LF: Re: MF JT9-2 report

[Rik Strobbe <rik.strobbe@kuleuven.be>]

Paul, all,

before I started programming SlowJT9 I did a "dry test" by generating JT9(-1) and JT9-2 WAV files with a known SNR and feeding these to the JT9 decoder of the WSJT-X suite (the JT9-2 WAV files after conversion to JT9). In the SNR range from -24 to -32 dB I generated 100 files per step, in 0.1dB steps (an automated process that over 12 hours).
The result was:

The 50% successful decodes treshold was -26.9dB for JT9(-1) and -29.3dB for JT9(-2), so 2.4dB in favour of JT9-2.

But is has to be said that just pure white noise was added (no static crashes etc...) and the signals were perfectly stable. No TX/RX drift, no ionospheric effects. All these might reduce the JT9-2 advantage.

Therefor test reports with (very) weak signals are most welcome.

73, Rik  ON7YD - OR7T

________________________________________
Van: owner-rsgb_lf_group@blacksheep.org <owner-rsgb_lf_group@blacksheep.org> namens N1BUG <paul@n1bug.com>
Verzonden: maandag 12 november 2018 13:15
Aan: 600MRG@mailman.qth.net; rsgb_lf_group@blacksheep.org
Onderwerp: LF: MF JT9-2 report

I decoded 4 stations on JT9-2 last night: KU4XR, N2END, WB4JWM,
K5DNL. KU4XR was visible many times on the waterfall without decodes
but he did have drift varying from 3 Hz / period to 8 Hz / period. I
never saw any decodes reported below -24 SNR, which if I understand
correctly is really -27. This is no better than JT9-1 but may be
just a matter of not enough data collected, no really weak signals
during the test period.

Let's keep the tests going! I need to see some weaker signals to
test the mode. As soon as I can get out to the antenna site to
install the variometer and transformer, I will join in for some
transmitting and QSOs.

This morning I tried to check for additional overnight decodes but I
found the program had exited (crashed) sometime during the night. I
looked in the program folder and found a decoded.txt file with just
one line in it. I guess that was the last decode before the crash?

73,
Paul N1BUG