Current Server Time: 15:06 (Central Europe)

#apertus IRC Channel Logs

2016/10/26

Timezone: UTC


00:01
BogdanXor
joined the channel
00:03
Spirit532
left the channel
00:05
BogdanXor
left the channel
00:37
DGMurdockIII
joined the channel
01:02
BogdanXor
joined the channel
01:06
BogdanXor
left the channel
01:17
Bertl
off to bed.
01:17
Bertl
changed nick to: Bertl_zZ
01:20
mithro
left the channel
01:21
mithro
joined the channel
01:55
jucar
left the channel
02:02
BogdanXor
joined the channel
02:07
BogdanXor
left the channel
02:51
arpu
left the channel
03:03
BogdanXor
joined the channel
03:04
arpu
joined the channel
03:08
BogdanXor
left the channel
03:15
DGMurdockIII
left the channel
04:04
BogdanXor
joined the channel
04:09
BogdanXor
left the channel
05:05
BogdanXor
joined the channel
05:09
BogdanXor
left the channel
05:27
aombk2
joined the channel
05:30
aombk3
left the channel
06:06
BogdanXor
joined the channel
06:10
BogdanXor
left the channel
07:00
Alex_Chooks
joined the channel
07:06
BogdanXor
joined the channel
07:11
BogdanXor
left the channel
07:52
Alex_Chooks
left the channel
08:07
BogdanXor
joined the channel
08:12
BogdanXor
left the channel
08:19
Bertl_zZ
changed nick to: Bertl
08:19
Bertl
morning folks!
08:24
niemand
joined the channel
08:37
niemand
left the channel
09:02
Bertl
off for now ... bbl
09:03
Bertl
changed nick to: Bertl_oO
09:08
BogdanXor
joined the channel
09:29
se6astian|away
changed nick to: se6astian
09:31
se6astian
good morning!
09:31
se6astian
its austrian national holiday
09:32
se6astian
and we celebrate by riding tanks around central vienna :)
09:45
niemand
joined the channel
09:51
niemand
Bertl_oO, on the new axiom I don't have the firmware with dual output anymore, could you reupload it please (or give me the address in case it's still there)
09:57
se6astian
how did the BBA go?
10:00
niemand
We haven't used the axiom
10:01
se6astian
why not?
10:05
niemand
Various reasons. Two of them: We could not test yet the color differences between the cams. And we would have needed one person to adjust settings during the event (at least in the beginning). Both are due to staff shortage.
10:05
niemand
We are now integrating the axiom as third camera in our running setup.
10:05
niemand
btw: live stream is here: https://streaming.media.ccc.de/pw16/pw1/
10:06
niemand
if the axiom goes online I will notify you :)
10:06
niemand
And color differences are not an issue here at the conference I guess
10:07
niemand
light intensity would have been no problem at all, i tested that yesterday
10:16
se6astian
great
10:24
niemand
How can I get the currently set exposure time?
10:27
niemand
ah, I remember that's not possible
10:31
alexML
niemand: take a snapshot without specifying the exposure (you need a small patch to cmv_snap3)
10:32
alexML
if (num_times == 0)
10:32
alexML
goto skip;
10:32
alexML
comment out these
10:33
illwieckz
left the channel
10:33
alexML
then, ./cmv_snap3 -2 -b -r > snap.raw12; raw2dng snap.raw12
10:33
alexML
this will print current exposure time without changing it
10:34
alexML
(just tried, not sure why the "goto skip" line is actually there; maybe Bertl_oO knows?)
10:40
se6astian
alternatively you can read the sensor registers defining the exposure time (see datasheet) and do the math to convert to seconds
10:40
se6astian
that would probably be something quite handy in general to have a small script for that
10:52
niemand
left the channel
10:56
Spirit532
joined the channel
11:06
illwieckz
joined the channel
11:27
niemand
joined the channel
12:19
se6astian
changed nick to: se6astian|away
12:38
jucar
joined the channel
12:51
intracube
left the channel
13:06
jucar
left the channel
13:26
niemand
left the channel
13:38
niemand
joined the channel
13:42
aombk2
left the channel
13:48
aombk2
joined the channel
13:53
niemand
It seems we need 4:2:2 ...
13:57
Bertl_oO
well, you have to implement that then
13:58
Bertl_oO
alexML: './cmv_snap3 -z -r' will simply output the sensor registers without changing anything or producing an entire raw
13:59
Bertl_oO
but of course, it would be simpler to access the sensor registers directly and just read (and interpret) the relevant ones
13:59
niemand
And I don't know if the card supports RGB too
14:01
Bertl_oO
look, I don't want to be rude, but I've spent a significant amount of time to adjust the firmware for 'your needs' in last week and you didn't manage to find somebody to 'operate' the camera at the BBA ...
14:01
Bertl_oO
if you need more features, please go ahead and implement them
14:03
niemand
I totally understand you and I'm too not amused to know these requirements toady. We try to get it working using an additional SDI->HDMI converter
14:03
niemand
so we can use it on another capture card
14:04
niemand
probably it works then. But its not clear that we get that converter today.
