chris/js8call
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merged master 8748

Browse Source
This commit is contained in:
Jordan Sherer
2018-08-05 11:33:30 -04:00
parent 8f8772f1bd
commit 62899069bf
1095 changed files with 31298 additions and 367679 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.svn/pristine/f6/f68f71bd3d3b816812122731ae2c2b9b2afca6ae.svn-base
-791
View File
@@ -1,791 +0,0 @@
WSJT-X ChangeLog
------------------------------------------------------------------
May 30, 2013: Version 1.0, r3323
--------------------------------
With this release of WSJT-X Version 1.0 we include a few (relatively
minor) enhancements in response to user requests, as well as some
program polishing and cleanup. Active program development will
continue, but new releases will become less frequent.
1. New option on the Setup menu: "Tx freq locked to Rx freq".
2. Double-click on a decoded "73" message now sets Tx5, rather than Tx6.
3. New keyboard shortcuts: Alt+1 through Alt+6 set the next Tx message
at the corresponding number.
4. PTT control via Ham Radio Deluxe has been imnplemented and tested.
5. "Tool Tips" are now provided for most on-screen controls.
6. Under Linux and OS X, listings of available audio devices and APIs
have been corrected.
7. Tab order among GUI controls has been cleaned up.
8. Updates to the WSJT-X User's Guide.
May 22, 2013: v0.99, r3297
--------------------------
1. CAT control via Ham Radio Deluxe is now available. For setup
details see item #5 at the top of page 6 of the updated WSJT-X
User's Guide.
2. Submodes JT9-5, JT8-10, JT9-30 have been de-activated. (As far as I
know, nobody was using them.) This action makes the program smaller
by some 150 MB and able to run effectively on some older computers.
3. Bizarre ordering of COM port numbers on the drop-down list has been
corrected, and suitable serial ports added to the list displayed in
Linux.
4. Gray bar between decoding periods now contains a dashed line.
5. Corrected a bug that prevented use of Setup | Configuration with
no existing wsjtx.ini file.
May 17, 2013: v0.95, r3278
--------------------------
1. Double-clicking on a decoded text line in "Band Activity" window
now copies the line into the "QSO Frequency" window if it was not
already there.
2. Option "Color highlighting in left window" removed from Setup
menu. Highlighting is now always done.
3. Positions of "QSO Frequency" and "Band Activity" windows have been
swapped.
4. F4 was restored to its previous use; F5 is now used to display
Special Mouse Commands
5. Small square between Band selector and Frequency readout was made
a control button. Orange indicates one-way CAT control from
program to radio, red indicates bi-directional control. Clicking
the orange button causes a one-time readout of dial frequency.
6. If Save=None, the last recorded file is deleted on program exit.
This prevents unwanted accumulation of files in the Save
directory.
7. Status-bar messages were re-arranged in a more logical order.
8. Tx signal report was added to wsjtx_status.txt (for JT-Alert)
9. More informative labels were placed on the "Tab 2" GUI controls.
10. Better default scaling for the "Cumulative" spectrum.
11. New algorithm for identifying JT9 signals to send to decoder,
resulting major improvements in decoder speed.
12. Bug fixes:
- Incorrect displayed frequencies for JT9-2 signals
- Infinitely repeated "Error rigOpen -1" messages
- User tries to open CAT control using busy or nonexistent serial port
13. Many updates to the User's Guide
May 2, 2013: v0.95, r3251
-------------------------
1. The "band change" function is executed whenever the Band combobox
is activated, even if the selected band has not changed.
2. The program does not set rig mode. That task is left to the user.
3. Time interval for polling rig frequency is now a user parameter on
the setup screen. I set mine to 1 second, which works fine with
the Kenwood TS-2000. Set it to 0 if you want no polling for
frequency (which means unidirectional CAT control from program to
radio). Choose something like 10 s for the K3.
4. Much new work on the WSJT-X User's Guide, which is approaching its
final form for Version 1.0. Please read it and tell us about
anything you find unclear or missing!
These changes address nearly all of the CAT issues found by a few
users -- those with K3, IC-746, IC-706, in particular.
One additional piece of advice when running WSJT-X in Windows: connect
and turn on the radio and any interface equipment before starting
WSJT-X, and exit the program before turning such equipment off.
April 29, 2013: v0.95, r3243
----------------------------
1. Now has bi-directionsl CAT control using direct calls to hamlib
functions. Highlights displayed dial frequency with red background
if frequency differs from nominal for the selected band by more
than 100 kHz. (Is there a more useful or logical condition to
flag?) Small red square between Band selector and Dial Frequency
display to indicate that CAT control is active. Mode is set to USB
on startup. (Note: CAT control diagnostics are presently rather
rudimentary, we're still working on this. Feedback is welcome!)
2. New controls on Setup | Configuration screen:
- RTS/DTR OFF -- needed by K2 and maybe other radios
- Data / Mic -- select CAT-control PTT type
- Test CAT Control -- try settings before returning to main screen
- Test PTT -- try settings for toggling PTT On/Off
3. Help menu now provides displays of Keyboard Shortcuts (quick access
via F3) and Special Mouse Commands (F4).
4. Option "Setup | Advanced | Allow multiple instances" enables the
use of more than one instance of WSJT-X for special applications.
(The program must be copied into and run from different
directories.)
5. No posts to PSK Reporter if band was changed during the reception
period.
6. Improved behavior of Tune button.
7. Improved inter-process communication between WSJT-X and JT-Alert-X.
8. Better interaction between "Report" spinner control and Tx messages.
9. Removed the NB checkbox and slider. (Has anyone found these useful?)
10. New buttons on main window: "Tx=Rx" sets Tx frequency to current
Rx frequency; "Rx=Tx" does the opposite.
11. Log QSO confirmation window is now "non-modal": you can keep it
open and still access controls on the main window.
12. Tab-order has been rationalized on most screens.
13. Dial frequency and mode written to file ALL.TXT.
14. Double-click on decoded line sets Tx message #3 if message has
the form "MyCall Call2 rpt".
15. Bug causing occasional, seemingly unpredictable program crashes
has been fixed.
16. The WSJT-X User's Guide is somewhat closer to being complete. User
feedback on the Guide will be most welcome. What is unclear? What
is missing?
April 17, 2013: v0.9, r3195
---------------------------
1. Sorry, the CAT control changes in r3187/3188 were a dismal failure
in many stations, and they introduced other bugs as well. This revision
goes back to uni-directional CAT control: the program can set the
radio's dial frequency and do T/R switching, but that's all. The band
setting is not reset on program startup.
2. Logic for the Tune button has been corrected.
3. For Linux compile-it-yourself enthusiasts: the interface to
PSK Reporter is now working undel Linux.
April 16, 2013: v0.9, r3188
---------------------------
1. CAT control now reads and follows changes in radio's dial
frequency. Readout gets red highlighting if radio is on wrong band.
On program restart, band is reset to the last selected band.
2. New "Tune" button generates an unmodulated carrier. Toggle button
a second time to turn Tx off.
3. Added labels at top of "Tab 2" and enlarged the text entry fields.
4. Fixed the broken logic for "Runaway Tx watchdog".
5. Fixed "Prompt me to log QSO" so that it no longer requires also
setting "ID after 73".
6. Additional changes of (eventual) interest to Linux users. Code for
sending spots to PSK Reporter now in place.
April 13, 2013: v0.9, r3166
---------------------------
1. Option to send Tx messages (highlighted in yellow) to the QSO window.
2. Prevent starting a transmission more than 24 sec into a Tx period.
3. "Setup | Options" changed to "Setup | Configuration".
4. Type Alt-V to save the most recently completed Rx file.
5. Fixed bug that truncated Rx messages to 16 characters.
6. Internal program changes that should provide better user diagnostics
when necessary at program startup.
April 11, 2013: v0.9, r3157
---------------------------
1. Maximum size of several window areas increased to accommodate system
fonts set larger than default.
2. New behavior of Erase button: click once to erase the left (QSO) window,
twice to erase both decoded text windows.
3. Keyboard shortcuts:
Alt-D: decode again at QSO frequency (same as clicking the Decode() button)
Shift-D: do another full decode in both windows
Alt-E: Erase()
Ctrl-F: Edit the free text message box
Alt-H: Halt Tx()
Alt M: Monitor()
Alt-N: Enable Tx()
Alt-Q: Log QSO()
Alt-S: Stop()
4. New Setup options: "Tx disabled after sending 73" and "Runaway Tx
watchdog".
5. Fixed bug in saving the "report received" for logging purposes.
6. Corrected the logic for "Runaway Tx watchdog".
7. Fixed bug that truncated characters 17 and 18 of decoded messages.
April 10, 2013: v0.9, r3151
---------------------------
1. Blank line between decoding periods is now in the right-hand
text window, where it should be.
2. Decoding range defined by fMin and fMax is now enforced.
April 9, 2013: v0.9, r3143
--------------------------
This minor release restores the decoding speed of earlier revisions
and corrects a bug that prevented sending CW ID.
April 9, 2013: v0.9, r3142
--------------------------
This version of WSJT-X has a number of significant changes. Please
read the following notes carefully. Also -- even if you are already
familiar with WSJT-X -- be sure to read the updated WSJT-X User's
Guide at
http://www.physics.princeton.edu/pulsar/K1JT/WSJT-X_Users_Guide.pdf ,
especially pages 3 and 4.
Changes since v0.8 r3118 include the following:
1. There are now two scrolling windows for decoded text. The left
window contains decodes only from close to the designated QSO
frequency. The right window includes signals over the full decoding
range.
2. An alternative set of controls is now available for generating and
selecting Tx messages. Some may find these more convenient to use
than the Tx1 through Tx6 message boxes.
3. A number of new user options are available on the Setup menu:
- Blank line between decoding periods (right window only)
- Clear DX Call and DX Grid after logging QSO
- Display distances in miles
- Runaway Tx watchdog
- Background colors for left window
- Double-click on decoded message sets Tx Enable
4. New or changed on-screen features
- "Tol" replaced by fMin and fMax on waterfall screen (see User's Guide)
- Spinner control for signal report
- On waterfall scale: green marker for Rx freq, red for Tx. blue
for decoding range
5. New behavior
- "CQ DX" is now treated properly when decoded line is double-clicked
- Message formate for compound callsigns (e.g., PJ4/K1ABC, G4XYZ/P)
are now handled correctly. (Some restrictions apply, and will
be spelled out in the completed User's Guide.)
- Decode button now causes a decode only at the specified Rx frequency.
- Click on waterfall spectrum sets Rx freq; double-click also invokes
decoder (as though Decode button had been clicked). CTRL-click moves
both Rx and Tx freqs.
- Amplitude at end of transmission is ramped down to prevent a final
key click.
6. The following bugs have been fixed:
- Logic error in decoder
- Certain non-standard Tx messages could cause a program crash.
- Certain (rarely used) messages did not pack/unpack correctly
April 2, 2013: v0.8, r3118
--------------------------
1. Improved interface to program JT-Alert, by VK3AMA.
2. The LogQSO confirmation dialog no longer blocks the GUI updating
process.
3. A blank line with gray background separates the decoded text lines for
each new invocation of the decoder.
4. New suggested default frequencies: 5.357, 18.104, and 24.918 MHz.
Be sure to edit these entries on the 'Default Frequencies' tab of the
Setup screen. (When you have done this once, the new values will be
remembered.)
5. The LogQSO button now does nothing is the 'DX call' entry field is
blank.
6. Several minor bugs were fixed.
March 27, 2013: v0.8, r3113
---------------------------
1. Bug fix: VOX control of T/R switching now works.
2. Potentially useful error messages now appear when CAT control
has failed.
3. Added an instruction on the Log QSO confirmation screen.
4. Clear the DXcall and DXgrid entries after logging a QSO.
March 26, 2013: v0.8, r3112
---------------------------
Edson Pereira, PY2SDR, recently became an active contributor to this
open-source project. Edson and I have been very busy over the past
few days! WSJT-X revision 3112 has many changes and new features.
1. The GUI layout has been adjusted and optimized.
2. CAT control is now operational, offering optional control of your
radio's dial frequency and T/R status. Go to the Setup | Options
window to select the necessary parameters.
3. CW ID has been implemented. You can have your ID sent after a fixed
time interval, or automatically when you transmit a "73" or free
text message.
4. Default dial frequencies are available for each band on a new tab
on the Setup | Options window. Please note: some of these
frequencies are probably wrong! You can edit them as needed.
(Please let us know if the original values are inconsistent with
actual practice on any band.)
5. Several new options appear on the Setup menu. Try them!
6. Azimuth and Distance information is displayed whenever a valid grid
locator appears in the "Dx Grid" box.
7. The decoder has again been adjusted for better compromise between
sensitivity and decoding time.
8. The User's Guide is out of date, and needs work. We hope to get to
that task soon.
9. Very important for some would-be users: WSJT-X now runs properly
under Linux. We haven't made a package yet, so for now you must
compile your own. If you don't know how, we hope to be set up
to make packages before too long.
10. If you know someone who might be interested in contributing to the
development of WSJT-X and related projects, please send him/her
our way! We're especially looking for someone interested in
producing packaged Linux distributions -- for example, *.deb or
*.rpm packages, but other programming help is also wanted.
As always: please report bugs, and don't be bashful about sending us
your feature requests!
March 22, 2013: v0.7, r3071
---------------------------
1. Correct a bug that (still) allowed display of previous decodes
when nothing new was decoded.
2. Add a user confirmation screen activated when you click Log QSO.
This lets you edit or add information before it is written to the
ADIF file.
3. Tx message macros and now available. Configure them on the Setup
window. They are invoked as a pop-up menu by right-clicking on the
Tx5 message window; then select the desired message by left-clicking
on the desired message.
March 20, 2013: v0.7, r3063
---------------------------
1. Add Frequency to the generated ADIF records.
2. Correct a decoder bug that led to duplication of previous output
when nothing new was decoded.
March 19, 2013: v0.7, r3061
---------------------------
1. Allow Windows COM port numbers up to 99.
2. Replace status files wsjtx_qrg.txt and wsjtx_txcall.txt with
a single file, wsjtx_status.txt.
3. Combine wsjtx_rx.log and wsjtx_tx.log into a single file ALL.TXT.
4. "Log QSO" now writes a file in ADIF format.
5. Starting to implement popup macros for Tx message #5.
6. Big improvement in decoding speed.
*** More changes to come! Please report any problems, especially
*** with the ADIF-format log.
March 12, 2013: v0.6, r3046
---------------------------
1. Decoded calls can now be uploaded to the PSK Reporter web site.
Check the box "Enable PSK Reporter" on the Setup screen, and go to
http://pskreporter.info/pskmap.html to see the spots. Be sure to
enter your "Dial Frequency (MHz)" at lower right of the Wide Graph
window. (Rig control features are yet to come...)
2. Added some interfaces to permit use with the program JT-Alert,
by VK3AMA. Look for this capability in the near future.
March 6, 2013: v0.5, r3038
--------------------------
1. Selection of Current/Cumulative/JT9Sync for the 2d spectral display
changed to a combobox.
2. Double-click on decoded text does not change frequency settings
if first decoded call is MyCall.
March 1, 2013: v0.5, r3026
--------------------------
1. The horizontal scale of 2d spectra (e.g., the "red curve") is now
correct when the user has selected FFT Bins/Pixel > 1.
2. Double-clicking on a decoded text line now selects the second
callsign independent of exactly where one has clicked on the line.
In addition, it sets the selected frequencies (both Tx and Rx) to
the frequency of the decoded transmission.
December 11, 2012: v0.5, r2791
------------------------------
1. Messages of the form "CQ DX K1ABC" are now supported.
November 30, 2012: v0.5, r2788
------------------------------
1. A bug was introduced when support for positive signal reports was
added. It could cause a program crash when certain free-text messages
were composed for transmission. The bug has been fixed.
2. In the slower JT9 sub-modes, the UTC listed on decoded text lines
has been changed to the start time of the Rx sequence, rather than the
time of the final minute.
3. The waterfall's "Auto Zero" button had no function, and has been
removed.
4. In previous revisions the installer put a number of DLLs into
the Windows system directory, normally C:\Windows\System32. This
revision installs the DLLs to the WSJT-X installation directory.
November 29, 2012: v0.5, r2786
------------------------------
1. In r2783, the companion program jt9.exe (started automatically when
you start WSJT-X) was a CPU hog for no good reason. This was an
oversight on my part, and the bug has been corrected.
2. The program should now run correctly if installed in a directory
whose name contains embedded spaces. (Under Vista and Win7, however,
it's still not a good idea to install WSJT-X into C:\Program Files,
because of restricted write permissions there.)
3. In r2783 and earlier, stopping a transmission by toggling to "Auto
OFF" would terminate Tx audio and release PTT almost simultaneously,
possibly hot-switching your T/R relay(s). This has been corrected so
that proper sequencing takes place.
November 28, 2012: v0.5, r2783
------------------------------
This revision has an unusually large number of changes relative to the
previous release, v0.4 r2746. These changes include:
1. PTT control via COM ports COM10 and higher is enabled.
2. Improved decoder performance: higher speed as well as better
chances of success. Moderate amounts of frequency drift are detected
and compensated. Computed S/N values are more reliable. Time offsets
from -2.5 to +5 s are now supported, which makes JT9 usable for EME.
(EME tests on 144 MHz have been successful, and performance on that
propagation mode appears to be good.)
3. Tx Frequency now tracks the selected QSO Frequency (unless you hold
down the CTRL key when setting QSO Frequency via mouse-clicks or the
F11/F12 keys).
4. Decoded text containing "CQ " is highlighted with green background;
text including "MyCall" is highlighted in red.
5. In previous versions, signal reports were required to be in the
range -30 to -01 dB. In v0.5 r2782 the range has been extended to -50
to +49 dB. There is backward compatibility for the range -30 to -01,
but reports in the range -50 to -31 and 0 to +49 will NOT be decoded
correctly by previous program versions. It is important to upgrade!
6. Items "Save Synced" and "Save Decoded" are now implemented.
7. UTC Date, JT9 submode, and a parameter related to the decoding
procedure are now included in file wsjtx_rx.log.
8. Editing of Tx messages (in any of the six Tx message boxes) is
complete when you hit "Tab" or "Return". The message is then parsed
and converted to the form in which it will be displayed if decoding is
successful. Free-text messages are trimmed to 13 characters and
highlighted with a pink background.
9. The most recent transmitted message is displayed in the right-most
label on the status bar. This can be useful if you have lost track of
where you were in a QSO.
10. By default, the program now starts with Monitor ON. An option on
the Setup menu allows you to select "Monitor OFF at startup".
11. Better scaling is provided for the red "JT9 Sync" curve. Note
that JT9 signals in the active sub-mode should appear in this plot as
a bump of width equal to the total signal bandwidth, with a narrow and
slightly higher bump at the left edge. The narrow bump is the
frequency of the Sync tone, which is defined as the nominal frequency
of the JT9 signal.
12. Basic QSO information is now written to file wsjt.log when you
click the "Log QSO" button.
13. The WSJT-X User's Guide has been updated.
14. Other known bugs have been fixed. There will probably be new
ones! When you find one, or if you know of any old ones that have NOT
been fixed, please send me email.
Summary of Present Status
----------------------------------------------------------------------
I believe that WSJT-X is now a stable and very usable program. Many
thousands of QSOs have been made with JT9-1, mostly at HF -- I have
made nearly 100, myself. Also a number of QSOs have also been
completed at MF, and successful tests have been made on 2m EME, etc.
A number of QSOs have also been made with JT9-2.
As far as I know the slower modes (JT9-5, JT9-10, and JT9-30) also
work correctly. (Certainly they do in my laboratory test setup.)
Most people will find these modes too slow for "everyday" use, and
they require high frequency stability. It remains to be seen whether
they will be widely used.
An alternative approach to obtaining improved sensitivity would be to
give the decoder an ability to average over several successive
transmissions. For example, the average of five JT9-1 transmissions
could reach a decoding threshold around -32 dB, only 2 dB worse than a
single JT9-5 transmission. Because of QSB, the shorter transmissions
may actually succeed in less total time. Stability requirements would
be those of JT9-1, much less stringent than those of JT9-5.
Program development is not finished, by any means. I will be grateful
for your feedback on performance issues, as well as your "wish-list"
of features to be added. As always, example recordings of files that
you think should have decoded, but did not, will be much appreciated.
November 16, 2012: v0.4, r2746
------------------------------
Changes from v0.4 r2731 include the following:
1. Valid signal reports are now generated by double-clicking on a
callsign in the decoded text window.
2. Consecutive spaces in a Tx message are now collapsed into a single
space.
3. Decoding speed is much improved, especially when strong (possibly
non-JT9) signals are present and "Tol" is set to a relatively large
value.
4. Scaling of the "JT9 Sync" plot (red curve) is more reasonable.
5. Layout of widgets on the main window has been improved.
6. Several minor bug fixes.
November 14, 2012: v0.4, r2731
------------------------------
A number of known bugs have been fixed, and the JT9 decoder is
significantly improved. Among other improvements, the program is now
much less fussy about timing issues.
November 6, 2012: v0.3, r2717
------------------------------
Changes from r2713 include the following:
1. A bug in the decoder that led to erratic behavior (failed decodes)
under certain conditions has been corrected. Decoding is now much
more reliable.
2. A valid algorithm is now used to calculate S/N values for received
JT9 signals.
3. The header format of recorded *.wav files has been corrected.
These files will now play correctly in Windows programs that expect
the standard header.
November 6, 2012: v0.2, r2713
------------------------------
Changes from r2711 include the following:
1. Updates to the Quick-Start User's Guide,
http://www.physics.princeton.edu/pulsar/K1JT/WSJT-X_Users_Guide.pdf
2. Double-click on waterfall now sets Tol to a reduced
(mode-dependent) value.
3. Tol is saved and restored on program restart.
4. A "digital gain" slider was added next to the green-bar audio level
indicator. With the slider at mid-range, the scale reads correctly in
dB above the least significant bit of 16-bit audio data.
5. There is now a test that rejects at least one type of data that is
sufficiently corrupt to cause Eddie's best friend, the message
"15P6715P67WCV".
6. Several minor tweaks to improve decoder performance.
7. The program now starts with Monitor OFF. You must click Monitor to
start accepting audio. For some types of testing, this may be an
advantage. This startup condition may be changed again, in the
future.
October 31, 2012: v0.2, r2711
-----------------------------
Three significant changes since r2706:
1. Three options are now provided on the "Decode" menu, controlling
the "depth" of the decoding process. For most purposes I suggest you
should use "Normal", but feel free to experiment with the others.
2. Decoding of multiple signals in one Rx interval has been improved.
3. Handling of strong signals has been improved.
October 309, 2012: v0.2, r2706
------------------------------
Changes since r2702 include the following:
1. The problem with "ghost" signals is fixed.
2. A problem causing very long decode times under certain
circumstances has been fixed. Please note: decode times on any recent
PC should no more than a few seconds!
3. I have re-directed the program's fatal error messages so they will
be sent to the command-prompt window from which you started the
program. Please send me full reports on any such messages you observe,
preferably with details on how to reproduce the problem.
#########################################################################
Some additional information ...
1. Yes, the JT9 modes require good stability in all system
oscillators. The present JT9 bdecoder does not attempt to track
frequency drifts. Such capability will be added, however. We have
been using digital modes for EME for nearly ten years now, at 144 MHz
and higher. There are more than 1000 WSJT users on EME, using all
kinds of rige. We have learned how to deal with reasonable rates of
drift. Surely if we can do these things at VHF, we can do them much
more easily at MF and LF.
2. If you're sure that you have seen degraded JT9 performance because
of frequency stability issues, don't just complain on the LF
reflector. Document your case and send me an example file with a
drifting JT9 signal. Making WSJT-X and JT9 better is partly YOUR
responsibility!
3. In other ways as well, test files are needed. I can make many
tests myself, but I can't foresee all the problems others will have.
That's what the "Save All" function is for! In these early tests,
always run with "Save All" checked, just in case you will want to
refer back to something that happened. You may want to send me the
file in question. You can always clean out your "Save" directory by
using "File | Delete all *.wav files in SaveDir". I need good
examples of signals that fail to decode for any unknown reason. Also
some good examples of atmospheric or other impulsive noise, for
testing the noise blanker.
4. I have added a page of "Hints for New Users" to the online WSJT-X
User's Guide,
http://www.physics.princeton.edu/pulsar/K1JT/WSJT-X_Users_Guide.pdf .
Please read it! ... and let me know if you find other operational
details of WSJT-X that need explanation. This will likely be
especially true for those not already familiar with older versions of
WSJT.
5. An operational suggestion: In many ways the different JT9 submodes
are treated as distinct modes. If you receive a JT9-x signal in a
different submode than the one you have selected, you won't decode
it. For this reason, if JT9 is to become popular we'll probably need
to choose one or two of the submodes for general use, and perhaps
assign a narrow slice of the band to each one. Note that "message
averaging" in the Rx software can make two or three JT9-2
transmissions as good as one JT9-5 transmission, with the advantage
that you will copy sooner if signals are better than required for
JT9-5. Message averaging is not yet present in the JT9 decoder... but
in future it can be. Again, we have dealt with such issues very
effectively on EME -- and can do so at MF/LF, for sure.
6. On the topic of CW, Beacons, WSPR, JT9, etc. I really don't
understand what all the fuss is about. Surely there is room for
everybody? Maybe I'm just too new here to understand? (Mal, is this
mostly just a matter of "Mal being Mal"???)
On the HF bands, the WSPR sub-band is just 200 Hz wide. If we did the
same on 630 m, the WSPR sub-band would take up less than 3% of the 7
kHz band. If that's too much, we could cut it in half, or even less,
and still have enough WSPR space. Moreover, a "slow WSPR", if
warranted, would require even less bandwidth. Similar comments apply
to JT9. The bandwidth of JT9 signals is significantly less than that
of CW, for comparable information rates. There should be enough
spectrum for both, even in our narrow MF and LF bands.
7. As for performance comparisons between JT9 and WSPR: WSPR is a
mature program, and its decoder has been optimized and tweaked over a
period approaching five years. You are playing with JT9 in infancy.
With help (as opposed to simple complaints) from users, it will
improve rapidly.
October 29, 2012: v0.2, r2702
-----------------------------
Changes since version 0.1, r2696 include the following:
1. Sample rate for audio output has been changed from 12000 to 48000
Hz. Tx audio may now be generated at any frequency from 500 to 20000
Hz.
2. The Decoder now tries to decode all synchronizable signals in the
"green zone", that is, within "Tol" Hz of the selected QSO
frequency. (Before, by default it decoded only the signal producing
the highest "sync" value. Other signals could be decoded by manually
setting the QSO frequency and reducing Tol as needed.)
3. The user's selected QSO Frequency is now saved and restored on
program restart.
4. The problem with re-initialization after changing sub-modes has
been fixed.
5. The problem (for some users) of not releasing PTT after end of a
transmission has been fixed.
6. The program now writes a log of all decodes to a file wsjtx_rx.log
in the wsjtx directory.
October 25, 2012: v0.1, r2695
-----------------------------
Initial version of WSJT-X (experimental WSJT) released for testing.
.svn/pristine/f6/f6bfb3bc33106bc58382b150ad91b1c4e6663b33.svn-base
-3123
View File
File diff suppressed because it is too large Load Diff
.svn/pristine/f6/f6cc5b8f55d08f451ed8cb903795022224ae1b5b.svn-base
-180
View File
@@ -1,180 +0,0 @@
subroutine osd174(llr,norder,decoded,cw,nhardmin,dmin)
!
! An ordered-statistics decoder for the (174,87) code.
!
include "ldpc_174_87_params.f90"
integer*1 gen(K,N)
integer*1 genmrb(K,N),g2(N,K)
integer*1 temp(K),m0(K),me(K),mi(K)
integer indices(N),nxor(N)
integer*1 cw(N),ce(N),c0(N),hdec(N)
integer*1 decoded(K)
integer indx(N)
real llr(N),rx(N),absrx(N)
logical first
data first/.true./
save first,gen
if( first ) then ! fill the generator matrix
gen=0
do i=1,M
do j=1,22
read(g(i)(j:j),"(Z1)") istr
do jj=1, 4
irow=(j-1)*4+jj
if( btest(istr,4-jj) ) gen(irow,i)=1
enddo
enddo
enddo
do irow=1,K
gen(irow,M+irow)=1
enddo
first=.false.
endif
! re-order received vector to place systematic msg bits at the end
rx=llr(colorder+1)
! hard decode the received word
hdec=0
where(rx .ge. 0) hdec=1
! use magnitude of received symbols as a measure of reliability.
absrx=abs(rx)
call indexx(absrx,N,indx)
! re-order the columns of the generator matrix in order of decreasing reliability.
do i=1,N
genmrb(1:K,i)=gen(1:K,indx(N+1-i))
indices(i)=indx(N+1-i)
enddo
! do gaussian elimination to create a generator matrix with the most reliable
! received bits in positions 1:K in order of decreasing reliability (more or less).
