Current Server Time: 03:10 (Central Europe)

#apertus IRC Channel Logs

2013/11/12

Timezone: UTC


02:18
Wescotte
left the channel
07:51
Sasha_C
joined the channel
08:03
se6astian
joined the channel
08:05
se6astian
good monring
08:06
Sasha_C
Greetings Sebastian
08:06
Sasha_C
How are you today?
08:20
se6astian
fine fine :)
08:20
se6astian
and you?
08:23
Sasha_C
I'm well, but my head feels horrible from a serious lack of sleep ;)
08:24
se6astian
then go to sleep! :D
08:25
Sasha_C
I will in a few hours. In regards to the next general meeting, could we schedule it for Tuesday night (around Wednesday 12am in Sydney) next week?
08:58
se6astian
doodle sent
09:00
Sasha_C
Thanks :D
09:17
intracube
joined the channel
10:20
Bertl
morning everyone!
10:22
Sasha_C
Good morning Bertl
10:23
se6astian
hello!
10:23
se6astian
I reviewed the lens mount part
10:23
se6astian
then tried to order from firstcut
10:23
se6astian
and on the final page of their order process
10:24
se6astian
"error processing your order" :)
10:24
se6astian
repeatedly
10:29
Bertl
hehe
10:29
Bertl
maybe a sign?
10:30
Bertl
btw, did you adapt the seal area?
10:31
se6astian
nope
10:32
Bertl
it might be a good idea to do that, so that we can actually seal the sensor off with a piece of plastic lateron
10:34
Bertl
i.e. just match the inner part to about 1mm around the glass aread, that should do the trick
10:36
Bertl
*area
10:40
se6astian
should we design the plastic piece as well already?
10:40
se6astian
you mean the palstic is put in from the "front" (lens side right)?
10:41
Bertl
I think the plastic sheet (or rubber or whatever) should go between the lens mount and the seal area of the sensor
10:42
Bertl
ATM, we have a gap, as the inner shape near the seal area is much larger than the seal area
10:43
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/lensmount_0002.png
10:43
Bertl
https://www.apertus.org/sites/default/files/alpha-lensmount02.jpg
10:44
Bertl
it is not that extreme as shown in the rendering, but it looks similar
10:51
se6astian
firstcut just called me
10:56
se6astian
ok have the lens mount open and will take a look at the seal area now
11:02
se6astian
how much overlap should the seal area/metal/rubber have from the sides?
11:02
Bertl
it should leave about 0.5-1mm space around the glass
11:03
Bertl
i.e. nothing should really need to touch the glass area
11:08
se6astian
so since the seal area has sharp edges
11:08
se6astian
should the inner aluminum "window" also loose the rounded corners?
11:12
Bertl
let me draw a picture :)
11:17
se6astian
ok :)
11:19
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/lens-mount-seal.png
11:20
se6astian
got it, what do the colors represent?
11:21
Bertl
the yellow area is the seal area, the red area is the glass
11:21
Bertl
and the blue overlay is how the lens mount IMHO should fit
11:21
Bertl
it is a little out of proportion to demonstrate the overlap
11:22
Bertl
the blue(lens-mount)-red(glass) distance should be at least 0.5mm
11:23
Bertl
the yello(seal area)-blue overlap should be at least 1mm
11:23
Bertl
(more is probably better here)
11:24
Bertl
the 'seal' (i.e. the plastic or rubber mask) we are going to add will be between the blue and yellow (z-dimension)
11:25
Bertl
and should probably have a cutout equal to the glass area
11:36
se6astian
hmm, doesnt make it much clearer for me sorry, will create an image as well :)
11:38
Bertl
we want the seal area to overlap with the lens mount bottom as much as possible without hitting the glass area (in the xy-plane)
11:39
Bertl
(that's what the seal area is for, sealing off the sensor :)
12:03
se6astian
https://cloud.gerade.org/public.php?service=files&t=88644302ffde4f6e5659627a3c8bbf32&download&path=/Axiom/lens-mount---seal-area-01.jpg
12:04
se6astian
option 1: window slightly outside seal area
12:04
se6astian
option 2 window slightly inside seal area (2mm overlap)
12:08
Bertl
with window you mean the rounded rectangle visible at the bottom of the lens mount, yes?
