Current Server Time: 22:49 (Central Europe)

#apertus IRC Channel Logs

2015/12/11

Timezone: UTC


00:52
Bertl
what's the reason for the +1 in the gamma(r+1) (and similar)?
00:53
Bertl
hmm, r is from 0-1 or from 0 to 4095?
00:53
Bertl
anyway, off to bed now ... we can discuss this later ...
00:53
Bertl
changed nick to: Bertl_zZ
03:59
troy_s
Never trust anyone that says YUV.
04:01
troy_s
:P
04:01
troy_s
You guys aren't dealing with analog for the love of all things good and pure. Use correct terms. YCbCr.
04:02
troy_s
se6astian|away: What is the illuminant source on the files?
04:03
troy_s
se6astian|away: I'm reasonably certain that the closer to D50 you can get the better as the XYZ swatch values are typically in reference to D50.
04:04
troy_s
(also curious how the hell you are encoding to YCbCr without knowing the colourspace of the footage. You kind of need that Y before you can derive Cb / Cr. :))
04:53
Bertl_zZ
changed nick to: Bertl
04:53
Bertl
back ...
04:54
Bertl
troy_s: I've completely given up educating folks about YUV vs YCbCr ...
04:54
troy_s
You shouldn't.
04:54
troy_s
It speaks volumes about what they are offering up as insight.
04:55
Bertl
regarding the colorspace, in this particular case it doesn't really matter, as the YUV/YCbCr is used as back box
04:55
troy_s
Correct.
04:55
troy_s
I get it.
04:55
troy_s
You can get 1:1 RGB out, it just makes it nonstandard.
04:55
Bertl
i.e. the values could be named ABC as well and the recorder use ABC 422 :)
04:55
troy_s
I knew where you were going. ;)
04:56
troy_s
Bertl: I will do a proper D65 version off of Sebastian's shots in a few days.
04:56
troy_s
Doing a crap load of commercials up to the holidays, so I hope to hit it Tuesday or so.
04:56
troy_s
And I will do the proper D65 CA on them.
04:56
Bertl
excellent!
04:56
troy_s
But it would help to know the illuminant.
04:57
troy_s
Because the data sheets for the CIE values is very typically D50
04:57
troy_s
Which means that our estimates will be slightly skewed if we aren't in D50
04:57
Bertl
check with sebastian/max (probably best in a few hours when they wake up, they can use a well known illuminant for some snaps I guess
04:58
troy_s
If we can, whoever is listening, D50 sets of exposures would rock it.
04:59
Bertl
not sure what equipment irieger|away has, but there is a chance for high quality snaps as well
04:59
troy_s
Then we can get a pretty good handle on some synthetic primaries for irieger|away 's twiddled registers.
04:59
troy_s
The sad part is there is no D50 source out there. 56k for sure.
05:00
Bertl
also, do not hold off to generate matrices _and_ 3D LUTs, we will probably implement a simple 3D LUT soonish :)
05:00
troy_s
Anyways... We can get a decent matrix. Doubly so if you use your wizardry to figure out how to get a perfect correction LUT.
05:01
troy_s
Oh great!
05:01
troy_s
OK will do.
05:01
Bertl
and do you have a place to store/upload them?
05:01
Bertl
i.e. something where they can be kept for some time?
05:02
Bertl
if not, we should arrange a place on ftp.apertus.org for you
05:02
troy_s
Sure. I have S3.
05:02
troy_s
But if you want a decent long term, perhaps Apertus.
05:04
troy_s
Anyways, I promise to crack it my first day off.
05:04
troy_s
Will bookmark Seb's new images.
05:05
troy_s
Bertl: That previous set are more than sufficient probably for a hack HDMI view.
05:05
troy_s
Bertl: And I want to pick your brain on how to feather a few 3D LUTs together.
05:06
troy_s
Bertl: You have to get cracking on that Piecewise soon... See how we can do in the latitude front.
05:06
Bertl
yeah, as I said, I got a decent output with just reducing the green channel by 16% i.e. unmodified sensor linear data, gamma corrected (0.8) looks quite nice
05:07
troy_s
Sure, but folks will be a little more interested in the colorimetry and latitude.
05:07
Bertl
so, a matrix must suffice to get usable output (not perfect though) especially if we use the linearization lut
05:07
troy_s
(At least probably 85% of the audience.)
05:08
troy_s
Bertl: For a quickie view, a dialed in matrix will be sufficient. Don't forget that we need to do a Bradford on chip.
05:08
troy_s
That will have some overhead in that you have to calculate (or use a lookup would be easier) of the various XYZ values for each white illuminant supported.
05:10
troy_s
(To XYZ + Bradford/White + XYZ into one matrix.)
05:10
Bertl
I'm not so much worried about the 3D LUT/matrix for the HDMI right now, most monitors/recorders (like for example the shogun) already have 3D LUT capabilities
05:11
troy_s
Yes.
05:11
troy_s
Most displays have the LUT ability.
05:11
troy_s
But you sort of are hand cuffed by the lowest common denominator.
05:11
Bertl
of course
05:12
troy_s
I just want to see it "done right" as opposed to some hackey dark mumbo jumbo of juggling values.
05:12
troy_s
Because I suspect the folks that care would care about seeing some sincere care and attention to that sort of thing.
05:13
Bertl
that's a lot of care :)
05:13
troy_s
It's 2X
05:13
troy_s
But nothing would be gained from a bunch of mumbo jumbo methinks. Charts and proper process don't lie.
05:15
troy_s
Anyways... Out for now. Bookmarking.
