From wbrisken@aoc.nrao.edu Wed Mar 10 16:12:10 2004 Date: Wed, 10 Mar 2004 15:08:30 -0700 (MST) From: Walter Brisken To: aips2-naug@zia.aoc.nrao.edu Subject: [Aips2-naug] Test report NAUG testing report imager, viewer, imagefitter Walter Brisken, Mar 10, 2004 This note reports on aips++ testing of multi-spectral window imaging and some aspects of the viewer. Note that these tests are not meant to be exhaustive, but attempt to excercise a data path used to extract coordinates from an astrometry data set. The functionality, usability and documentation quality for the tested features are listed below. Note that these are in the context of the data set I used and may not apply to all data sets. The standard grades [A, A/E, I, N, U] are given. ************* Executive Summary ****************** 1. Imaging with multiple spectral windows: functionality: A usability: A documentation: A Basically, it just worked. supplying, as per the documentation, a selection of spectral windows to image to both setimage and setdata allowed imaging of the desired spectral windows. Comparison with an image made with classic AIPS shows no perceivable differences for a simple test case. 2. Viewer blinking: functionality: A/E usability: A documentation: A/E Blinking is the second of two animation modes, which is selected in a very intuitive way. I tested blinking on two images : one made with AIPS IMAGR and the other with aips++ imager. The imaging was made as identical as possible. The biggest difference between the two was a 1 pixel offset in the coordinates. When blinking, the image pixels stayed aligned which meant that the image (and coordinate system) shifted back and forth by 1 pixel with each blink. Although the two images were registered, a given world coordinate blinks to a different position on the screen. Documentation is fine, but it would be nice to have a link from the top-level viewer URM page to the more practical hands-on documentation. Will discuss with David King when he returns. 3. Imagefitter: functionality: A/E usability: A documentation: A/E The data sets used to test the above were natural data for tests of Imagefitter. A direct comparison with AIPS JMFIT was performed. I have lots of experience with JMFIT and more-or-less agree with the positional uncertainty that is reported. Imagefitter produces positional uncertainties a factor of ~20 smaller than JMFIT. I submitted a bug report to this affect before (AOCso04226), with a "non-reproducable" resolution. I believe the problem is real, thus the score of A/E. A detailed bug report will be submitted in the future. For simple 1-component fits, the interface is very nice. For more complicated fits, I need to review the documentation. Documentation: There is a do-nothing link on the page (Link to Viewer in the Synopsis section). Documentation is generally pretty good for this tool. ************* Test details ***************** parallax<51>% source /aips++/weekly/aipsinit.csh /home/parallax2/big/BC120DLib1 AIPS++ Version 1.9 Build 461 Use calibrated 4-IF 2-Pol gated pulsar data from VLBA: include 'ms.g' mset := fitstoms(fitsfile='IB1.SPL2',msfile='ib1.ms') mset.summary() NORMAL: MeasurementSet Name: ib1.ms MS Version 2 Observer: BC120 Project: Observation: VLBA(10 antennas) NORMAL: Data records: 320100 Total integration time = 6863.58 seconds Observed from 12-Feb-2004/17:47:54 to 12-Feb-2004/19:42:17 NORMAL: Fields: 1 ID Name Right Ascension Declination Epoch 1 B2053+36 20:55:12.23 +36.34.22.69 J2000 NORMAL: Data descriptions: 4 (4 spectral windows and 1 polarization setups) ID Ref.Freq #Chans Resolution TotalBW Correlations 1 1413.97MHz 1 7500 kHz 7500 kHz RR LL 2 1507.97MHz 1 7500 kHz 7500 kHz RR