! reliability will not be strictly decreasing because column re-ordering is needed
! to put the generator matrix in systematic form. the "indices" array tracks
! column permutations caused by reliability sorting and gaussian elimination.
do id=1,K ! diagonal element indices
do icol=id,K+20 ! The 20 is ad hoc - beware
iflag=0
if( genmrb(id,icol) .eq. 1 ) then
iflag=1
if( icol .ne. id ) then ! reorder column
temp(1:K)=genmrb(1:K,id)
genmrb(1:K,id)=genmrb(1:K,icol)
genmrb(1:K,icol)=temp(1:K)
itmp=indices(id)
indices(id)=indices(icol)
indices(icol)=itmp
endif
do ii=1,K
if( ii .ne. id .and. genmrb(ii,id) .eq. 1 ) then
genmrb(ii,1:N)=mod(genmrb(ii,1:N)+genmrb(id,1:N),2)
endif
enddo
exit
endif
enddo
enddo
g2=transpose(genmrb)
! The hard decisions for the K MRB bits define the order 0 message, m0.
! Encode m0 using the modified generator matrix to find the "order 0" codeword.
! Flip various combinations of bits in m0 and re-encode to generate a list of
! codewords. Test all such codewords against the received word to decide which
! codeword is most likely to be correct.
hdec=hdec(indices) ! hard decisions from received symbols
m0=hdec(1:K) ! zero'th order message
absrx=absrx(indices)
rx=rx(indices)
s1=sum(absrx(1:K))
s2=sum(absrx(K+1:N))
xlam=7.0 ! larger values reject more error patterns
rho=s1/(s1+xlam*s2)
call mrbencode(m0,c0,g2,N,K)
nxor=ieor(c0,hdec)
nhardmin=sum(nxor)
dmin=sum(nxor*absrx)
thresh=rho*dmin
cw=c0
nt=0
nrejected=0
do iorder=1,norder
mi(1:K-iorder)=0
mi(K-iorder+1:K)=1
iflag=0
do while(iflag .ge. 0 )
dpat=sum(mi*absrx(1:K))
nt=nt+1
if( dpat .lt. thresh ) then ! reject unlikely error patterns
me=ieor(m0,mi)
call mrbencode(me,ce,g2,N,K)
nxor=ieor(ce,hdec)
dd=sum(nxor*absrx)
if( dd .lt. dmin ) then
dmin=dd
cw=ce
nhardmin=sum(nxor)
thresh=rho*dmin
endif
else
nrejected=nrejected+1
endif
! get the next test error pattern, iflag will go negative
! when the last pattern with weight iorder has been generated
call nextpat(mi,k,iorder,iflag)
enddo
enddo
!write(*,*) 'nhardmin ',nhardmin
!write(*,*) 'total patterns ',nt,' number rejected ',nrejected
! re-order the codeword to place message bits at the end
cw(indices)=cw
hdec(indices)=hdec
decoded=cw(M+1:N)
cw(colorder+1)=cw ! put the codeword back into received-word order
return
end subroutine osd174
subroutine mrbencode(me,codeword,g2,N,K)
integer*1 me(K),codeword(N),g2(N,K)
! fast encoding for low-weight test patterns
codeword=0
do i=1,K
if( me(i) .eq. 1 ) then
codeword=ieor(codeword,g2(1:N,i))
endif
enddo
return
end subroutine mrbencode
subroutine nextpat(mi,k,iorder,iflag)
integer*1 mi(k),ms(k)
! generate the next test error pattern
ind=-1
do i=1,k-1
if( mi(i).eq.0 .and. mi(i+1).eq.1) ind=i
enddo
if( ind .lt. 0 ) then ! no more patterns of this order
iflag=ind
return
endif
ms=0
ms(1:ind-1)=mi(1:ind-1)
ms(ind)=1
ms(ind+1)=0
if( ind+1 .lt. k ) then
nz=iorder-sum(ms)
ms(k-nz+1:k)=1
endif
mi=ms
iflag=ind
return
end subroutine nextpat
.svn/pristine/f6/f6e38c6484c3ea8d7af87ac70c54562deaa6d755.svn-base
+160
View File
@@ -0,0 +1,160 @@
subroutine decode65a(dd,npts,newdat,nqd,f0,nflip,mode65,ntrials, &
naggressive,ndepth,ntol,mycall,hiscall,hisgrid,nQSOProgress, &
ljt65apon, nexp_decode, &
bVHF,sync2,a,dt,nft,nspecial,qual,nhist,nsmo,decoded)
! Apply AFC corrections to a candidate JT65 signal, then decode it.
