Normally you can assume 5 characters per word, that factor is is reasonably constant - the average word is 5 letters

 

The number of bits per character  is more complicated to quantify.  In  CMSK for example,  where the character length varies based on the probability of the letter's occurence in typical messages, the effective number of bits per character is less than for ASCII encoding, say around 5 or 6 bits per character average for the whole alphabet.   With a heavily source encoded scheme - like WSJT with a small alphabet to start with, - it is possible to get right down to less than 5 signalling bits per character.  But particularly in Varicoded data, there are more bits added in framing and synchronisation, so call it total of 7 bits per character complete

 

If you then take note of the error correction overhead, which doubles up the number of bits sent for each message, and add in another 17% of synchronisation bits , you end up with a net rate of something like 17  bits per character for the error corrected message.  So, CMSK8 (8 bits per second) equates to about 8/17 characters per second, or 30 characters per minute, or 6 words per rminute - which is rather faster than the 3.75 rate quoted in the table - so I've missed something somewhere?

 

CMSK8 is 8 times faster than CMSK1, but its not the whole story.   If the link is too weak to support the faster speed, but the slow one gets through - which is then faster ?   CMSK8 that may work tomorrow, if you're lucky, eventually;  or will work CMSK1 now?

 

Andy
www.g4jnt.com