Many swear by AC believing it's more natural; there's less circuitry in the signal path, and it's cheaper on hardware too. Others believe that DC is the only way to go - not only should it remove any hum but it's cleaner too removing the harmonic fringes of the 50Hz AC (or 60Hz depending on where you live).
But even in the DC camp there are different ways to go. You could try passive DC with lots of high current chokes and big capacitors, and LCL is said to sound the best and means there isn't a whacking great electrolytic cap strapped across the filament. It's not so easy to set the voltage though.
Or you could try a current source (leave the voltage sources for your indirectly heated valves). These can be fairly simple, or fairly complicated. Fortunately there are some options available if you want to have a go but don't know how to design a good one. If you want a ready made option then the Tentlabs modules will do the job admirably. DIY Hi Fi Supply used to have their own modules too, but they look to have stopped making them. Anyway, just wire 'em up, set the voltage by twiddling a screwdriver and away you go.
Or if you like to build stuff (and save a few quid) then another option is Rod Coleman's regulators. Rod has been developing his modules for the best part of a decade now I'd guess and has been supplying them for maybe three years. Send him some money and you'll get a compact PCB and all the components to populate it, plus instructions to build and test them. You just need to supply a raw DC power supply, and advice is given in the instructions. Rod is contactable by PM on the DIYAudio forum.
I've built three pairs of Rod's modules before for my 26-10Y-300BXLS monster and each one improved the sound over the passive DC I was using before. So when I finally started the GM70 amp I'd been meaning to build for about three years it was an obvious "fit and forget" choice to build a pair of Rod's modules again. And here's a series of photos of the build sequence.
First I like to lay out all the bits, check them off, and mark them up with which component they are
So start with the bare board
And then the smallest components first - the resistors. And then the first capacitor.
Then the variable resistor used to set the voltage
Then the first two transistors, to the same height as the variable resistor
Next the second capacitor
Then the third of the transistors, again to the height of the variable resistor and adjacent capacitor
Next the thermister which stabilises the current temperature (in position R11) (note that only the higher current GM70-type modules use these)
Then I add the sense resistors as I prefer to mount the two 3 legged chips when I assemble the heatsink. I find this the best way for me to get a good job, but it is a bit fiddly soldering the chips as access is a bit tight. Note that the resistors need to be stood off the board by at least 14mm as they get very hot.
So here's the module mounted on the heatsink - a piece of aluminium angle with the PCB connected by standoffs. Note the use of mica insulators as the higher current regulators need insulating. Make sure you use some thermal paste on both sides of the micas - I didn't have any to hand at the time but retrofitted later. Note also a zener fitted across the chip pins, a little fiddly to solder. This is some added protection in case the valve isn't plugged in when the power is turned on.
And here's the finished module with additional heatsink. As I breadboard I like to use them as self contained modules, which is why I mount them on the aluminium angle. If I ever do get around to putting them in a box they can then be mounted to the metal chassis for additional heatsinking. The finned heatsink does a very good job of taking the heat away from the regs and I intend to mount the heatsink above the chassis eventually. (The heatsinks I bought in Japan for about a quid each...)
Here are the two regs in service. Rod supplies instructions on how to test and commission the regs, and as you can see from the DMM it's easy enough to set the voltage. I like to set it just under the nominal voltage and backed it off just a little from the voltage shown.
You might just be able make out the additional capacitor across the raw DC supply - this is a 1000uF cap that decouples the reg from its power supply when connected by longer wires, if it's in a separate chassis for instance. This is a rather nice amp I have to say, one I should write about some time.