[Amps] RE: 4CX250B Screen Supplies

[Will Matney]

Peter,
Yes, your correct and there have been many other tests and 
investigations. I have a university test here somewhere that talks about 
IMD even being caused by each component that's used in the complete 
supply system due to dissimilar metals. When you think about it. it can 
happen. They were looking into anything that could cause IMD and found 
this, which was really not to help you and I but Ma Bell, Sprints pin 
drop, and Uncle Sam. Who, in my opinion, footed the bill for the tests 
funding. At least those tests are made public most of the time. The very 
worst problem of IMD is the 120 Hz ripple that's left from a power 
supply. Everything else is a minor cause except for the tube design 
itself and the circuit placement/shielding, and bypassing. When 120 Hz 
ripple hits that PA tube, it's just a very large mixer amplifier. Keep 
in mind that the control grid supply can be infected with 120 Hz too, 
especially if it's grid driven! So why super clean the screen when we're 
feeding a dirty supply to the control grid (tetrodes and pentodes) and 
mixing it there too! In my opinion, there's always going to be some but 
we can cut it back with proper filtering. The only "cleaner" way would 
be to go back to battery supplies like in the 1920-1930's or so. In a 
way, it would be a lot better if the power companies would have fed us 
square wave instead of sine wave power but as Paul Harvey says, "That's 
another story" =)

Will Matney


Peter G3RZP wrote:

"Some professional tests use a very narrow spacing between tones - as low as 30Hz - to simulate the syllabic nature of voice, and to exercise the power supply regulation. Another approach, which, if I remember correctly, has been advocated here by Tom, W8JI, is to use a three tone test with two of the tones very closely spaced. As Steve, G8GSQ says, the screen current varies at a frequency dependent upon the tone spacing.

Because of the peak to average nature of a voice signal, it isn't easy to relate the performance on a two tone test with that under voice conditions. The advantage of a two tone test is that it does (or should!) give a number that can be reproduced in testing by other people. In the days of analogue FDM telephone circuits, one testing method developed was to load up the audio channel with noise to the correct PEP: then a a narrow slot in the frequency domain was inserted in the noise, and the noise power in this slot at the demodulated tx output was measured. This gave a good  approach to the real IMD performance under practical loading conditions, but did look at the results of an aggregate signal from many individual voice channels. However, for a single channel, because of the lack of components at syllabic rate, it isn't so good. Apparently, for ISB work with voice on one sideband and multiple data channels on the other, it is a testing method used extensively by the Russian military, according to a guy I see at ITU meetings from the Russian administration.

Personally, especially for anything that can have negative screen current, I like the idea of an electronic  shunt regulator. To ensure that the screen can't over dissipate, this is fed from a constant current source capable of  little more than the maximum rated screen current. As FETs are pretty cheap ( and things like 6L6s and 807s even cheaper if you have a good stock!) this seems the way to go.

73

Peter G3RZP"