Current Server Time: 06:50 (Central Europe)

#apertus IRC Channel Logs

2021/12/20

Timezone: UTC


03:09
fredy
joined the channel
05:25
Bertl
off to bed now ... have a good one everyone!
05:25
Bertl
changed nick to: Bertl_zZ
08:41
se6astian
good day
12:34
Bertl_zZ
changed nick to: Bertl
12:34
Bertl
morning folks!
12:35
vup
good morning :)
17:00
se6astian
MEETING TIME
17:00
se6astian
who is here?
17:01
Bertl
is here ...
17:02
se6astian
seems we will be quick today again :)
17:02
se6astian
any news Bertl ?
17:02
Bertl
not really
17:02
se6astian
right
17:02
se6astian
so quick updates from me:
17:02
se6astian
new rubber feet from manfred arrived: https://twitter.com/ApertusOSCinema/status/1472557170701094918
17:03
se6astian
new recorder enclosure 3d print parts (we are at version 0.2 ) have been produced and I am putting the recorder together to test as we speak
17:03
se6astian
recorded the CCC talk last week together with Max
17:03
se6astian
and he is in full swing editing already
17:03
se6astian
as the deadline to submit it is today
17:04
se6astian
with all the inserts to add we might not make it but 1 or 2 days shall not be an issue hopefully
17:04
cscar
joined the channel
17:04
se6astian
anuejn completed new additions to the axiom recorder to do realtime debayering of A+B frames from HDMI stream
17:05
se6astian
so thats what I am getting ready to test currently
17:05
se6astian
directly on the lattepanda recorder hardware
17:06
se6astian
had a chat with vup/anuejn yesterday about workflow decision and the plan now is to apply darkframe, dcnu frame, gainframe in realtime in the recorder before displaying/saving the raw footage
17:06
se6astian
directly in dng then though
17:07
se6astian
there is still the question if one would like to save the "pure raw" footage before the corrections
17:07
se6astian
what we have in raw12 currently basically
17:07
Bertl
I'd definitely opt for keeping raw data
17:08
se6astian
well dng is also raw :)
17:08
se6astian
I guess the user will have the choice in the end
17:08
Bertl
any noise correction could potentially be done in a better way if there is more cpu power available
17:09
Bertl
so we should not do any correction which can not easily be undone
17:09
se6astian
as the axiom recorder can already read raw12 sequences now so it would be possible to apply the corrections later (not from hdmi source but filesource) as well
17:10
Bertl
note: it's perfectly fine to do any corrections on-the-fly for visualization and we can even do them in the stored 'raw' data if the process is reversible
17:11
se6astian
there was also the question if the raw12-viewer is obsolete now and I concluded that the axiom-recorder has a slightly different goal so I/BAndiT1983_ will continue working on it
17:11
se6astian
Bertl: agreed
17:11
vup
is a bit late
17:12
vup
Bertl: for these simple corrections it should also be pretty doable to just save a bit of extra data and make it reversible
17:12
se6astian
vup: we discussed yesterday if the corrections are reversible but concluded: "not entirely" right?
17:12
vup
maybe that could also be a option
17:13
vup
se6astian: well only darkframe, dcnu frame and gainframe should be reversible (apart from maybe rounding errors)
17:13
se6astian
even better
17:14
vup
more interesting is the dark column based row noise filtering, but worst case I think we just need to save 2xheight extra values to make it reversible
17:16
se6astian
I think thats it from my side
17:16
vup
(very important in making it reversible is of course using the same *frames to undo the correction, but we probably can just save a checksum of them or something like that to catch mismatches)
17:17
se6astian
saving darkframe/dcnu/gainframe with the footage would sound like a sane idea
17:17
vup
maybe saving them on the camera? So you can not easily loose them
17:18
vup
(and they are probably to some extend camera specific, so you need to match them to the right camera anyways)
17:18
se6astian
true
17:20
se6astian
anuejn: it works! realtime raw playback is running on the lattepanda
17:20
vup
(Bertl: also yes in general I completely agree with you about only doing corrections we can (easily) reverse if we only save a corrected image)
17:21
se6astian
motion creates weird block effects though, maybe from downscaling
17:21
se6astian
is that expected?
17:22
vup
Hmm block effects? when anuejn and I tried it last time it looked fine, but there some color artifacts which looked like the wrong frames were getting combined
