View Single Post
Old 17th Mar 2012, 9:22 am   #12
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,799
Default Re: Valve Linear Amplifier project

Nice photo of the 4CX250 distributed amplifier!

In case anybody is interested, these things were a British invention that sort of languished until a paper was published at Stanford. Look up Ginzton, Noe and Hewlett. Yup *that* William R Hewlett.

The distributed amplifier plays two nice tricks: The Gm of all the bottles add, making for mor gain than one device would give, and the ability to drive a conveniently low output impedance directly without transformation. secondly, the inductors forming the transmission line prevent all the anode capacitances ganging up in parallel to wreck your bandwidth.

You pay a few prices for all this though. There's a lot of heater power which is a seasonal benefit/penalty. And everyone will tell you that the termination at the 'inside' end of your anode line soaks up half your output power. They're right, sort of.

First note that a distributed amplifier needs TWO transmission lines. One for the anodes and one for the grids. The two lines can have different characteristic impedances, but they MUST have the same time delays between valves.

The amps use not real transmission lines, but approximations built out of lumped components. This gives them an upper cutoff frequency. Think of them as tapped filters with equal value inductances and capacitances per section. The lowpass response is handy in an HF linear amplifier, but a nuisance if you want max bandwidth. Modelling the line with more sections of smaller L and C values increases the BW. Ultimately you can build a real transmission line inside a vacuum tube and the Travelling Wave Tube (TWT) or 'twit' to its friends is born.

One thing rarely mentioned is that distributed amplifiers have a band where the power wasted in the anode line load is nulled. If you think of all the signal paths from input to output, they are all the same length... go so many stages along the grid line, go through a bottle and go the rest of the stages along the anode line to the output and the total number of filter sections is the same. This is why both lines/filters must have the same time delay per stage. Everything arrives at the output in phase.

Not so for the input to the anode line load resistor. Each different route has a different delay, and things arrive with staggered phases. They partially cancel and the wasted power is reduced. Great!

I played around with one of these beasties and found that I could design something with a cutoff above 30MHz and with the waste power in the line termination kept down to a few percent of output provided the frequency stayed above 1.8MHz. It needed 30 stages to do it (a problem for 4CX250Bs, but no problem for cheap SMPS power MOSFETS)

To get lower distortion, I redid the thing with a different time increment, still 30 devices, but distributed on a pair of output lines configured in push-pull.

You don't have to do the maths. Visit the Linear Technology website and download their free copy of SPICE. It's very powerful, it's not crippled with any artificial limits and it's free. A bit of hacking around and you can create models for your favourite valves, or you can find them done by other enthusiasts on the support websites. It's a great tool for playing 'What if' and unlike building hardware for silly ideas, no-one else need know and ask embarrassing questions.

CHeers
David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now