ACT: Matlab front-end for the Acoustic Toolbox written by Alec Duncan from the Centre for Marine Science and Technology at Curtin University.
atWinPII_f95. Version for a Windows, Pentium II machine. The Makefiles are in a sort of Unix style. If you download cygwin (publicly available on the web) you get a sort of pseudo-Unix window in which you can just type 'make all' and 'make install' to recompile and install.
atLinuxPII_f95. Same as above but with binaries compiled under Linux.
BELLHOP: A beam/ray trace code
KRAKEN: A normal mode code
Fortran compilers differ in whether record lengths should be declared as bytes or words. To ensure the programs run, I declare lengths in bytes. If your compiler uses words as the default, look for a compiler switch to override that so that your files are not unnecessarily large.
Some of the programs in this package use direct
access I/O and need to know a record length. Many systems do not
maintain this information as part of the file info so the code
writes that information directly into the file. Then the file is
opened twice. First it is opened with a dummy record length to
read the actual record length from the file. Then it is re-opened
with the correct record length. On machines that do
know the correct record length, that first open may generate an
error message and will have to be removed. The RECL specifier should then also be removed from
the second open.
The Twersky-scatter option is problematic because it calls mathieu.f and slbessel.f which don't adhere to a rigorous Fortran standard. To avoid this problem, the current distribution has this option disabled.
If you want to put it back in, you will need to link to twersk.f instead of twerskfuse.f. These routines call mathieu.f which needs to be compiled with a switch to initialize all variables to zero. Most compilers have some such switch.
Plot packages come and go. However, the output
formats of the various models are all fairly simple to work with.
The distribution includes Matlab scripts which I currently use for plotting.
If you don’t have Matlab, then these
will at least provide an example of how to read the file formats.
To run the Acoustics Toolbox, the executables
need to be in your path. The process for setting your path depends
on your operating system. I can't stay current with the options
but here are some things that have worked in the past:
DOS: add the following to your autoexec.bat to make the scripts executables in your default path:
SET PATH=%PATH%;C:\AT\SCRIPTS;C:\AT\BIN
Windows 2000 (and perhaps NT): right click on the “My Computer” icon and select “Properties”, “advanced”.
Windows XP: go to
control panel/Performance and Maintenance/System/Advanced and look
for the pushbutton labled
“Environment Variables”. There you can edit the system variable
“Path”.
If you’re having problems, verify your path includes these by typing “path” from a DOS window. Typically you run KRAKEN by typing something like “kraken pekeris”. at the DOS command line, or kraken( 'pekeris') at the Matlab command line.
On Unix/Linux/MacOSX systems you can ensure that path info is set in Matlab at start up by editing .../bin/matlab to add something like: source ~/.profile
where .profile is the shell start up script
that initializes your path.
Finally, if you use Matlab then you just need
to use the "Set Path" option under the "File" menu. If you include
all of the Acoustics Toolbox in your Matlab path, then the Matlab
scripts find the location of the binaries using the Matlab "which"
command.
Once you have the basic code working, you might
want to run some of the test batteries in at/tests to completely
verify your installation. The m-file, runtests.m
runs the complete suite of tests using the Fortran binaries. The
m-file runtestsM.m does the same thing, using the Matlab
versions of BELLHOP, SCOOTER, BOUNCE, and SPARC.
(Check also the individual subdirectories for updates on individual components of the Acoustics Toolbox.)
December 1997
Minor changes have been made to the package for improved f77 to f90 portability. Fortran 90 seems to treat parameters differently in terms of precision so some of those changes are important for the most accurate results.
July 1999
The whole package has been converted to f90/95 standard with free-format lines and taking advantage of dynamic allocation. The code is more concise and portable. Dynamic allocation means that you will not have to worry about dimensional limits--- the size of the problem you can solve is limited only by RAM on your system.
The change to dynamic allocation has forced a small change in the input structure. Previously an input format was often used with a number of items in a vector followed by the actual vector. For instance, the vector of receiver depths is read this way.
In the new version the number of items in the
vector must appear on a separate line so that the code can read
it, allocate the space, and then read the vector contents.
May 2009
The Matlab version of BELLHOP has been replaced
by a new version that traces one beam at a time. The ealier
version traced all the beams in parallel for efficiency; however,
the logic is too complicated to maintain.
December 2009
Converted most "char foo*5' references, which
are deprecated in the new Fortran, in favor of "char (len=5) ::
foo".
January 2010
Updated the volume attenuation from the 1967
Thorp formula to the more recent one in JKPS. Converted all the
Fortran models so that they would determine the file names from
the command line arguments. This eliminates the need for DOS,
Unix, or Matlab scripts to move files around.
