This is the WSJT Home Page by K1JT

You are at ../updates.htm

Last update January 26. 2006

Go back to main page

Current full version: WSJT 5.9.3

Changes in WSJT 5.9.3: January 26, 2006
This is a minor maintenance release.

1. New Help screen provides a list of available suffixes and add-on DXCC prefixes.

2. The occasional error message about "images do not match" has been trapped and should no longer occur.

3. It is no longer necessary to enter one's own grid locator with the first two letters in upper case and the last two in lower case.

4. Initialization of the PTT line to the "OFF" state has been fixed.

Changes in WSJT 5.9.2: January 16, 2006


1. Thread priorities have been adjusted for smoother operation.

2. The JT65 decoder has been given improved immunity to garbage data
(birdies, QRM, etc). It exhibits better performance on strong
signals and provides more accurate estimates of their S/N.

3. The FSK441 decoder produces less on-screen gibberish when
processing mouse-picked decodes.

4. The JT6M decoder now makes better use of Freeze and Tol. You can
set the value of "Freeze DF" by using the Right/Left arrow keys.
(This feature is also useful in JT65 mode.)

5. On-screen font sizes can be set by using Windows Notepad to edit
the file If your screen has resolution greater than
1024 x 768, or if you have old eyes like mine, you may want to
increase the font sizes from 8 and 9 points (first three lines of
the file) to, say, 9 and 10 points.

6. A simulator mode is now built into WSJT for testing purposes. It
is presently most useful in JT65 mode. By entering, say, "#-22"
in the text box for Tx6, you signify that the program should
generate its Tx audio files with the signal embedded in white
gaussian noise, 22 dB below the noise power in a 2.5 kHz
bandwidth. You can direct this signal into a second computer
running WSJT, for example to test the decoder or to practice
operating in JT65 mode. You can even have the two computers "work
each other" in a simulated QSO, although changing messages of
course requires operator action.

7. Dividing lines are now provided on the waterfall display between
spectra corresponding to wave files read from disk.

8. The PTT line is explicitly set low on program startup.

9. The F10 key brings up the SpecJT screen (if it was hidden) and
toggles foreground and focus status between the WSJT and SpecJT

10. You can use the Alt-F and Alt-Z keyboard shortcuts to toggle
"Freeze" and "Zap" on and off.

11. "Accelerated decoding" has been removed from the Setup menu. In
effect, this option is now always ON.

12. Text windows are now cleared when switching between modes.

13. Linux and FreeBSD versions (see below) offer PTT control via
parallel port, as well as serial port. They offer sound support
via ALSA and OSS.

Bug Fixes:
1. The use of non-threadsafe code for FFTs could cause occasional and
unpredictable program crashes. Fixed.

2. A bug in the JT65 decoder could (rarely) cause large errors in the
reported level of strong signals. Fixed.

3. The program could be made to crash by trying to read a very short
wave file. Fixed.

4. "Save None" now works as it should.

Please note:
When entering your grid locator on the Setup->Options page, use upper
case for the first two letters and lower case for the last two. For
example, for K1JT the locator is FN20qi.

For Curious Users, and Especially for Programmers

WSJT is no longer a one-person effort, and the program no longer runs
only under Microsoft Windows. WSJT is now a full-fledged Open Source
project, with an active working group making contributions to future
development. Source code is now stored in a public repository under
control of a version control system called "Subversion," or SVN. You
can learn more at

The first significant result of the group effort has been to create
versions of WSJT that run under the Linux and FreeBSD operating
systems. Porting WSJT to the Macintosh platform should be
straightforward, but has not yet been done.

If you are interested in testing and using WSJT on your own Linux or
FreeBSD system, we'd like to hear from you. Please note that the
present Linux and FreeBSD versions are intended mainly for
programmers. You need to know your way around these operating
systems to be able to install them.

If you feel that you can usefully contribute to the future development
of WSJT on any platform, we would also like to hear from you! We
could use help with documentation and website maintenance, as well as
actual programming.

The present WSJT working group consists of:

Diane Bruce, VA3DB
James Courtier-Dutton
Bob McGwier, N4HY
Jonathan Naylor, ON/G4KLX
Stewart Nelson, KK7KA
Joe Taylor, K1JT
Kaj Wiik, OH6EH

Changes in WSJT 5.9

1. JT65 decoding has been made faster and significantly improved in other ways. Three new options appear on the Decode->JT65 menu: "Fast", "Normal", and "Exhaustive". The program is most sensitive if you choose "Exhaustive". Choosing "Normal" will make decoding slightly less sensitive, but the loss is not great, and decoding can be twice as fast. The "Fast" setting is faster still, but can be less sensitive by 2 dB or more in some cases. If you have a 1.5 GHz or faster computer, use "Exhaustive". With a slower computer you may want to experiment with the other settings.

