The PIC/DDS beacon source is now running, and it's being tested  on air for the first time with FSTW-120, -300 and now -900 

There's just have one issue to sort out.   I generate the Gaussian glide by using just two adjacent symbols and adding their weighted components based on a lookup table, sampled at 64* the sample rate.

Doing it like that means each session starts half a symbol late.   For the faster modes this doesn't matter, it's less than a second on DT.  On FSTW-900 the resulting delay of half a symbol is 2.8 seconds - which the decoder seems able to cope with but on -1800 the delay, DT,  becomes 5.6 seconds which doesn't work.

I can sort this by doing the job properly, and adding weighted versions of three symbols, last, this and next.  A few more lines of code, but easy enough to do

and.. as I write ... just checked WSPRNET, and there's a decode of the first -900 signal off air.   It works !