// Status=review

The JT65 protocol was described in a {jt65protocol} in 2005; details
of the JT9 protocol are presented in the next section of this Guide.
To users already familiar with JT65, the most striking difference
between the two modes is the much smaller occupied bandwidth of JT9:
15.6 Hz, compared with 177.6 Hz for JT65A.  Transmissions in the two
modes are essentially the same length, and both modes use exactly 72
bits to carry message information.  At the user level the two modes
support nearly identical message structures.

JT65 signal reports are constrained to the range –1 to –30 dB.  This
range is more than adequate for EME purposes, but not enough for
optimum use at HF and below.  S/N values displayed by the JT65 decoder
are clamped at an upper limit –1 dB.  Moreover, the S/N scale in
present JT65 decoders is nonlinear above –10 dB.  

By comparison, JT9 allows for signal reports in the range –50 to +49
dB.  It manages this by taking over a small portion of ``message
space'' that would otherwise be used for grid locators within 1 degree
of the south pole. The S/N scale of the present JT9 decoder is
reasonably linear (although it’s not intended to be a precision
measurement tool).  With clean signals and a clean nose background,
JT65 achieves nearly 100% probability of correct decoding down to 
S/N = –22 dB and about 50% at –24 dB.  JT9 is about 2 dB better, 
achieving 50% decoding at about –26 dB.  Both modes produce extremely 
low false-decode rates.

Early experience suggests that under most HF propagation conditions
the two modes have comparable reliability.  The tone spacing of JT9 is
about two-thirds that of JT65, so in some disturbed ionospheric
conditions in the higher portion of the HF spectrum, JT65 may do
better. JT9 is an order of magnitude better in spectral efficiency.
On a busy HF band, we often find the 2-kHz-wide JT65 sub-band filled
wall-to-wall with signals.  Ten times as many JT9 signals can fit
into the same frequency range, without overlap.