RE: SSTC Reliability

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Jim Lux" <jimlux-at-earthlink-dot-net> 

At 08:22 AM 11/9/2004 -0700, you wrote:
>Original poster: "Aron Koscho" <kc5uto-at-wt-dot-net>
>Hi Steve,
>
>
>As Terry and Steve W. have said, it would have to be thoroughly over
>designed. Also, a project of that magnitude would not be inexpensive.
>
>However, I do believe it's possible given the right amount of
>time/money/expertise. There are several key factors to getting more than
>two feet out of a traditional SSTC such as frequency, component ratings,
>heat dissipation, and circuit protection. When I first achieved 36" with
>7kw the system became unstable and nuked its self in less than two runs.
>After adding fairly complex protection and a large heatsink/fan it
>became more reliable. Still, the best topologies for a large system
>would seem to be the "DRSSTC". Regardless of the chosen topology more
>work needs to be done concerning coil efficiencies, peak currents, and
>component cooling.
>
>Bottom line, it can be done and hopefully some will do it:)

I think the difference between SSTCs and sparkgap TCs (SGTCs, to use YAA) 
is sort of like the difference between designing and building prototype 
tube and solidstate high power RF amps.

With a tube, it takes a fair number of joules (more than 1, usually) to 
damage the device, which has physically large and massive components (grid, 
plate, etc.), so that momentary hiccups or instability during test don't 
kill the device.  Also, the tube operates at a higher temperature (don't 
see many silicon devices running at Tj of 300C, do you?), so it has farther 
to go from "normal operating point" (few hundred degrees) to "destruction 
temperature" (melting point of copper/glass/etc).

With a solid state device, it takes a lot less energy to kill it, and, 
because the devices are physically smaller, thermal management is more of a 
challenge.  It's also harder to probe the circuit to see what's going on, 
and, unlike a tube, where you typically have a few seconds to see that 
something is wrong and shut the power off, with modern high speed SS 
devices, it's dead in the first few half cycles. Partly this is because the 
line between operating and destruction is quite fine.  Doesn't take much 
energy to heat something from 100C (operating Tj) to 200C (destruction Tj) 
(Ask my officemate trying to get those Ixys/DEI kilowatt FETs to work in 
the eval board)!

Moral of the story is that you need more test equipment and a bucketload of 
spare parts to do high power SS design, because the cost of an oops is a 
destroyed part.