On 11/24/2018 5:57 PM, Bill Somerville wrote:
On 25/11/2018 01:51, Dana Myers wrote:
Audacity has long been my tool of choice for that. Not really sure I'd say
"degrade",
since 8ks/s is perfectly adequate for the usual <= 3KHz FT8 audio; downsample
is true, but that's not degradation.
73,
Dana K6JQ
Hi Dana,
the degrade step I was referring to was the application of the propagation model by the application. Agreed the resampling
artefacts should be small in comparison, but not insignificant due to the non-integral rate conversion factors (8:12 and 12:8),
if the best resampling algorithm parameters are selected in the resampling tool.
FWIW, my very simple test (generate a CQ and record that via Virtual Audio
Cable,
then downsample to 8Ks/s, process with PathSim 1.0, upsample the result to
12Ks/s),
showed me that 2.0rc4 decoded a "CCIR Good, SNR -10dB" 77-bit FT8 signal as
-19dB.
Rolling back to 1.9.1 and using a 75-bit FT8 signal, 1.9.1 decoded the signal
with the
same CCIR path, I saw an SNR -7dB signal decoded as FT8 -16.
In both cases, this is the best decode performance I saw.
In a very broad way, v2.0rc4 seems to perform better with the 77-bit frame.
Anyone interested in these results in welcome to reproduce them, I'll
explain how if contacted.
73,
Dana K6JQ
_______________________________________________
wsjt-devel mailing list
wsjt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
