[hpsdr] Transmitter Linearisation

[Chris Bartram]

Just a quick comment on the use of linearisation techniques. Putting it in a 
very few words, it's much harder than it seems! 

I doubt whether it will be possible to use any form of conventional feedback, 
in SDR systems based on PCs, as the processing delays, due to processing 
latencies, limit the achievable bandwidth. That's just simple(?!) control 
theory.

Predistortion is a distinct possibility, but it will have to be highly 
adaptive, as power amplifiers have surprising unstable transfer 
characteristics with power level and temperature. It will be an IPR nightmare 
if any one wants to make a lineariser commercially. Most of the obvious 
techniques were patented around a decade ago, and depending on the 
juristiction, have rather more than that to run... 

I have seen a prototype UHF amplifier with adaptive predistortion operating 
several dB into compression, and simultaneously giving ~ -60dB third-orders, 
with 5MHz modulation bandwidth, but that used _very_ heavy duty 
quasi-real-time DSP. 

It's possible to do quite well with analogue processing. I think there are 
free-market integrated cartesian feedback loop chips now available, they'd be 
capable of giving -60dB third orders with amateur-type power amplifiers. I 
was once involved with a 200W pep CFBL AM transmitter for avionic 
applications, which at 90% mod. index had only three outputs visible on a 
spectrum analyser with 80dB dynamic range: two sidebands and a carrier! 

CFBL isn't the only analogue feedback technique. It's not necessary to use a 
zero frequency IF, and bandpass feedback can be made to work reasonably 
easily - it's simpler to set-up, and has less IPR clutter.  EER is certainly 
a runner but I've always felt that EER is a bit more difficult to to do than 
a CFBL or bandpass feedback.

Vy 73

Chris
GW4DGU