Re: [wsjt-devel] Using WSJT-X Decoding Technique with RTTY

[Bill S. <g4...@cl...>]

On 03/01/2019 15:33, Frank Kirschner wrote:
> Many years ago, I had a home-brew RTTY TU. It had two 88 mH torroids 
> in the tuned circuits and a differential amplifier as a detector. 
> Occasionally, the received signal would fade below the noise, but the 
> Kleinschmidt teleprinter would keep printing the decoded text. I have 
> tried several current software packages for decoding RTTY signals, and 
> all of them require quite strong signals to decode at all, and very 
> strong signals to provide 100% print.
>
> I was wondering if the decoding method used in WSJT-X for FT8 could be 
> ported to a RTTY decoder. I realize that FT8 has a lower symbol rate, 
> which provides more time to integrate, and incorporates redundancy, 
> which improves accuracy. But the theoretical lower limit for decoding 
> RTTY is something like -5 dB S/N, and the current programs are nowhere 
> near that.
>
> If RTTY facilities could be integrated with WSJT-X, not only would the 
> decoding be better, but the log scanning functions of WSJT-X and 
> JTAlert would be better, significantly improving contest operation.
>
> I confess I haven't looked at the decoding techniques used by RTTY 
> software, but I suspect, based on performance, that it could be improved.
>
> 73,
> Frank
> KF6E
>
Hi Frank,

this is well covered ground, some existing Baudot decoders have very 
good performance with built in weak and fading signal models to optimize 
them.

The techniques used to decode FT8 are based around a block message 
format that includes parity bits and checksums for advanced forward 
error detection and correction, Baudot is a character based stream code 
with no forward error correction capability. RTTY was designed for an 
era where automation was mechanical and quite limited, just having a 
code that could be used to modulate a transmitter and be printed by a 
receiver was about the sum of the technical design. RTTY's *only* 
similarity with FT8 is that it is a frequency shift keyed constant 
amplitude signal. Even the simplest addition of a single parity bit to 
detect, but not correct, single bit errors would need a fundamental 
non-backwards compatible change to the Baudot code.

73
Bill
G4WJS.