12:08
Bertl
that should completely overlap the seal area (that's the purpose of the seal area)
12:08
Bertl
but it should not overlap the glass area
12:09
Bertl
(and to be on the safe side, I'd keep a distance of 0.5mm to the glass area)
12:10
Bertl
btw, your 'cloud' has the wrong mime data for jpegs (and maybe other images as well) forcing a download
12:12
se6astian
with window you mean the rounded rectangle visible at the bottom of the lens mount, yes? yes
12:13
se6astian
ok, so option 2 is looking "right, right?
12:13
se6astian
I just make sure the distances are accurate to the glass area (0.5mm)
12:14
Bertl
yes and keep them round (like the glass are)
12:14
Bertl
*area
12:14
Bertl
keep the window round in equal distance to the glass area - is what I meant
12:16
se6astian
understood
12:17
se6astian
the datasheet does not specify the glass area dimensions, but I think we measured it from your scans?
12:19
Bertl
7.1 contains the data
12:19
Bertl
28.7x27.8
12:20
Bertl
where the seal area is 34.7x33.8
12:21
Bertl
but double checking with the scans won't hurt
12:21
se6astian
in the scans the glass extendes further outside than the groove in the ceramic
12:24
Bertl
note that there are two steps in the groove
12:25
Bertl
actually two horizontal and three vertical
12:25
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/cmv12k_close_001.png
12:26
Bertl
on the top, you see the gold pins in the first and second step
12:26
Bertl
and the gold border on the third, inner step
12:26
Bertl
on the left you see the bar-code like stripes on the first level
12:26
Bertl
and the gold on the second (which is the same level as the third from top)
12:27
Bertl
the glass are seems to be inside the first groove
12:27
Bertl
(this also matches with the schematic drawing in the datasheet 7.1)
12:32
se6astian
I see, perfect
12:43
se6astian
done
12:43
se6astian
https://cloud.gerade.org/public.php?service=files&t=88644302ffde4f6e5659627a3c8bbf32&download&path=/Axiom/F-mount%20to%20CMV12k-V04.pdf
12:43
se6astian
please verify
12:45
se6astian
should I create STL for you as well?
12:46
Bertl
nah, looks good to me
12:46
Bertl
ah, one thing
12:46
Bertl
the base is now 1mm larger than originally designed or 1mm on both sides?
12:52
se6astian
59mm total width
12:53
Bertl
okay, that is 1mm larger which is perfectly fine
12:54
se6astian
yes
12:55
se6astian
uploaded new file to firstcut
12:55
se6astian
will take a bit again until its reviewed
12:55
se6astian
did you have time to cut the enclosure parts yet?
12:56
Bertl
not yet, was busy during the weekend with all kind of stuff
12:56
Bertl
but should happen in the next few days
12:56
Bertl
(i.e. definitely before we get the sensor)
12:57
Bertl
btw, you didn't bring the power adapter with your zedboard, yes?
12:57
Bertl
(last time in the lab)
12:57
Bertl
s/in the/at the/
12:58
se6astian
nope, left it at home, you want it?
12:58
Bertl
no problem at the moment, just checking
13:05
se6astian
ok
13:09
se6astian
new firstcut quote is here
13:09
se6astian
since we cut away a bit less its 5 € cheaper per piece ;)
13:11
PhilippeJ
joined the channel
13:12
se6astian
hi PhilippeJ
13:13
se6astian
Surface Finish: As-Milled (tool marks visible)
13:13
se6astian
is that a problem for us?
13:19
Bertl
maybe, but we can put it in the tumbler I guess
13:19
Bertl
we are talking black POM, yes?
