Wild_Bill wrote:
....
I was thinking that the primary windings may be better than the
secondaries, but you likely know better, and the number of turns on the
secondary side is much greater than the primary.

It's basically a volts-amps choice on the supply. High voltage/low amps
using the secondary; low voltage/high amps if the primary. The winding
ratio is about 17:1 (2000:120), so for the same amp-turns, you'd need 17
times the current using the primary as using the secondary.

...
If one had a significant number of identical MOTs, doubling either
winding onto the E-core may be possible.

I did this once to get an isolation transformer: I put 2 primaries on
the same core. It took a long time to find 2 MOTs with the same winding
size. They are very closely fitted, so there is not much margin. I now
have 14 extra MOTs as a result of that search (i.e., the 14 are ones
that didn't match anything).

A second aspect is the mating surfaces of the E-core sections of the
core to a separate piece of metal.
I'm thinking that the most effective coupling of the two surfaces would
need to be very precise for the most effective magnetic transfer to the
steel bar.. possibly a very close tolerance smooth/flat surface that a
surface grinder might provide.
Attachment of the E-cores could then be held securely to the steel bar
by tack welding the E-cores in position.

I agree. But the MOTs are already pretty flat - they don't want to have
air gaps either. I would fasten them in place & then surface grind
(Blanchard would be good enough).

I believe that winding the coil to fit directly on the lower section bar
would be the best solution, for maximum efficiency.

I have wound a couple of windings for customized MOTs and it's not
something that I want to do a lot of.

Bob