On Feb 24, 2012, at 3:38 PM, Dave AA6YQ wrote:
> It's parameters were chosen empirically to
> optimize decoding of a -10db SNR (AWGN) .wav file produced by Moe AE4JY's
> PathSim application, with results assessed by Alex VE3NEA's RTTYcompare
> application. This configuration yields a 16% character error rate (CER),
> compared to the standard profile's 41% CER, the hypersensitive profile's 21%
> CER, and the EU1SA profile's 53% CER.
As seen from Figure 4 of the Extended Nyquist Filter article that I just posted
a few minutes earlier, the Raised Cosine and Matched filter both comes up with
better than 13% character error rate for a -10 dB SNR. So there is another
level of improvement you can easily make if you can modify the FIR filter
coefficients in MMTTY. You should be able to get 1/5 fewer errors at -10 dB
SNR, and very likely even better improvement ratio than that when SNR becomes
better (i.e., you get to 0.1% error rate sooner).
(Too bad that in this day and age, we cannot post charts on the RTTY reflector.
:-).
By the way, I am not arguing for always using the optimal filter that is
designed for AWGN. They are fantastic during the late evenings on 20m when
conditions are "quiet" and the band appears to be "dead" and a really, really
weak signal appears.
However, for signals exhibit multipath (when you start noticing selective
fading, or even worse, can hear flutter), such a narrow filter will produce
more errors than other schemes. But that is yet another topic that we can
discuss at a later date :-).
In any case, an extra 5 dB of antenna directivity (or changing to use better
modulation modes like MFSK16) will beat any demodulator optimization one can
make for RTTY :-).
Incidentally, the VE3NEA chart shows MMTTY at nearly 50% character error rate,
so MMTTY "standard" setting must have improved (to 41%) since Alex made his
measurements.
>From what I can tell, TrueTTY comes very close to the ideal (Raised Cosine or
>Matched Filter) at -10 dB SNR, but it flares out a bit when the SNR becomes
>better. This is typical for a filter that has some residual ISI -- it could
>be coming from some preceding BPF stages, causing the combined filter to no
>longer be Nyquist. Other than that, I suspect TrueTTY went "by the graduate
>level textbook" when it comes to demodulator design.
73
Chen, W7AY
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|