I learnt after blowing a couple of dozen FETs. V=L di/dt (calculus :) di/dt is say 40A switched off in 100ns, a very big number. L is all the stray induction, a little number. A little number times a very big number is a big number (advanced calculus :) So V could be big - could be goodbye FET. The output capacitance of the FET helps here though. So keep wire lengths short minimise stray induction. Make the design as compact as possible. Resistors especially any low value resistor used for current sensing must be low inductance types. Wire wound implies a coil.
Use a current inrush limiter across the mains to avoid the switch on thump from big mains transformers. (Farnell Components lists these with the thermistors)
Bridges and half bridges are much better than pushpull designs. Push pull is like a see-saw. The fulcrum is the centre tap to +ive. Now pull one end down to the ground and the other end goes up! So choose FETs that can comfortably handle double the supply voltage.
Think about transformer saturation - more voltage - more primary turns, lower frequency more primary turns. In push pull - the two sides are never completely equal - over a number of cycles the core becomes magnetised in one direction and saturates. Use lower flux densities and current limiting. Good news - its heavenly once it works properly - I haven't blown any fets for a year - due to withdrawal symptoms I'm now messing with another design.