Altering Leginon statistics calculation¶

• in dmsem.py, changed the mean and sd calulations to use 8* decimated for super res, and 4* decimated for counting (4096 x 4096)
• testing in IDLE on Gatan pc indicated that sd calculation takes ~ 1 sec for super res, fast for 4* decimated super res, and instant for 8 * decimated
• 4* decimated mean differed from full buy ~0.2%, 8* decimated differed by ~0.8%
• would make sense to do same decimation on gain corrected images (arraystats.py on leginon pc)
• implemented changes on Krios and Arctica (test on Arctica, implement partway through MJ session 21aug12b after 8 pm aug 13)
• is faster on both by ~1 s

NYU Krios measurements¶

• Krios prior: mean time between esn images is 11.17s, sd=0.43s (n=52)
• Krios after: mean time between esn images is 10.05s, sd=0.29s (n=60)

• approximate 10% savings in esn time
• change implemented in arraystats.py only affects esn images taken in manual and stat calculation for zlp
• arraystats.py does not have access to gui, so information about decimation is printed in terminal

NYU Arctica measuremnts¶

• test grid
• esn before:mean time = 11.05s, sd=0.51, n=58 (session 21aug13a)
• esn after: mean time = 9.96s, sd=0.38 (test session 21aug13b
• enn after: mean time = 7.62s, sd=0.5, n=21 (test session 21aug13b)
• enn after: mean time = 7.67s, sd=0.48, n=21
• again 10% improvement in super res speed