There's more on the Heathkit IP17 here:
http://www.sgitheach.org.uk/ip17.html, including the schematic and manual. Heathkit got a lot out of only six tubes - two paralleled 6L6GC for the pass element, a 6BH6 to drive them and for feedback regulation, two 150V stabiliser tubes and a double diode to rectify the supply to the stabilised -300V rail.
However the complexity was the custom transformers. They would be uneconomic to reproduce for a one-off build unless you are equipped to wind your own and are experienced at doing so. The filament supply one is pretty simple, it only needs five 6.3V secondaries (actually three + a CT 12.6V winding), and two of those are for the user's load. It could easily be replaced with two off-the-shelf transformers. Its the main HT transformer that's the P.I.T.A. Its got three secondaries: 175V for the 6L6GC G2 (screen grid) bias supply, that's referenced to the cathodes, 210V for the raw unreg 600V HT rail, via a voltage doubling rectifier (*STEAL* that idea!), and 600V CT for the raw -390V that feeds the regulated -300V bias rail.
Worst case, a 6L6GC needs less than 20mA of G2 current, so in a modern design it would be acceptable to use an isolated flyback converter to derive a 50mA G2 bias supply, possibly with a floating IC regulator to further stabilise it, and the same could be done for the -300V bias supply. That's probably going to be significantly cheaper than the custom transformer. A 220V:220V isolating transformer would probably be close enough for the main HT supply, though it would be preferable to use one with both 220V and 240V primary taps so it could be tapped down a bit.
@Chris Leyson,
No, I was referring to Hero999's sim. However, there is an alternative if you want to sweep a resistor using the .dc command - use a load resistor but edit its value to
R=I(Vctrl) to make a
behavioral resistor, then create a voltage source
Vctrl to control it. Set the series (internal) resistance to
-1 ohms
(Yes the NEGATIVE resistance is intentional to get positive current through the source for positive source voltage), and of course, ground one side of it.
Another approach is not to short the voltage source or set its Rser, but instead label its output node, and change the behavioural resistor expression to refer to that node voltage. However, you then have two names to keep consistent.
One can then sweep Vctrl in a .dc command.
.dc dec Vctrl 10k 10 covers the same range as Hero999's original .stepped sim.
There's a pitfall for the unwary: You'd think you could simply sweep a shorted current source, but LTspice deletes shorted current sources before generating the netlist so the .dc command will always bitch that it cant find your control source!
One fix is don't short the source, instead use a 1 ohm resistor, but that's more complexity