Progress on new shop

The only thing in the KL10 that needs three-phase (not counting any external peripherals) is the H760/H761 power supply for the ECL. The H760 "raw supply" takes about 7kW of 120/208 three-phase to produce about 12.6V 500A, which the H761 then regulates down to nine or ten -5.2V 35A and four -2.0V 35A supplies for the logic and termination (about 2.1kW). All of the other power supplies in the KL10 are single-phase supplies. The standard wiring distributes those among the three phases. There is a bank of 13 130,000 uF capacitors between the H760 and H761, so it's probably OK for the input to be somewhat noisy. The inrush current to the H760 is hundreds of amps on each phase for more than a second. The mains wiring is 75A/phase. I've got an untested design for a 12-phase 35A switching regulator board using the LTC3731 to replace the -5.2V and -2.0V linear regulators of the H761, which would then still use the H760 and three-phase input, but at much higher efficiency. That's the minimal-modification way to reduce the power requirement. Using a normal buck controller chip as a non-isolated negative buck controller is a bit tricky, since they are all designed for positive use, but I got some technical assistance with the design from a Linear Technology apps engineer, and it checks out in simulation. I'd want to build one and bench test it extensively with a programmable load before I'd actually use it in a KL10. If you didn't want to use three-phase, it would probably be best to replace the entire H760/H761 power supply system with single-phase mains input switchers. I'd reccommend using nine or ten 5.2V 35A and four 2.0V 35A isolated switchers, so that they can be referenced for negative operation. Power supplies with sense inputs are a must since there is non-trivial voltage drop in the cabling. The ECL needs very clean power, which is probably part of the reason DEC used linear regulators. Back then switchers tended to be pretty noisy. I'm not sure whether a typical modern switcher is clean enough. My regulator design uses 12 phases to reduce the ripple.