I recently acquired a Pharmacia Biotech EPS 3500 XL power supply for 'next to nix' and thought a mini-teardown might be of interest.
It can generate up to 3500V at 100W and monitor the current to within a microamp. Unfortunately the shrouded 2mm connectors for the outputs are expensive unobtainium, so I haven't (yet) been brave enough to run it up to its full capacity. Fortunately some multimeter probes reach nicely, so its good up to 1000V.
The instrument was designed and built in Sweden and dates from the mid nineties. Apparently it was designed principally for 2D electrophoresis using isoelectric focusing.
I also have an EPS 300 from the same manufacturer. With a 300V maximum output it is rather less exotic, although I could probably do a teardown if anyone is interested.
Here is the EPS 3500 XL feeding 1000V to a multimeter. Meter input impedance is 10 megohms, so the 100uA measured draw is spot on:
The back of the unit. The fan is a Panaflo FBA08A12W, curiously the exact model used on an HP E3631A PSU which I am currently attempting to repair:
The main PCB, proudly designed by Kenneth Lindh. The large power MOSFET at the back is a 2SK1359:
And the backside, with some appropriately tall standoffs:
The processor, a Hitachi (now Renesas) H8/532. The 8 pin chip is a TL7705ACP voltage supervisor. Other ICs on the board are the firmware ROM, an LT1244 switching controller, an LM339 comparator and two TL074 op amps:
The voltage multiplier section. The diodes are BYV26E 1000V ultra fast avalanche types. Electrolytics are all Nichicon 400V 10uF:
The final output stage, linking -1750 and +1750 volt sections with a 4kV 3n3 ceramic. The wires to the output terminals with the ferrite beads attached are stiff and somewhat springy: