Current Server Time: 23:24 (Central Europe)

#apertus IRC Channel Logs

2013/11/09

Timezone: UTC


22:45
Bertl
off for a nap ... bbl
23:23
Sasha_C
joined the channel
23:28
Sasha_C
left the channel
23:36
rexbron_
Bertl, any changes?
23:36
rexbron_
:)
00:46
Bertl
back now ...
00:47
Bertl
rexbron_: had a nap :) well, I got HDMI data output (not sensor data but memory data) almost finished, there seems to be a minor bug somewhere which I have to debug
00:48
rexbron_
Does the hardware support hdsdi currently?
00:49
Bertl
nope, it is a zedboard
00:50
rexbron_
But does it have 2 free pins? You could bit bang it ;-)
00:51
Bertl
yes, that I could do, even with up to 500MHz :)
00:51
rexbron_
Ah lol. So 6ghz hdsdi isn't in the cards?
00:52
Bertl
http://www.zedboard.org/sites/default/files/Avnet%20ZedBoard%20Brochure%20English%20Version.pdf
00:52
Bertl
here that you have an idea
00:56
rexbron_
Cool
01:08
Bertl
the gigabit ethernet is basically the fastest existing interface
01:09
Bertl
and it currently can do about 450-500Mbit/s (probably a driver issue)
01:10
Bertl
and the XC7Z020 doesn't have any gigabit tranceivers (from the FPGA side)
01:11
Bertl
the XC7Z030 ist the first in the zynq series to feature high speed serial ports
01:13
Bertl
so all a little limited, but the price of the ZedBoard is reasonable and it is well documented
06:13
Bertl
off to bed now ... have a good one everyone!
06:33
Sasha_C
joined the channel
07:36
se6astian
joined the channel
07:36
se6astian
morning
07:42
dmj_nova
morning se6astian
07:57
se6astian
anyone ever bought or sold bitcoints before?
07:57
dmj_nova
se6astian: nope
08:16
Sasha_C
left the channel
08:23
dmj_nova
se6astian: what changed on the meeting?
08:23
se6astian
points added to agenda
08:24
se6astian
and I converted the time to more timezones
08:24
se6astian
as only a handful of people accepted the appointment so far
10:05
se6astian
time to go to the supermarket
12:02
se6astian
back
12:45
Bertl
morning everyone!
12:57
se6astian
hello
13:10
se6astian
just talking to konstantin
13:10
se6astian
he will send me source files of the lens mount (privately) :)
13:12
Bertl
good :)
13:13
se6astian
I have it :)
13:14
Bertl
\o/
13:14
se6astian
and I gained some insight into optical sensor design/manufacturing
13:15
se6astian
its a bit tricky to do an "open source" design
13:15
se6astian
or more tricky to call it that
13:15
se6astian
because the software to design such sensors is EXTREMELY expensive
13:15
se6astian
so sharing files without access to the software is kind of pointless
13:16
se6astian
even if you could call sharing that files "open source"
13:17
se6astian
and the files contain like a "picture" of each silicon layer
13:17
se6astian
where you could easily have 20 layers or more
13:17
se6astian
to manufacture you need to create a stencil mask for each layer
13:17
se6astian
which costs around 50.000
13:17
se6astian
konstantin didnt remember if that was per layer or for a certain amount of layers
13:18
se6astian
but there are many very expensive steps in such a process
13:18
se6astian
so even if you have these files under and open license
13:18
se6astian
it will cost you a shitload to use them
13:19
se6astian
so for me the question is what exactly could be open with a sensor design
13:19
se6astian
it could be documentation
13:19
se6astian
or interface definitions
13:20
se6astian
also the final files that contain the layers are somewhat optimized for the machine they are used for already
13:21
se6astian
so it could be seen like a binary for a specific architecture if we translate that into the software world
13:22
se6astian
what do you think?
13:22
se6astian
does that explanation make sense?
13:25
Bertl
explanation makes sense and confirms what I basically assumed ... i.e. an open hardware sensor doesn't make that much sense :)
13:27
troy_s_
Greets all.
13:28
Bertl
hey troy_s_!
13:28
troy_s_
How are you Bertl?
13:29
Bertl
fine so far, thanks for asking! and you?
13:30
troy_s_
Blah. Trying to kill a weekend sitting in a mall's coffee shop.
13:30
Bertl
why not work on axiom then? :)
13:32
troy_s_
Work is a relative thing. ;)
13:32
troy_s_
Bertl: How goes engineering land?
13:33
Bertl
fine so far, I didn't expect this phase to be ultra fast (it's a lot of boring compiling and testing)
13:34
troy_s_
Bertl: Would love to see a shot of that IT8
13:35
Bertl
even with incandescent/halogen lighting?
13:35
Bertl
(i.e. without proper illumination as you would call it)
13:38
troy_s_
Bertl: If you have a relatively decent 100 watt tungsten I am betting it will be consistent across 100 watt tungestens.