2. In JT65 mode, double-clicking on the waterfall (SpecJT window) or on the red curve (main window) sets "Freeze DF" to the selected frequency, turns Freeze ON, sets Tol to 50 Hz, and invokes the decoder. Using this feature, you can quickly decode a transmission at several different values of DF. I find this feature to be *extremely* useful.

3. The range of DT values searched to establish synchronization has been doubled, now extending from -2 to +10 seconds. The reported values of DT are more accurate, as well. You should normally expect EME signals to have DT in the range 2 to 3 seconds, but the program will now synchronize properly even if DT is well outside this range.

4. WSJT now offers the ability to correct for errors in soundcard input and output sampling rates. Numbers displayed in the first panel of the status bar (at lower left of the main screen) give the ratio of actual sample rates for input and output to the correct value, 11025 Hz. The numbers should stabilize within about one minute after program startup. If they fall in a "safe" range between about 0.9990 and 1.0010, you have a good sound card (at least in respect to sampling frequency). You can then leave the entry fields "Rate In" and "Rate Out" on the "Setup -> Options" page at their default values, 1.0.

If your soundcard gives one or both numbers well outside the safe range, you should enter the displayed errant numbers as "Rate In" and/or "Rate Out" on the Setup->Options page. This needs to be done only once; subsequent changes in the last decimal place of the displayed values are not very significant, and can be safely ignored.

The result of this procedure is that your Tx signal will be "trimmed" so that your tone spacings in time and frequency are more nearly correct. In addition, your digitized Rx signals will be adjusted so that the software can properly interpret them.

This trimming is an important procedure. Some recent sound cards produce sampling error factors as low as 0.9932 or as high as 1.0068. If uncorrected, such results can degrade your S/N in WSJT modes by 2 dB or more.

If one of the measured sample rates differs from the corresponding value specified for "Rate In" or "Rate Out" by more than 0.1%, a red warning label will appear just below the graphical area on the main screen.

5. The graphical display of information obtained during JT65 decoding has been enhanced. As before, a red line illustrates the maximum correlation between the pseudo-random sync tone pattern and the received signal at each value of frequency offset, DF. A blue line shows the correlation at the best DF, plotted as a function of time offset, DT. If a shorthand message is detected, two new lines colored magenta and orange replace the red and blue lines. The new lines illustrate phase-resolved spectra measured in each of the two phases of the shorthand square-wave pattern. A properly detected shorthand message will show a peak in the magenta curve, followed at a specified distance by a peak in the orange curve. The correct
locations of the two peaks are marked by small yellow ticks. Unlike the alternating shorthand message tones, birdies will appear with approximately equal amplitudes in the magenta and orange curves.

6. For the convenience of temporary DXpeditions, a new JT65 feature permits use of add-on DXCC prefixes that are not in the published list of supported prefixes. Both stations in a QSO must enter the required prefix (for example, PJ8 or FS) in a box on the Setup->Options page. The effect will be to temporarily add the entry to the table of supported prefixes.

7. The Setup->Options page has new entry fields labeled "Source RA" and "Source DEC". You can enter the current right ascension and declination of a radio source to be used for system calibration, or perhaps a pulsar or a deep space probe that you wish to detect. The program will display (on the Astronomical Data screen) the current Azimuth and Elevation of the specified object at your station. The source Azimuth and Elevation are also written every second to the file azel.dat, in case you have automated tracking capabilities that depend on this information.

8. For contest-style operations, the Setup->Options menu has an item labeled "F4 sets Tx6". If this item is checked, when you hit F4 to clear the To Radio box the program will turn Freeze OFF and set the Tx message number to 6.

9. To facilitate the coming release of the full source code of WSJT under the GNU General Public License, the proprietary soft-decision Reed Solomon decoder has been removed from WSJT proper and made into a separate executable module, KVASD.EXE. This change is transparent to the user, and the full benefit of the soft-decision decoder is still available. An open source hard-decision decoder is also provided; it's what you get when you select the "Fast" JT65 decoding option.

