All during the year of 2015 I worked on a design for an an all-transistor bench power supply that I could use to diagnose and build tube electronics. The idea was a fully contained unit that offered an HV supply following the Maida topology, but adjustable from 10V to 480V at 300 mA, a negative bias supply adjustable from 0V to -100V at 30 mA, a filament supply for both 6.3AC and 12.6VAC duty, and bullet proof--ability to survive short circuits on any of the supplies as well as unconditionally stable. Those are tall orders for a weekend engineer. In my original work, the high voltage supply caused me some problems. No matter how I tried, I consistently blew up the power transistors, whether they were mosfets or BJTs. I was at a dead end, and being burned out on the project, I put it aside for a year. What I wanted to try was swapping out the power transistors for big power pentodes, but would a Maida-style regulator still work? Since I built the whole thing on modular boards, all I would need to do is cook up a new board for the power tubes, figure out how to power the filaments and screens, and figure out how to interface the regulator board to the power tube board. So for the past several months I've been building a pass tube array board and a screen supply board. The whole Maida regulator approach is a floating design, so to make the pentodes work in place of the transistors I would also need to float the filament and screen voltages. In the original Maida circuit, Interfacing the LM317 regulator to the power transistors requires the use of a Zener diode to force enough of a voltage drop across the 317 so that it can regulate. Maida originally used a 6.2V Zener I believe, so that the 317 always had 6.2V of headroom with which to regulate. But since power tubes run the grids at a good number of volts less than the cathode, the needed voltage drop is already implicitly provided, so I did away with the Zener all together. My prototype never runs the grids at more than 20V below the cathodes, therefore there is never more than 20V across the 317, so I felt this was safe. (Note a higher screen voltage may exceed the 37V limit of the 317.) Under short circuit conditions, the 317 flops out of regulation, but who cares? It's a short circuit, not useful for real world situations anyway. I've now got the prototype mocked up on my bench and for the past several days I've been beating on it. Where power transistors fail, vacuum tubes shine! One torture test I tried was a metal rasp to implement a short circuit by connecting ground to one end of the rasp, and dragging the HV lead across the teeth of the rasp. In my previous attempts with the power transistors, this test would immediately blow out the transistors, usually with a mighty bang, and take out the LM317 in the process. But not with the pentodes, a few sparks were flying across the teeth of the rasp, but the 6550s were cool cucumbers, dutifully carrying out their responsibilities. Here is a picture of the test bed with the HV boards shown. Boards top left to top right are screen supply board, power tube board, and regulator board. The shot shows the power supply delivering 270 VDC across a 5K load (54 mA). I'm using three 6550 tubes with a semi-regulated 116V screen supply. Screen regulation is not strictly necessary I think, but it was easy to provide and helps the supply deliver more constant current under varying loads. Current limiting is achieved through the use of the screens of the pentodes themselves. Not only do they allow the tube to swing to near zero volts voltage drop so the regulator can efficiently deliver high output voltage, but they implicitly limit current too. It's not perfect current limiting (current limit changes slightly based on output voltage selected) as the pentode curves are not quite flat, and the 6550 curves have a kink down near 100V plate, but hey, close enough, as long as the thing doesn't go up in smoke under a short circuit test! I chose the screen voltage quite carefully. Several parameters needed to be met simultaneously: a) never exceed max screen dissipation under any circumstances, b) never exceed max plate dissipation under any circumstances, c) never exceed 37V drop across the 317, d) maximize efficiency by having each tube deliver as much current as possible while conforming to the first three conditions. At the screen voltage I chose, each tube contributes a max of 100 mA to the total output current, so 3x tubes gives 300 mA max current. At the raw HV voltage I chose (510 VDC), under short circuit conditions each tube dissipates 42 watts, which is design-center max dissipation for a 6550. I built the board for 4x tubes, but the power transformer I'm using only supplies 250 mA on its HV winding. I think I can drive it a little over that without problems, so I am limiting my unit to 300 mA via three 6550 tubes. The regulator will work with anywhere from one to four 6550 tubes installed. If you want 100 mA max current, remove all but one 6550 tube, etc. I could have built a variable screen supply to be able to limit current from the front panel, but this is more work than I am willing to do right now. Up next is more testing. There might be certain conditions under which the supply will oscillate, so I need to investigate that and determine how to address those cases, if possible. Under normal resistive only loads, the supply is superbly stable and quiet--noise is down under 1 mV. I think you sort of expect a very quiet output given the use of the LM317, since it rejects something like 60 dB.