13:21
se6astian
yes: Acetal Copolymer - Black (TECAFORM AH)
13:23
Bertl
okay shouldn't be a problem then, what would be the difference for a perfect finish?
13:23
Bertl
(price difference that is)
13:24
se6astian
there is no option
13:24
se6astian
its just a notice
13:24
Bertl
okay, so nothing we can do about it anyways :)
13:24
se6astian
yes :)
13:24
se6astian
and ordered!
13:28
se6astian
and sent another reminder to the guy who wants to sell his nikon f65
13:29
se6astian
he must be very happy he found a buyer as he has not replied to me for a week so far after I said I wiould buy it
13:29
Bertl
what about the second person?
13:36
PhilippeJ
hello !
13:36
Sasha_C
hi Philippe
13:39
se6astian
good idea I ll contact the other f65 sellers as well now
13:41
se6astian
done
14:05
Bertl
off for a nap ... bbl
14:07
dmj_nova
left the channel
14:10
dmj_nova
joined the channel
14:40
se6astian
PhilippeJ, I updated the doc for tomorrows meeting with cmosis
14:40
se6astian
is there anything else you want to prepare/discuss before meeting them?
14:47
PhilippeJ
well
14:47
PhilippeJ
let me check
14:48
PhilippeJ
about them being featured
14:48
PhilippeJ
on the crowdfunding campaing
14:48
PhilippeJ
do we ask for something in return ?
14:48
PhilippeJ
would they like to partner with us somehow ?
14:49
PhilippeJ
Also, se6astian
14:50
PhilippeJ
(sorry postedtoo fast
14:50
PhilippeJ
)
14:50
PhilippeJ
can you explain me how the sensor is attached
14:50
PhilippeJ
when you say aluminium
14:50
PhilippeJ
is it the sensor socket ?
14:51
se6astian
I would not ask for something in return, maybe they will feel creative themselves
14:51
se6astian
like backing the campaign when the time comes
14:52
se6astian
the sensor is screwed directly to konstantins lens mount
14:52
se6astian
which is aluminum and in the new version I just ordered a hard plastic
14:52
PhilippeJ
ok other side
14:52
PhilippeJ
sensor mount
14:53
PhilippeJ
something to lock it down and avod vibrations for instance
14:53
PhilippeJ
?
14:56
se6astian
on the other side (back) is the andon socket
14:57
PhilippeJ
yep
14:57
PhilippeJ
ok I'll ask for general recomendations
14:57
PhilippeJ
at least I know what we do
14:57
PhilippeJ
andon socket : what material is it ,
14:57
PhilippeJ
?
14:58
Sasha_C
left the channel
15:05
se6astian
hmm, I am afraid we dont know exactly
15:05
se6astian
some kind of plastic
15:05
se6astian
latest crowd funded camera: http://www.indiegogo.com/projects/panono-panoramic-ball-camera
15:06
se6astian
just wired the money to firstcut after they called me again to check if we really have no vat id :)
16:05
PhilippeJ
se6astian, is the lens mount much cheaper with hard plastic ?
16:14
se6astian
yes
16:34
se6astian
I need a bit of help with regular expressions
16:34
se6astian
I have a function:
16:34
se6astian
$txt = preg_replace('"<!--import-ignore-start-->[\\s\\S]+<!--import-ignore-end-->"', "", $txt);
16:35
se6astian
that removes anything between the import-ignore tags
16:35
se6astian
but now the problem seems that I have multiple of such tags in the content
16:35
se6astian
and instead of removing the content between each of them, the result is that all other content between the first opening and last ending tag are stripped
16:43
Bertl
back now ...
16:45
se6astian
I bet you are good with regex :)
16:45
Bertl
regular expressions are not the best choice to match strings
16:45
se6astian
http://regexr.com?375iq
16:46
Bertl
what exactly do you want to do?
