Ok, responding to your post above. Easy stuff first.
On the snubber, it's intended to get rid of some ringing and slow down the output edge foolery during deadtime. Since very fast output edges wind up coupling back into the gates, you are trying to kind of slow the thing down -- rise and fall times can otherwise be very fast when driving an inductive load. In other words, tuned below resonance a little, which is where all mine work best. As I noted elsewhere, I'll often put at least some of the primary turns on the other leg of the core to get a little more leakage L, makes things much easier to drive.
Of course, this depends heavily on the magnetics you've got, and sometimes the easiest thing to do is play with or rewind the primary winding, it's usually easy and not many turns.
If you're seeing no effect with those numbers, there can be a couple of reasons -- and you should see it even at low inputs. One possibility is that your resistor itself is really inductive, so the fast edge coupled in via the 1k pf cap simply doesn't see a load. Some WW resistors are really bad that way, and it's getting hard to find big carbon ones these days, which are much better. Of course, you can make a big one out of some smaller ones. You'd like to see a few watts in the resistor (at full voltage), and if it's not getting warm -- it's either really inductive or it needs to get smaller, and perhaps the cap a little larger. You probably don't want to tune the RC below or too close to the 3rd harmonic of the base frequency, or it will simply eat too many watts.
I want you to try something for me here. Short your probe, and hook the scope to some big square wave (such as a bridge output) and look at the resulting trace -- I want to see if your scope can really be floated on a big SQ wave, because if you really can, heck, I want one, I've never seen one you could really do that with without a lot of spurious stuff making it into the trace when you should see nothing (note, by doing this you make a tiny one turn inductive pickup, so don't do it near a wire carrying a lot of current). If that's happening, then some of the other strange things I see in your pictures are explainable that way.
Some of the gate drive waveforms look very-very strange -- I am trying to understand why there is stuff there during the on time for example. Nothing in the chip is happening, nor should there be anything happening on the output during that time, so where's all the junk (rather than a flat top) coming from? See above scope test -- you gotta be where you can trust your tools pretty absolutely.
Your "high gates ringing" might be almost entirely due to a scope issue here" It's actually sort of hard to make that happen unless the output is ringing hard and coupling back there via the fet capacities, so I find that slightly suspicious -- a possible measurement artifact that isn't really happening. It's actually pretty safe in this circuit to assume the top drives look pretty much like the bottom ones once your bootstrap cap issues are taken care of, and they should be at this point. Your extra diodes may be causing a little of this by allowing more outside the rails stuff.
I know the book says don't use the fet body diodes, but I do, and they are not all that bad these days with the more modern parts we have. They may cause a slight bit more of fet heating, but if you've got thermal headroom, it's not a big deal. The fact that they are a little bit slower than the best diodes you can get isn't all bad -- helps control the output waveform some, though the snubber is
supposed to be doing that job. But it's not necessarily bad to share that job around (spread the pain/heat).
Though perhaps not needed in either of our projects, this trick is often used in cases where the self resonance of the transformer is just too low.
- Tuning trick
Here, L1 is used to raise the effective resonant frequency of the magnetics. You can only take this so far without running into big losses, because it's the secondary doing the resonating and running it above that frequency means resistive losses in that winding as it tries to drive its own capacitive load, but it might get you "enough" without being too bad. Obviously, the inductor has to not saturate at full drive for this to be good, and it has to be not very lossy itself, so it's a big chunk of extra expense. The series L-C is tuned to resonate at the drive frequency (which is about the same as the new net output resonant frequency). This does two things for you. One is you get the capacitor to keep from having the magnetic walk in the main core if there's a bit of DC on the bridge, with the series L2 canceling out the capacitor's reactance, and the series L2 also acts like a higher impedance back to the bridge for both the fast edge on your desired signal, and any blow back from arcing on the output. You don't want this to be very high Q, else the voltage at the junction of L2 and C1 will go wild, and may even need to make L2 delibearly lossy (or put some R across it, but that's going to eat some power). Wouldn't be surprising to accidentally make many times the normal signal voltage at that point, so you have to be a little careful choosing the L2/C1 values from the infinite number of combinations that create a particular resonance. Perhaps Cliff will chime in and mention some good rough rules of thumb here, as I don't normally do this and don't have a lot of experience with doing it this way.
There are several ways to help with burning fets and chips during arcing situations (but it's best to not arc of course). One is the zeners I mentioned above. MOV's generally aren't "hard" enough, you have to go way overvoltage on them to get them to do very much, and you probably don't want to choose them so low voltage as to draw current when things are normal.
Spark gaps, though, are fast and can take a certain amount of pounding, so if you can find some that don't "go" at your full rails voltage, then hooking them between each output and each rail is good (eg you need 4 total). You could try zeners there too, but I've not had great luck with them, and it's hard to find ones that will take really big peak currents, no point in putting anything there that will fry (and fail shorted!) itself. You can count some on the big filter caps and a simpler catch diode circuit to keep the outputs inside the rails, and you just might find some fast ones that turn on before or at the same time as the fet body diodes without needing the one in series with the fet drain. Most big electrolytics do have some series impedance, though, which is why you often see the use of a bunch of smaller ones in parallel -- this results in less ESR and other stuff, not to mention better cooling. You'll find your filter caps getting warm under some tuning conditions due to the ESR and the catch diodes shoving quick peak current pulses back into them.
I would be trying to run the chip as close to the 15v as you can. I like using a 3 terminal regulator (7815) for that job, so I don't have to depend on heating the zener in the chip, or a resistor outside it -- the 7815 is a lot easier to shove on a heatsink anyway. The internal zener is a little above this, so you should be fine, but in my board I made provision for a 50-100 ohm series R from the 7815 to the chip just in case. You get a free current limit this way too, a nice thing to have anyway.
Even if you don't want the ability to have a constant current output (I really like having that, by the way, which is why I'm going to try and go down the path suggested by JohnF, with a current fed topology and a current limiter in the input switcher) you do need some kind of fast acting way to shut this down under arc/short conditions.
I'd suggest a current sensing resistor in one of the main supply rails, a series R to an optocoupler (else a big spike will just fry that) and using the output of the optocoupler output to drive the chip shutdown pin, for probably the simplest way of doing that. Yes, it's annoying to have to power cycle to reset the thing -- but it's more annoying to have to change out parts. On my fusors, it takes awhile of "conditioning" before little sparks and arcs stop happening of the sharp edges of the grid, and the "arc warning" light on my Spellman goes constantly for awhile. I had to disable its shutdown for that or I'd never get through that phase which happens pretty much every time the system has been up to atmosphere, and especially whenever I put in a new grid.
Especially until you get really good gas control (and that will take some doing, as I've found) a fusor can easily get into a kind of avalanche mode, where the more current that goes in, the more gas gets ionized, so it draws more current, and eventually winds up as a near short circuit. A larger value ballast helps there, but you can't make it so large as to completely solve that problem unless you're getting it so big it's going to need a fan and so on -- and when not needed, it wastes a lot of power when it's big valued.
There is a link to a very nifty reactance/resonance slide rule here. I use this all the time, it's a big time-saver when calculating snubbers, resonance problems, an RC lowpass filters, to name a few. It's worth the hassle to make one from the images there.
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.