[RTTY] Improved RTTY decoding

[Shoppa, Tim]

I've been analyzing received contest audio looking for strong correlators, telltale bit patterns, that help in non-realtime RTTY decoding.

Many transmissions begin with LTRS LTRS which is a very good correlator, but if replies are stepping on each other, this is often lost. So I have been looking for patterns that happen in the middle of a callsign or exchange, to help me pick up after-the-fact and regain synchronization.

There are a couple of really super strong correlators that are very common and help me pick out, non-realtime, callsigns and exchanges more effectively from noise:

FIGS(digit from zero to 9)LTRS happens in almost every callsign. There are some oddball callsigns with two or more digits, especially in RTTY WPX, that do not show this pattern. Sometimes this shows up in a couple of canned macros that send reports as 5NN but are not callsigns.

FIGS 599 SPACE or FIGS 599 HYPHEN happens in many common exchanges.

Any of the above help me a lot to identify the where we are in the "decoding flywheel" that others here have mentioned. Being multi-character sequences they also help a lot in determining the stop bit length.

Also, of course, my own callsign N FIGS 3 LTRS Q E is common in my exchanges recorded with me. I can imagine, say, taking the entire dictionary of most common RTTY contest callsigns, and doing a correlation with every one, but computationally this might just be ridiculous.

My observations that funky WPX-contest-specific callsigns are hard to decode probably applies to humans experienced with "normal" callsigns in CW and phone too!

There are a handful of stations with apparently completely random stop bit times that my correlators give no match to.

I do not know how to apply any of my observations to realtime decoding yet.

Tim N3QE