Feeding a Full-Bridge from a Boost Converter?

When you have a large step-up voltage and a fair amount of power to transfer a full H bridge is the best way because your turns ratio is the least required of all the topologies. So thumbs up for that decision.

Also, when it comes to this sort of application, controlling the DC level and sticking with 50:50 squarewave control is not only simpler but more efficient. I've tried PWM but I've had resonance problems (leading to extreme losses and over-heating) with the high-turns secondary and had to despatch circuits to the graveyard. So, in my opinion thumbs up on the fixed duty cycle, variable DC method for control.

So, do you boost 12V to 48V and reduce turns ratio by 4:1 or go straight for the 12V control. Your turns ratio from a 12V supply is going to be based on a primary input of 24Vp-p yielding 700Vp-p with a turns ratio of about 30:1. If you used a little bit of capacitive resonance on your output winding to peak the voltage transfer, you might happily find that 25:1 will do for input voltages down as low as 10V.

My conclusion is that I'd stick with the full step-up from the 12V buck regulator because it's likely to be more efficient. Also, under no-load conditions you might find that you need to "buck down" to maybe 2 or 3V - that's maybe 20% of 12V - how would you get 20% of 48V from a boost - it would be switched-off and spluttering and you'd find that on very light loads you may not be able to control the voltage low enough to prevent the output dc rising significantly above 350Vdc.

I've designed products for similar voltage and power ranges. The answer to your question is: absolutely it's feasible, but in your case it may not be necessary.

The reason you can't regulate the voltage by adjusting the pulse width on your transformer is the capacitor on the transformer secondary. I haven't worked out all the math, but if you put an inductance between the secondary winding and the filter cap, I think you'll find that the system regulates exactly as expected. You'll find that your secondary now closely resembles a standard buck converter.

Now, doing that may cause other concerns. The recovery kicks on the secondary diodes may become prohibitively high, depending on the diodes you're using. That's what stopped me, but I was running 8kW at 600 volts, so you may not have that problem.

My solution was to run two stages, more or less as you describe: big dumb isolator stage, followed by a regulator stage. In my case it made more sense to have the transformer stage run directly off the low voltage, and then have the regulator at the higher voltage; having two stages that have to run those high currents would have increased my losses significantly. You may also want to consider that, if you stick with the two-stage architecture.