-
Q: Why was GRASP password-protected? (It isn't anymore.)
A: I'm not sure. This was recommended to me early on by people
with more experience than I had. They said that this would make it
easier to prevent the following scenario: someone downloads the software,
hacks it up, obtains unreasonable results, then says "GRASP is a bunch
of crap, look at the crazy results that it gives."
-
Q: How do I install GRASP?
A: Read the detailed GRASP installation instructions (currently section
2.5 of the manual). Note that you can download the manuals from our
main GRASP page without downloading the entire package.
-
Q: Can I get a hardcopy of the manual?
A: Yes, but you have to print it/reproduce it yourself!
You can find instructions on printing the manual in 2-up format (4 pages/sheet of
paper) in Section 2.2 of the manual. The doc
subdirectory of GRASP contains the following files:
-
manual.pdf a Portable Document Format version of the manual (has
nice hyperlinks!)
-
manual.ps a PostScript Format version of the manual
-
manual.dvi a Device Independent File version of the manual (viewable
with xdvi if you have TeX installed). Note that we use version 0.9 of teTeX
with the hypertex aware tools, so our .dvi file has links for all the references
and web site in it.
-
manual_bw.ps the black and white pages (only) of the manual, in
PostScript Format
-
manual_color.ps the color pages (only) of the manual in PostScript
Format
The recommended way to print the manual is: print manual_bw.ps
on a black and white laserprinter, and print manual_color.ps on
a color printer. Then take the whole pile, and send it to your local
copy center with instructions to print N spiral bound two-sided copies
with plastic covers. Be warned: copying the color pages is expensive...
-
Q: Can I buy a hardcopy of the manual?
A: Yes. Send $75/copy in the form of a check to the UWM
Physics Department. Send it to:
GRASP manual
Business Office
Physics Dept
University of Wisconsin
PO Box 413
Milwaukee WI 53201 USA
The delivery may take several months: we do print runs about once every
three months, and at the next print run you'll get copies of the most recent
version. The price is high partly because of the cost of reproducing
the color pages. (No, we're not making a profit on this...)
-
Q: What systems does GRASP run on?
A: Any standard POSIX/UNIX system. It's been tested on Sun Solaris
2.*, HP Unix, DEC Alpha OSF, DEC Alpha Linux, Intel Linux, Intel Paragon,
SGI Irix, and IBM SP2. Serge Droz has also ported GRASP to Macintosh.
At one point I heard that some people at Gainsville were also trying a
Win95 port.
-
Q: What other software do I need?
A: The detailed answer can be found in section 2.4 of the GRASP manual.
To summarize, you need
-
an ANSI-C compiler
-
the Numerical Recipes in C library (this can be obtained here)
-
to use frame-format data you will need the LIGO/VIRGO frame libraries from
here
-
to use the inter-channel cross correlation code in GRASP you will
need the public-domain LAPACK library
(linear algebra)from here
-
To display "time-frequency" maps on the screen you will need the
public domain Mesa library from here
We highly recommend that you also obtain the following public-domain items:
-
the graphing program XMGR from here
-
the public domain MPI implementation MPICH from here
-
the LIGO/VIRGO frame libraries from here
-
the public domain optimized FFT package FFTW from here
-
Q: After installing GRASP, I don't seem to be using optimzed FFT's,
just the slow stuff from Numerical Recipes.
A: In order to use optimized FFT's, (we suggest FFTW) you need to first
go into the appropriate src/optimization directory such as src/optimization/fftw
and type make. This creates an optimized object something.o
to link to. After this you can carry out (or repeat) the GRASP installation.
This is worth doing: FFTW is roughly five to ten times as fast as the Numerical
Recipes FFT. (Sorry Bill, it's true. Yours is easier to understand,
at least!)
-
Q: I want to try doing some analysis of data from the Caltech 40-meter
instrument. Can you send me some data tapes?
A: No. To work with the data, you need to (1) be a member of
the LIGO Scientific Collaboration (LSC) and (2) reach a specific agreement
with the LIGO Laboratory about what you intend to do with the data.