10. In WSJT 5.8.6, if the value of "Freeze DF" (as displayed in the Status Bar) differs from the sync tone frequency by more than "Tol", shorthand decoding was suppressed even if Freeze was not checked. This is a bug, and it has been fixed.

11. Earlier versions of WSJT had a bug that could cause the "Zap" function to notch out a valid sync tone. Fixed.

12. The Help screens called up by F1 and Shift-F1 have been updated. Be sure to read these screens: they contain many operational conveniences that you may not have discovered!

13. At scrolling speed 5, the time labels and "minute separator" lines were displayed erratically and the CPU load was excessive. Fixed.

14. Signal strength measurements above -20 dB were formerly compressed and significantly underestimated. This has been fixed.

15. Decodings of the average of many properly synchronized transmissions would sometimes go from "good" to "bad" after approximately 8-12 transmissions. This was a bug, and it has been fixed.

16. Several bugs in the FSK441 decoder have been fixed. Both automatic decoding and mouse-picked decoding have been improved.

17. Changing WSJT modes now sets Auto to OFF, Tol to 400, and the Tx message number to 1.

18. The generated audio for CW ID in FSK441 and JT6M modes has been moved to 440 Hz, to avoid possible confusion with the other tones used in these modes.

19. Readout of "Rx noise" on the main screen is now highlighted in red if the level is outside the range -10 to +10 dB.

20. The Monitor button is no longer highlighted in green while you are transmitting.

21. No attempt is made to decode if the Rx level is very low -- for example, if your receiver is turned off.

22. If the Grid box does not contain a valid locator, readouts of azimuth and distance are suppressed.

23. Keying of the audio tone to produce Morse code has been softened to suppress key clicks.

24. Transmitted messages recorded in the file ALL.TXT are now identified as to mode, and shorthand transmissions are noted as such.

25. A number of other very minor bugs have been fixed.

Changes in WSJT 5.8.6

1. Audio input and output has been modified in a way that accommodates certain soundcards (e.g., SB Live!) that did not work correctly with WSJT 5.8.3.

2. New item on Setup->Options menu to select whether GenStdMsgs forces Tx message number to 1.

3. Status of all selectable items on Setup->Options menu is preserved on program restart.

4. If a CQ is transmitted in JT65 mode, the Sked box is automatically unckecked.

5. In v5.8.3, entering the same callsign and locator information in MyCall and ToRadio/Grid could cause the program to freeze.  Fixed.

6. If MyCall includes an extra prefix, as in 4X/ZL1RS, the standard JT65 messages should not include a grid locator.  Fixed.

7. The "ms" parameter has been removed from the Soundcard status readout at bottom left.  Separate sample-rate factors are now displayed for audio input and output, but only if "Enable diagnostics" is checked on the Setup menu.

8. The SpecJT screen may now be made invisible by clicking on "X" in the upper right corner.  To restore it to visible status, click on View->SpecJT on the main screen.

9. Decoded text lines in JT6M mode were sometimes too long, causing end-of-line wrap-around.  Fixed.

10. Some diagnostic messages printed to console window have been removed.

11. The values of S, Sync, Clip, Zap and NB are now preserved when th program is terminated and restarted.

12. Version 5.8.3 was unable to read back its own recorded wave files. Fixed.

13. A programming error in the JT65 shorthand message decoder has been fixed.  Under certain conditions, this error could cause false decodes of shorthand messages.

14. The logic of file saving commands Save Last, Save decoded, etc., has been corrected.

15. Wave files read from disk will now produce spectral plots on the waterfall display if Monitor is OFF.

16. The CW ID feature has been implemented.

17. The mapping of signal levels to pixel colors and its dependence on settings of Brightness and Contrast controls has been changed so as to improve sensitivity to very weak signals.

18. In v5.8.3, changing Dsec would create erroneous results for the displayed soundcard sample rate factor.  Fixed.

19. Running in JT65 mode with Dsec>0 caused transmission errors (including a gap in transmitted tones at t = 38-41 s), and the resulting transmission was unreadable.  Fixed.

20. Undesired resizing of main screen could occur when a long FSK441  message was transmitted.  Fixed.

21. The "yellow line" displayed in the graphical area in JT6M mode was computed incorrectly in version 5.8.3.  Fixed.

22. When running at speeds 1-5, the waterfall spectrum may optionally be "flattened" to remove rolloff at edges.  To enable this feature, check "Flatten spectra" on the SpecJT Options menu.


