Hi again,
So as promised a few days back I have recreated a schematic I have seen use d to create a low power 3.3v logic supply from 84-264vac and using a PWM dr ive for 24v relay coil. I think it can be extended to encompass your higher voltage range.
formatting link
Although it looks at first glance that it works on DC it will not since it switches the unfiltered rectified input to charge the storage cap C1 only d uring the leading and trailing portion of the input waveform.
The design works most efficiently when C1 is small enough for the ripple vo ltage V3 to be large, of course it must not droop so low that it causes the linear regulator U1 to dropout. I saw it used with D1 at 39V R1-330k, R2-2
2k, R3-68k, R4-2.2k, C1-22u, Q1-3904, Q2-STN0214, D2-1n4148. All of which m eant the C1 voltage swung between 35v and 6v. Dissipation is mostly in R4 b ut I think a half watt device was sufficient since the action is more switc hing than linear.
It could be that Q2 was actually two in a Darlington configuration but when I return from travelling next week I can investigate further.
The application I saw drove the 24 v relay coil with pulses of the unfilter ed rectified input as I show in the sketch and the microcontroller drove at a duty cycle that varied with input voltage. Although I drew a mosfet I th ink they actually used a high voltage npn bjt, maybe same device as Q2.
I suppose you might also want to consider increasing the C1 voltage to enco mpass the solenoid 48v and drive the solenoid more conventionally from dc. The drawbacks then are that the C1 value will have to very large (eg 820uF) to support the solenoid current for 50ms and that it will take some time t o recharge enough to support another activation, also dissipation in U1 wil l be higher.
This should be very easy to simulate on lt-spice.
piglet