To:      Users of WSJT
From:    Joe Taylor, K1JT
Subject: Beta release of WSJT Version 1.8.0
Date:    March 3, 2002

Beta Release 1.8.0
Prešvercal S53X

------------------


This is a major upgrade from beta release 1.5.4 (also known as 1.54).
There are no significant changes to the FSK441 meteor scatter mode;
JT43 has been replaced by an entirely new mode called JT44, which has
similar goals but much better performance.

You can download v1.8.0 from 

  http://pulsar.princeton.edu/~joe/K1JT/BETA180.ZIP

and install it in the usual way for all WSJT updates.  Save your old
copies of WSJT.EXE and WSJT1.DLL, in case you need to go back to them.
Then copy the files extracted from the ZIP file into the locatiions
where WSJT.EXE and WSJT1.DLL had been.  Upon initial startup, be sure
to enter your station parameters in "Setup | Options" and be sure that
a valid period (usually 30 seconds) is entered in the T/R Period box.

Unlike JT43, the new JT44 mode is *not* compatible with PUA43.  JT44
transmits 22-character messages and uses 30-second TR sequences.
Slightly more than half of the transmission time is devoted to sending
a sync tone at a fixed frequency of 1270.5 Hz.  This tone is keyed on
and off in a pseudo-random pattern designed to permit accurate time
and frequency synchronization.  In practice, the program seems to lock
easily onto signals as weak as -20 dB relative to the receiver noise
in a 2500 Hz bandwidth.

A fixed-length 22-character message is sent in the remaining 50% of
each transmission interval.  This means that 1.5 dB of theoretical
sensitivity has been sacrificed to the goal of having robust time and
frequency synchronization, which seems to be a good trade-off.
Characters may be selected from the same 43-character "alphabet" of as
that used in FSK441, PUA43, and the now-obsolete JT43 .  Each of the
43 characters has its own specific tone.  The tones are spaced from
each other by about 10.77 Hz, and together with the 44th tone (the
"sync" tone) they run from 1270.5 up to 1755.0 Hz.  This narrow spread
means that even if TX and RX are offset by up to 600 Hz in either
direction, the full range of tones can still be included in your 
radio's SSB passband.  

Each tone interval lasts for exactly 2048/11025 seconds, and the
entire 22-character message is repeated three times in a single
transmission.  Transmitted tones start about 1 second into the
30-second TX interval, and they finish about 25.1 sec later.

One of the big improvements of JT44 over JT43 should be its ease of
synchronizing two stations properly in time and frequency.  The
program searches for time sync over the range -1 to +4 seconds.  A new
parameter called DT measures the time offset between computer clocks.
The allowed range for DT is asymmetric about zero so that the extra
EME delay of up to +2.7 s can be accomodated without problems.  In
like manner, the program searches for frequency offset over the range
-600 to +600 Hz.  This range should be enough to accommodate most of
today's radios, at least up through the 144 MHz band.  There should be
no need to use RIT or to "tune the other guy in" to achieve a small
DF.  As long as you're within +/- 600 Hz, the program should handle
everything automatically.

The mouse doesn't do anything in JT44 mode.  In fact, this is pretty
much a "hands off" mode, as presently implemented.

There is no message averaging at present, except that an average of
the last four characters of the 22-character message is presented at
the right edge of each line of decoded text.  This simple averaging
facilitates getting the necessary RRRs to complete a difficult QSO.
Since the time and frequency synchronizing seems to work well, I will
soon put real message averaging back into the code.

With the code as presently distributed, messages seem to decode
reliably down to signal strengths of -15 to -18 dB.  I probably made
the "gibberish filter" too hard too please; if the signal is too weak,
you'll get no line of decoded text.  I will probably put in a
user-selectable print limit soon.

In the graphical display area, the green line is the same as in FSK441
mode: a plot of signal strength versus time over the 30-second RX
interval.  The red line summarizes the attempted synchronization with
respect to frequency offset.  The horizontal scale runs from -600 to
+600 Hz, with yellow tick marks at the top indicating 200-Hz
increments.  The blue line is the corresponding information for time
synchronization.  For this purpose the scale runs from -1 to +5
seconds, with the yellow tick marks denoting 1-second intervals.  The
blue line does not go all the way to the right-hand edge because the
maximum positive time offset is only about 4 seconds.

