[Amps] reasonable design

[John Lyles]

I have followed recent threads that go back and forth about cooling transformers and use of switch mode technology for HV. I have two practical examples that I will share that demonstrate that these things can go either way, there is no hard rule these days, not to mention individual preference and acceptable design risk. I am not suggesting that I did the right things, but after lengthy testing and use, the decisions were the right ones. I know that we hams can be quite stubborn to try new things, and often good old heavy iron technology is just fine. These are just examples, nothing to force anyone to be uncomfortable with a choice. 

First, I needed several tetrode filament power supplies for 100 kW at 2.8 MHz. Tubes had 15 Volt AC center tapped filament transformers, with 320 Amps RMS load. Since they had big filaments, they could not be switched on quickly or even in step start fashion, instead needing 8 minutes of linear ramped voltage on turn on and off. Conventional designs would use a big variac with a stepper motor drive. I disliked anything with moving wiping contacts that must last for years, as it needs repeated maintenance. I could have found a solid state AC oscillator to generate a sinewave? I chose a saturable reactor design and enlisted the late Carl Seivers of SNC in Oshkosh. He told me back in the 1980s that he made many of them, although there are a lot few companies now making them. Saturable reactors have a DC flux on one winding (or 2) and AC passing through the other. They run hot, by design. Insulation is rated for it, so there is no reason for concern. The SR is essentially in series with
the 480 VAC power line to the filament transformer.  I was mounting this in a large chassis, along a small linear ramping supply for DC bias, a logic card and some relays and breakers. I HAD to have a fan to remove some of the heat buildup, that would otherwise have cooked my circuits nearby. The fan I chose was a very reliable Pabst, and it has an additional thermal cutout mounted on the SR that will detect a dead fan and prevent power until it is fixed. At the time I was nervous about having to add a fan to a unit to help cool the effects of big iron, and then the clap trap logic to protect in case the fan died. Its been in service in two power amplifiers since 1998 now, without any problems or failures besides replacing the mechanical running time meters when they give up. Had it used variacs, I wouldn't be saying the same thing. 

In more recent work, I had a filament power supply that needed 20 VDC at 1000 Amps, for a bigger tetrode at VHF. Note i said DC, not AC here. This would have been 3 phase primary, and needs rectification. The SR got too complicated, followed with a big rectifier and filter. Carl was no longer around at SNC. I talked to several companies who make big iron, and there was a lot of risk due to their experience (lack) plus the cost of the copper and iron. Everyone wanted to make an SCR controlled power supply. We chose one, ran it for a year, and it was a POS. I won't name the company, but reliability was low, not so much the power conversion but the controls were very susceptable to noise, RFI, transients when the HV capacitor vault crowbarred, etc. I spent a lot of time cussing it when it would trip off with no warning, and slam off my expensive tube's filament. The answer was staring in my face, but I didn't want to use it due to the old fears about switchers. True that in the 1970s
and 80s, they were just moving towards MOSFETs and later IGBTs to improve over old BJT circuits. I did select a nice switcher (not so easy at this power, needed to parallel two commercial units) and it ramps up beautifully, never faults *yet* and has run for half a year. And it is really efficient. 

I use switcher supplies for screen, control grid and filament, now, and am very satisfied with COTs designs. 

I would still use iron if I could, in places with simplicity prevails, more immune to RFI and transients in a multi-megawatt RF system, but cost, size, cooling are important factors to realize for new designs too. I took the risk of using switchers there, and was pleasantly surprised with how well it worked out. 
73
John 
K5PRO