Changes between Version 61 and Version 62 of MD10.refstack.20130715

Timestamp:

Sep 10, 2013, 8:43:47 PM (13 years ago)

Author:

Mark Huber

Comment:

--

MD10.refstack.20130715

@@ -305 +305 @@
 
 '''PROBLEM -- new 1D convolution code produced unexpected excess rejection and cannot be used for the refstack''' -- investigate further below
+
+
 == 1D Excess Rejection ==
 
@@ -335 +337 @@
 == Odd Features in 1D version ==
 
-|| input warp || 1D conv stack || 1D conv mask || unconv .num. || 1D .num. || unconv .exp. || 1D .exp ||
+Odd streaks seem more common in 1D convolved stacking when should be plenty of inputs to reject pixels.
+
+|| unconv stack || 1D conv stack || 1D conv mask || unconv .num. || 1D .num. || unconv .exp. || 1D .exp ||
 
 [[Image(oddpoorstreak_nanexp.png,500px)]]
@@ -341 +345 @@
 Appears that the .exp. files for the >30 input stacks are mostly NAN. Suspect the compression had a problem. Has been happening for over a year. Added a .EXP.NOCOMP option (one already existed for EXPWT) and if a stack is made uncompressed (as is with deep stacks) then this is used.
  * filesize change 933 KB -->163 MB, a large change without compression, but only for deep stacks (MD ref and deep types only) -- 
-== 1D convolution v3 OK ==
+
+
+== 1D convolution v3 Rejection OK ==
 
 r35925 appears to produce similar rejection to the original ISIS method, and includes a fix the the other rejection issue found in the night stack sample below. Target PSF for convolved stack is now set to the max FWHM of the input warps that pass all input cuts with an additional epsilon of 0.1 pixels.
@@ -383 +389 @@
 
 ----
+
 = Night Stacks and PV2 =
 Configuration modifications for PV2 and 1D convolution should probably be applied to nightly stacks as well as other additions