Current Server Time: 18:41 (Central Europe)

#apertus IRC Channel Logs

2014/03/09

Timezone: UTC


23:09
intracube
left the channel
00:21
jerknextdoor
joined the channel
00:32
jerknextdoor
left the channel
01:36
ThatCantBe
left the channel
04:20
Bertl
off to bed now ... have a good one everyone!
04:22
troy_s
left the channel
04:25
troy_s
joined the channel
04:41
se6astian
joined the channel
04:41
se6astian
good early morning
08:51
jucar
joined the channel
09:32
jucar
left the channel
10:08
sebix
joined the channel
10:10
sebix
left the channel
10:11
niemand
joined the channel
10:17
niemand
Hi all! Have you thought of presenting the project at the linuxwochen vienna in may? http://linuxwochen.at/
10:19
sasha_c
joined the channel
10:21
Bertl
morning folks!
10:21
Bertl
niemand: not really, but it might make sense, thanks for the suggestion
10:26
niemand
it does perfectly fit to the focused topics (open software and hardware), you can reach a quite big local audience. I would really appreciate if you could show up there :)
10:36
se6astian
hi niemand, I presented the project at the linux wochen 2 years ago if I remember correctly
10:36
se6astian
would be a good idea to do an update
10:37
se6astian
and at the linux wochen linz 3 years ago
10:57
danhanes
joined the channel
11:15
Bertl
niemand: do you have a good contact to the event orginization? Might simplify the procedure for us.
11:16
niemand
yes, I do :)
11:17
Bertl
excellent then!
11:17
niemand
that means i'm not in the official core team. But i'm helping them where possible and i meet them regularly, next time on next friday
11:18
Bertl
btw, I see you know tuxlover? :)
11:21
niemand
hm, no I don't?
11:22
niemand
afk for making lunch
11:23
Bertl
ah, probably just coincidence then, no problem
11:47
sasha_c
left the channel
12:17
intracube
joined the channel
12:19
rainer_vie
joined the channel
12:57
rainer_vie
left the channel
13:05
rainer_vie
joined the channel
13:06
rainer_vie
left the channel
13:07
rainer_
joined the channel
13:27
gcolburn
joined the channel
13:42
rainer_vie
joined the channel
13:43
rainer_vie
left the channel
13:43
rainer_
left the channel
13:44
rainer_vie_
joined the channel
14:27
danhanes
left the channel
15:07
gcolburn
left the channel
15:33
troy_s
Makes not much sense se6astian
15:33
troy_s
The chart and software will deliver the best values. The issue is the data.
15:34
troy_s
So fudging a grade is _no_ different, aside from the fact that it is a worse approximation.
15:34
Bertl
hmm? context? :)
15:35
troy_s
Again, the issue is the data, not the matrix. This can be confirmed by looking at the 3D LUTs Bertl and I futzed with the other day.
15:35
troy_s
The recent list post.
15:35
troy_s
Anyways... kind of goofy. Better is to simply sort out the data.
15:37
Bertl
ah, okay, well, I guess the idea is to be able to get some 'proper' footage to prepare for funding
15:38
troy_s
Then solve the data
15:38
troy_s
The issue is the data, and no grader can fix that
15:38
Bertl
working on it, but I think it will take a while till we can get a perfect profile
15:38
troy_s
The LUT / matrix result will _always_ skew where the chart is struggling which is again, _data_
15:38
troy_s
I think your clipscale is not a bad idea.
15:39
Bertl
well, it can be done already, even without the planned LUT before the matrix
15:40
Bertl
by creating the matrix only for the 'linear' range and then clipping/scaling the result via gamma LUT
15:51
rainer_vie_
for later on if the HDR function is enabled, it would be cool to have a logarithmic LUT as well
15:52
Bertl
not sure that would make any sense
15:52
Bertl
we get 12bit max from the sensor, we have a 12bit LUT (12->16bit)
15:53
Bertl
you can already do any possible mapping with 1/16th the resolution
15:53
rainer_vie_
ah yes forgot that its made on the sensor... so the 15 stops will be mapped into the 12 bits there with the kneepoints
15:53
Bertl
precisely
15:53
rainer_vie_
got you
16:23
jucar
joined the channel
16:36
troy_s
rainer_vie_: Log only makes sense in bandwidth challenged contexts.
16:37
troy_s
rainer_vie_: And there technically aren't any knees. It is a linear diagonal until you hit either A) noise floor or B) random data ceiling (see discussions for issues with that upper range)
16:38
troy_s
Bertl: Totally agree - just use the linear segment.
16:38
troy_s
Bertl: Anyways... not fussed, just seemed silly as hell.
16:38
troy_s
(and a gross misunderstanding of what color is and why you need the chart version)
16:41
troy_s
(getting a "decent" looking image has little to do with the validity of a "decent" 709 transform. Anyone can take abstract raw junk data into a compositor and get to someplace different.)
16:42
Bertl
rainer_vie_ was talking about the HDR modes
16:42
rainer_vie_
so the final raw data will have 16bits? because of 12>16bit LUT?
16:43
Bertl
the final data has 16bit, but is then converted to HDMI atm
16:46
rainer_vie_
ok fine... only thinking on filesizes and bandwidth. makes it sense to convert to 16bit when the sensor delivers only 12 regarding writing it to SSD later on?
16:47
Bertl
I think the 12bits aren't even real on the Cmosis sensor :)
16:48
Bertl
so I doubt we will store all the 16bit generated atm, but it keeps the error low, i.e. we might even add a few bits during conversion
16:50
troy_s
Bertl: Whatever ends up HDR mode, keep close track of the TRC for inversion into compositing apps etc.
16:58
se6astian
time to leave
16:58
se6astian
bye
16:58
se6astian
left the channel
18:00
rainer_vie_
going off as well... cu
18:01
Bertl
cya
18:01
rainer_vie_
left the channel
18:07
trakstar
joined the channel
18:36
niemand
left the channel
18:52
trakstar
left the channel
18:55
niemand
joined the channel
18:58
fsteinel_
joined the channel
19:01
stavad
joined the channel
19:07
fsteinel_
left the channel
19:55
intracube
left the channel
20:33
niemand
left the channel
20:48
Bertl
off for a nap ... bbl
22:45
gcolburn
joined the channel
22:45
gcolburn
left the channel