[daip] IMAGR problem with MeerKAT data {External}

Marc Verheijen verheyen at astro.rug.nl
Mon Mar 28 09:03:04 EDT 2022


Dear daip, dear Eric,

I am trying to reduce some MeerKAT L-band spectral line data with AIPS
to have some fiducial results to be compared with other (unexpected)
software outcomes.  For this purpose, I have upgraded to the 31DEC22 TST
version of AIPS to replace the older 31DEC17 version. 

The UVFITS file (created with casa using exportuvfits) is 224 GB in
size, comprising 2016 channels:

AIPS 1: Image=Shyama    (UV)         Filename=SHY-A6      .EXTCAL.   1
AIPS 1: Telescope=MeerKAT            Receiver=MeerKAT
AIPS 1: Observer=Tirna De            User #= 2626
AIPS 1: Observ. date=22-SEP-2021     Map date=26-MAR-2022
AIPS 1: # visibilities   3667381     Sort order  TB
AIPS 1: Rand axes: UU-L  VV-L  WW-L  TIME1  SUBARRAY  INTTIM  ANTENNA1
AIPS 1:            ANTENNA2  SOURCE
AIPS 1: ----------------------------------------------------------------
AIPS 1: Type    Pixels   Coord value     at Pixel     Coord incr   Rotat
AIPS 1: COMPLEX      3   1.0000000E+00       1.00  1.0000000E+00    0.00
AIPS 1: STOKES       2  -5.0000000E+00       1.00 -1.0000000E+00    0.00
AIPS 1: FREQ      2016   1.3265022E+09    1009.00  2.6123047E+04    0.00
AIPS 1: IF           1   1.0000000E+00       1.00  1.0000000E+00    0.00
AIPS 1: RA           1    23 37 56.800       1.00       3600.000    0.00
AIPS 1: DEC          1    20 26 26.300       1.00       3600.000    0.00
AIPS 1: ----------------------------------------------------------------
AIPS 1: Coordinate equinox 2000.00
AIPS 1: Rest freq      0.000         Vel type: OPTICAL wrt YOU
AIPS 1: Alt ref. value  0.00000E+00  wrt pixel 1009.00
AIPS 1: Cals applied:    1 DOCAL   1 DOBAND   0 DOPOL
AIPS 1: Maximum version number of extension files of type NX is   1
AIPS 1: Maximum version number of extension files of type FQ is   1
AIPS 1: Maximum version number of extension files of type HI is   1
AIPS 1: Maximum version number of extension files of type AN is   1
AIPS 1: Maximum version number of extension files of type CL is   1
AIPS 1: Maximum version number of extension files of type SU is   1

After a minimal external calibration, I ran IMAGR on channels 1000-1010
without any cleaning.  It seems to work well initially but it dies of
'unnatural causes' when working on the second channel:

