I have finished a series of simulations looking at the performance of several modes that seemed appropriate for extended keyboard to keyboard rag chew QSOs. I was looking at modes that offered a throughput of about 40 wpm so they could keep up with a reasonable typist with a bandwidth of no more than 500 hz.
I used PathSim to measure accuracy of text transmission under white noise and CCIR 520-2 "Poor" simulated propagation conditions. I measured text accuracy over at least seven minutes of text for each data point. The graph can be found at http://mysite.verizon.net/wz7i/modeimages/Digital%20Modes%20Poor%20Condx.png The methodology, including software packaged used, is outlined at http://mysite.verizon.net/wz7i/digitalmodes.html Summarizing, I arrived at the following SNR (db) for a character error rate of 5%: AWGN "Poor" DonimnoEX8 -15.3 -3.1 MFSK16 -14.7 -8.5 PSK31 -13.2 -0.8 Contestia500/16 -14.0 -9.2 RTTY -9.1 +3.7 I probably need to look at Olivia 500/4 These data confirm my prejudice about the excellent performance of MFSK16. With the extended low tones implemented in several packages, the mode is not difficult to tune. A couple things surprised me. I would have expected DominoEX to do better under poor propagation. Another surprise is the difference in performance between different software implementations of a given mode. A software program may have excellent decoding performance with one mode and then have performance with another that is not competitive. The above numerical data would vary a good deal if different decoding software were used. So if you find operating with a given mode frustrating, don't discard it without trying another program. I hope that with RSID some of these excellent modes will find greater use. The web site may well have errors so if you find something surprising, please let me know so I can check things. I don't want to mislead anyone. Wes, WZ7I www.wz7i.com