Cataloging R&D


1. Border size

I cataloged night 20050401 with border=100, border=50 and border=20 . It was run in heplx48:/opt/pi/dev/pisys/daq/ndir/data/20050401/.

DATABASE
#STARS
#MEASUREMENTS
<sigma_mag>
CONCLUSIONS
20050401_100
95567
2086221
0.1195

20050401_50 101593
2233062
0.1189
rather not worse then border=100 , just more stars due to bigger FOV used
20050401_20 109141
2327833
0.1229
rather not worse then border=100 , just more stars due to bigger FOV used

Wnioski : wiecej miejsca bedzie zajmowac niz do tej pory o jakies 12% . It seems that just new stars apear and those which are already with
bigger border do not change very much their magnitude error, but I will write a script to check distribution of sigma_20-sigma_100 where
I will plot difference of magnitude error after border size is decreased from 100 -> 20 for stars existing in both databases
I created script comparing stars in to databases run :
       cd heplx48:/opt/pi/dev/pisys/daq/ndir/data/20050401/anal
       compare_stars.pl > out_100_to_20 &

gave result that for 12678 stars in database 20050401_100 only 85 had worse sigma_mag in database 20050401_20 ( with border=20 pixels )
so in fact there is no visible decrease of quality for stars in center when border is smaller.
I decided that it can be border=20 as final setting and I will re-catalog 2004_2005 on heplx48 ,but the fact is that it requires more disk space ...

2. Median and quarter difference ( 3/4-1/4) instead of mean and sigma

Comparison of sigma_mag and magnitude calculated as mean and sigma , calculated as meadian and 3/4-1/4
Small field comparison mean and sigma, median and 3/4-1/4 with any number of measurements

Small field comparison mean and sigma, median and 3/4-1/4 with no_measurements>=80
In smaller fields : mean vs median
Sigma calculated as (3/4) - (1/4) is has a little bit higher values , but it seems to be better at least according to marek - I think good solution will be
to use spare fields sigma_field and tmp_value as new sigma and median magnitude