NB: properly synchronized data will show a narrow spike in the red
curve and a somewhat broader spike in the blue curve.  A spike in the
red curve only probably means you have a birdie in the passband.

All suggestions for more or better operational features will be very
welcome!  

Please send email to [email protected].

			-- 73, Joe, K1JT

Beta Version 1.5.4
------------------
Can be downloaded from

  http://pulsar.princeton.edu/~joe/K1JT/BETA154.ZIP

Its improvements over v1.52 (see below) are in the COM port handling.
Entering zero for the COM port number now disables all T/R switching
through the COM port.  

Beta Version 1.5.2
------------------

Version 1.52 introduces some improvements in the JT43 mode.  They are
minor in a sense, but they do seem to make a big difference in getting
the program to find the corect DF and "lock up" on a received signal.
A number of additional improvements are sketched out in my notebook
but I have not yet had time to implement them.

Owing to screams of protest after its removal, the "Save Last" button
has been restored on the main screen.  

The JT43 test file WD4KPD_011028_2150.WAV is no longer bundled into
the ZIP file; instead, you can download it separately from the URL

  http://pulsar.princeton.edu/~joe/K1JT/WD4KPD_011028_2150.WAV

Version 1.52 looks only for standard message #1 when doing its
frequency search.  More precisely, it looks for your callsign at the
beginning of the message and the ......... filler at the end.  Once
you have found the other guy by seeing some recognizable decoded text
on a line in the upper window, you should note the DF, hit the "Clear
Ave" and "Decode Again" buttons to set the average message equal to
the correctly (or partially correct) decoded line.  You can continue
averaging the message from there.  It usually helps to freeze the DF,
once you have it right, by clicking on the blue peak.  But you may
need to click again every few minutes in order to follow a slow
frequency drift.  I am working on ways to improve this.

My on-the-air experinces that best exemplify the capabilities of this
mode have been with W8PAT on 2 meters.  John is 400 miles from me, and
our respective signals are too weak to be able to work ordinarily on
tropo.  We work very easily in JT43 mode, however, and we can reduce
power quite a bit (I think we went down to 15 Watts or so, but I can't
find my notes) and still work easily.  Beyone 600 miles or so, it is
generally easier to work by using underdense meteor pings and the
FSK441 mode.


*Please*, if you have gained some experience using the JT43 mode, send
me email summarizing how it has worked for you.  What version(s) have
you used?  What distances have you worked successfully and tried
unsuccessfully?  What seems to work best for you in determining the
other guy's DF, and keeping the program properly locked?  Have you
made good QSOs when you could not hear (or could barely hear) the
other guy's signal, so that a CW QSO would have been impossible?

This is new territory, so the more feedback I get, the more I should
be able to improve the program!

I append below the full text of the first JT43 announcement.

				-- 73, Joe, K1JT


###################################################################
			Beta Version 1.50
###################################################################

Important!
----------

Please read all of the following before upgrading to WSJT v1.5.

Preliminary Background
----------------------

This note is to bring to your attention some major new capabilities
being incorporated into the WSJT software package.  As you know,
FSK441 -- the mode that brought WSJT its wide initial popularity -- is
designed for high speed meteor scatter work.  Such communications rely
on the peaks of very brief signal enhancements.  For good performance
you need the highest possible throughput of information in pings that
last only a fraction of a second.  In the present context, these pings
can be classified as "strong" signals -- by which I mean that you can
hear them above the receiver noise in a 2500 Hz passband.

For some time I have hinted at future implementation in WSJT of a
weak-signal mode based on PUA43.  This mode attacks the opposite end
of the spectrum of signal strengths and degree of variability.  Bob
Larkin, W7PUA, designed the PUA43 mode to take advantage of a
moderately slow transmission rate, 0.5 characters per second, together
with coherent and incoherent integration techniques.  PUA43 can yield
excellent copy on signals 20 dB or more below the level of receiver
noise.  The mode works well with signals that are more or less steady
in amplitude, at least for periods of some minutes.  Appropriate
propagation modes would include troposcatter, ionoscatter and EME.

I have not yet decided exactly what the weak-signal slow mode in WSJT
should be called.  For the present I am tentatively naming it JT43, a
name that has the advantage of plainly recognizing PUA43 as an
original source of ideas, while also making clear that the mode
implemented in WSJT is not the same in all respects.

