WSJT-X/doc/user_guide/implementation.adoc

// Status=review
.Transmitting

Immediately before the start of a transmission _WSJT-X_ encodes a
user's message and computes the sequence of tones to be sent.  The
audio waveform is computed on-the-fly, with 16-bit integer samples
sent to the audio output device at a 48000 Hz rate.  Generated signals
have continuous phase and constant amplitude, and there are no key
clicks.  The transmitter's power amplifier need not be highly linear.

.Receiving

_WSJT-X_ acquires 16-bit integer samples from the audio input device
at a 48000 Hz rate and immediately downsamples the stream to 12000 Hz.
Spectra from overlapping segments are computed for the waterfall
display and saved for the decoder at intervals of 3456/12000 = 0.288
s, half the JT9 symbol length.

.Decoding

At the end of a reception sequence, about 50 seconds into the UTC
minute, received data samples are forwarded to the decoder.  In JT9
and JT65 modes the decoder goes through its full procedure twice:
first at the selected Rx frequency, and then over the full displayed
frequency range.  Each decoding pass can be described as a sequence of
discrete blocks.  Details of the functional blocks are different for
each mode.  In dual-mode JT9+JT65 operation on computers with more
than one CPU, decoding computations for the two modes are done in
parallel.

The basic decoding algorithm for JT65 mode is described in the 2005
{jt65protocol} paper.  The following list summarizes the corresponding
algorithm for JT9 mode. Blocks are labeled here with the names of
functional procedures in the code.

[horizontal]
+sync9+::    Use sync symbols to find candidate JT9 signals 
            in the specified frequency range

Then, at the frequency of each plausible candidate:

[horizontal]
+downsam9+::  Mix, filter and downsample to 16 complex 
            samples/symbol

+peakdt9+::   Using sync symbols, time-align to start of JT9 symbol 
            sequence

+afc9+::    Measure frequency offset and possible drift

+twkfreq+::   Remove frequency offset and drift

+symspec2+::  Compute 8-bin spectra for 69 information-carrying
            symbols, using the time- and frequency-aligned data;
            transform to yield 206 single-bit soft symbols

+interleave9+:: Remove single-bit interleaving imposed at the
	    transmitter

+decode9+::   Retrieve a 72-bit user message using the sequential
            Fano algorithm 


+unpackmsg+:: Unpack a human-readable message from the 72-bit 
            compressed format

Decoding of clean JT9 signals in a white-noise background starts to
fail below signal-to-noise ratio -25 dB and reaches 50% copy at -26
dB.

With marginal or unrecognizable signals the sequential Fano algorithm
can take exponentially long times.  If the +sync9+ step in the above
sequence finds many seemingly worthy candidate signals and many of
them turn out to be undecodable, the decoding loop can take an
inconveniently long time.  For this reason the step labeled +decode9+
is programmed to "`time out`" and report failure for a given signal if
it takes too long.  The choices *Fast | Normal | Deepest* on the
*Decode* menu provide the user with a three-step adjustment of the
timeout limit.

Decoding in JT4 and WSPR modes is basically similar to that for JT9.
The same K=32, r=1/2 sequential Fano algorithm is used for all three
modes.  Other details such as tone numbers, symbol lengths,
synchronizing scheme, and subroutine names are distinct for each mode.
The WSPR decoder now uses a two-pass algorithm.  Decoded signals are
reconstructed and subtracted from the received waveform, after which a
second decoding pass takes place.