13:38
troy_s_
(little like quartz, assuming not at edges of lifespan)
13:38
se6astian
I gotta bake a cake now ;)
13:41
Bertl
what I have is a simple light bulb as used for roomlighting (100W)
13:44
troy_s_
Bertl: That is an order of a magnitude better than the information available now. :)
13:47
troy_s_
an 100 watt tungsten halogen should be around 3000
13:47
troy_s_
generic incandescent slightly warmer. 2700+
13:52
Bertl
okay, then I'll prepare a shot
13:52
troy_s_
rexbron_: ^^
14:10
Bertl
will take a little, I have to finish the current code/compile/test session first, but I should have something this evening
14:10
Bertl
(probably have to grab some groceries before as well)
15:07
se6astian
gotta pick up my girlfriend now, its her birthday ;)
15:08
se6astian
left the channel
15:46
Bertl
hmm, new problem, the chart is _very_ reflective
15:47
Bertl
i.e. I probably have to find a way to illuminate it from two sides
16:06
rexbron_
changed nick to: rexbron
16:08
rexbron
cool
16:09
rexbron
Bertl: Are you bouncing the light?
16:10
Bertl
currently no, I have placed the lamp 30cm right of the lens system
16:11
Bertl
that seems to work somewhat, but I probably need a second one to even it out
16:11
rexbron
troy_s_: http://bmcuser.com/showthread.php?6422-BMCC-2-5K-RAW-frames-converted-using-libraw Boom, dng decoding using libraw ;)
16:13
rexbron
Bertl: you can use either a white sheet or a slightly opaque shower curtain to diffuse the light and reduce specular reflections. Not the most elegant thing, but this isn't a rigorus engineering test is it?
16:14
Bertl
the interesting part is that diffusing the light doesn't really help
16:14
Bertl
the chart is so specular that the background (table and so) is reflected :)
16:14
Bertl
but I'm uploading the first shots
16:18
rexbron
nice!
16:23
rexbron
Bertl: I shot a low budget feature on the BMCC. There are interesting workflow problems to overcome when working with image sequences
16:24
rexbron
in particular, file system performance becomes an issue when a days footage, in our case about 1TB, was 400,000 files.
16:24
Bertl
hmm, what filesystem?
16:25
rexbron
NTFS
16:25
rexbron
things improved slightly when I turned off 8.3 name creation
16:25
Bertl
well, who uses NTFS in serious work? :)
16:26
rexbron
Bertl: Unfortunately I do because Blackmagic doesn't support Davinci on linux without the $30k hardware interface
16:26
rexbron
:P
16:26
rexbron
that may change however as with v10, there is 0 difference between the gui's across all platforms
16:27
rexbron
we ended up going to tar archives, as those had much higher write performance at the expense of having to wait 10 min to open them back up
16:28
Bertl
I can imagine that a tar archive is faster and more efficient than NTFS :)
16:28
rexbron
we also produced 30TB of raw files for 8 days of shooting (3 backups)
16:28
rexbron
and that is at 23.98 :P
16:29
rexbron
the trade off with uncompressed raw is storage. For a film with a 8k budget, we spent almost 1.5k on hard drives
16:30
rexbron
Without 3tb drives, it would have been even worse
16:30
troy_s_
Bertl: Part of that reflectance should be by design.
16:30
troy_s_
Bertl: A broader source will reduce it.
16:32
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/RAW/IT8_incand.raw16.xz
16:32
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/RAW/IT8_incand.raw8.xz
16:33
rexbron
Bertl: is that a compressed file or is xz some file format I've just never heard of?
16:33
rexbron
nm
16:33
Bertl
yes, xz is just a compressor, more efficient than bz2
16:34
Bertl
and don't forget, the .raw* are really raw and the .raw16 is padded at the LSB side with zeroes
16:34
troy_s_
16 is 16 bit?
16:34
Bertl
(from 12 to 16 bits)
16:35
rexbron
Bertl: lol, I'm going to need to figure out how to read this file
16:35
troy_s_
gotcha
16:35
troy_s_
so 12 bit + 0 pads
16:35
troy_s_
I am not at home so cant dickery with it
16:35
Bertl
yes, basically XXXXXXXXXXXX0000
16:35
rexbron
src >> 2
16:35
troy_s_
Bertl: I take it it is stored in integer?
16:35
Bertl
rexbron: >> 4 :)
16:36
rexbron
Bertl: whats the recomended workflow to open said raw file right now?
16:36
troy_s_
rexbron: Watch the seemingly simple integer to float converts. there are more than a few articles that show how screwed up a seemingly simple integer conversion can go bad.
16:36
troy_s_
rexbron: load an oiio buffer?
16:36
Bertl
I can either display it in gray, with 'display -size 4096x3072 -depth 16 gray:<file>
16:36
troy_s_
rexbron: Export to EXR?