05:16
Bertl
bookmarking?
05:16
troy_s
The new files.
05:17
troy_s
Looking forward to getting the results.
05:17
Bertl
okay, have fun :)
05:18
intracube
joined the channel
07:01
tezburma
joined the channel
07:08
alexML
Bertl: this +1 is because array indices start from 1
07:09
se6astian|away
changed nick to: se6astian
07:10
cbohnens|away
changed nick to: cbohnens
07:10
se6astian
good morning
07:16
se6astian
troy_s: the colorcharts were illuminated with two high CRI daylight fluorescent tubes
07:23
intracube
hi se6astian
07:24
intracube
se6astian: do you know if the Beta clips in the recent team talk went through some motion interpolating?
07:25
intracube
there are some strange artefacts in a few places
07:25
intracube
https://goo.gl/photos/RDKc3mUVZUzFHwaE7
07:25
intracube
https://goo.gl/photos/rkxuo4ZAaJ5abZc86
07:30
se6astian
yes apparently max retimed the 4K highspeed clips with 105% by accident, premiere didnt show it in the previews but did the pixel interpolation then on export....
07:32
intracube
ah ok
07:37
se6astian
but I think only the box sequence is affected
07:40
intracube
yep, it looks like it
08:55
LordVan
joined the channel
09:24
irieger|away
changed nick to: irieger
09:27
irieger
troy_s: Bertl: We have quite a few things to play here. There is currently a short film production which has most of the lights but they are back soon.
09:28
irieger
Currently I use two Arri Lite 750W tungsten lights to have a known source with a Full CTB correction (gets visually relatively even color channels)
09:29
irieger
I'm still trying to optimize black offset and get rid of some row noise and stuff thought.
09:29
irieger
The plan is to get some sufficient sensor params the next days and do a good round of color meassurement and tests next week.
09:30
irieger
The short film is back monday so I'll have Kino Flos, HMI to get some different light sources captured.
09:31
irieger
Bertl: My temperature sensor order should finally arrive today or tomorrow so I might get back to you about wiring and reading them the next days.
09:31
Bertl
okay, no problem
09:53
Bertl
qick poll: which cables do you prefer for the Early Beta (power cables)
09:54
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/BETA/cables_apertus.jpg
09:54
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/BETA/cables_chs_schrenk.jpg
09:54
Bertl
note: the colors are a little off, the purple looks similar to the OSHpark PCBs
10:03
irieger
I would go for the first variant with black
10:03
Bertl
i.e. the braided black cables
10:04
Bertl
s/cables/cable/
10:04
irieger
yep. Braided just looks nice I'd say
10:15
tezburma
left the channel
10:25
Bertl
off for now ... bbl
10:25
Bertl
changed nick to: Bertl_oO
10:26
alexML
Bertl: can you tell me the bit order for what you call Raw Mode 01 and 02?
10:26
alexML
se6astian told me the order for Raw Mode 01, but after fixing some bugs in my simulation, I get results similar to Raw Mode 02
10:28
Bertl_oO
https://github.com/apertus-open-source-cinema/beta-software/blob/master/cmv_hdmi3/top.vhd#L2930
10:28
Bertl_oO
note that mode 6 and 7 has a bug (2x blue instead of red, blue)
10:28
tezburma
joined the channel
10:29
Bertl_oO
also note that mode 1 and 2 are the two single frames for mode 3 (for testing)
10:29
alexML
thanks
10:29
alexML
I have these samples:
10:29
Bertl_oO
so not really a raw mode per se (only half of it)
10:29
alexML
Raw_Mode_01_00282.tif Raw_Mode_01_00283.tif
10:29
alexML
Raw_Mode_02_00418.tif Raw_Mode_02_00419.tif
10:29
Bertl_oO
mode 01 is probably #3
10:30
Bertl_oO
and mode 02 is probably #5
10:30
alexML
that means, when "011" and when "101" ?
10:30
Bertl_oO
yup
10:30
alexML
cool
10:32
alexML
btw, there are 2 HDMI frames, but under one "when" label I only see one of them
10:33
alexML
ah, it's handled in the when/else block
10:34
alexML
understood, so #3 is what sebastian described to me
10:35
alexML
* is an overloaded operator that interleaves the bits? (I guess it's not the standard behavior in vhdl)
10:36
alexML
and & is bit concatenation, MSB first?
10:49
alexML
seems to work
10:49
alexML
so, I guess the samples were mislabeled
12:39
troy_s
irieger: Does the institute have a color temp meter that you could use to gel the lamps down a bit?
12:39
troy_s
(Assuming they are HMI)
12:45
LordVan
left the channel
12:58
se6astian
changed nick to: se6astian|away
13:00
se6astian|away
changed nick to: se6astian
13:08
irieger
troy_s: We have one or two Minolta or whatever color temp meters and also a cheapy spectrometer
13:10
irieger
(UPRtek MK350D
13:11
irieger
There is also an i1 Pro somewhere which I'm sure I could get to have another "eye"
13:13
troy_s
Fine. Easy to get close to 5000k with the Minolta and some 1/8 O.
13:38
irieger
Yes, shouldn't be a problem
15:08
se6astian
changed nick to: se6astian|away
17:33
se6astian|away
changed nick to: se6astian
17:33
LordVan
joined the channel
18:20
LordVan
left the channel
19:10
tezburma
left the channel
19:24
Bertl_oO
off for a nap ... bbl
19:24
Bertl_oO
changed nick to: Bertl_zZ
20:59
slikdigit
left the channel
21:41
se6astian
time for bed
21:41
se6astian
good night
21:41
se6astian
changed nick to: se6astian|away