In general, "high end" sound cards offer no advantages when used with WSJT.  Motherboard AC-97 compliant sound systems are cheap and work well.  If you do need to buy a sound card for use with WSJT, my advice is to get a simple one.  You do NOT need 8-channel surround-sound, wavetable synthesis, special effects, etc.  Those features are for games and listening to music, and they will be wasted on WSJT.

Likewise, you do not need 24-bit A/D and D/A conversions. Specifications having to do with signal/noise ratio are quite
irrelevant to use with WSJT, as you should never be operating in a regime where A/D quantizing noise (or any other noise generated in the sound card) contributes significantly to the system S/N.

If you have a choice, get a card that offers a *native* sampling rate of 44100 or 11025 Hz, or both.  (Unfortunately, it is often very difficult to tell from the manufacturer's literature whether this capability is present or not.)  If native sampling at 11025 Hz is available -- or if the manufacturer has at least provided a well-designed resampling capability -- the soundcard sample-rate factors (displayed by WSJT in the bottom left corner, if "Setup->Enable diagnostics" is checked) should both be very close to 1.0000.

Changes in Version 5.8.3

I am aware that some users -- maybe 5 or 10 percent? -- have been unable to use version 5.8.1 because of erratic audio input levels.  I do not know whether this problem has been solved in 5.8.3, because I have been unable to reproduce the problem on any computer that I have access to. It appears to be a problem peculiar to certain sound cards or Windows configurations.

If you are among those who have experienced the Rx audio problem, please be sure to read the section headed "KNOWN PROBLEMS", a few screens down in the above link.

If you experienced the problem but found a solution, please post the solution to this reflector!

As always, I will be happy to receive your comments on the new release.

Note that I will probably not answer email until some time next week -- after I have recovered from the ARRL VHF QSO Party, this weekend.

With best wishes,
            -- 73, Joe, K1JT

Changes in Version 5.8.1

Beta release 5.8.1 is the first open release of an entirely new version of WSJT.  After a period of beta testing, the program will become known as WSJT 6.0.  I will refer to it here as WSJT6.

Many features of WSJT6 will be familiar to present users of WSJT4.x. However, the user interface and other real-time portions of the program have been entirely re-written, so you may find a few things that look different or work differently than before.

WSJT6 is a multi-threaded program.  This architecture permits much better timing control and much better sharing of the CPU among the program's many tasks.  A real-time waterfall is provided, optimized for each of the WSJT modes.  You can measure and set the Rx gain in real time.  You can decode FSK441 pings immediately after hearing them. 

The decoders for JT65 and JT6M are essentially the same as those in recent versions of the program.  (I have many plans for enhancements to the decoders, but these must wait for future updates.)  The FSK441 decoder has intentionally reverted back to approximately version 3.8.1, because it has been shown that recent versions did not perform quite as well.

WSJT 5.8 is a stable release.  Extensive tests by a small group of early beta testers have already shown that it works well.  However, it contains many thousands of lines of new code, and most likely that code has some bugs.  Nevertheless, I think you will find many operational advantages to using it.  Over the past 2.5 months, all of my own MS and EME QSOs have been made with the new version.

If you decide to participate in the beta testing period, please accept the responsibility of reporting any bugs that you find, as well as telling me of features you would still like to see.  I would appreciate hearing about new features that you like, as well.


1.  I suggest installing WSJT6 to a new directory such as C:\Program
    Files\WSJT6.  After installation you will probably want to copy
    your version of the callsign database, CALL3.TXT, into the new
    WSJT6 directory, replacing the one that is supplied.

2.  When you start the program you should get three windows on your
    screen. One has a black background and I will call it the "console
    window"; it is mainly used for debugging messages.  You can
    minimize it and generally ignore it.  You should, however, look
    there for messages if the program crashes (see also items 9 and
    11, below).  The other two windows are a Spectran-like waterfall
    and the more-or-less familiar WSJT window.  It is no longer
    necessary or desirable to run Spectran simultaneously with WSJT.

3.  FSK441, JT65, and JT6M are all present and functional. I have made
    many QSOs with FSK441 and JT65, so I know they are working well.
    JT6M has been tested somewhat less, but I have seen no problems as
    yet.  EME Echo mode and the CWID feature are not yet implemented.

