I graduated from college back in the 1994, and even then we were admonished to
avoid magnetics whenever possible. Of course, these days I know better, but
as a result my academic coverage of IF transformers was non-existant.

I'm now trying to make up for that transgression. :-)

I've done a fair amount of reading and have a good understanding on how IF
transformers work, how they should be modeled, how to build them, etc. (Most
of the books that address this in detail are from the 1970s or older, it
seems...) I still have a few questions, though, that I'm hoping a few of the
older reads could help me out on. They a

1) The really big 450kHz IF transformers you see in tube sets... why did they
wind the coils in the form of "pancakes" rather than "the usual way"
(single-layer coils)? Is it just a consequence of needing lots of turns (to
get enough magnetizing inductance) but, for the coupling coefficient desired,
finding that you'd end up with, e.g,. a foot-long tranformer if you only used
a single layer?
2) I can readily see why you'd want a center-tapped primary, or a primary
with, say, a tap 10% "up" as a small feedback winding, but why do you get such
things as an IF transformer with 103 and 50 turns on the primary (on either
side of the tap) and then 27 turns on the secondary? (E.g.,
http://www.mouser.com/catalog/specsheets/XC-600014.pdf ). None of my books
address this, and the only thing that looks close on the web is this article:
http://hem.passagen.se/communication/ifcan.html . Is his conclusion, "by
tapping the transformer the Q value increases" the main reason?
3) Sticking a parallel capacitor on the primary to resonate out the
magnetizing inductance makes sense to me. I'm a little less clear on parallel
capacitors on both the primary and secondary -- a double-tuned arrangement.
Hagen's "Radio Frequency Electronics" assigns leakage inductance to the
secondary and then converts the resonating capacitor in parallel with your
load resistance back into a series circuit and, voila!, you now have a series
RLC circuit so clearly bandpass behavior... but this approach implies that you
could just use a *series* resonating capacitor on the secondary instead. Is
that correct? (I am aware that there are a handful of commonly used
transformer equivalent circuit models, you can transform magnetizing or
leakage inductances and losses from primary to secondary or vice versa at
will, etc.)
4) Anyone have pointers to good books or articles that ideally discuss some
actual design examples of the more complicated cases (weird primary turns
ratios, double-tuned circuits, etc.)? -- The ones I've found so far as the
simpler single-tuned case, just center-tapped, etc.

Thanks a lot... I appreciate the help here!

---Joel