Steven Franke
701d517e6e
For fst4sim, use Lorentzian fading spectrum when fspread is negative.
2021-01-04 15:00:03 +00:00
Steven Franke
b087328fc1
Don't waste time on excess BP iterations when doing K=50. Decode a little deeper.
2021-01-04 15:00:03 +00:00
Steven Franke
8079d35561
Append decoder data to file fst4_decodes.dat instead of overwriting.
2021-01-04 15:00:02 +00:00
Steven Franke
89cbd177df
If file decdata is present in the data directory, then write detailed decoder data to file fst4_decodes.dat in the same directory.
2021-01-04 15:00:02 +00:00
Steven Franke
812bd29e54
FST4W: Use K=66 for first OSD decode attempt and for updating fst4w_calls.txt. Use K=50 for 2nd attempt.
2021-01-04 15:00:02 +00:00
Steven Franke
fc3e24f696
More work on K=50 decoding.
2021-01-04 15:00:01 +00:00
Steven Franke
65ff3ce05e
Remove some debug prints.
2021-01-04 15:00:01 +00:00
Steven Franke
11b7aa9429
Implements decoding of FST4W messages as (240,50) crc-less codewords.
...
By cascading the full 24-bit crc generator matrix with the (240,74) LDPC code generator, create a (240,50) generator that is used to decode with approximately 1 dB better sensitivity than the (240,64) with 14-bit CRC approach that is normally used. This approach treats the CRC bits as additional parity bits and provides no means for identifying incorrect codewords. All codewords on the list generated by the OSD algorithm have CRCs that match the CRC of the message payload. Codewords are validated by unpacking the message and comparing the unpacked message with the list of stored callsign/grid pairs stored in the fst4w_calls.txt file.
2021-01-04 14:59:56 +00:00
Bill Somerville
ac6a6d0cd7
Fail wsprd if data directory specified not accessible
...
Also fixed trivial memory leak on error exit.
2020-12-03 13:12:12 +00:00
Bill Somerville
3134dd8a60
Removed incorrect pointer declaration, tnx to Rud, K5RUD
2020-12-03 12:20:14 +00:00
Bill Somerville
de0af5a2ed
Merge branch 'feat-outgoing-udp-interface' into develop
2020-11-10 20:09:48 +00:00
Steven Franke
e0888eb70c
Restore wsprsimf and dependencies.
2020-11-06 09:33:44 -06:00
Bill Somerville
1a223f5c45
Remove unwanted file and directory
2020-11-03 20:29:25 +00:00
Bill Somerville
731fe53a1e
Reasonable default values for NB setting for command line jt9
...
Use the -X command option to set single decode mode and to set the NB
level.
-X 256*(n+3)+s
Where s=32 selects single decode mode, and 'n' is the equivalent NB
setting as used in the WSJT-X UI. Default for both is zero in FST4(W)
mode if no -X option is passed. E.g. -X 544 selects single decode with
NB=-1 (256*(-1+3)+32).
2020-10-26 20:20:52 +00:00
Bill Somerville
5efe448a9c
Cosmetic change
2020-10-23 03:32:22 +01:00
Bill Somerville
308d815fb5
Add --freq-tolerance (-F) jt9 option
2020-10-23 03:24:19 +01:00
Bill Somerville
90b3fa2a46
Merge branch 'release-2.3.0' into develop
2020-10-13 16:51:40 +01:00
Bill Somerville
6026944e33
Workaround for some memory issues, possibly thread stack sizing
...
Reduced maximum thread team size for MSK144 candidate discovery to 4.
2020-10-07 13:12:20 +01:00
Bill Somerville
ead587ee55
Fix up out of date common block sizing
2020-10-04 22:52:51 +01:00
Bill Somerville
203e0da5d6
Fix up out of date common block sizing
2020-10-04 00:45:56 +01:00
Joe Taylor
9200c3e27a
Add NB = -3% option to cycle through NB=0, 1, 2, 3, ... 20%.
2020-09-30 13:32:03 -04:00
Joe Taylor
0ab3e5116f
Fix several flaws in the loop-over-NB logic in the FST4 decoder.
2020-09-19 10:08:42 -04:00
Joe Taylor
2af01ebaa1
Fix a flaw in the loop-over NB logic. There are more flaws!