16:47
Bertl
btw that page seems to load forever
16:48
se6astian
basically this: "some content that should stay <remove-start>some content that should be removed<remove-end> some more content that should stay <remove-start>some more content that should be removed<remove-end> and some content that should stay"
16:48
se6astian
I only want to have the content to stay at the end
16:50
Bertl
what file contains those tags? i.e. what syntax/semantic does it follow?
16:50
Bertl
html? xml? unknown?
16:53
se6astian
html
16:56
PhilippeJ
Ok see you guys !
16:59
Bertl
http://regexr.com?375j3
16:59
PhilippeJ
And for your enjoyement : openrisc linux emulator in javascript : http://jor1k.widgetry.org/
16:59
PhilippeJ
left the channel
17:03
FergusL
Hi here!
17:03
se6astian
haha, the "?" did the trick
17:03
se6astian
Hi FergusL
17:03
se6astian
many thanks Bertl
17:03
se6astian
I am just testing Gabes DNG converter
17:03
se6astian
it works
17:03
se6astian
I am looking at the picture you took of the IT8 chart in ufraw
17:04
se6astian
though the colors are "anywhere"
17:05
se6astian
actually it looks like the bayer pattern pixel order is not set correctly
17:05
FergusL
Oh the dng converter is done! Great!
17:05
FergusL
Maybe it's an int/float issue
17:06
Bertl
well, I uploaded a png as well IIRC
17:06
FergusL
Depends what dng expects
17:08
se6astian
yes I am looking at the png right now
17:08
se6astian
there is no position I can turn the whitebalance sliders to to get colors anywhere need what the png looks like :)
17:08
se6astian
*near
17:08
FergusL
Can I take a look at the code ? (haven't looked at this channel in days)
17:09
Bertl
yes, I'd be interested in the code as well
17:09
se6astian
https://github.com/apertus-open-source-cinema/CMV12000_DNG_Writer
17:09
se6astian
you need numby as well
17:09
se6astian
sudo apt-get install python-numpy
17:18
se6astian
the self_10ms.raw16 file works fine though
17:18
se6astian
IT8_incand.raw16 doesnt
17:20
Bertl
yeah, I know why, it is because the reader doesn't check that the image is Y flipped
17:20
Bertl
so the bayer pattern changes
17:20
se6astian
ah yes!
17:21
Bertl
for the older images it was RGGB, now it is GBRG
17:22
se6astian
emailed gabe
17:26
gcolburn
joined the channel
17:27
Bertl
welcome gcolburn!
17:27
se6astian
hurray, the man of the hour :)
17:27
Bertl
I wonder why :)
17:27
gcolburn
saw the email
17:27
se6astian
to get rid of the fixed-pattern-noise we should try this approach: https://en.wikipedia.org/wiki/Flat-field_correction
17:28
gcolburn
so if I understand correctly should y columns are completely reversed, so instead of RGGB its reading GBRG?
17:29
Bertl
se6astian: yup, that is what is planned
17:29
Bertl
gcolburn: the latest image is y-flipped, so the bayer pattern changes
17:30
Bertl
this will, as a matter of fact, be the default, but it is recorded in the sensor registers
17:30
gcolburn
is that an image other than the test one I was uploaded with the source?
17:30
Bertl
we won't be able to do X flips because we only have 32 channels
17:30
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/RAW/IT8_incand.raw16.xz
17:32
gcolburn
i'm looking up the mirroring in the spec right now
17:32
Bertl
(so what I'm trying to say is, that we can do RGGB and GBRG, but GBRG will be more common for normal images)
17:33
gcolburn
okay. you can specify either in the DNG, and you can also specify mirroring
17:33
Bertl
with y-flip enabled (sensor registers) the image is 'normal'
17:34
Bertl
with y-flip disabled, naturally the image is flipped
17:34
Bertl
(i.e. mirrored)
17:34
Bertl
this is the result of the coordinate system the sensor uses
17:37
gcolburn
if we embed the registers in the DNG private data we could put the correct tag in
17:38
Bertl
I'd definitely add the registers to the dng, we will need them later, but for now, a simple change to the other pattern should suffice
17:45
FergusL
hi gcolburn
17:48
gcolburn
I'll look at the new file and try to add the registers when I can
17:49
FergusL
wow you wrote DNG.py yourself I guess ?
