Intro:

     The fits files generated by the mock spectrometers had an error in the recorded positions for alfa's outer beams (1,2,3,4,5,6). The error has been there since the start of mock spectrometer data (nov08). The problem was reported by the galfacts group in nov09. The errors occur in the cimafits as well as the psrfits files.

    The outer beam positions are  computed by:

• The outer beams follow an ellipse with:
• semi-major axis (azlen0 = 329.06 asecs great circle).
• semi-minor axis (zalen0 = 384.005 asecs great circle).
• the beam angle (with 0 rotation) is then added to the alfa rotation angle. This angle is then used to project azlen0,zalen0 to the location of this feed.
• After the feed offsets are computed, they are added to the az , za of the central pixel.

The mock position error(s):

• The alfa rotation angle was being added to the beamAngleWith0DegRotation.
• It should have been subtracted because of the "creative" definition of the coordinate system.
• so rfeed_offaz, rfeed_offza is wrong in the fits header.
  
• When adding the beam offsets to the central pixel you need to go from great circle azimuth offset  to little circle azimuth.
• The program was using  az=   azbeam0 + rfeed_offaz*sin(za). It should have been az=azbeam0 + rfeed_offaz/sin(za) .. This was just a dumb typo.

Drift scans through continuum sources at meridian.

• drift scans using cima fixed azimuth drift routine. A 120 second drift was used.
  
• Position at the meridian.
• 3 sources were used:
  
• J2052+365 offset 20 secs of  ra from source. alfa rotation angle=60
  
• J2109+355 offset 30 secs of ra from source. alfa rotation angle=-60
  
• J2130+050 offset 30 secs of ra from source. alfa rotation angle=-60
• alfa rotation angles of +/- 60 were used so that the source would drift through:
• alfa rotation angle=+60: bm3,bm0,bm6
• alfa rotation angle=-60:bm1,bm0,bm4
  
• The mock spectrometer ran with a 172 MHz band centered at 1450 MHz
• The wapps ran with 100 MHz band centered at  1450 MHz.
• 1 second integrations were done.

• Page 1: total power vs sample
• The different beams total power is color coded.
• The wapp and mock data are over plotted. They line up pretty well except:
  
• The mock spectrometers are being started 4 seconds after the wapps.
  
• Top: J2052+365 alfa rotation = +60
• The source drifts through beams 0 and 6 (we missed beam3). This is as expected
• Middle: J2109+355 alfa rotation = -60
• Source drifts through beam 4,0,1 as expected
• Bottom: J2130+05 alfa rotation = -60
• Source drifts through beams 4,0,1 as expected.
• Pages 2-4 total power vs ra:
• The total power is plotted vs ra. 
  
• Wapp data is plotted with + , the mock data is plotted with a diamond.
• The different beams are color coded.
• The top frame has the data as read from the file
• the bottom frame corrects the ra positions from the mocks.
• Page2: J2052+365 alfa rotation=+60
  
• Page3: J2109+355 alfa rotation=-60
• Page4: J2130+050 alfa rotation=-60
• Once the rotation angle is reversed for the mock data, the source from the wapp and the mocks have the same ra.
  
• Page 5-7 Ra/dec offsets from central  pixel
  
• The top frame shows the wapp ra,dec offsets from beam0
• The bottom frame shows the mock ra,dec offsets from beam 0 (as read from the header).
• Page5: J2052+365 +60
  
• Page6: J2109+355 -60
  
• Page7:J2130+050  -60
• J2130+-50 ra/dec offsets differ from J2109+355 because the az has swung around by 180 degrees.
• Page 8 (wapp-mock) ra,dec positions after mock correction.
• This has the wapp-mock ra,dec difference after the mock correction.
• black in ra direction, red is dec direction.
• The difference is because the model offset is being evaluated a little differently in the two programs.
• Page 9: plot  differences for various header variables.
• The colors  black,red,green are the 3 sources at +60,-60,-60 rotation angles.
  
• Top: wapp - mock encoder_time. The mocks are interpolating the encoder_time (and positions) to crval5 while the mocks use the time as it is sent (usually +/- 20 milliseconds from the tick).
• 2nd/3rd: the (wapp-mock) model  offsets. They differ because the model offsets are evaluated a little differently in the two programs.
• frame 4: alfa beam offsets. This is the unrotated offsets for each  beam. They are identical.
• frame 5: alfa beam az offsets after rotation
• frame 5: alfa beam za offsets after rotation.
• The variables that have trouble are the rotate feed offsets in az,za.
  

When the bug was fixed:

• 17nov09: The sin(za) error was corrected.
• cimafits Version    2.00 -> 2.01
• psrfits hdrver keyword    3.4 -> 3.41
  
• All observations with alfa rotation angle=0 are correct after this date
• 28jan10: The alfa rotation angle bug was fixed:
• cimafits keyword Version 2.02 -> 2.03
• psrfits    hdrver keyword   3.42 -> 3.43
• Note ver 2.02 aofits, 3.42 psrfits are the same as 2.01, 3.41. They still have the rotation error.
  

What needs to be fixed in the old data:

• Positions for all outer beams. this includes
• crval2,3  glon,glat,  beam_offraj,beam_offdecj,map_offra/dec
  
• azimuth,elevatio,map_offaz/za
• rfeed_offaz/za
  
• par_angle
• Program to fix the positions.
• I've written a program: fixmockhdr_cimafits_09.c (in svn/pdev/pdev/fixhdr_09) to correct the aofits data.
• I haven't made an exportable version yet since it uses  a number of files from here at ao (dut1 file mainly).
  
• After running the program, it will update the version keywords so that the correction will not be tried twice (although it shouldn't matter).
  

Summary:

• Alfa outer beams have position errors.
• beam offset  was added to central beam az  via    azbeam0 + rfeed_offaz*sin(za).
  
• should have been azbeam0 + rfeed_offaz/sin(za)
• Alfa rotation angle was being added in backwards (only affected observations with alfa_ang != 0).
• Versions with fixes
  
• cima fits.
• VERSION = 2.01 , VER_DATE=2009-11-17
  
•  fixed 1/sin(za)  this fixes all data with alfa rotation angle=0
• VERSION = 2.03 , VER_DATE=2009-01-28
• This fixed the alfa rotation angle problem.
• The VERSION, VER_DATE keywords are in the bintable header
  
• psrfits
• HDRVER = 3.41
• fixed 1/sin(za)  this fixes all data with alfa rotation angle=0
• HDRVER = 3.43
• This fixed the alfa rotation angle problem.
• the HDRVER keyword is in the main header.
• The mocks start 4 seconds after the wapps. The time stamps in the headers are correct.