This is an excellent and insightful evaluation. I am also very interested
to see how GRITTY evolves.
However, after thousands of (mostly) contest QSOs, I have seen many
instances of each of the following cases, using multiple decoders on the
same audio stream:
1. 2Tone prints clearly when MMTTY and the DXP38 show random or no
characters.
2. MMTTY prints clearly when 2Tone and the DXP38 show random or no
characters.
3. The DXP38 prints clearly when 2Tone and MMTTY show random or no
characters.
In my first few hours of using 2Tone, I naively concluded that it was
"superior" to MMTTY and the other decoders. Indeed it was superior for the
specific band and signal conditions I was experiencing at the time. Some of
those experiences were cases where there was no audible signal in my
headphones and there was no visible signal on the P3, but the call sign of a
station answering my CQ was printing multiple times, perfectly. I sent my
exchange and received a perfectly printed exchange, again without audible or
visible evidence of a signal. That, in itself, was so "other worldly" that
I had to conclude 2Tone was superior!
Then, after encountering a wider variety of band and signal and activity
conditions, I had to conclude that each decoder is superior in certain
specific situations. The value of using multiple, different and parallel
decoders is clear. Perhaps GRITTY will become an additional valuable
decoder.
Ed W0YK
_____________________________________________________
-----Original Message-----
From: RTTY [mailto:rtty-bounces@contesting.com] On Behalf Of David G3YYD
Sent: 12 April, 2015 08:03
To: rtty@contesting.com
Subject: [RTTY] GRITTY
Been evaluating GRITTY against 2Tone on 20m. I am listening to GM station
and 2Tone is giving perfect copy while GRITTY is all over the place often
with wrong callsign and at times giving numbers in place of letters in the
callsign. This is with the receiver bandwidth set at 1400Hz. Normally I
would use 500Hz but such a narrow bandwidth is not compatible with GRITTY.
The signal to noise ratio that GRITTY indicates is dependent on the receiver
bandwidth. 350Hz bandwidth on the K3 and GRITTY says the signal to noise is
40+dB even when there is no signal in the pass band. The band is not very
busy but in a contest I am not sure what it would do. This is not good as it
probably fools the Bayesian system. Yet need to have a narrow receiver
bandwidth when a very strong within the 3KHz bandwidth dominates the
receiver upsetting the decode.
Using receiver bandwidth narrower than 3KHz bandwidth fools the squelch
system so that it is continuously unsquelched.
Further listening to a pile up calling a DX station GRITTY stopped decoding
as it squelched while 2Tone was getting complete callsigns. As the pile up
thinned out then GRITTY started decoding but often got the callsign wrong
even though 2Tone was copying them OK.
Another one I noticed a station sent a lot of RYs, which was initially
decoded as RYs but then Bayesian turned it into a string of 46s. I have seen
similar on other letter combination where the decode was correct and then
Bayesian changed the case to the wrong case.
Interestingly enough I had a go at something similar a couple of years ago
and abandoned it in the end because it caused more errors than it corrected.
I concluded that under some conditions it was of value but on balance it was
not. It will be interesting to see if Alex can improve the system.
73 David G3YYD
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|