IMAGR1: Task IMAGR  (release of 31DEC22) begins
IMAGR1: Using temporary OBOXFILE = 'HOME:OBOXtemp20220327.123824'
IMAGR1: UVGET: doing no flagging this time
IMAGR1: Not calibrating  gains  bandpass  polarization
IMAGR1: Create SHY-A6      .IMAGR .   1 (UV)  on disk 48  cno    1
IMAGR1: Beginning channel 1000 through 1000 with  1 IFs
IMAGR1: IMACPY: Copied  3667381 visibilities to be imaged
IMAGR1: QINIT: did a GET  of      5120 Kwords, OFF   17521741164347
IMAGR1: UVWAIT: begin finding uniform weights
IMAGR1: UVWAIT: Average grid weight  3.767E+03
IMAGR1: UVWAIT: Adding temperance S  7.534E+02
IMAGR1: UVWAIT: begin applying uniform or other weights
IMAGR1: UVWAIT: Sum of weights in  9.993E+06 and out  2.471E+07
IMAGR1: UVWAIT: Noise is increased by a factor 1.255 due to weighting
IMAGR1: UVWAIT: Average summed weight  3.767E+03 over    3667381. vis
IMAGR1: Create SHY-A6      .IBM001.   1 (MA)  on disk  1  cno    5
IMAGR1: Create SHY-A6      .IIM001.   1 (MA)  on disk  1  cno    6
IMAGR1: GRDFLT: X and Y convolution type = SPHEROIDAL
IMAGR1: GRDFLT: X and Y parms =  3.0000  1.0000
IMAGR1: GRDFLT: convolution function sampled every 1/100 of a cell
IMAGR1: GRDMEM: Frequency 1.326267E+09 Hz
IMAGR1: Field    1 Sum of gridding weights = 2.623673E+08
IMAGR1: Field    1 Beam min =  -50.5 MilliJy, max =    1.0      Jy
IMAGR1: Fit Gaussian FWHM =   17.549 x    9.210      arcsec, PA=   -1.1
IMAGR1: CLBHIS: minimum component 0.304 of current peak
IMAGR1: Field    1 min =   -3.0 MilliJy,max =   39.2 MilliJy
IMAGR1: No Cleaning requested
IMAGR1: Peak residual in window  39.231 less than limit 105.000 MilliJy
IMAGR1: Beginning channel 1001 through 1001 with  1 IFs
IMAGR1: UVGET: doing no flagging this time
IMAGR1: UVWAIT: begin finding uniform weights
IMAGR1: UVWAIT: Average grid weight  3.767E+03
IMAGR1: UVWAIT: Adding temperance S  7.534E+02
IMAGR1: UVWAIT: begin applying uniform or other weights
IMAGR1: UVWAIT: Sum of weights in  9.992E+06 and out  2.471E+07
IMAGR1: UVWAIT: Noise is increased by a factor 1.255 due to weighting
IMAGR1: UVWAIT: Average summed weight  3.767E+03 over    3667381. vis
IMAGR1: GRDMEM: Frequency 1.326293E+09 Hz
IMAGR1: Field    1 Sum of gridding weights = 2.623580E+08
IMAGR1: ZMI2: DELAY 1
IMAGR1: ZMI2: DELAY 2
IMAGR1: ZMI2: REQUEST FOR BYTES 4194305 THRU = 6582272
IMAGR1: ZMI2: BEYOND EOF = 4194304
IMAGR1: ZMIO: OPER=WRIT LUN= 18 BLKNO=      4097 8-BIT-BYTES=   2387968
IMAGR1: ZERROR: IN ZMI2 ERRNO = 22 (Invalid argument)
IMAGR1: GRCMEM: WRITE ERROR  3 ROW   584
IMAGR1: OUVIMG: MAKING BEAM BEAM field number 0001
IMAGR1: OUVIMG: FROM UVdata work object
IMAGR1: CLNINI: ERROR OPENING CLEAN process object
IMAGR1: CLNUV : ERROR CLEANING CLEAN process object
IMAGR1: Deleting UV work file:
IMAGR1: Destroyed  1 extension files of type AN
IMAGR1: Destroyed  1 extension files of type FQ
IMAGR1: Destroyed UV image file: catno=      1 disk=48
IMAGR1: Purports to die of UNNATURAL causes
IMAGR1: cuby 31DEC22 TST: Cpu=    137.3  Real=    171  IO=    340613
Note: The following floating-point exceptions are signalling: IEEE_UNDERFLOW_FLAG IEEE_DENORMAL

This behaviour is reproduceable.

I'm running AIPS on a Linux machine with 256 GB of RAM and a 132 TB
RAID-6 file system.  Do you have a suggestion for where to look for a
solution to this problem, or for some further tests I could do to
isolate the cause of this problem?

Many thanks for your consideration!

 Kind regards,

 Marc


-- 
*************************************************************************
* Prof.dr. Marc A.W. Verheijen                                          *
* University of Groningen                                               *
* Kapteyn Astronomical Institute                                        *
* P.O. box 800                                                          *
* 9700 AV  Groningen                Phone : +31-50-363-4077             *
* The Netherlands                   Email : verheyen at astro.rug.nl       *
*************************************************************************



More information about the Daip mailing list