4.  To start the real-time spectral display, click Monitor.  Normally
    you can leave Monitor on all the time.

5.  There are two ways to set the Rx Audio gain.  You can call up the
    Windows mixer as before, using the "Rx Volume Control" item on the
    Options menu.  There is also a digital gain control near the
    bottom right of the waterfall screen.  You should aim for around 0
    dB, as before -- but with 16-bit audio sampling it is no longer
    very critical.  Use the "S-meter" at the lower right of the
    waterfall display, or the familiar box labeled "Rx noise" at
    bottom center of the main screen.

6.  You should be able to make FSK441, JT65, and JT6M QSOs more or
    less as usual.  In FSK441 and JT6M you will want to run the
    spectral display at speed "H1" or "H2" (speed is selected at the
    top of the waterfall screen).  (The "H" means horizontal
    scrolling.)  Scrolling speeds "5" and "H2" use a lot of CPU time,
    so you may want to avoid them unless your computer is pretty fast.
    I generally use speed "1" or "2" for JT65 and "H1" for FSK441 and

7.  When running FSK441 in the horizontal scrolling mode, the spectral
    display shows current data in the top half and the previous Rx
    sequence in the bottom half.  The most recently decoded sequence
    is shown also on the main screen, as in version 4.x.

8.  You can decode FSK441 pings right away by clicking on any of the
    2-dimensional spectral displays, or the accompanying green lines.
    You can click on the top half, the bottom half, or in the main
    screen's graphical area.

9.  If you have more than one sound card, you can select the desired
    one.  Look at the startup messages in the console window. You
    should see a list of the available Audio devices and information
    about which one has been selected.  If you wish to change the
    selection, enter the desired device numbers on the Setup->Options
    screen, then terminate and restart the program.

10. If your display has resolution 1024 x 768 or less, you may prefer
    to resize the waterfall window so that only its top portion
    remains visible.  The two main WSJT windows may then be kept
    visible without overlapping.

11. Some program crashes can kill the console window so that you can't
    read the error messages.  If this happens, and if the crash is
    repeatable, open a Windows Command-Prompt window; CD to your WSJT6
    installation directory, and start the program from there by typing
    "WSJT6".  With this startup procedure, any subsequent post-mortems
    will remain visible.  Please report any such messages to me!

12. You may find decoding to be slightly slower than with v4.9.x.  I
    have not yet spent any time optimizing the new code for speed; it
    will get faster when I turn attention to that task.  If you have
    an older computer you may wish to check the menu item
    "Setup->Accelerated decoding", which will suspend updating the
    waterfall during the decoding process.

13. Be sure to look at the screens called up by function keys F1 and
    Shift-F1.  These screens are also available from the Help menu.
    They list some useful keyboard and mouse commands that you might
    otherwise overlook.

14. Be sure to explore all the menus and on-screen controls, and try
    out the commands listed on the help screens.  Until I can find
    time to write a new manual, this is the best way to learn about
    some of the new features.

15. In JT65 mode, a horizontal green line on the frequency scale shows
    the range of frequencies that will be searched for a sync tone.
    You can set the "Freeze DF" value by clicking on the main screen's
    red curve (as in WSJT 4.x) or by clicking on the waterfall with
    the shift key held down.  If "Freeze" is checked, vertical green
    ticks will mark the selected sync-tone frequency and the
    corresponding frequency of the highest data tone.  Red tick marks
    denote the frequencies of the RO, RRR, and 73 shorthand messages.

16. WSJT6 can read and process WAV files produced by earlier versions.
    The converse is not true, however, because earlier versions of
    WSJT are not equipped to read the 16-bit data files produced by

17. When you click "Log QSO," a line with date, time, HisCall,
    HisGrid, frequency, and mode is added to the file WSJT.LOG in the
    installation directory.

18. Every second, a short file named "c:\azel.dat" is updated with
    time, moon and sun coordinates, frequency, doppler, and doppler
    rate information.  This file could be used by other software to
    make your antenna track or your radio follow doppler changes.

19. In the lower left corner of the main screen you will see a message
    of the form "Soundcard: x.xxxx", where x.xxxx is a number close to
    1.0000.  This number is the ratio of the soundcard's measured
    sampling frequency to the nominal value, 11025 Hz.  The displayed
    value should stabilize after the program has been running for a
    minute or so.  If you see values less than about 0.9990 or greater
    than 1.0010, please let me know about it and tell me what kind of
    computer and sound card you are using. 