2020-09-18 15:52:33 -04:00
Joe Taylor
f0ed93cdd0
In the NB-loop, don't cycle around a decode attempt for the npct=0 pass.
2020-09-18 11:45:28 -04:00
Joe Taylor
52bdd57e57
Implement NB=-1%, NB=-2%. Fix a startup problem with WideGraps's fMax value.
2020-09-18 11:30:23 -04:00
Joe Taylor
375a869a51
Revert "Revert "Very basic code (including some diagnostics) for "try all NB settings". Will remove it again.""
...
This reverts commit 033cc65d08
2020-09-18 09:01:51 -04:00
Joe Taylor
033cc65d08
Revert "Very basic code (including some diagnostics) for "try all NB settings". Will remove it again."
...
This reverts commit bcf7f36b9c
2020-09-17 19:30:07 -04:00
Joe Taylor
bcf7f36b9c
Very basic code (including some diagnostics) for "try all NB settings". Will remove it again.
2020-09-17 19:28:55 -04:00
Steven Franke
7d58df4cc1
In FST4 mode with Single Decode not checked, move candidates within 20 Hz of nfqso to the top of the list.
2020-09-17 14:22:38 -05:00
Joe Taylor
21dc6a5c59
Connect the FLow and FHigh limits for FST4 decoding.
2020-09-16 17:23:59 -04:00
Steven Franke
221ede2903
Remove some unused variables.
2020-09-14 13:07:07 -05:00
Steven Franke
b49a90f530
Remove a redundant array.
2020-09-14 13:03:33 -05:00
Steven Franke
1b59d9dc8c
Eliminate redundancies from the calculation of sequence correlations.
2020-09-14 12:42:32 -05:00
Steven Franke
ca0804450b
Remove some redundant code.
2020-09-14 09:07:45 -05:00
Steven Franke
3d8c8e3dd7
Merge branch 'develop' of bitbucket.org:k1jt/wsjtx into develop
2020-09-14 09:00:35 -05:00
Steven Franke
98d52e35ac
Speed up FST4 decoding.
2020-09-14 09:00:30 -05:00
Joe Taylor
a1baaebee9
Decoder should reject data with rms < 3.0 over first 15 seconds.
2020-09-12 13:46:09 -04:00
Joe Taylor
263675cac4
Fix an oddball result with i*2 numbers: abs(-32768)=-32768.
2020-09-12 09:00:39 -04:00
Steven Franke
71fdcd1119
Silence a compiler warning.
2020-09-10 16:11:07 -05:00
Steven Franke
2dcde590df
Use widegraph limits for noise baseline fit. Limit signal search to within the widegraph limits.
2020-09-10 14:58:10 -05:00
Joe Taylor
47fcddcb50
Send nfa, nfb to fst4_decode().
2020-09-10 14:59:52 -04:00
Joe Taylor
01a1688b32
FST4 and FST4W decoder: get freq range from nfqso and ntol; remove BCD submodes, i.e. hmod=2, 4, 8.
2020-09-05 13:14:40 -04:00
Joe Taylor
f24d15b16e
Eliminate the FST4/FST4W submodes with hmod > 1.
2020-09-04 15:15:30 -04:00
K9AN
f066936043
Remove an unused variable from fst4_decode
2020-08-29 19:55:23 +01:00
Steven Franke
5ca81a6507
Use 3rd order polynomial fit to estimate the noise baseline. The
...
polynomial fit is done over 400 Hz bandwidth for T/R periods longer
than 15s, and over approx. 600 Hz (10 times the signal bandwidth) for
T/R period of 15s.
2020-08-29 19:55:23 +01:00
Steven Franke
d82b9f5b0e
Speed up decoder by eliminating some complex multiples in sequence detection loop. Add timer calls for doppler spread calculation and sequence detection loop.
2020-08-29 19:55:22 +01:00
Steven Franke
ecaca6af9f
Fix argument list in call to fet_fst4_bitmetrics.f90
2020-08-29 19:55:22 +01:00
Steven Franke
b9328b96c9
Tweaks to update the diagnostics that are written to fort.21.
2020-08-29 19:55:22 +01:00
Steven Franke
e02850ae5a
Streamline fst4_decode. Add timer for downsampling.
2020-08-29 19:55:22 +01:00
Steven Franke
782c779392
Reconfigure to optimize decoder for MF/LF (high coherence) channels.
2020-08-29 19:55:21 +01:00