July 2010
Made all type conversions, e.g. complex to real
or double to single, explicit. Expanded use of EXPLICIT NONE to
ensure all variable types are called out. Further expansion of the
lengths of variable names to take advantage of Fortran 2003.
Expanded use of structure-variables. Additional changes as
described in the read.me files for the various models. This is a
fairly significant revision in terms of the evoloution from
Fortran 77 to Fortran 2003. However, there are virtually no
changes to the input and output files.
April 2011
Greatly expanded use of INTENT( IN, OUT,
etc.) to clarify the source code. Continued changes in variable
and subroutine names to make them more readable as the Fortran
2003 standard permits. Use of explicit dimension sizes in variable
declarations to make it easier for the compiler to detect
incompatibilities. Change to SHDFile format for the pressure field
to allow multiple x-y coordinates for the source.
October 2011
Added an option for a source beam pattern in
KRAKEN and SCOOTER. The pattern is read from a file, following the
same procedure as in BELLHOP. The beam pattern is actually
introduced after the KRAKEN or SCOOTER run by processing the mode
file (FIELD and FIELD3D) or the Green's function file (FIELDS).
Currently the beam pattern is implemented by shading the
wavenumber spectrum. As a result the beam pattern must be
symmetric with respect to the horizontal (only the part of the
pattern with positive angles is used). This capability has also
been implemented in the Matlab versions of FIELD and FIELDS;
however, in FIELD it current only works for the range-independent
case.
March 2012
The new Matlab changed how it did shell escapes
(to run things at the command line level) so that piping of stdin
and stdout no longer worked. To get around this FIELD and FIELDS
have been modified so that they read from a user specified file
instead. The field parameter information is now read from
'root.flp' instead of field.flp, where 'root' is the root of a
filename passed to the code. This change has been made in both the
Fortran and Matlab versions. In addition, to make the Matlab more
compatible with the Fortran version, the various output files
(SHDFIL.mat, GRNFIL.mat, etc.) now use individually named files
(root.shd.mat and root.grn.mat, etc.).
February 2013
A bug was fixed in detecting the end of the SSP for a given layer.
Instead of testing for exact equality of the SSP depth to the
bottom-depth of the layer, I had previously checked for agreement
to within a tolerance. The tolerance was scaled based on the
bottom depth of the layer so for a bottom depth of zero (in an
aeroacoustic problem) the tolerance was 0 and the bottom was not
detected.
February 2014
Checks have been added to ensure certain vectors are monotonically
increasing where that assumption is critical to the codes.
November 2014
The Matlab routines that read altimetry and bathymetry files were
not clearing old data on subsequent calls. This occured only when
the flat boundary was selected and the routines were supposed to
return dummy boundary information. Also, the ranges were not being
converted to meters from kilometers. It effected SimplePE runs and
probably the Matlab version of BELLHOP as well. These routines are
not used by any of the Fortran implementations and the error often
had no effect on the Matlab codes.
June 2014
Sometime around January of 2010 the Thorpe formula for volume
attenuation was modified at the request of a user. The modified
formula was slightly more accurate. The change was verified in
BELLHOP; however, there an incorrect variable name was used in the
crci routine used by KRAKEN and SCOOTER. That error mean the added
Thorpe attenuation was never added. This is corrected in the new
code. The volume attenuation was typically most important for
higher frequency sources, which in turn were more often treated in
BELLHOP. However, the error can be important for low frequency
sources at very long range.
July 2017
The package has been modified to allow broadband runs for most of
the models. Thus, for example, you can give a vector of
frequencies to KRAKEN or SCOOTER and produce a single pressure
file containing results for each of the frequencies. This has
required a small change to the file formats for modfils, shdfils,
grndfils, etc. In addition, all the Matlab codes have been
modified to allow the '/' convention used in the Fortran versions.
This option allows a user to either specify a full vector or just
specify the upper and lower limits terminated by '/'. It then uses
those limits to create a vector of equally spaced points.
November 2017
Updates and fixes to the noise routines in at/Matlab/noise. A
routine has been added to construct the noise covariance matrix
directly from the shdfil. The shdfil contains the complex pressure
field vs. range and depth and the noise routine can use such files
produced by KRAKEN, KRAKENC, SCOOTER, and BELLHOP.
Why do I get an end-of-file in
reading the very last line of the envfil?
You left off a carriage return/line feed after the last line, or other appropriate end-of-line marker for your system.
Why do I get a segmentation
violation?
You ran a case that required an outrageous
amount of storage. The code is supposed to catch all such cases by
detecting an error in the allocation of dynamic variables (or
limits for the few arrays with fixed dimensions). However, we see
cases where the compiler simply does not handle huge arrays
properly.