Instructions about who to contact and how to go about this can be found
on the LSC
web page page and in the GRASP manual. After you have reached
an agreement with the LIGO Lab, the LIGO Document Control Center (DCC)
will send you data tapes. To contact the LIGO Document Control Center (DCC),
email Linda Turner.
-
Q: I am having trouble running the InstallGRASP script
on my SGI IRIX machine. What's wrong?
A: The problem seems to be with the shell. Do:
setenv SHELL bash
before running InstallGRASP
[Thanks to Adrian Ottewill
and Scott Hughes.] [By the
way, bash is a public-domain shell, which does not come pre-installed so
you need to download and install it. To do this, use a site at Illinois
which mirrors most gnu software: type 'ftp ftp.cso.uiuc.edu', login as
'ftp', and give your email address as the password. 'cd pub/gnu', type
'bin' (to make sure you do your download as a binary file), then 'get bash-2.02.tar.gz'.
This will give you the source code for bash.]
-
Q: I would like to see how file X changed between GRASP 1.3 and 1.65.
What do I do?
A: Download the CVS repository and use the CVS diff command or log
or history command to study the changes. (NOT YET IMPLEMENTED, BUT
COMING SOON...)
-
Q: On my Pentium computer, all the routines which call the Frame Libraries
to read frame files crash. What do I do?
A: For Frame Library version 3.60 and older there is a bug for managing
long integers on Pentium computers. Long integers are not used in newer
frames. You are probably using frame files that were generated with an
older version of the frame library. A quick way around this is to take
your frame files to a non-Pentium computer and use the file copying routine
(FrCopyFile) from an up to date version of the Frame Library to
copy your files. The FrCopyFile executable file is included in
the Frame Library. For example if C1-X was the offending frame
file in the directory /data
SunOSprompt> cd /frame/v3.60/SunOS
SunOSprompt> FrCopyFile /data/C1-X /data/C1-X $LEVEL
would copy the file using the compression level $LEVEL. The
copied files should now work on your Pentium computer.
-
Q: I would like to see the very latest development version of file X.
What do I do?
A: The entire
GRASP DEVELOPMENT FILE TREE
is now available. These are the absolute most current versions of
every file - not release code! If it's broken, don't complain.
-
Q: How can I determine what version of the FRAME files I have?
A: Here are two different methods for doing this:
-
First Method
Use the FrDumpFile utility included with the FRAME library to examine
a single file:
dirac> /usr/local/frame/v3.50/SunOS/FrDumpFile C1-94_11_15_06_17_57
| more
...
History records:
(Wed Jan 21 15:30:24 1998) FrameWrite:Jan 19 1998 v3.50
(Wed Jan 21 15:30:24 1998)
Translation carried out by:
login: ballen
user: ballen
directory: /disk3/14nov94.1.frame
datapath: /disk1/14nov94.1
translation program name: /home/ballen/GRASP/bin/translate
source code name: translate.c
translation date: Wed Jan 21 15:30:24
1998
...
This shows that the frame was written with version 3.50 of the Frame
library, that it was created (translated from old data) on Jan 21, 1998.
-
Second Method
Look within a given frame file for the history structures which
are ASCII strings. To do this (on any UNIX or LINUX-like machine)
choose any FRAME file and give the command:
strings C1-94_11_15_06_17_57 | more
and hunt through the output until you find a listing showing
the version of the Frame library used to write the files and the time at
which the files were created (written)
...
FrameWrite:Jan 19 1998 v3.50
Translation carried out by:
login: ballen
user: ballen
directory: /disk3/14nov94.1.frame
datapath: /disk1/14nov94.1
translation program name: /home/ballen/GRASP/bin/translate
source code name: translate.c
translation date: Wed Jan 21 15:30:24
1998
...
-
Q: What's wrong with (some of the) Frame Data files from November 1994?
And what should I do about it?
A: There were three separate problems that occured in writing
FRAME data files tapes that were sent to the LIGO Document Control Center
(DCC) and then distributed. None of these problems are serious (and
there are simple work-arounds). Here is a description of these problems:
-
File names off by two hours.