20. Callsigns for Swaziland (prefix 3DA0) can now be used in standard
    JT65 messages, and they will provide the full "deep search"

21. The box labeled "NB" enables a software noise blanker.  If your
    receiver already has a good noise blanker, this may be of little
    use; if it does not, you may find this one better than nothing.
    It can be helpful when short, impulsive noise spikes are present.

22. The "QRN" parameter of older WSJT versions has been combined with
    the "Clip" parameter.  In FSK441 mode, Clip=0 corresponds to the
    old QRN=5.  If you want more FSK441 immunity to summertime QRN,
    increase Clip above 0 just as you would have increased QRN above

23. The "B" and "C" submodes of FSK441 have not been implemented.  As
    far as I could tell, they were little used.

Let me call your attention to the online WSJT Forums hosted by DK5YA at  If you provide information and post questions about WSJT 5.8 there, it will help others as well as yourself -- and may help to reduce the load on my email inbox. 

I do, of course, want to hear from you directly if you have found a problem with the new program version or suggestions for its improvement.

With best wishes,

                -- 73, Joe, K1JT

Changes in Version 4.9.8
Version 4.9.8 is a minor maintenance release. The changes are:

1. Bug in JT65 decoder could (rarely) cause a correct decoding by the Reed-Solomon decoder to be "overruled" by an incorrect decoding from the Deep Search decoder. Fixed.

2. Bug in JT65 Reed-Solomon decoder could cause program to crash under certain rare conditions. Fixed.

3. The standard "CQ" message generated for a callsign having an extra prefix or suffix should not include a grid locator. Fixed.

4. Automatically generated CW messages for stations with callsigns longer than 4 characters were improperly truncated. Fixed.

5. Bug in JT65 decoder could occasionally cause one or two extra letters to be appended to a correctly decoded callsign. Fixed.

6. Grid locator entered in "Report" box in FSK441 mode (as used by some in North American meteor scatter contest) would cause a crash on program restart. Fixed.

Changes in Version 4.9.7
The latest update is WSJT Version 4.9.7. Like Version 4.7.0, it is able to run simultaneously with the latest version of Spectran (by I2PHD and IK2CZL), which is included with the 4.7.0 distribution. This combination of programs provides an excellent real-time spectral display in addition to all the features of WSJT.

Changes in Version 4.9.6

1. WSJT 4.9.5 fails to decode some files that have relatively high S/N and good Sync level. This is a bug, and has been fixed. For this reason alone, you should definitely upgrade to Version 4.9.6.

2. New optional message formats are provided for conveying and responding to signal reports. It has been permissible for some time to send, for example,
VK7MO K1JT -24
(The number after the minus sign must have two digits and must be in the range -01 to -30.)

I plan to implement a quick way of copying the measured strength of a decoded transmission into TX message #2, when desired. This is not yet done in v4.9.6, however. You must edit the TX messages by hand if you use these formats.

Enhancements in version 4.9.6 now allow you to send messages like the ones listed below. Both stations will need to be running v4.9.6 in order for these to work:

3. Decoding by the deep search algorithm has been extended so as to include messages of the types discussed in item 2.

Changes in Version 4.9.5

1. Full support for long callsigns like ZA/PA2CHR and G4ABC/P is now provided. When using such a callsign prefix or suffix, do not include a grid locator in your transmitted message.

2. If you double-click on a callsign in the decoded text window, and if the word preceding the callsign is "CQ", then TX message #1 will be selected after the messages are updated. Otherwise, TX message #2 will be selected.

3. So that you will be aware of what is happening, the background color in the TX message box turns red whenever a message you have entered is "non-standard" and will be sent as 13 characters of plain text.

4. Items related to decoding have been removed from the Setup |Options screen and replaced by a new menu labeled Decoding. Here you may now select "No shorthands" for FSK441 and several options for JT65 decoding.

5. In case you are upgrading directly from WSJT v4.7.0 to v4.9.5, a callsign database file has been included as CALL3A.TXT. If you do not already have a file CALL3.TXT, you should rename the supplied file to CALL3.TXT. Otherwise, you will probably want to ignore the supplied file, since you will have made additions to your own copy.

6. The frequency of program crashes (for example, after a long period of monitoring) is much reduced, possibly to zero.

Please, if you encounter a received wave file that reproducibly cause WSJT to crash, send it to me.

Changes in Version 4.9.2

1. In CW mode you can now set the desired T/R period by using the text box provided. This feature did not work properly in v4.9.1.

