I won't go into the theory, so this is an inaccurate representation of what we can see, rather than calculate the cross sectional area of the core, and area of the loops etc etc.... I aint that smart.
So I will explain it incorrectly but for the same result.
Looking at the three stack above.. look at the length of wire in each turn in each of the primary and secondary ( black wire secondary here).
Because the primaries will be parallel, nothing changes there, and they can be considered as a single winding , three times the current of a single one... just as we would expect.
Each primary has only 1/3rd the total length/height of the "new" torroid... ( only talking about the length in the core area, forget the rest).....
Now look at the black secondary, its length is three times the length of each of the primary bits in the core, and exposed to the same field all the way through..... so it is effectively three electrical turns compared to the primaries for each real physical turn... so in effect the turns ratio of any one of the primaries to the single secondary is the same as in the original if we divide the secondary by three.... as all three primaries are parallel, this holds for all of it.
More properly we would look at the core area.... the primaries are only wound around 1/3 the core than the secondary is wound.. so the secondary will exhibit three times the EMF than a single turn of the primary would do.... so the secondary needs to be divided by three to keep the "turns ratio" the same, as they are not in the same environment.... one of those explanations should get the ahha reflex going.
Your turns ratio that you are talking about only holds true if the turns are on the same core... here we have pulled a shifty ... and the core of the secondary is three times the size of any primary..... so different frames of reference really.
We would do this to not have to rewind the whole thing. Some folks don't want to unwind the 200 turn inner core, and so this helps them get some better efficiency... we can see that a full rewind allows us to use 1/3 the number of inner turns too, and then the same physical ratio as the original will be in place as they will share the same cores this time..... but look at all the wire you save then... gets worthwhile to do the whole rewind...... trust me.... I'm an idiot....
Here it is in real life
First the two transformers:
Then we wrap them together with tape after pulling off the 130v winding ... keeping only the 240v winding on each core:
Then wind your secondary for 48v for two cores... note the number.
Then we do it with three cores... and look at the drop in turns of the primary LV winding now
Actually that has full rewind of 1/3 the 240v original, and 1/3 the LV original I would have needed.
Frackers.....
I would try to get the 15kw driver and a complete 15kw power card.. the heat sinks don't weigh very much anyway, but would be a pain to reproduce properly.... I have seen those boards advertised last week on the USA site I think.... they were not obvious, and may take some ebay skill to find.... that way it is an off the shelf 7-8kw unit.
edit: found it
http://www.ebay.com/itm/Main-Board-Control-Board-for-LF3000-15000w-pure-sine-wave-power-inverter-/121372307145?pt=LH_DefaultDomain_0&var=&hash=item1c425a6ec9....................oztules