A QUICK look at the steam you need...
Based on running at 1000rpm at 20psi - which is a complete guess - I reckon the volume of steam per crank rev. = 0.884 cu.in. which needs the boiler to boil 1.22 cu,in./minute:
Now I don't know if I am interpreting the books correctly, as I have not found an explanation of the steam table I have written anywhere.... but I think that needs about 0.9kW of heat. I would allow 1.2 kW of heat as the minimum burner as there will be losses, so "BIGGER IS BETTER". You could manage with a smaller boiler - say 2 1/2" diameter x 6 in long if you are sure you won't need more than slow free-running the engine. - That is "Large Mamod" sized, with the simplest meths burner.
But I would propose a 3in diameter horizontal boiler, 5 or 6 in long, and a single up-and-back loop for steam drier/heater, with 3in by 5 in. ceramic heater beneath it, using a number 12 or maybe 16 jet - with butane fuel - will easily power the engine. Or this could be a simple vertical boiler of the same dimensions and a 3 in diameter burner, but then you would not manage a no 16 jet and may find it limited to do more than just "free-running". (These are what I have already for similar sized engines).
With most engines, the "as new" friction is higher than "run-in" so it is nice to have reserve on the boiler & burner. And particularly with Compound engines a superheater is very beneficial, as that way the steam doesn't all turn to hot water in the first cylinder. (With the remaining "expansion" cylinders
acting as a water pump!). And you need to lag connection pipes between boiler and engine - which should be one size larger than you think and as short as possible - otherwise the steam is half-used just getting to the engine before it gets to warm the cylinder!
A lot of fun. Keep in touch and I'll advise with what I know.
Incidentally, if anyone else wants to teach me more - I'd love to learn! (especially the "energy to steam" calcs). There is a lot of physics in steam plant - but a lot of "black art" as well. And loads of people with experience know better than engineers, because they have been there and done it. I am just trying to find the right way to make the numbers meet reality...
K2