JT43 is closely modeled on Bob Larkin's sub-mode called PUA43-C28.  It
uses frequency-shift keying with 43 tones spaced at intervals of 37.5
Hz.  All messages are 28 characters long, and they are transmitted at
0.5 characters per second during the first 56 seconds of a UTC minute.
Each encodable character translates to one of the 43 possible tones,
but the encoding algorithm includes a pseudo-random frequency shift
that changes with each minute of the UTC day.  This strategy is an
important and highly effective method of combatting low-level birdies
and other narrow-band interference.  

In some other respects JT43 differs significantly from PUA43.  Some
enhancements were necessary because the run-of-the-mill radios used by
most VHF operators do not have the degree of frequency accuracy and
frequency stability available with the home-made DSP-10 transceiver
and its specialized hardware (see Bob Larkin's articles in QST for
September, October, and November 1999).  In its initial
implementation, JT43 accommodates frequency offsets between
transmitter and receiver up to +/- 300 Hz.  Instabilities at the level
of 5-10 Hz or drifts of a few Hz per minute should be acceptable as
well, with some loss of sensitivity.  These relaxed specifications
should help to make the JT43 mode usable with most existing equipment.
I hope and intend that it will be possible for a station using JT43 to
work a station using PUA43-C28, so in that sense the modes will be
compatible.  Right now I have no way of testing that capability, as I
do not have access to a DSP-10.


Changes in the FSK441 Features of WSJT
--------------------------------------

1. The meteor-scatter features and capabilities of WSJT have not been
changed significantly.  Nearly all features of the user interface and
program operation will be completely familiar to existing users.

2. The "Save Last" button has been removed.  I don't think anyone used
it very much; all too often, the file you might have wanted to save
had already been erased, a few seconds earlier, when the next
recording interval began.

3. The "Save All" function has been changed from an on-screen checkbox
to a item on a new "Save" menu.  I will probably add a second
selection on that menu called "Save Decoded" that will save only those
files producing some decoded text.


Introduction to the JT43 Mode
-----------------------------

To copy a CW signal you need to be able to hear it in real time, and
this means that the signal needs to be at least several dB above the
noise level.  The human ear/brain combination can narrow the effective
passband to something like 200 Hz, permitting a good operator to copy
CW at something like 6-8 dB below the noise in an SSB transceiver's
2500 Hz passband.

JT43 uses coherent integration techniques to achieve bandwidths in the
1-10 Hz range, thereby gaining another 13 to 23 dB in sensitivity
beyond CW.  JT43 is a "slow" mode, but not "ultra-slow": it achieves a
respectable throughput of 28 characters per minute by encoding every
transmitted character as a distinct tone sent for exactly 2 seconds.
Each detectable signal element becomes a full character, rather than a
"dit" or "dah" in a Morse code sequence.  JT43 also provides for
incoherent integration to increase sensitivity even further.  For
every doubling of integration time, theoretical sensitivity is
increased by another 1.5 dB.  In other words, if you are able to copy
a signal 20 dB below the noise in one minute, theory says you should
be able to reach -21.5 dB in two successive one-minute transmissions
of the same message, and -23 dB in four minutes.

What will these things mean in practical circumstances that include
imperfect frequency calibration, frequency drifts, weak birdies and
occasional QRN?  What gains in sensitivity will actually be achieved?
I very much look forward to learning the answers to these and other
questions about JT43!  Right now what is needed is more on-the-air
experience.  You will be interested, however, that my early tests have
been extremely encouraging.  For example, ten days ago I recorded a
number of one-minute transmissions from WD4KPD over a 360 mile path on
2 meters.  David was running 40 W and we both have single yagis, his
at 25 feet and mine at 45 feet.  His signal was barely audible on QSB
peaks, and at best would have required many repeats to be copyable on
CW.  In many of the one-minute transmissions I heard nothing at all
from him.  Nevertheless, with JT43 the copy was very close to 100%
over 40 full minutes of testing.

If you can usually work another 2 meter station at 350 miles on CW,
you should be able to work a similar one at 500 miles with JT43.
There should be no need to wait for tropo enhancements, aurora, or
whatever.  Beyond that distance, of course, WSJT with the FSK441 mode
will get you out to 1100-plus miles by using meteor scatter.  With a
combination of the two modes, you should be able to work half the
continent with a modest station -- that's around 500 grid squares, if
you are not on the coast!  Similat arguments apply to the other VHF
bands.