LL 3 1593.97MHz 1 7500 kHz 7500 kHz RR LL 4 1683.97MHz 1 7500 kHz 7500 kHz RR LL NORMAL: Antennas: 10 ID= 1-10: 1=BR, 2=FD, 3=HN, 4=KP, 5=LA, 6=MK, 7=NL, 8=OV, 9=PT, 10=SC NORMAL: Tables(rows): (-1 = table absent) MAIN(320100) ANTENNA(10) DATA_DESCRIPTION(4) DOPPLER(-1) FEED(10) FIELD(1) FLAG_CMD(0) FREQ_OFFSET(-1) HISTORY(556) OBSERVATION(1) POINTING(10) POLARIZATION(1) PROCESSOR(0) SOURCE(0) SPECTRAL_WINDOW(4) STATE(0) SYSCAL(-1) WEATHER(-1) NORMAL: mset.done() Attempt to image IFs 1 and 4: include 'imager.g' im:=imager(filename='ib1.ms', compress=F, host='', forcenewserver=T) im.setdata(fieldid=1, spwid=[1,4]) NORMAL: By selection 320100 rows are reduced to 160050 (as expected) im.setimage(nx=512, ny=512, cellx=[value=0.0012, unit="arcsec"], celly=[value=0.0012,unit="arcsec"], stokes="I" , mode="mfs" , fieldid=1, spwid=[1,4]) im.weight(type="briggs", rmode='norm', robust=0) im.clean(algorithm='clark', niter=50, gain=0.1, threshold="0.0 Jy", displayprogress=F, model='ib1.cln',fixed=F,complist='',mask='',image='ib1.im', residual='ib1.resid',interactive=F,npercycle=50) This message appears: WARN: Did not get the position of VLBA from data repository WARN: Please do inform aips++ to put in the repository WARN: Frequency conversion will not work -> Defect reported (AOCso04758) v := image('ib1.im') v.view() The aips and aips++ images with the same data cuts give rise to the same source position and flux density and very similar noise patterns in the residuals. Question : how does the robust value in aips and aips++ compare? The input value for aips ranges from -5 to 5 and for aips++ is between -2 and 2. I used 0 for both aips and aips++ to simplify the comparison. Now blink the aips image and aips++ image Wow, the images are almost identical. I used a clean box in AIPS, but not in aips++, which probably makes most of the difference. Blinking at 6 fps. One image seems to stay up for a slightly longer amount of time than the other. Two images with slightly different coordinate offsets appear to rock back and forth, even though they are presumably registered. Gaussian fit results: 1. AIPS image with JMFIT: parall> JMFIT1: Peak intensity = 6.9405E-03 +/- 4.40E-04 JY/BEAM parall> JMFIT1: Integral intensity= 9.1416E-03 +/- 9.20E-04 JANSKYS parall> JMFIT1: X-position = 243.182 +/- 0.1092 pixels parall> JMFIT1: Y-position = 256.980 +/- 0.2748 pixels parall> JMFIT1: RA 20 55 12.2286668 +/- 0.000010877 parall> JMFIT1: DEC 36 34 22.692876 +/- 0.00032973 2. AIPS image with imagefitter: Peak Flux : [0.00697997 0 0 0] ([4.82791e-05 0 0 0] ) Jy/beam Int Flux : [0.0090682 0 0 0] ([6.27229e-05 0 0 0] ) Jy RA/DEC : 20:55:12.2287 +036.34.22.693 (3.51901e-05 arcsec 1.4154e-05 arcsec) [243.174 256.976 1 1] absolute pix 3. aips++ image with imagefitter: Peak Flux : [0.00681704 0 0 0] ([4.7126e-05 0 0 0] ) Jy/beam Int Flux : [0.00873858 0 0 0] ([6.04095e-05 0 0 0] ) Jy RA/DEC : 20:55:12.2287 +036.34.22.693 (3.33216e-05 arcsec 1.40269e-05 arcsec) [244.12 256.846 1 1] absolute pix Excellent agreement on fit values, but the uncertainties don't match. They differ by a factor of 20 for Dec! I will file a more detailed bug report when I have time to properly investigate this. Note that for precision astrometry, it will be essential for more digits to be presented in the imagefitter coordinates. This defect was filed before (AOCso02865) and marked as resolved (by adding more precision), but I believe the additional digits were added to the flux fields, not the spatial location fields. New defect filed: AOCso04757. _______________________________________________ Aips2-naug mailing list Aips2-naug@listmgr.cv.nrao.edu http://listmgr.cv.nrao.edu/mailman/listinfo/aips2-naug