This is my old linearly regulated argon laser power supply:
Fortunately I could use an old industrial case with meters and heat sinks ready-to-go. It is quite hefty and heats up a room quite quickly; there is nothing as ineffiecient as a linear regulator without transformer... note the fan-cooled 1400W ballast resistor hanging from the rear.
It runs directly off 230VAC and delivers up to 8A at 160VDC, current regulated; this means it dissipates more than 1KW (here the simple circuit ca 500kb, adapted from Sam's laser FAQ). It uses 10 linear pass bank transistors BUS24B in parallel. In the mean time I have rebuilt this into a much more efficient and elegant multi-use switching power supply.
What is important for linear regulators is the ``safe operating area'' (SOA) of the transistors used. The point is that the max current rating at elevated CE voltages can drop much faster than what the max Ptot may suggest; this is because the limiting factor at voltages higher than ca 10-20V is not Ptot, but the ``second breakdown'' effect. This is particularly pronounced by high-voltage junction transistors, even if they have a very high Ptot and current rating near 10V, say. This is demonstrated by the following examples, most of which I have used over times in linear regulators:
It is thus very important to check the data sheets for the SOA, however not all manufacturers include it. Mosfets are generally immune to second breakdown and are just Ptot limited, however mosfets are not so well-suited to be used in parallel when in linear mode (though excellent in switched mode).
For linear regulators, cooling is a non-trivial task even with a step-down transformer to 110VAC. They must get get rid off several hundreds of watts (say, 50V*10A=500W) - the best solution is a tunnel cooler of this kind:
I have a good source for such and interested people plz feel free to ask !
(*) I have a lot of those, unused new old stock, selling them for 2Euro per piece (minimum 10).