[Radio Wrangler]

SSB and CW filters are sometimes made in the 'RF' frequency range, but they are placed in the IF section of receivers.

I have an Eddystone EA12, rather a glorious set, but an early example of a radio for SSB. It uses L-C filtering in the shape of a lot of IF transformers, but to be able to get LC filters narrow enough for SSB, the centre frequency of the filter has to be low. So the set is a double superhet with the final IF at 100kHz, chosen to make that LC SSB filter possible. The IF bandwidth is actually variable. A front panel control turns a shaft that via levers and linkages moves one coil in each IFT closer or further from its neighbour, thus varying the coupling. When you move it to the very narrowest extreme, a microswitch closes and brings a quartz crystal into play for the very narrow bandwidth wanted for CW. 100kHz was a common IF frequency for higher performance communications receivers.

So the mode you want to work sets the bandwidth you need. Then looking at the different filter technologies, you find there is a happy region of IF frequencies you can make good filters over for each mode and for each filter technology.

Quartz crystals have very, very high Q compared to the best we can do with LC circuits. This is both a blessing and a curse. They can make some of the narrowest filters, but they are very difficult to use for making wider filters. Narrower and wider are terms that depend on the centre frequency.

You can make quartz filters on 455kHz for CW and SSB bandwidths, but the crystals are large and therefore somewhat expensive. The big-three Japanese makers of amateur radio opted for a 9MHz IF for many of their sets (though the Americans did it first). SSB crystal filters are cheaper there, CW are a bit more difficult but not too bad. The big saving was that for an amateur band only receiver, it could be a single superhet, and their accountants could celebrate (probably with a new set of golf-clubs).

But their customers wanted general coverage receive in their HF sets. Drat! There went the single superhet architecture. So they mixed up to a high IF somewhere in the 35 to 70MHz region where new sorts of overtone crystal filters were becoming affordable, then they mixed back down to 9MHz and it was business as usual from there onwards and they could use their existing crystal filter parts. You may find these receivers called 'upconverters' because their first IF is higher than the highest input frequency they can tune to.

You should now be beginning to see that filters rule receiver design. You need filters to do various things and the available and affordable technologies limit what sort of frequency you can do them at. Consequently signals in receivers are mixed up and down willy-nilly to suit the filtering edicts. Oh, and those mixers require additional filters to block their images.

Enough for tonight. Give me a nudge and the next time I'm sleepless, I'll go over mixer images and the whole business of whether they are linear or non-linear.

David