14:04
Bertl_oO
https://github.com/apertus-open-source-cinema/beta-software/tree/master/cmv_hdmi3
14:08
aombk2
left the channel
14:08
Bertl_oO
alexML: the gain firmware has to wait till I (or somebody) write a test firmware to figure out why the zynq doesn't match the simulation behaviour
14:09
Bertl_oO
but if you like, I can upload the new row_col_noise entity for investigation
14:11
alexML
no rush, I'm currently trying to figure out some matrix math
14:13
alexML
(trying to find cam_xyz to implement kelvin white balance)
14:15
aombk2
joined the channel
14:15
Bertl_oO
the best approach would be to first linearize the relevant portion with the linearization LUTS (maybe directly adjusting the white balance for the sensor)
14:15
Bertl_oO
then identify the primaries and use that to map to XYZ
14:15
Bertl_oO
(should be easy when the data is basically linear)
14:20
aombk2
left the channel
14:23
alexML
sure, easy in theory, but troubleshooting it takes a lot longer than expected...
14:26
aombk2
joined the channel
14:27
Bertl_oO
as always ÌÌ...
14:30
aombk2
left the channel
14:32
se6astian|away
changed nick to: se6astian
14:36
aombk2
joined the channel
14:55
aombk2
left the channel
14:57
niemand
what "monitors" do you use to test the hdmi output?
15:01
aombk2
joined the channel
15:04
alexML
niemand: I use a SmallHD DP6
15:08
niemand
thanks. I think I need such a device^^
15:09
Bertl_oO
the atomos devices work too, but they require a special mode
15:10
niemand
but are very expensive
15:15
niemand
left the channel
15:16
se6astian
the DP6 isnt that much cheaper
15:16
se6astian
I created a task for the exposure time script so we do not forget it
15:16
se6astian
https://lab.apertus.org/T682
15:19
intracube
joined the channel
15:22
aombk2
left the channel
15:28
aombk2
joined the channel
15:30
niemand
joined the channel
15:32
aombk2
left the channel
15:46
BogdanXor
left the channel
16:26
Spirit532
this is one of the problems with apertus in my opinion
16:26
Spirit532
it needs an internal recorder
16:26
Spirit532
but the microzed doesn't have high-speed transcievers, so no SATA
16:26
Spirit532
(no HS sata anyway)
16:33
se6astian
help make it happen :)
16:42
BogdanXor
joined the channel
16:47
BogdanXor
left the channel
17:04
Spirit532
se6astian, not much I can do unless you propose a several-thousand-dollar budget to switch from the microzed to something that has a 7015, or an entirely custom board
17:04
niemand
I probably can help with implementing some more resolutions and subsampling later
17:04
Spirit532
7015/7030 anyway, but even a 7015 should be plenty for reading the cmv12k sensor at full speed into ram, and at about half speed into an ssd
17:05
niemand
Spirit532, then we can encode it to an h.264 stream too^^
17:05
Spirit532
what's stopping you from doing that with the 7010/20?
17:05
Spirit532
aside from proper IP being ridiculously pricy
17:06
niemand
Spirit532, you mean encoding on the fpga or in the zboard? (I don't know much details about the hardware)
17:06
Spirit532
on the fpga
17:07
Spirit532
it should have enough grunt to implement an h264/5 encoder if I remember the specs correctly
17:08
Spirit532
but the microzed is a rather terrible choice for an integrated camera
17:08
niemand
okay, doing this on the fpga is a great idea, but hard to do
17:10
Spirit532
I'm going to design a camera based on the 7015, probably a module from trenz electronics or the bora from DAVE
17:10
Spirit532
but not using the CMV12k, it's just far too expensive for me
17:11
Spirit532
I'm going to use the CMV2k sensor instead
17:40
Bertl_oO
well, there are plenty of closed cameras out there, go ahead and create another one
17:43
BogdanXor
joined the channel
17:48
BogdanXor
left the channel
17:55
niemand
Is there any solution to update all scripts?
18:12
Spirit532
Bertl_oO, what good is an open camera that requires a ton of closed hardware to make it work? :P
18:21
Bertl_oO
what good is another proprietary solution?
18:22
Spirit532
who said it will be proprietary?
18:22
Spirit532
and even if it is, I want to implement all features sought after in a camera of this price range(~$1k), rather than dividing it into silly models like socanikon do
18:23
Spirit532
that's what makes hobbyists differ
18:23
Spirit532
we do what we want, we aren't driven by profit charts and the responsibilities before a board of 850 directors and 300 CEOs
18:25
Spirit532
different*
18:30
Bertl_oO
looking forward to your FOSS/OH camera based on a Trenz or DAVE module ...