use jt65_mod
use timer_module, only: timer
parameter (NMAX=60*12000) !Samples per 60 s
real*4 dd(NMAX) !92 MB: raw data from Linrad timf2
complex cx(NMAX/8) !Data at 1378.125 sps
complex cx1(NMAX/8) !Data at 1378.125 sps, offset by 355.3 Hz
complex c5x(NMAX/32) !Data at 344.53125 Hz
complex c5a(512)
real s2(66,126)
real a(5)
logical bVHF,first,ljt65apon
character decoded*22,decoded_best*22
character mycall*12,hiscall*12,hisgrid*6
character*27 cr
data first/.true./,jjjmin/1000/,jjjmax/-1000/,cr/'(C) 2016, Joe Taylor - K1JT'/
save
! Mix sync tone to baseband, low-pass filter, downsample to 1378.125 Hz
call timer('filbig ',0)
call filbig(dd,npts,f0,newdat,cx,n5,sq0)
if(mode65.eq.4) call filbig(dd,npts,f0+355.297852,newdat,cx1,n5,sq0)
call timer('filbig ',1)
! NB: cx has sample rate 12000*77125/672000 = 1378.125 Hz
! Check for a shorthand message
if(bVHF .and. mode65.ne.101) then
call sh65(cx,n5,mode65,ntol,xdf,nspecial,sync2)
if(nspecial.gt.0) then
a=0.
a(1)=xdf
nflip=0
endif
endif
if(nflip.eq.0) go to 900
! Find best DF, drift, curvature, and DT. Start by downsampling to 344.53125 Hz
call timer('fil6521 ',0)
call fil6521(cx,n5,c5x,n6)
call timer('fil6521 ',1)
fsample=1378.125/4.
call timer('afc65b ',0)
! Best fit for DF, drift, and dt. fsample = 344.53125 S/s
dtbest=dt
call afc65b(c5x,n6,fsample,nflip,mode65,a,ccfbest,dtbest)
call timer('afc65b ',1)
dtbest=dtbest+0.003628 !Remove decimation filter and coh. integrator delay
dt=dtbest !Return new, improved estimate of dt
sync2=3.7e-4*ccfbest/sq0 !Constant is empirical
if(mode65.eq.4) cx=cx1
! Apply AFC corrections to the time-domain signal
! Now we are back to using the 1378.125 Hz sample rate, enough to
! accommodate the full JT65C bandwidth.
a(3)=0
call timer('twkfreq ',0)
call twkfreq65(cx,n5,a)
call timer('twkfreq ',1)
! Compute spectrum for each symbol.
nsym=126
nfft=512
df=1378.125/nfft
j=int(dtbest*1378.125)
call timer('sh_ffts ',0)
c5a=cmplx(0.0,0.0)
do k=1,nsym
do i=1,nfft
j=j+1
if(j.ge.1 .and. j.le.NMAX/8) then
c5a(i)=cx(j)
else
c5a(i)=0.
endif
enddo
call four2a(c5a,nfft,1,1,1)
do i=1,512
jj=i
if(i.gt.256) jj=i-512
s1(jj,k)=real(c5a(i))**2 + aimag(c5a(i))**2
enddo
enddo
call timer('sh_ffts ',1)
call timer('dec65b ',0)
qualbest=0.
qual0=-1.e30
minsmo=0
maxsmo=0
if(mode65.ge.2 .and. mode65.ne.101) then
minsmo=nint(width/df)-1
maxsmo=2*nint(width/df)
endif
nn=0
do ismo=minsmo,maxsmo
if(ismo.gt.0) then
do j=1,126
call smo121(s1(-255,j),512)
if(j.eq.1) nn=nn+1
if(nn.ge.4) then
call smo121(s1(-255,j),512)
if(j.eq.1) nn=nn+1
endif
enddo
endif
do i=1,66
jj=i
if(mode65.eq.2) jj=2*i-1
if(mode65.eq.4) then
ff=4*(i-1)*df - 355.297852
jj=nint(ff/df)+1
endif
s2(i,1:126)=s1(jj,1:126)
enddo
nadd=ismo !### ??? ###
call decode65b(s2,nflip,nadd,mode65,ntrials,naggressive,ndepth, &
mycall,hiscall,hisgrid,nQSOProgress,ljt65apon,nexp_decode, &
nqd,nft,qual, &
nhist,decoded)
if(nft.eq.1) then
nsmo=ismo
param(9)=nsmo
nsum=1
exit
else if(nft.eq.2) then
if(qual.gt.qualbest) then
decoded_best=decoded
qualbest=qual
nnbest=nn
nsmobest=ismo
endif
endif
if(qual.lt.qual0) exit
qual0=qual
enddo
if(nft.eq.2) then
decoded=decoded_best
qual=qualbest
nsmo=nsmobest
param(9)=nsmo
nn=nnbest
endif
call timer('dec65b ',1)
900 return
end subroutine decode65a