2. Under some conditions using the "Add" button to edit information in the file CALL3.TXT would cause a program crash with the message "Run-time Error #53". Fixed.

3. Using the double-mouse-click on a callsign in the decoded text window will now set the active Tx Message to Tx2. I believe this will be most commonly what is desired, and will be an added convenience for random JT65 operation.

4. The "Sked" box remained visible on the EME Echo screen, covering part of the RIT box. Fixed.

5. Two numbers are now made available if you have checked the"Aggressive decoding" option. These numbers appeared without explanation at the end of each decoded text line produced by version 4.9.1, and you may have wondered what they were. The first number is 0 or 1 according to whether the soft-decision Reed Solomon decoder has failed or succeeded. The second number represents a confidence level on a 0-10 scale for messages decoded using the "deep search" algorithm. Anything under 3 is questionable; messages rated 6 and above are unlikely to be wrong, unless you are processing "garbage" data containing strong birdies, QRN, etc. In that case, you are on your own.

6. A bug was introduced when implementing the "Aggressive decoding" check box. This bug caused a stray "OOO" flag to be sometimes displayed even when no signal was present and synchronization had not been achieved. Fixed.

Changes in Version 4.9.1

1. A programming error in version 4.9.0 prevented the "deep search" portion of the JT65 decoder from detecting some messages that include the "OOO" signal report. The bug has been fixed, and consequently the extra 4 dB of sensitivity will become available for those messages.

2. The callsign database, a file named CALL2.TXT in version 4.9.0, has been converted to a comma-delimited format and is now named CALL3.TXT. As has been true in the past, you should maintain your own copy of this file according to your own needs. New calls may be added to the file using the "Add" button of WSJT, and you can edit the file directly with the Windows NotePad program. I apologize for the fact that if you have already edited CALL2.TXT extensively, you will need to do so again. The good news is that the programs WSJT, MoonSked (by GM4JJJ) and Tracker (by W7GJ) will now use the same database file, CALL3.TXT.

3. Two new JT65 check boxes have been made available to the user: one labeled "Sked", located on the main screen, and one labeled Agressive decoding" on the Setup | Options screen. Check "Sked" to signify that you are trying to work a known station; the deep search decoder will then look only for your own call and the one displayed in the "To Radio" box. Check "Agressive search" if you want to see all messages found by the deep-search decoder, even if the confidence level is moderately low. Leave this box unchecked if you prefer to see only decoded messages that have been assigned a relatively high confidence.

4. The duration of T/R sequences in CW mode defaults to 60 s if the Band is 50 MHz, 150 s if 432 MHz, and 120 s otherwise. However, an on-screen box now allows you to override the defaults and set any

Changes in Version 4.9.0

On the outside WSJT version 4.9.0 looks nearly the same as its recent predecessors. A "CW" entry now appears on the Mode menu. This is presently a "transmit only" mode: it sends standard EME-style messages at 15 WPM, by keying a 1500 Hz audio tone, and it takes care of the timing and T/R switching for you. Receiving is left up to you, the operator. For me, this combination makes CW EME QSOs relaxing and enjoyable. Presently the program uses 2.5 minute sequences if you are on 432 MHz, 2 minutes on 144 MHz, and 1 minute on 50 MHz. (If you want to run with 1-minute sequences on 144, set the band indicator to 50 MHz.)

Double-clicking on a callsign in either one of the decoded text windows will cause that callsign to be copied into the "To Radio" box. The call will then be looked up in the database and will be inserted appropriately into the transmit message boxes Tx1 and Tx2. This feature is designed to facilitate random JT65 operation by making it easy to call a station you have just copied calling CQ, or responding to your CQ.

The most significant program enhancements are those made to the JT65 decoder. It has been transformed into a multi-layered procedure that takes better advantage of the structured nature of JT65 messages and the substantial computing capability that most WSJT users have in their hamshacks.

In version 4.9.0, if the initial JT65 decoding effort fails then deeper searches are attempted using an entirely different approach. The result is a net gain of about 4 dB over a wide range of circumstances. My JT65 digital simulator, which has accurately predicted the performance of previous versions of the software, correctly decodes about 50% of simulated Rx files with the v4.7.0 decoder at a signal level of -24 dB. With the v4.9.0 decoder, it correctly decodes more than half of the simulated data files at -28 dB. This very substantial improvement means that JT65's message-averaging facility will be needed much less frequently than with earlier versions of WSJT. Most of the time, if the transmission synchronizes properly, it will also decode properly.

