Electronic – SMPS input voltage

The 2N6782 N-channel FET is rated at 100v drain to source at a maximum 3.5A and 15W max power. I picked a through-hole part because the surface-mount part would be difficult to solder by hand.

To drive the FET, you can use the LT4440-5, which takes a logic level in and can drive an N-channel MOSFET switching up to 60 or 80V depending on the version. The supply current is negligible, only a few µA. The driver only comes in a surface mount part (SOT 23-6 or MSOP).

The only drawback is it requires a supply voltage (VCC) between 5 and 15V (thanks to the OP for pointing out this variant of the LTC4440 with a lower minimum supply voltage). Since you have only a 3.3v supply, you could use either a voltage doubler or a simple boost converter (such as the MIC2141) to supply the VCC voltage.

Both the driver and voltage doubler/boost converter should only require a few ma (if that) from your 3.3v rail.

The FET and driver should work up to 1 MHz, maybe a little higher.

The AC to DC part will be simple, just use bridge or full wave rectification and one or possibly several filter caps in series, it depends on how much current you need at that voltage. Then, use a buck converter to scale down the voltage to what you need. You might have an issue with having a buck converter with the full voltage range and current output you need, depending on whether or not you want to use a single IC switching solution or build up your own. Is it critical that the voltage on the output changes immediately? Can it lag a few milliseconds?

The tricky part is trying to scale the voltage on the output. Since supplies that use switching use "sensing" on the output to ensure stability and proper output voltage as well as setting the voltage on the output, you can use this to manipulate the voltage to what you want. The trick is figuring out how you want to do the math to compare the input and output voltages. You could use a voltage divider on the input and output to scale the voltage to something usable by an op-amp, have it go into a buffer first then have it compare the two voltages and use the error signal to drive the switching circuit (likely through a transistor if you need the voltage to be high on the sense input). The problem is you will need to tweak the resistor values of both the input and output dividers to get it to represent the proper output voltage across the whole range. The advantage is that it's closed loop feedback so you will get basically instantaneous response to voltage changes. Also you will need to bypass the crap out of the op-amps and most likely use the input buffers with a low pass filter to prevent wild oscillations or fluctuations in voltage which could be... very very bad. Also be wary of the switching signal possibly adding some noise, hence the low-pass filtering.

The other solution for controlling the voltage is simply using a microcontroller to compare the voltages, do the math, output a certain voltage (which will likely drive a transistor which in turn drives the voltage sense on the switching circuit) which then changes the output. The disadvantage is the programming, sensitivity to voltages if something goes horribly amiss, and response time of the microcontroller. The advantage is you can read in actual voltages and set up parameters you want to control, or even control the rate you want the voltages to change on the output etc. Offers some flexibility to you.