Enough advertising for now!  Suffice it to say that I think JT43 could
be a very interesting mode.


Operational Details for the JT43 Mode
-------------------------------------

1. Meteor-scatter contacts need to be synchronized so that one station
is listening when the other one is transmitting.  A second or two of
clock error is not devastating, however.  With JT43 the requirements
are much more stringent.  Your computer clock needs to be synchronized
with UTC to within half a second, and preferably a tenth of a second
or better.  If you have web access, one of the best ways to achieve
accurate timing is to download a free utility program called
"Dimension 4" from http://www.thinkman.com/dimension4, and run the
program before starting any sked.  If you are fortunate enough to have
continuous web access, just leave Dimension 4 running all the time,
and your computer clock will remain accurate to about 0.05 seconds.

2. JT43 needs to be very tightly scheduled in its internal operations,
as well.  As a consequence the only way to transmit is to toggle "Auto
Period On", leaving all T/R switching to the computer.  Transmission
and reception periods always start at 00 seconds of a UTC minute, and
the T/R period is always 60 seconds.  For sequence ordering I strongly
suggest that the same convention be used as with meteor scatter, so
that (in North America) the westermost station transmits "first,"
taking the even minutes (00, 02, ...) of each hour.

3. At 56 seconds into a reception period, WSJT does a computation that
defines the audio file for the following minute's transmission.  T/R
sequencing begins at 58 seconds into the minute.  All switching
transients should have died out by 00 seconds, when the next
transmission or reception interval begins.

4. To enter into the JT43 mode, select menu item "Mode | JT43".  You
should then see several changes in the familiar WSJT screen.  The
large box for decoded text will divide itself into two parts, and the
default standard messages change to a format better suited to JT43
communication.  In addition, the status of several controls changes
from enabled to disabled ("grayed out"), or vice-versa.  The purposes
of a few other buttons become different, as well.

5. A word about the default message formats for JT43 will be useful
here.  Neither the message formats used in meteor-scatter QSOs nor the
logical "flowchart" used to determine when to step to the next message
is well suited to a slow integrating mode like JT43.  The
signal-to-noise ratio of a JT43 conversation builds up during
successive receiving periods, so it does not make much sense to send
an ordinary signal report or to delay sending part of the essential
information (other than "Rogers") until sometime in the middle of the
QSO.  I have therefore defined standard messages that contain both
callsigns and the transmitting station's 4-digit grid locator,
followed by "filler characters" to pad the message to 28 characters.
Message #1 in a QSO sequence uses period (.) as the filler character.
Message #2, to be sent when you have copied message #1 from the other
station, uses "R" as the filler.  Message #2 still contains both calls
and your grid, because even though you have copied his information, he
may not yet have copied yours.  When you received "RRRR..."  from the
other station you proceed to message #3, which contains both calls
followed by "73 73 73 73 73".

These standard message formats help to give JT43 its very important
capability of acommodating an unknown frequency offset between
transmitter and receiver, even when the offset is much larger than the
37.5 Hz separation between tones.  The WSJT program knows that in a
minimal QSO the first part of a standard message sent to you should
contain your callsign, and the last part should contain either
".........", "RRRRRRRRR", or "73 73 73 73 73".  A full search is made
over +/- 300 Hz of frequency offset for messages of one of these
forms, and a "combined spectrum" of the transmitted message is
computed relative to each expected message.  At the end of each
reception period the combined spectrum is plotted as a blue line in
the large plot area, just above the green line that is familiar to
WSJT users.  Yellow tick marks at the top indicate 100 Hz intervals
from -300 to +300 Hz, left to right.

6. When you receive a properly decoded standard message, the blue-line
spectrum will show a narrow peak that looks something like a ping on
the green line.  Its location relative to the center of the plot area
indicates the frequency offset DF.  For example, if DF = 100 Hz, you
should see a peak in the blue line under the first yellow tick to the
right of center.  If the received text is recognizable you may click
on the blue spectral feature, effectively "freezing" DF to a value
corresponding to the location of the pointer.  The frozen DF value
will appear in the last status bar panel, at the lower right of the
screen.  Subsequent received files will no longer need to be searched
to find the correct frequency offset.  To re-enable offset searching,
you can set DF back to zero by clicking the "DF 0" or "Reset
Defaults".