17:50
gcolburn
yeah. its been a good learning experience to understand TIFF/DNG. To implement the full spec as a reader would be a lot of work. luckily we just need to write DNG files :)
17:51
FergusL
yes, very well done !
17:51
FergusL
can be reused elsewhere as well
17:53
gcolburn
does anyone have software that can read Cinema DNG?
17:53
gcolburn
when I was reading the spec, one implementation is just a sequence of DNG files numbered appropriately
17:54
FergusL
I too have made a small python script that reads the raw files, but I didn't think the struct module would work, but struct.unpack indeed is great
17:54
intracube_
joined the channel
17:54
FergusL
hi intracube
17:54
intracube_
hey FergusL
17:55
intracube
left the channel
17:56
intracube_
changed nick to: intracube
17:59
gcolburn
my wifi connection is too slow where I'm at (not at home), so its taking forever to download the latest image. you could try this to change the bayer pattern. at the end of the CMV12000toDNG file before the write call paste this:
17:59
gcolburn
rawIFD.getTag('CFAPlaneColor').value[0] = 1 #G
17:59
gcolburn
rawIFD.getTag('CFAPlaneColor').value[1] = 2 #B
17:59
gcolburn
rawIFD.getTag('CFAPlaneColor').value[2] = 0 #R
18:01
Bertl
only 3 entries?
18:02
gcolburn
yeah
18:03
Bertl
nah, all red/blue now
18:04
FergusL
maybe by reading the file backwards ?
18:04
gcolburn
oh wait. there is a 4th
18:04
FergusL
I had issues with the Y inverted (or not) pages
18:04
gcolburn
rawIFD.getTag('CFAPlaneColor').value[3] = 1 #G
18:05
gcolburn
Does adding that fix it?
18:05
dmj_nova
se6astian: meeting with cmosis?
18:07
Bertl
gcolburn: not really
18:07
gcolburn
so the TIFF 6.0 spec supports defining the orientation(mirroring), but DNG states that TIFF/EP is not compatible with it, so we'd have to decide whether to try that route, or always just rearrange the data
18:08
Bertl
we cannot rearrange the data
18:08
Bertl
that will change the pixel positions/meaning
18:09
Bertl
we can always flip it, but that will require a mirror tag
18:09
Bertl
i.e. the original images were all mirrored
18:21
Bertl
rawIFD.getTag('CFAPlaneColor').value[0] = 1 #G
18:21
Bertl
rawIFD.getTag('CFAPlaneColor').value[1] = 0 #R
18:21
Bertl
rawIFD.getTag('CFAPlaneColor').value[2] = 2 #B
18:21
Bertl
rawIFD.getTag('CFAPlaneColor').value[3] = 1 #G
18:21
Bertl
this gives something which could be correct
18:22
Bertl
the colors are still terribly off IMHO
18:24
gcolburn
are you setting the white balance?
18:24
Bertl
no, I'm using display to display the image
18:24
gcolburn
i don't have any code to determine it, so its using whatever was in the original Canon image with its profile
18:24
gcolburn
that's probably why then
18:25
Bertl
good then add the tags above for now and we should be fine
18:25
gcolburn
what i do is I open it in a raw converter and select a neutral gray area and it looks beter
18:25
Bertl
btw, why can't we remove the canon info?
18:25
Bertl
and e.g. replace it by a neutral one?
