1. History
   a) KATpoint delay model :  https://github.com/ska-sa/katpoint 
       David Terret, Jun 6th 2019

   b) KATPoint details of antenna/delay input parameters and 
       executing Katpoint
      Ludwig Schwardt, 24th Jul 2019.

   c) calc11 code :  The ALMA version of Calc publicly available at
	https://bitbucket.sco.alma.cl/projects/ASW/repos/control/
                browse/thirdPartyPackages/Calc-11
        (Benjamin Mort, David Gorden, Ralph Morson, 29th Jul 2019)

   d) calc11 wrapper added (producing NULL delay values), code+wrapper on SKA gitbub at
       https://github.com/ska-telescope/delay-testing, Mark Ashdown ( 2nd Aug)

   e) Checked calc11 code, fixed to prodcuce some reasonable delay numbers for one
      instant of randomly chosen time, and dummy parameters hardcoded.
      (Sanjay Kudale, 6th Aug)

   f) Continued to modify calc/Katpoint/GMRT to run for many antennas and 
         longer time duration, details below.
      (Sanjay Kudale, from 6th to 28th Aug, 2019)

2. Comparison between different delay models.

  A. Comparison between GMRT & Calc11 is done, steps and main points to note.

    i.      Antenna positions of GMRT anntenna are taken as starting point.

    ii.     C02 antenna of GMRT taken as reference antenna.

    iii.    MJD calculated for 24 hours with 60 second resolution, time 
		input to calc11 model.

    iv.     Leap second correction not enabled for calc11. Current leap 
		second = 37 can be enabled for next level of comparison.

    v.      GMRT antenna coodinates are astronomical coordinates (XYZ-) GPS 
		measured and astronomically confirmed and corrected. Geocentric 
		ITRF coordinates of GMRT antennas are used in calc11

    vi.     calc code modified to run for all GMRT antennas for range of 24 hours.

    vii.    GMRT delay model code is not yet separated, it is integral part 
		of correlator code. Hence, delay model generated during actual 
		observation was extracted from observation data/log.  Note 
		gmrt delay model is generated at every 0.671 sec.

    viii.   For two separate days (16th Aug, 21st Aug) delay model is 
		extracted for comparison.

    ix.     With proper time adjustment, UTC vs IST (GMRT time for delay 
		calculation) compared calc & GMRT delay models. Dealy models 
		compared on 16h Aug, for 5 hours duration and on 21st Aug for 3.8 hours.

      Results of comparison:

      i.    Reasonable match between calc11 & GMRT for whole range of times 
		mentioned for both days.

      ii.   An average offset about 1 nano-second on shortest baseline and 
		of about 100 nano-second on longest baseline is seen (checked 
		at some sample time). However these numbers are approximate.
            
      iii.  More precise comparison is to be done by using more appropriate 
		inputs of time, antenna coordinates (e.g. geocentric coordinates 
		were used to calc, where as astronomically corrected coordinates 
		were used for GMRT). Besides, fixed delay used in observation 
		need to be corrected carefully, although I have made quick 
		correction for it. For calc11 fixed delay correction is not done.


  B. Comparison between GMRT & KATpoint models is in progress. Steps and main points to notes

   i.    All GMRT related parameters same as mentioned in case of calc11 comparison.

   ii.   KATPoint delay model : used GMRT antenna coordinates in ENU format.

   iii.  KATPoint : All fixed delay values made zero.

   iv.   KATPoint : Extra-delay to be added in each signal path made explicitly zero.

   v.    KATPoint : Generated delay model for 24 hours, with 60 second resolution. 
		Same time range used as in calc11 case.

   vi.   Used GMRT delay model from observation (as mentioned in calc11 case) 
		exactly the same model.

   Results of comparison :
   i.    Delay change across time is showing expected variation.

   ii.   For some reason, yet to be find out, there is large offset 
	   in time that it need to be shifted (either in GMRT model or 
	   KAT model) to get match between two delay models. With 
	   this offset in time adjusted, the delay models shows desired match
           with each others, however, not as accurate as in GMRT-calc11 case.

   iii.  An average difference between KAT and GMRT models is about 
	   10 nano-seconds on shortest baselines,  and grows to about 
	   350 nano-seconds for longer baselines.


3. Future steps :

  A. Basic code functionality check is done, Carry out robust 
	comparison between KATpoint and GMRT models.

  B. To understand and/or nail down the differences in various models.

  C. Coding for calc11 simiar to Katpoint library.