pulsars

SIGPROC

PSRFITS2FIL

psrfits2fil converts AO psrfits files from the pdev spectrometer to sigproc filterbank program. This standalone routine will probably one day be incorporated in the sigproc filterbank program.

psrfits data taken with the mock pdev spectrometer is scan based. During a single scan one or more files are written to disk until the end of the scan. Sequence numbers will increment by 1 for each file in a scan. The sequence number for the first file of a scan is a multiple of 100.

Conversion of psrfits->sigproc.fil format is scan based. The routine will read all of the psrfits files from a scan and output a single sigprog.fil file.

Calling psrfits2file:

The options/filenames have been switched to normal unix format. The options come first preceded by - and then followed by the input filename:

Usage: psrfits2fil -{options} <psrfitsfile>

OPTION	ARG	DESCRIPTION
-fcent	freqMhz	band center frequency. (def: take from header)
-filterout	-	write to stdout (def: disk file: "inputfile".fil)
-floats	-	write floating-point numbers. (def: same type as input)
-flip	-	flip the frequencies. (def: used header info)
-headerfile	-	write header to ascii file (head). No header in data stream
-headeronly		write ascii header to file "head" and then exit. No data output.
-help	-	print this message.
-odir	outputDir	output directory to use if disk output. (def: . )

Output data:

by default output is written to a file in the current directory. The output filename is the same name as the input file after replacing the trailing .fits with .fil:

If current directory is /share/psrdat
inpfile=/share/pdata1/pdev/x101.20090113.b0s0g0.00500.fits
outfile will be: /share/psrdat/x101.20090113.b0s0g0.00500.fil

The -filterout option will cause the data to be written to stdout.
All of the files of the input scan is written to a single output file.
By default the sigproc header is written to the start of the .fil file.

-header will write the sigproc header to the file ./head in ascii. The output data will not include the header (it will just have data).

By default the datatype of the output data matches that of the input data (normally 2 byte shorts). the --floats will force the output data to be floats.

Examples:

Output to a single file with same format as input:

psrfits2fil /share/pdata1/pdev/x101.20090113.b0s0g0.00500.fits

Output to a single file as floats. Leave the header in the output file:

psrfits2fil -floats /share/pdata1/pdev/x101.20090113.b0s0g0.00500.fits

Send data output thru a filter. Write the header to disc:

psrfits2fil -filterout -floats -headerfile\ /share/pdata1/pdev/x101.20090113.b0s0g0.00500.fits | rest of filter

Disc locations:

Program\

executable: /share/megs/phil/svn/pdev/pdev/psrfits/sigproc/psrfits2fil.

It is currently compiled for 32 bits but also runs on galfas1.

Source: /share/megs/phil/svn/pdev/pdev/psrfits/sigproc/psrfits2fil.c
Compiling:

Look at makefile . Need to set INCFITS=, LFITSD=..
need to make sure that the cfitsio lib you link in is the same architecture (32/64) as the executable you are creating.

Data files:

see /share/phil/sigproc. Look at the two Readme for what data was taken.
The psrfits files for the test data are located in /share/pdata1/pdev/
Look at /share/megs/phil/x101/090113/makefil.sc for a script that converted the 090113 files.

Notes:

If the input files (.fits) are > 2gb then I've seen the the cfitsio open routine abort. The problem occurs if you run nfs version 2 on linux. The open tries to position to the end of the file and it looks like the nfs call can't handle a int > 2gb (signed 32 bits). nfs version 3 doesn't seem to have the problem. The problem also doesn't occur if you read the file locally (no nfs read).

cfitsio can complain of different versions. The cfitsiolib is a shared object. The cfitsio code wants to make sure that the library version on your system matches the version used when psrfits2fil was compiled (CFITSIO_VERSION 3.03 as of 27jan09).

Changes from duncs psrfits2fil used at gbt:

got source from dunc for gbt version psrfits -> .fil

updates to work with pdev spectrometer psrfits format

differences in psrfits format

added 2 bintables after the primary header and before the subint table
extname: AOGEN .. generic ao if/lo setup info
extname: PDEV .. pdev spectrometer config info
These bintables have keyword values but no rows..
extname: SUBINT the psrfits data table. I added a STAT column (after the data column). This holds the stat info for each spectra: seqNum, numAccums, blanking info.

program differences:

psrfits.h

added struct: SUBINT_COL. It maps column names to column number of subint bintable
few other variable additions

read_psrfits.c:

psrfits_open

parameterized number of digits in filename seq nun.
grab info out of PDEV bintable needed for blanking.
keep track of bytes in statblock each row, fix fold mode to use nbits to compute databytes in row.
on first call map colnames to column numbers (needed for read).
only malloc the data buffers on the first open.. No sense doing it every row since the size can't change...

psrfits_read_subint

define args for cfitsio that are longlong to be longlong.
change readcol routines to use the colNums we got in open
changed the read of the subint header info to only occur on the first read. Sigproc isn't using any of this info that changes...
switched to read byte,short,int,float,doubles. original
version only read bytes..

psrfits2fil.c

options now come before filenames (normal unix convention).
flip is now an option -flip
fcentMhz is now an option: -fcent cfrFrqMhz
added options: filterout,floats,headerfile,odir,help (see list above).
by default the output filename is the same as the input filename
change filename parsing to just grab the seqnumber (parse from the end backwards).. don't bother with all the other fields in the name.
if -header write ascii file. Make it look like the output of the header command.
Got rid of -o.
process 1 row at a time.
key off of the datatype to determine what the input/output data types are (and -floatOut).
use a macro to work with all the different datatypes.
include blanking in the row processing. If blanked then scale so that the mean of all spectra are the same. If ints then scale and then round.
use header scanlen to only output good data
free the alloced data once at the end

Pulsar programs