18:26
gcolburn
we can do that. I just need to look at the calibration array details and set it to neutral
18:26
gcolburn
when I completely removed them photoshop and lightroom couldn't recognize the file
18:27
Bertl
I see, okay, so probably next task
18:27
gcolburn
for auto-white balance you can just run it through dcraw, which just has one source file to compile: http://www.cybercom.net/~dcoffin/dcraw/
18:27
Bertl
and please try somehow add the registers as binary blob
18:27
se6astian
dmj_nova, meeting with cmosis? yes, philippe and oscar will meet them tomorrow
18:27
Bertl
s/try/try to/
18:27
dmj_nova
ah, cool
18:28
Bertl
gcolburn: so that we can extract that information lateron
18:28
se6astian
you can look at the Elphel jp4 (jpeg raw variant) to DNG converter for reference:
18:28
se6astian
https://github.com/apertus-open-source-cinema/jp4tools/blob/master/JP4toDNGconverter/source/elphel_dng.c
18:30
se6astian
or this alternative implementation based on the adobe sdk
18:30
se6astian
https://github.com/apertus-open-source-cinema/jp4tools/blob/master/Movie2DNG/src/dngwriter.cpp
18:30
se6astian
line 262+ deals with bayer patterns
18:37
FergusL
(gcolburn: why python, if I may ask ?)
18:39
gcolburn
I tried to get the adobe sdk to compile and had problems with deprecated functions in 3rd party libraries
18:39
gcolburn
their sdk would probably be the best route if one wanted to read any DNG. but for writing I figured it would be simpler to just write some code
18:43
gcolburn
I'll definitely look at their code though. that would be a handy reference. thanks.
18:46
se6astian
the adobe sdk is rather slow as the author of the above code noted
18:47
se6astian
the first link of the two is based on a patched libtiff to write dng
18:47
se6astian
much faster
18:47
se6astian
but inconvinent as it requires an outdated libtiff to patch to compile the code
18:49
gcolburn
yeah. I started on a C++ version but I haven't had time to work on it. prototyping in python has been helpful
18:50
FergusL
a C++ version of the dng writer ?
18:52
gcolburn
yes
18:52
gcolburn
how import is it to be able to write the DNG without seeking?
18:52
gcolburn
important
18:52
Bertl
I'd say you can do that with a memory buffer in python
18:53
Bertl
just create the image with seeks first and then dump it to stdout
18:53
FergusL
you can map the raw file to memory and then seeking would be faster ? or is that irrelevant to the problem ?
18:53
Bertl
irrelevant, we are talking output
18:54
gcolburn
yeah I could try that
18:54
FergusL
ok
18:54
Bertl
gcolburn: is there a dng specific compression for the data?
18:55
Bertl
(lossless compression I mean)
18:57
gcolburn
I believe it supports lossless jpeg
18:57
gcolburn
"Two Compression tag values are supported:
18:57
gcolburn
•Value = 1: Uncompressed data.
18:57
gcolburn
•Value = 7: JPEG compressed data, either baseline DCT JPEG, or lossless JPEG compression."
18:58
gcolburn
the spec includes thumbnails though, so I need to check that applies to the raw data, not just thumbnails
18:58
Bertl
hmm, okay
18:58
Bertl
so we stick with .dng.xz for now I guess :)
19:00
gcolburn
yeah. I'd probably need to find a library to do the compression.
19:01
se6astian
forget the compression, its mapped to 8 bits, not worth the trouble to save a few MB of space
19:02
Bertl
hmm?
19:03
se6astian
compressed DNG uses JPEG with higher bit depths remapped to 8 bits
19:03
Bertl
25222907 Nov 12 19:24 CMV12000_DNG_Writer/img.dng
19:03
Bertl
14164816 Nov 12 19:24 CMV12000_DNG_Writer/img.dng.xz
19:03
Bertl
this is what I'm talking about
19:04
se6astian
ah, I see :)
19:04
Bertl
i.e. we want to compress the DNGs somehow
19:04
Bertl
either the data within the dng format (according to some dng spec)
19:04
se6astian
bash script?
19:04
Bertl
or the entire file, which can happen in the unix pipe, yes
19:07
FergusL
what do you mean by "seeking" ?
