Quoted from G-P-E:
I have always believed that there is no substitute for big iron (big transformer) for the solenoid loads.
I know these are older posts ... and this may have been addressed elsewhere in the thread ... and I am not being snarky or stupid ... just trying to contribute.
All a big power transformer adds to a system is cost. They're also a waste of power since you have to rectify that voltage. Moreover they're heavy ... they add significant mass to the pin, so that means they're more expensive to transport. Moreover, when linear supplies fail, things are guaranteed to blow up ... most switching power supplies fail open ... that's a good thing since it won't destroy anything.
Powering coils in a pin with 48V driven by a switching power supply is fine so long as one takes the exact same precautions that have been taken since solid state pins were introduced to the masses. The diode across the inductor (in their example, the inductor represents a motor) shown in the video is no different than the ones on pinball coils. Forget those in a pinball machine with a linear supply and things will still blow up regardless of it being powered with a big ass transformer. The phenomena described in the video is identical why one needs diodes across a coil.
One can run into problems if a switching power supply output is driving a highly inductive load ... something like a long, long wire can have enough inductance to cause a problem ... some care has to be taken there of course (I suspect that is why they are using TVSes at the input of the regulators). You wouldn't run into this problem with a linear supply, but I think the benefits of a switching power supply far outweigh that risk.
One final note. 48V driving a switching regulator that regulates a voltage down to 5V or lower isn't that big of a deal these days. Switching regulators are VERY reliable ... far more reliable than linear regulators in my opinion since far, far, far less heat is generated. Heat is public enemy #1 to semiconductors in a working system. Switching regulators are extremely efficient ... far less heat is given off as a result.
Most networking equipment designed to "5 9s" (that's nerd speak for 99.99999% reliability) are designed to be powered by a 48V source (keep in mind I am talking about the big ass telecom switches/routers, not the stuff you have on your desktop ). The same voltage is standard on a lot of robots as well. Most boards (or "adaptors" as the hip and trendy call them for some stupid reason) for these kinds of systems will regulate that 48V down to 12V or 5V ... then, depending on power requirements, use additional switching or linear regulators to get the ridiculously low voltages (like 2.5, 1.8, 1V or less) to power CPUs, ASICs, FPGAs, DSPs, memories, and other devices (almost all of the time switching regulators are used).
That being said, what I don't like about 48V being routed everywhere in a pin is safety ... that's a relatively high voltage to have running everywhere in a pin w/o any safety interlock being possible (if you lose 48V, you lose the ability to debug anything ).
Switching regulators aren't perfect of course ... most problems are due to designer error though ... and most of the problems are EMI related ... if one isn't careful with the layout of the regulator on the PCB, you will see nasty EMI spikes right around the regulator's switching frequency.
If it were 1995 or so, I wouldn't be too crazy about switching regulators being used to drop 48V down to 5V ... however, these days, it's a no brainer. The technology is very mature. It is nothing like the switching power supply Atari used in some games (I, Robot comes to mind ... some Firefoxen ... and I think a couple of others) .