Current Server Time: 23:40 (Central Europe)

#apertus IRC Channel Logs

2022/01/15

Timezone: UTC


01:56
aombk
joined the channel
02:30
aombk2
joined the channel
02:33
aombk
left the channel
02:37
aombk
joined the channel
02:40
aombk2
left the channel
03:34
Bertl_oO
off to bed now ... have a good one everyone!
03:34
Bertl_oO
changed nick to: Bertl_zZ
04:06
aombk
left the channel
06:53
aombk
joined the channel
12:52
Bertl_zZ
changed nick to: Bertl_oO
14:30
aombk
left the channel
14:33
aombk
joined the channel
15:51
vup
se6astian: nice
15:51
vup
whoops: MemoryError: Unable to allocate 63.2 GiB for an array with shape (1023, 3840, 2160) and data type float64
15:58
anuejn
:see_no_evil:
15:59
vup
the one numpy improvement I need is actually already implemented, but for some reason it stalled in 2019: https://github.com/numpy/numpy/pull/13263 :(
16:04
anuejn
sad :(
17:20
vup
well adhoc numba snippts will need to do the job for now
17:32
se6astian
oh no :P
17:42
vup
ok
17:42
vup
so this is a per pixel mean over the 1023 darkframes se6astian captured, with the mean on the red channel and the standard deviation on the blue channel: https://files.niemo.de/darkframe_pdfs/per_pixel_mean_std.png
17:43
vup
and this is that image with every two consecutive rows side by side: https://files.niemo.de/darkframe_pdfs/per_pixel_mean_std_wide.png
17:44
vup
(the colors are normalized such that the mean over the whole image minus 3 times the standard deviation is zero, and the mean plus thre times the standard deviation is one)
17:45
vup
also when looking at the column wise means again with the two rows side by side, it seems like the pattern i observed in steps of 32 columns is actually a pattern in steps of 128
17:45
vup
which would matchup exactly with the readout sequence
17:45
vup
https://files.niemo.de/darkframe_pdfs/column_wise_averages.png
17:46
se6astian
very interesting!
17:46
vup
Bertl_oO: ^
17:55
vup
also the fft of the residual column averages (substracting the per pixel mean from each darkframe, then averaging over the columns, then fft, then average over all darkframes) is pretty interesting: https://files.niemo.de/darkframe_pdfs/column_residual_noise_fft.pdf
17:55
vup
(again putting two consecutive rows side by side)
17:57
vup
compare this to the same, but doing rows instead of columns: https://files.niemo.de/darkframe_pdfs/row_residual_noise_fft.pdf
18:01
vup
se6astian: how hard would it be to capture a full width sequence ie 2048, again maybe a stack of 1024?
18:01
se6astian
not hard at all, just requires different configuration
18:01
vup
nice, I think that wolud be pretty interesting
18:01
vup
especially with enabled dark columns
18:01
se6astian
I am in linz currently though on mobile internet, but will do on Monday
18:02
vup
sure no hurry
18:02
se6astian
noted in calendar
18:03
vup
also from what I can tell, the new settings did not really make any difference in the column noise
18:03
vup
nice
18:03
se6astian
right
19:41
Bertl_oO
vup: yeah, I suspected that there is some read out effect ... glad that we could link this