We don't like Mag Amps in the power industry because they generate a lot of harmonics. And reducing line harmonics is a big deal these days. For a 100 KW power supply I'd just eat the extra costs (from the power Co). At 5 MW the harmonics are a little harder to eliminate.
I really like the Digitec stuff because it is a standard buck converter - thus low losses. And if you need more power you just add another phase (or ten) to the power supply. At $.25 a watt the cost is not too bad for all the wonderful goodness you get. And the best thing about it for now? COTS!
But hey - a 40 MW DC supply for $50K or even $100K It could be well worth it. BTW are you sure it is 40MW and not 40KW? The Cu in the transformer alone has to be worth more than the shipping.
Well the specs say this...
Serial Number = RBR64991
Manufacturing date: 1965
Voltage: 67 / 12.47 kV
MVA Rating: 30 / 40 MVA
Phases = 3
Core and Coil = 68,880 lbs
Case = 34,000 lbs
Oil Gallons = 3780
Oil Weight = 28350 lbs
PCB Report = 16 ppm
I think a smaller transformer could be found eventually.
Jeeze - mag amps. I haven't had serious contact with them since my Navy days. Talk about solid state. They were a fad for a while in switching regulator circles. Think MagAmps at 100 KHz. You don't see that sort of thing advertised much in the power magazines these days. Every one wants to go solid state. IGBT is where the big power action is these days. I saw one rated at 6,500 volts and 600 AMPs. Now of course you have to derate them for actual use - 4,000 volts and 400 Amps is about right. About 30 of those and you are up to being able to control about 4 MW. Of course keeping the semiconductors cool is a trick. Twenty five of them fed DC and you have 100,000 volts at 400 Amps at your disposal. There is your 40 MW.
I think the real deal is going to work out like you say. I'm just postulating a poor boy Mad Scientist version of a Test reactor that can prove the concept works or doesn't.
In any case, this is sorta what I was talking about.
The Load carrying tube will probably have to be made, and may have to be a bank of several tubes, and even then might have to be shunted with a very large resistor. The whole thing requires a MicroController with AD and DA conversion etc. If the thing turns out to be non linear ( a very likely prospect ) it's easier to just write code to deal with it than it is to work out the headaches in hardware.
There are several design issues with this approach, but I believe they can be resolved with enough number crunching.