Quadratic DC-DC converters?

10 watts over the full range?

Most of the papers are paywalled, but this one

looks like a boost feeding a boost. Inefficiently. Output can't go much below input.

If an intermediate rail is useful, two separate converters sound sensible.

High-gain Cuk converters can be very efficient by charging up inductors in parallel and discharging in series into the link capacitance, with the usual disadvantage that the output is inverted. The stock isolated Cuk isn't quadratic but can fix the inversion issue and the transformer doesn't have to be 1:1, can step the gain up that way, with further advantage it doesn't have to store energy and can be planar or something if needed.

For when it has to be above or below the input this buck/boost looks interesting:

If you need that 5:1 anyway just do it this way.

45:1 at a fixed frequency won't work without skipping cycles which you probably don't want. Our HV sources are guaranteed between 500 and 5000V, they work much below 500 though - but I would not put out such a spec. If you have to do it with 5:1 input range I would suggest you make the frequency input voltage dependent so the load regulation does not need to cover much more than 10:1.

Sepic is the non-inverting equivalent buck-boost. That might do what the OP wants. It's easy to Spice.

The trend seems to be to brag about needing one mosfet, with no limit on how many diodes that costs.

You could do it with a standard converter, then handling the wide range with oversized output capacitor and (you probably need that anyway due to feedback stability at worst case point)

A little wiggle on the dutycycle is ok as long as the overall stability is ok

Regards

Klaus

The general idea of the high-ratio converters is to get the duty cycle away from the extremes of the range where unless the comparator, switch, and diodes very fast the nice output gain and control loop equations become less valid as they're struggling to keep up.

A forward or flyback can do this too but at high turns ratio the primary side switch gets pummeled.

Some extra diodes is OK if they can slum it thanks to the extra gain, but the high-side switch in that particular one doesn't enthuse me. There are lots of variations on this stuff this high-gain Sepic uses two low-side switches, one diode, one L, one coupled L, cool:

But as usual it's not documented particularly well, they fudge a control loop in Simulink who knows what the transfer function of that structure is.

Thanks. I simulated some by Maksimovic and Cuk and although they worked I was not too impressed.

piglet

Yeah, many times multiple mosfets all switched from same gate signal is going to be more efficient and cost only cents more.

piglet

Thanks, I'd like to keep it agile if possible so a very large cap is lower down my list of options :)

piglet

Thanks, good tip.

piglet

Right, diodes have a lot more voltage drop than saturated mosfets.

SEPIC might work for you.

On a second thought I may be overcomplicating things suggesting to vary the frequency. You will still need filtering for the lowest frequency, so may be just go as low as it takes so you can do the 1:45 PW variation. 10W is not much, we have mosfets and drivers nowadays which can switch really fast... unlike during the 80-s when we had to deal with storage times of a few microseconds etc.

Seems like it would be hard to get worse than cascading converters for appreciable power outputs tho