17:22
se6astian
will take picture
17:23
vup
oh with the newest version you can also save dngs (but now while viewing the preview), so would be interesting to check if they are present there aswell
17:23
se6astian
cool
17:23
se6astian
will test
17:23
vup
unfortunately I think the color filter array is not configured correctly in the dngs currently
17:23
se6astian
what is the syntax?
17:24
vup
well anuejn did not push the changes :(
17:25
se6astian
ah
17:25
se6astian
right
17:25
vup
(just noticed it)
17:26
se6astian
https://cloud.apertus.org/index.php/apps/gallery/s/bMYjZA5QgqL7w9f
17:26
se6astian
last image is moving hand
17:26
vup
that looks interesting
17:26
Bertl
that's nice ;)
17:27
se6astian
you noticed the penguin immediately didnt you? :D
17:27
vup
I don't think that has anything to do with the downscaling
17:27
se6astian
debayering?
17:28
Bertl
the downscaling could be responsible for the pattern
17:28
Bertl
but the color effect seems to be motion related
17:30
Bertl
as it seems to be blue-green vs red-green, mismatched frames could be the reason
17:31
vup
oh yeah se6astian: do you get any output on stdout?
17:31
se6astian
yes, but nothing that looks like an error
17:31
vup
(it should print a message if it detects a mismatched / dropped frame)
17:32
vup
hmm
17:32
se6astian
using gpu: intel...
17:32
se6astian
mesa-intel warning: performance support disabled...
17:32
se6astian
it did report 2 slipped frames
17:33
se6astian
but not related to the motion artifacts
17:33
se6astian
but I noticed similar artifacts on the raw12 sequences recorded when played back with the axiom recorder
17:33
se6astian
when there was motion
17:33
vup
oh interesting
17:33
se6astian
should I capture a video
17:34
vup
when we tested it on saturday we also had some motion related color artifacts, but they looked a bit different, but until those are fixed, I would just assume its the same artifact you are seeing
17:34
se6astian
my screen is rather low resolution
17:34
se6astian
they appear much finer and less severe when the screen is fullhd like the material
17:35
se6astian
so that maybe explains the differences to what you are seeing
17:35
se6astian
this lcd is 1024x600 IIRC
17:35
se6astian
anyway, anyone else who wants to report/share something at the meeting?
17:35
vup
yeah maybe, we can try with screen scaling and see what happens
17:36
vup
sure I can quickly report some bits
17:36
vup
So I finally did the finishing touches on https://github.com/apertus-open-source-cinema/pcbs
17:36
vup
its now live at https://apertus-open-source-cinema.github.io/pcbs/
17:37
vup
atm it just generates the preview images, the gerbers and pdf version of the schematics
17:37
se6astian
nicely done, forgot to mention I added the link to many pcb related wiki pages already: https://wiki.apertus.org/index.php/Special:RecentChanges
17:37
vup
nice
17:37
se6astian
and updated our OUI PR request accordngly
17:38
se6astian
https://github.com/openmoko/openmoko-usb-oui/pull/37
17:38
vup
if anybody thinks it would be interesting, one could also add a interactive boardview / bom thingy like this: https://files.niemo.de/axiom_beta_main_board_v0.38_r1.2_ibom_eagle.html
17:39
se6astian
yes!
17:39
vup
ok :)
17:39
se6astian
does the conversion work automatically or did you have to drag it through kicad manually?
17:39
se6astian
its beautiful!
17:39
vup
the one I linked can be generated using a ulp script from eagle
17:40
se6astian
great
17:40
se6astian
would be a nice addition to the pcb page IMO
17:40
vup
sure should not be hard to add
17:40
vup
to add a new board one should just need to drop a `.brd` and a `.sch` into this folder: https://github.com/apertus-open-source-cinema/pcbs/tree/main/boards
17:40
cscar
I'm off to the groceries. I'll read the rest when I'm back. Great progress! See you all later!
17:40
vup
(best with accompanying .license files)
17:41
vup
and if the naming scheme of the other pcbs is followed it should work out nicely
17:41
vup
thats it from me
17:43
se6astian
many thanks
17:43
se6astian
right then, off to dinner as well
17:43
se6astian
MEETING CONCLUDED
17:43
se6astian
many thanks everyone!
17:53
Bertl
changed nick to: Bertl_oO
18:31
anuejn
hey, changes for dng recording are pushed
18:31
anuejn
tho it is not really stable at the moment
18:53
se6astian
cool
18:58
elafon
left the channel
18:58
elafon
joined the channel
19:00
elafon
left the channel
19:00
elafon
joined the channel
21:07
cscar
left the channel