[RTTY] PSK31 is faster (Was FD RTTY Question)

[aflowers at frontiernet.net]

Apologies for the provocative title, but the lamenting of PSK31's perceived slowness during the FD festivities prompted me to crunch some actual numbers.  The question I thought to investigate was "how much slower PSK31 is vs. RTTY in various contest-like exchanges?"  I'm giving my results for discussion.  For the record, I have never participated in in a PSK31 contest, so this is all back-of-the-envelope math and not based on any kind of actual contest experience.  I have other thoughts regarding PSK31 in contesting, but this message is strictly about information transmission speed.

I took several messages that are typical of RTTY contests.  For all messages I assumed a CR/LF at the beginning of the message.  RTTY is sent with USOS on, 1.5 stop bits at 45.45b.  PSK31 is sent in lowercase (more on this below) at 31.25b.  (Given the recent thread on character encoding in email you may want to cut and paste the table below into a fixed width editor to get columns to line up right.)

Message, Bits PSK31, Bits RTTY, Time PSK31, Time RTTY, PSK31 advantage

<CR><LF>cq k0sm k0sm cq <EOT>        98        172.5                3.14        3.80        17.37%
<CR><LF>cq aa5au aa5au cq <EOT>        98        187.5                3.14        4.13        23.98%
<CR><LF>aa5au andy ny ny <EOT>        104        172.5                3.33        3.80        12.31%
<CR><LF>aa5au 4a wny <EOT>        72        150                2.30        3.30        30.19%
<CR><LF>aa5au 599 678 <EOT>        122        180                3.90        3.96        1.42%


It turns out its really quite hard to engineer a message where 31.25b PSK31 is slower than 45.45b RTTY!  Even numeric exchanges, which are quite long in varicode end up being at parity or better.    Replacing the CR/LF with a single leading space makes PSK31 even faster.

Because of the variable-length encoding in PSK31 your own callsign could potentially take longer to send.  This is a function of how many "unusual letters" you have in your callsign.  Here are some examples, although it should be pointed out that "wa5zup" breaks nearly even with RTTY if you include leading and trailing spaces (which you would in a message), which are only one bit in varicode:

n6ee        17        45        0.54        0.99        45.06%
aa5au        27        52.5        0.86        1.16        25.20%
k0sm        27        45        0.86        0.99        12.74%
4x6z        35        60        1.12        1.32        15.16%
wa5zup        41        60        1.31        0.99        -32.51%


This begs the question as to why PSK31 is slow.  How many people are using their RTTY macros in all caps for PSK31?  PSK31 averages about 10 bits for capital letters, and exactly 6 bits for lowercase.  One way to change this behavior would be to have contesting software always send lowercase, and if one wanted, always print uppercase on the screen for legibility.  I can't imagine a contest where case would matter, and you get instant speed improvement and better copy without having to change your beloved macros.  Like USOS for RTTY, forcing lowercase transmission would a kind of hack to make things go a little faster and improve copy, and people who don't know any better still reap the advantage of lower-case transmission by default.  This seems to this armchair observer to be a much better solution to increasing the baudrate (and thus losing SNR) by switching to PSK63 or whatever I to overcome poor utilization of the character table.

I hope this is helpful!

Cheers,

Andy K0SM/2