# HVPS overall design
The design of the high voltage power supply is a little unconventional, but
by convention, most are. In order to simplify the transformer design, I'm going
with a hybrid structure, with an unregulated H-bridge inverter fed by a
regulated buck converter, all powered by the system's 110V supply.

- The **buck converter** reduces the 110V input to around 50V, to allow
regulation of the outputs by adjusting this voltage. If it looks unusual,
that's because this buck converter is upside down — that's just done to
make it easier to control, with a low-side MOSFET and current sense resistor.
The inverter already requires gate drive transformers regardless of where it
sits on the rails so there are no real drawbacks to this approach.
- The **inverter** uses a MOSFET H-bridge to create a 100Vp-p square
wave from the output of the buck converter. A frequency divider creates its
25kHz drive waveform from the 50kHz buck converter drive. This keeps them
in phase lock, preventing beat frequencies from developing between them.
- The **transformers and rectifiers** multiply the square wave by a ratio, then
rectify it. The transformers are kept as simple as possible so there is plenty
of space for good isolation; both transformers must handle 2000V (for the
filament transformer, in the form of isolation voltage). +8kV is developed
by an external multiplier pumped by the 4kVp-p AC output from the
transformer.
- The **control circuit**, shown below (not present in the block diagram),
generates all drive waveforms and provides the control loop to regulate the
output voltages.
## Control circuit

The control circuit is based around a [UC3843](https://www.ti.com/product/UC3843)
PWM controller. This old chip provides a synchronizable oscillator, a voltage
reference, error amplifier, current-mode sense comparator, and an output
capable of driving a MOSFET directly.
Regulation is taken from the +2kV output, through a voltage divider and into
the voltage sense input of the UC3843. An external frequency compensation
circuit and slope compensation circuit (latter not shown) keep the control
loop stable.
The 50kHz PWM signal from the UC3843 is fed to a
[4013 flip-flop](https://www.ti.com/lit/ds/symlink/cd4013b.pdf) configured as
a frequency divider, which produces a 25kHz square wave at 50% duty cycle
regardless of the pulse width driving the MOSFET. This signal is buffered
by a gate driver IC, producing a strong 24Vp-p output that is then
isolated and split by a pair of gate drive transformers and sent to the gates
of the H-bridge.
## Transformers
Two transformers are used by this circuit:
### High voltage transformer
(Detailed posts to come on the transformer construction!)
The high voltage transformer is a fully custom transformer, with a single
primary and a single secondary, wound on a pair of
[Fair-Rite 9277012002](https://www.digikey.com/en/products/detail/fair-rite-products-corp/9277012002/8594176)
U-shaped cores. It uses a sectioned winding for the secondary, with the full
2000V output split into six 333V sub-windings separated by pieces of insulating
material. This reduces the voltage stress seen by the insulation of the magnet
wire to no more than 167V across one layer of enamel (remember that there are
two layers of enamel between two adjacent wires).
The filament transformer does not produce high voltages, but it needs to be
sufficiently spacious to isolate the primary from the secondary to a working
voltage of 2000V. The current plan is for this transformer to be a modified
[Signal HCTI-1000-2.4](https://www.digikey.com/en/products/detail/signal-transformer/HCTI-1000-2-4/7362991)
toroidal inductor, with a secondary added using high voltage wire. Whether the
ferrite material is suitable for this use remains to be seen.