16:38
Bertl
or convert it with bayer2rgb to an rgb file and display that as rgb
16:38
rexbron
IIRC, rawtherapee was the recommended app by Aaton before it they went under
16:42
rexbron
for reference Bertl, here is the cinemaDNG spec http://download.macromedia.com/pub/labs/cinemadng/cinemadng_p1_spec_091009.pdf
16:42
rexbron
not sure if it is the most up to date
16:43
troy_s_
Bertl: Problem is that there are no known primaries.
16:43
troy_s_
Bertl: So ideally going to float and transforming to primaries would be optimal.
16:43
rexbron
http://wwwimages.adobe.com/www.adobe.com/content/dam/Adobe/en/devnet/cinemadng/pdfs/CinemaDNG_Format_Specification_v1_1.pdf
16:44
troy_s_
rexbron: I think dumping the ints to OIIO then using OIIO to convert to floats is safest.
16:45
rexbron
troy_s_: very likely
16:46
rexbron
also for reference, DNGs genereally include two 3x3 matrixes to get from cameraRGB to XYZ under CIE A and CIE D65 (or D50 can't remember)
16:48
troy_s_
rexbron: Hate that.
16:48
troy_s_
What CT is Illuminant A?
16:48
troy_s_
(D50 is riddled through ICC ville. Sucks)
16:49
rexbron
~2856 CCT
16:49
troy_s_
Typical tungsten house light.
16:49
troy_s_
Grr.
16:49
rexbron
http://en.wikipedia.org/wiki/Standard_illuminant#Illuminant_A
16:50
rexbron
troy_s_: that does represent the bottom end of most encountered incandencent sources
16:52
troy_s_
rexbron: Sure. Just an arbitrary and IMO irrelevant value for working with a blasted camera.
16:53
troy_s_
rexbron: Upside is that I guess we can calculate an almost native matrix using Bertl 's test. Lulz.
16:58
rexbron
troy_s_: that's what most standards are. Arbitrary.
17:03
se6astian
joined the channel
17:05
se6astian
back
17:24
Bertl
wb, send congratulations to your better half from me :)
17:26
Bertl
after all, she managed to age suddenly by another year ...
17:41
troy_s_
rexbron: Most good ones are contextual. Tell me why incandescent for a camera?!?!
17:41
rexbron
only available artificial light source at the time the standard was created?
17:42
troy_s_
rexbron: Right... so tungsten 3200 for the past 100 odd years of cinema didn't exist.
17:42
rexbron
CIE B and C are noon light and daylight simulations with liquid filters over a CIE A source I think
17:42
troy_s_
idiocy
17:42
rexbron
troy_s_: that was created by Kodak lol
17:43
rexbron
3200 was where their emultion worked best for tungsten stocks
18:03
se6astian
Bertl, thanks will do :)
18:10
troy_s_
rexbron: proper tungsten burns very close to that. a photographic certainly does.
18:11
troy_s_
rexbron: But even still... I would have thought a newer standard would merely provide an open ended transform
18:24
rexbron
open ended transform?
18:30
aombk
left the channel
18:30
aombk_
joined the channel
18:33
dmj_nova
se6astian: as far as open hardware sensors, you're probably not going to do better than open documentation and a standardized interface
18:35
se6astian
yes, thats what I concluded as well
18:35
se6astian
still not a bad product I think
18:35
dmj_nova
but I suppose that's a bit like dedicated GPUs. For most purposes, the important thing for most isn't that they can built their own modified unit, it's that they can make it do new things
18:36
se6astian
but tbh its pretty much exactly what kodak/truesense does already as well
18:36
se6astian
they even give you the full datasheet without an NDA
18:37
se6astian
an image sensor can't be taught to do new things though once its produced
18:39
dmj_nova
though, to be fair, if it were somehow possible to put the software in more hands, that might change the underlying design process
18:40
dmj_nova
I doubt that will happen soon though
18:40
dmj_nova
and we certainly don't have the developer power for that
18:41
se6astian
which software?
18:44
dmj_nova
design software for sensors
18:46
dmj_nova
most engineers that design such things don't necessarily need access to the manufacturing facilities. On the other hand, most hackers likely don't have a matching skill set
18:52
se6astian
according to konstantin design software for sensors costs around 1 million dollars per license, there are like a handful of companies that are able to buy that software, using it is quite a science on its own
18:53
se6astian
I dont think we can help to spread that software :)
18:53
dmj_nova
se6astian: no, I don't think so
18:54
dmj_nova
the only thing possible would be to replace it with something new
18:57
aombk_
left the channel
20:20
se6astian
hosa we received our first bitcoun donation!
20:21
se6astian
*bitcoin
20:21
se6astian
0.04 BTC = 15$
20:54
rexbron
Cool
21:08
troy_s_
rexbron: Just a guideline for the transform. Maybe a triplet that defines XYZ white balance and a 3x3 matrix that describes the transform?
21:09
troy_s_
rexbron: That would permit arbitrary white balance (and even a color shift skew) as well as permit choice of transform etc. on how to get there.
21:20
se6astian
time for bed
21:20
se6astian
see you!
21:20
se6astian
left the channel