18:30
Spirit532
once I finish the thermal camera, I'll start pouring money into it
18:44
BogdanXor
joined the channel
18:49
BogdanXor
left the channel
18:52
niemand
left the channel
19:45
BogdanXor
joined the channel
19:49
BogdanXor
left the channel
19:59
niemand
joined the channel
20:06
niemand
The black offset is something I haven't done previously in my attempts
20:06
alexML
by default, after RCN calibration, it should be 128
20:07
alexML
if you shoot dark frames, it usually is
20:07
alexML
however, once you shoot real images, the actual black level may change
20:07
alexML
so if you get clipped shadows, try lowering it
20:10
alexML
my current setup: ./gamma_conf.sh 0.4; ./set_wb.py 1 1.5 2.8 1 128 1; # tungsten light
20:10
Bertl_oO
the 128 make perfect sense for raw recording, but for HDMI output, you probably want it to clip on both ends
20:10
Bertl_oO
i.e. linearize the sensor response to the full range
20:11
alexML
yes, the 128 in set_wb.py subtracts it
20:11
alexML
so the output is from 0 to white (I think)
20:12
alexML
I'd like to run a test with the Shogun (which interprets 16-235) to compare to what I see here
20:12
alexML
se6astian: do you still have it around?
20:21
niemand
left the channel
20:26
se6astian
yes its connected to the beta
20:36
se6astian
just tell me what parameters/commands I should run and what to capture
20:43
niemand
joined the channel
20:44
alexML
if you have an IT8 chart and a tungsten light, run the command I've pasted above, with set_wb.py from https://dl.dropboxusercontent.com/u/4124919/bleeding-edge/apertus/set_wb.py and record a single frame in regular HDMI mode (not UHD) and a raw12 snap; the beta should be RCN-calibrated
20:46
BogdanXor
joined the channel
20:50
BogdanXor
left the channel
20:53
Bertl_oO
alexML: can you check if there is any column related variance on the back column data?
20:54
Bertl_oO
i.e. what I would like to know is if we should separate even and odd columns when averaging them
20:57
alexML
column-wise, I didn't notice any patterns, but odd and even rows have different behaviors
20:59
Bertl_oO
yeah, okay, so even/odd column doesn't matter for the black columns at all
20:59
alexML
right, I didn't use this property at all
21:00
alexML
median of 16 values would probably be best (as it will not be influenced by hot pixels), but I guess it's a lot harder to compute than mean, on FPGA
21:01
alexML
what may cause surprises is the transition from black pixels to active pixels (when the light is strong, border black pixels may become brighter as well)
21:02
alexML
so probably it's best to ignore them and average e.g. 6 + 6 columns
21:02
Bertl_oO
really?
21:04
Bertl_oO
that would suggest that the sensel discharge even with the electronic shutter closed, very interesting
21:06
Bertl_oO
do you have any data to back that up? I'd love to ask the CMOSIS folks why this happens
21:07
Bertl_oO
s/why/how/
21:09
alexML
well, I know for sure it happens on Canon (an image with the sun near the edge of the frame)
21:09
alexML
also, the 8th column is brighter than the others on CMV
21:10
alexML
and I can try to reproduce the troublesome scene on the CMV as well
21:10
niemand
left the channel
21:11
alexML
let me check if it reacts to a desk lamp
21:13
Bertl_oO
does the canon sensor allow to switch the black columns on and off?
21:13
Bertl_oO
because it might be, that they simply covered the black columns on the filter, which would result in the effect you describe
21:14
alexML
I know how to tweak the size of the top black bar (smaller or much larger), but no idea about the left one (which caused the issue)
21:18
se6astian
I am off to bed
21:18
se6astian
good night
21:19
se6astian
changed nick to: se6astian|away
21:33
alexML
Bertl_oO: this is a scene that caused trouble in Canons: https://dl.dropboxusercontent.com/u/4124919/bleeding-edge/apertus/black-columns-scene-light-test/borderlamp.jpg
21:33
alexML
CMV black columns [1:8, end-7:end]: https://dl.dropboxusercontent.com/u/4124919/bleeding-edge/apertus/black-columns-scene-light-test/black-columns-borderlamp.png
21:34
alexML
and from some dark/gray/clipped frames:
21:34
alexML
https://dl.dropboxusercontent.com/u/4124919/bleeding-edge/apertus/black-columns-scene-light-test/black-columns-darkframe.png
21:34
alexML
https://dl.dropboxusercontent.com/u/4124919/bleeding-edge/apertus/black-columns-scene-light-test/black-columns-grayframe.png
21:34
alexML
https://dl.dropboxusercontent.com/u/4124919/bleeding-edge/apertus/black-columns-scene-light-test/black-columns-overexposed.png
21:36
alexML
so there is some influence, but more like a global effect; on Canons, it was much more obvious and only on the lines with the strong light source
21:36
alexML
(the test image for the Canon is with the sun in the frame; I can look it up on request)
21:46
BogdanXor
joined the channel
21:53
BogdanXor
left the channel
21:53
alexML
anyway, if you average the first and last 6 columns, those seem to have no influence from the scene
22:07
Bertl_oO
okay, great!
22:49
BogdanXor
joined the channel
22:53
BogdanXor
left the channel
23:49
BogdanXor
joined the channel
23:54
BogdanXor
left the channel