Further I had to assume that for all submodes the transmissions start a 1 seconds after the minute, as it is for JT9(-1).

The check version 0.9.10 I did 2 things:

-

1. Fed the soundcard input with a 600Hz sine tone while SlowJT9 is running in JT9-1 mode, with audio saving on (that way the audio files used to feed the JT9 decoder are saved on disk).

In this audio file I could see a nice sine at 1333Hz, as expected (with the original parameters the conversion rate between JT9-2 and JT9-1 is 15360/912 = 2.222, so 600Hz becomes 2.222*600 = 1333Hz). This is at least a strong indication that the JT9-2 to JT9(-1) conversion is done correctly.

For those interested in the details: due the the odd conversion rate, each 2.222 JT9-2 samples now have to be averaged to 1 JT9-1 sample, I will explain in a separate mail how I did this.

2. Recorded the JT9-2 signal generated by SlowJT-9, added (a lot of) noise to it and then played it back tp SlowJT9. I did this for the both JT9-2 version. For both versions I had a sloid copy down to -30dB. But as said earlier: it was just a short test. Good enough to constate that the decoding of the original JT9-2 is more or less as good as the earlier version, but certainly not good enough to confirm the (theoretical) S/N gain of about 0.5dB.

About the frequencies: the maximum frequency that the JT9 decoder will handle is 5000Hz. This means that for JT9-2 the maximum frequency is 5000/2.2222 = 2250 Hz (but SlowJT9 has a general upper limit of 1400Hz to keep out of the WSPT band).

For JT9-5 the maximum frequency is 5000*6912/40960 = 843 Hz and for JT9-10 it is 5000*6912/82944 = 416Hz!

An important remark: I just noticed that the baurdrate conversion is SlowJT9 is OK, but not the entire timing. As a JT9-2 transmission now also starts at 1 second, after conversion to JT9(-1) it will start at 0.45 seconds (1/2.2222), thus an error of 0.55 seconds. The JT9 decoder should be able to handle this without too moch loss. But for JT9-5 and JT9-10 the error will become larger!. I will correct this in the next version of SlowJT9.

73, Rik  ON7YD - OR7T

________________________________________
Van: [email protected] <[email protected]> namens N1BUG <[email protected]>
Verzonden: woensdag 21 november 2018 13:47
Aan: [email protected]; [email protected]; rsgb_lf_group
Onderwerp: Re: LF: SlowJT9 update (v0.9.10): tuning issue solved

Hello Rik, all,

Yesterday I could do some tests with my QRP Labs U3S transmitting,
SlowJT9 receiving. I used a lot of attenuation between U3S and TX
antenna. Then there is another 30 or 40 dB lost between the TX
antenna and RX antenna. The result was that in JT9-1 I had a signal
which was most of the time -27, occasionally -26 or -28.

I then switched to JT9-2. I confirm SlowJT9 will decode these
transmissions using the old baud rate. SNR was now mostly -29, so I
am not sure this is reporting correctly. SNR should have remained
the same when changing modes.

I then switched to JT9-5. I confirm SlowJT9 will decode this OK. SNR
was now mostly -30, again does not seem logical to me.

More tests are needed to see if I always get the same SNR results on
the different modes! There was not enough time to run enough cycles
of each mode to eliminate random chance.

I tried JT9-10 but I could not get any decode. I assume for this
mode the audio frequency must be under 500 Hz? I tried 470 Hz but
nothing. The signal was clearly visible on the waterfall but did not
decode.

Later, K3MF upgraded to the new version and made some JT9-2
transmissions on 2200m. His signal was strong as usual, but I could
not decode him at all. I could confirm that his transmission was
occupying the correct amount of time for the slower baud rate, but
it could be decoded.

That is all for now. I will continue my SNR tests between the submodes.

73,
Paul N1BUG