You will get better performance from the new JT65 decoder if you understand a few things about how it works. The following is a very brief description; more complete technical details will be forthcoming then I find time to write it all down.

JT65 is capable of transmitting and receiving 2^72 (about 5 x 10^21) distinct user messages. Instead of sending the minimum number of 72 information bits needed to to convey any one of those distinct messages, the program actually sends 63 six-bit "symbols" for a total of 378 bits in each transmission. The 302 extra bits comprise the powerful forward error correction (FEC) capability of the JT65 mode, allowing the system to function reliably with signals far below the audible threshold.

One of the first tasks of the JT65 decoder is to measure the signal level at each of the 64 data-tone frequencies during each of the 63 data intervals in a transmission. The program must then decide which one of the possible 2^72 messages was most likely the one sent. The procedure is necessarily probabilistic in nature. The best decoder will go as far down into the noise as possible, but it must also know when to give up so that it produces few false decodes.

The total of 2^72 distinct messages is far too many to permit each one to be tested individually against the received signal. However, an important characteristic of the Reed-Solomon FEC code used in JT65 is that well-defined mathematical algorithms can be used to direct the decoder toward the most likely candidate messages, based on the available signal information. A mathematical inversion of the code is made possible by the organization of the redundant information contained in the 306 extra bits.

The new JT65 decoder goes far beyond the capabilities of normal Reed-Solomon decoders. If the standard decoding procedure fails to produce a high-confidence solution, the program proceeds to search explicitly for each one of a number of messages that it considers likely or plausible on other grounds.

Nearly 2^28 (over 250 million) different callsigns can be accommodated in each of the two callsign fields of a JT65 message. Once again,  this is far too many to permit an exhaustive search for them all. Consequently, the "deep search decoder" takes the callsigns listed in the file CALL2.TXT (located in the user's WSJT directory) as being the most likely alternatives in the message's second field. A correlation algorithm is then applied to find out if one of these calls and its associated grid locator are present, combined with either "CQ" or the receiving station's callsign in the first field. High-confidence matching of this kind can be accomplished down to about -28 dB on the WSJT scale, in a single transmission, with a very low error rate.

The bottom line is that for any arbitrary callsign the new JT65 decoder performs at least as well as the one in WSJT version 4.7.0. Message averaging works just as it did before, and if you are listening in to a "third party" QSO between two other stations, the sensitivity will be the same as in version 4.7.0. However, if a station that is listed in the file CALL2.TXT is calling CQ or is calling you, your sensitivity will be about 4 dB better on average.
Please note that the decoder is given no information whatsoever about what station you may be trying to work. Its heart is "as pure as the driven snow," even if you are working a sked. However, the decoder does make the educated guess that the callsign of the transmitting station is more likely to be one listed in CALL2.TXT than some other callsign constructed at random.

The program always attempts to decode a purely arbitrary message first. Failing that, it will look more deeply in the noise for the presence of a message that includes the callsign of a station listed in the database file.

It is no accident that the algorithm just described bears close resemblance to the thought processes (conscious and otherwise) that we use to copy very weak CW by ear. Familiar combinations like CQ and one's own callsign are always easier to dig out of the noise than random combinations of characters. Callsigns that we have seen or heard before are more easily recognized than arbitrary calls generated at random. The new JT65 decoder behaves similarly, except that it is kept fully in the dark about who you are trying to work.

A final note: the file CALL2.TXT replaces the file CALLSIGN.TXT used by earlier versions of WSJT. The format has been changed to permit extended callsigns such as those sometimes used by DXpeditions, for example ZA/PA2CHR. The name of the database file has been changed so as not to "break" an earlier version of WSJT that you may wish to keep available. Full support for extended callsigns (i.e., calls with an extra prefix or suffix) is planned for a future version of WSJT.

Changes in Version 4.7

4.7.0 Includes version 2.0 (build 213) of Spectran, by I2PHD and IK2CZL. WSJT and Spectran can run simultaneously.
If you have already installed WSJT, you can upgrade to the current version by downloading and executing UPD470.EXE. Versions 4.0 and later will not overwrite an installed version of WSJT 3.x. You may, for example, have versions 3.8.1 and 4.6.1 installed at the same time, which will give you access to JT44 as well as the newer JT6M, JT65, and FSK441 modes.