By convention, the frame file names were constructed from the (UTC)
time of the data. For example, the first run from November 14, 1994
began at approximately 22:17:57 local California time (actually 22:17:36
local time, see GRASP manual section 15 on time standards!), which is November
15, 1994 06:17:57 Universal Time (since California PST is 8 hours behind
GMT at that time of the year). Hence the label of the frame file
containing this data should be C1-94_11_15_06_17_57. But an error
in the GRASP translate program used to create these frame from
old-format LIGO 40-meter data meant that the file names were only correct
when the translation was done on a machine in the California PST time-zone.
Frames that were translated in Milwaukee had file names that were off by
two hours, for example C1-94_11_15_04_17_57. This is not very
serious - note that none of the GRASP package makes use of the time stamps
implicitly contained in the file names (assuming that they are monotonically
labeled) so that these incorrect file names had no effect on GRASP.
This error affects frames written between January and September 1998. [Thanks
to Bernard Whiting for
noticing this!]
-
Times stored in UTC-Unix format rather than GPS format.
Starting with version v3.23 of the Frame format, internal time stamps
should be stored in GPS format rather than in Unix-C format. Up to
82 microseconds (see the GRASP manual "Time Standards" section), these
differ by 315964811 sec. This is
315964811 sec = 3600 sec/hour x 24 hours/day x (365 days/year x 8 years
+ 366 days/year x 2 years+ 5 days + 11 leap secs)
which is ten years (including two leap days in 1972 and 1976) plus
five days. In Unix-C time, the instant t=0 is 00:00:00 UTC Jan 1,
1970. In GPS time, t=0 is 00:00:00 UTC Jan 6, 1980. However Frames
written with v3.50 of the frame library, before September 1998, incorrectly
contain time stamps in Unix-C format. Note that this problem is slighly
complicated because the FrDump and FrDumpFile utilities included with the
Frame library, and the Frame library itself contained slight bugs for versions
less than or equal to 3.70. This mean that these utilities print out the
range of times covered by the frame with an error of 1 or more hours!
However if these utilities show that your 1994 data set frames are from
2004, then the frames that you are working with are written in UTC-Unix
rather than GPS format!
-
Incorrect time offset between local time and Universal Time.
The frames contain a structure which stores the offset (in seconds)
between local time at the detector site and Universal Time. This
should be -8x3600 sec since California PST differs from GMT by 8 hours
in November. However in Frames written before September 1998, this
time offset is incorrectly set to -9x3600 sec = 9 hours.
There are several possible work-arounds for these problems:
-
Write to the LIGO DCC, and request a new Frame data tape (written after
Sept 1, 1998). These frame files have names like: C1-468915466.F, in other
words they are labeled with GPS time labels not with UTC time labels. The
most current version of the 40-meter data being distributed by the DCC
was written with Frame Library Version 3.72 on Oct 13/14, 1998.
-
Live with it. Unless you need correct or accurate time-stamps/file
labels, these errors don't matter.
-
The latest versions of GRASP (1.7.1 and greater) contain a modified frame
interface, which correctly detects if the times have been stored in the
Unix-C rather than GPS format, and automatically corrects these times (after
issuing a warning message to the user).
-
Q: There seems to be a missing directory/files on my frame data tape(s).
What's going on?
A: The LIGO 40m prototype detector data has been distributed
in the frame format on Exabyte tapes. There have been two different releases
of the data. The first release was in February, 1998 and the data
was divided on two Exabyte tapes. Each tape contains part of the data.
Tape #1 contains:
14nov94.1.frame
14nov94.2.frame
18nov94.1 .frame
18nov94.2.frame
Tape #2 contains:
19nov94.1.frame
19nov94.2.frame
19nov94.3.frame
19nov94.4.frame
20nov94.1.frame
20nov94.2.frame
The frame data corresponding to 20nov94.3.frame was missing and
is not on either of the two tapes.
The second release of the data took place in Oct 1998 and it
contains the complete data; all the data is contained on one Exabyte tape.
If you have one of the earlier tape sets with a missing directory, please
send them back to Linda
Turner in the LIGO Document Control Center (DCC) and ask her
to replace it with one of the more recent one-tape-has-all-data tapes.
If you have questions about any of this, please contact R.
Balasubramanian.