Graham Armitage wrote ...

Wow - this is fantastic. Quite a project you have taken on. Having all this intel in one place will be invaluable. While your knowledge far exceeds mine, I would be happy to proof-read and provide feedback. Were you thinking of adding sections for Secondary Coil and Top-Load - size ratios, wire guide, materials etc. ?

Sigurthr wrote ...

You should put in about the pulse width, tank voltage rating, and switch current peak ratings as well (all three are mathematically related). This is a subject not well covered in most resources and a major stumbling point I ran into a while back. I eventually wrote up a simple C++ calculator program to do the math for me, which has proven indispensable when roughing out design parameters. (I can link it if requested, but don't want to hijack your thread).

The short of it is if you choose too low of a voltage or current rating you'll severely limit the maximum pulse width to the point that you cripple your system.

Uspring wrote ...

A very nice writeup
In the part about IGBTs you calculate the duty cycle by dividing burst length by burst period. Is this correct, since the IGBT is only on during half the RF period?

Wolfram wrote ...

Excellent initiative, finally all the information on DRSSTC construction is collected in one place.

I'm reading through it at the moment, and I'll write comments here as I go along.

For the busbar chapter, where you talk about switching spikes, I don't fully agree with your formula. The formula you give will give you the AC voltage ripple on the bus caused by the inductance, but the switching spikes can be much worse than this. The switching spikes are given by the same formula V = L*dI/dt, but with dI/dt given by the IGBT switching time and the conducted current at the switching instant. The conducted current at the switching instant is given by Ipk*sin(switching phase angle). I'm not entirely sure if there are other mechanisms at play as well, but I think this is the major one.

It could also be good to mention potential losses from eddy currents induced into materials near the primary and primary wiring.

This is hard to find good information about, but copper tubing is normally phosphor deoxidized, so it has significantly lower conductivity than pure copper.

Phosphorus is the most commonly used deoxidant for copper but does have a deleterious effect
on the conductivity of the copper which will be around 92% IACS at a phosphorus content of
0.015%, reducing to about 78% at 0.05%.

Luckily, when used for high frequency stuff, the skin depth increases with lower conductivity, so the problem is not as bad as it first seems.

Excellent point about the switching losses being mostly uncorrelated with switching times.

Do you have a diagram of the internal circuit of the IGBT? It could be useful since you talk about the internal MOSFET, which is not explained elsewhere.

I really like the discussion on IGBT switching speed, some very good points on a topic where there otherwise is a lot of confusion.

Assuming that switching losses will decrease with increasing gate voltage is maybe not realistic.

I'll have a look through the rest later, but generally there's not much to criticise, it's very solid work and well explained.

Sigurthr wrote ...

I'm reading through your guide so far, Mads, looks great. Might I recommend you change "Vripple" to "Vtransient" on the BusBar page to avoid confusing inexperienced readers?