In a message dated 12/24/01 9:58:25 PM Eastern Standard Time,
[email protected] writes:
<< I have decided to add an extra space after each 7FSK pair. This does slow
down the mode by a third, but it allows easier synchronization for the RX as
well as QSK to watch between characters on the TX site, and it gives the PA
a
rest. >>
Thanks, Markus. Based on the sample of 7FSK as posted a couple of days ago,
it struck me that making this mode machine-decodable would require (1)
knowledge of the absolute timing of the start of the message; (2) a way to
convey to the decoder some reasonably precise knowledge of which frequency
(which FFT bin) represents the "1" or the "7"; and, (3) reasonable faith that
the separations of the seven frequencies as actually transmitted conform
pretty closely to the expected values.
These three requirements also come into play for visual decoding when the
signal is noisy or subject to short-term fading. It is a modulation scheme
in which even the human eye-brain system appears susceptible to threshold
effects.
A "stop bit" such as you have introduced substantially eliminates point (1)
above for eye or machine. It will also help with points (2) and (3) for
visual decoding, I suspect, but those two points may need to be addressed
further for machine decoding.
Just speaking speculatively here, it would seem two other factors should be
considered before fixing upon a 7FSK "standard." One is that 7 is not a
power of 2, and therefore we are losing some potential efficiency in the
event this method is ever adapted for machine decoding. An eighth level
would let each spot frequency become a symbol representing three bits of a
six-bit word. It isn't absolutely necessary if we're only going to copy
visually, but some error correction capability could then be built in for
later machine use.
Second, the coding scheme could be made more efficient. Right now, the code
is very much on the order of the A=1 B=2 C=3 technique of youthful schoolyard
secret agents. One might legitimately argue that since all characters are
represented by two symbols (and now, also a space), there is no payload
efficiency to be gained by ordering the relationship between characters and
symbol-pairs in any particular way. True for payload. But it is also
legitimate to consider the efficiency with respect to visual decoding, which
can be improved by letting the most commonly used characters help the viewer
establish the maximum span of the FSK and its increments.
Just as Morse code is partially optimized by representing the most common
English letter (e) with a single dit, the next most common (t) with a single
dah, etc., we could make it a little easier on the eye by letting "e" be
frequency pair 1-8, "t" be represented by 8-1, "a" by 1-2, "o" by 8-7, "i" by
1-3, "n" by 8-6, and so on.
Now, I realize the traditional "etaoin" sequence may not in fact be the most
common characters found in proper amateur QSOs. In that case, we would want
to assign whichever is the most commonly used character to be our widest
span, and let next-most-common characters emphasize the top and bottom
frequency extremes for the sake of humans decoding visually.
Meantime, the inclusion of a stop bit in the current 7FSK scheme does slow it
down somewhat, but it's still faster than 2FSK based on Morse character
encoding (and which is itself faster than straight QRS having the same dot
length).
73,
John
|