7. Decoded text appears in the usual scrolling window and also in the
smaller window below the first one.  The second window displays the
"average" message, and it will typically increase in accuracy as a
message is repeated in subsequent minutes.  Pressing "Erase" clears
the main text area, as in the FSK441 mode.  Pressing "Clear Ave"
clears second text box and the averaging buffer, so that averaging
can start over.

8. Some information provided on the decoded text lines has slightly
different meaning than in FSK441 mode.  Since receiving intervals
always start on an even UTC minute, the time is given in HHMM format.
The parameter denoted as "T" gives the total integration time for the
decoded message; this number is always 1 in the upper window. "Width"
gives the width of the selected spectral feature in the blue-line
diaplay.  Typically it should be 10 Hz or less, and in general smaller
is better.  (Very rapid QSB, doppler shifts, oscillator drifts, and
other effects can cause the width to increase.)  The next column gives
relative signal strength in dB -- supposedly relative to the total
receiver noise, but I haven't calibrated the numbers yet.  DF is the
relative offset between transmitter and receiver, exactly the same
meaning as in FSK441 mode.  It says nothing about which oscillator is
"off" in calibration, but it should give the measured difference to an
accuracy about equal to width/2.  The decoded text comes next, with a
full 28-character message.  Finally, at the right-hand side under the
label "Freq" for the smaller plot area, a single character is listed
indicating which one of the standard filler characters has been
detected.  Messages #1 and #2 are shown as "." and "R", while message
#3 is shown as "7".  A fourth filler character is also allowed based
on the last part of the message being left blank.  The blank filler is
indicated by the "_" (underscore) character.

9. The decoding parameters W, ST, and Tol used for meteor scatter are
de-activated in JT43 mode.  Parameter S is interpreted to mean
"smoothing" in JT43.  All computed spectra are smoothed with a running
average of S frequency bins, with one bin nearly equal to 1 Hz.  In
general you should leave S set at a low value, say 3 or 5, for maximum
sensitivity (it is always an odd number).  If you see messages decoded
with W larger than about 5 Hz, you may get improved copy by increasing
S to slightly larger values.

9. As presently implemented there is one glaring deficiency in JT43
that will slow down QSOs.  Suppose station A transmits first.  Station
B may copy the transmission perfectly, but will not know that the copy
is good until his own first transmission has already started.  You
can't change a TX message once the transmission is started, so station
B needs to wait a full 2 minutes -- his transmission, followed by station
A's transmission -- before he can send message #2.  I will improve
this situation by the next release.


Request to Beta Testers
-----------------------

The JT43 mode in WSJT v1.5 is early beta-release software.  The new
mode has had very little on-the-air testing.  Your comments,
criticisms, and suggestions will be greatly appreciated!  As usual,
please send them to [email protected].


And Finally ...
---------------

Here's what to do if you want to take part in beta-testing the JT43
mode of WSJT v1.5

1. Download the file BETA150.ZIP.  You will need to point your
   browser explicitly to 
   http://pulsar.princeton.edu/~joe/K1JT/BETA150.ZIP, and note
   that it is case sensitive.  The file is "invisible" and has
   no evident link on the WSJT download page.
2. Unzip the downloaded file in a convenient directory such as C:\TEMP.
3. Make backup copies of your installed files WSJT.EXE, WSJT1.DLL,
   and WSJT.INI, or move them to another directory.
4. Delete your present file WSJT.INI
5. Copy the new versions of WSJT.EXE and WSJT1.DLL into the 
   directories where the previous versions were.
6. Start WSJT and re-enter your station information on the "Setup | 
   Options" page.
7. You should now be operational with WSJT version 1.5.  

8. I suggest that you use File | Open to read in the test file
   included with the beta distribution, WD4KPD_011028_2150.WAV, while
   you are still in FSK441 mode.  (You will first have to rename the
   file to the name just given; the zip utility probably mangled it
   to something else.)  You will see that it is an ordinary
   looking recording of receiver noise; you'll hear nothing but noise
   if you play it back, and the waterfall spectrogram shows no signal
   either.  Then switch to JT43 mode and press "Decode Again".
   David's signal gives 100% copy (at DF=45 Hz), some 12 dB below the
   receiver noise.
				-- 73, Joe, K1JT
-----------------------------------------
za kkdxc S53X marec2002