19:10
Bertl
the code currently writes the header, then the data and then updates the header information according to the data (simplified)
19:11
Bertl
to do that, it needs to move back (seek) to the beginning, which is not possible in a pipe
19:11
gcolburn
when you seek you tell it which byte offset in the file to go to.
19:11
FergusL
okay, got it :) thanks guys
19:12
FergusL
yes, I use seek in my py script as well
19:12
gcolburn
to write it without seeking, i would have to calculate all offsets to where the tag values and image will be in advance, and then write them out at the beginning. sometimes TIFF stores the value in the tag, and sometimes the tag points to the value, so pre-calculating all this would take a bit of handling different cases
19:13
FergusL
seems much clearer to me now, I understand what you were discussing now indeed
19:13
FergusL
I think Bertl suggestion is great
19:16
gcolburn
I got the orientation tag to be interpreted by photoshop
19:16
gcolburn
here's the code you can append at the end:
19:16
gcolburn
orientationTag = TAG()
19:16
gcolburn
orientationTag.tag = 274
19:16
gcolburn
orientationTag.type = 3
19:16
gcolburn
orientationTag.value = np.zeros(2,dtype=np.uint8)
19:16
gcolburn
orientationTag.value[0] = 4
19:16
gcolburn
orientationTag.value[1] = 0
19:16
gcolburn
orientationTag.count = 1
19:16
gcolburn
dng.ifd.tags[274] = orientationTag
19:17
gcolburn
this sets the orientation: orientationTag.value[0] = 4
19:17
gcolburn
3 = The 0th row represents the visual bottom of the image, and the 0th column represents
19:17
gcolburn
the visual right-hand side.
19:17
gcolburn
4 = The 0th row represents the visual bottom of the image, and the 0th column represents
19:17
gcolburn
the visual left-hand side.
19:17
gcolburn
there are other options as well
19:18
Bertl
k, great!
19:18
gcolburn
1 is default
19:22
FergusL
you can create arbitrary files with mmap in python but you need the exact length
19:22
FergusL
or use resize() before each write() call
19:24
FergusL
(http://docs.python.org/2/library/mmap.html)
19:29
Bertl
stringio provides a file like class in memory
19:30
Bertl
i.e. it can be used like a file, but is held in memory, and can, at the end, simply be dumped to stdout
19:30
gcolburn
I've got to head out for a bit...
19:30
Bertl
k, cya!
19:30
gcolburn
i'll look into those options
19:31
gcolburn
left the channel
19:31
FergusL
looks better
20:05
Sasha_C
joined the channel
20:08
rexbron
Bertl: for the love of all that is holy, do not use the 8bit jpeg compression aspect of DNG. It destroys images.
20:10
rexbron
Bertl: the Blackmagic Pocket camera just got it's raw firmware update with lossless compression, likely a standard algo like the ones used in EXR compression.
20:13
Bertl
do not worry, we have no plans to use that :)
20:14
rexbron
:) phew
20:15
dmj_nova
yeah, not doing 8 bit jpeg compression
20:15
dmj_nova
(well we can't guarantee that nobody will add that as an option)
20:16
dmj_nova
because we really can't guarantee anything won't happen that its technically possible for a user with the right skillset to implement
20:17
rexbron
dmj_nova: I'm fine as long as it's not forced upon the user
20:17
dmj_nova
but we're definitely doing lossless raw
20:18
dmj_nova
I'd say lossless compression is a likely feature though
20:19
dmj_nova
which is basically the same principle as a zip file or a tarball, just using entropy encoding
20:20
gcolburn
joined the channel
20:21
Bertl
rexbron: the beauty of open source/open hardware is that the user is not forced to anything
20:22
rexbron
dmj_nova: you do also need to make sure that any non-spec compression has vendor support, the largest of which is BMD and Adobe, or better yet copy what they do ;)
20:22
Bertl
even if the community or the developers do not support what you want or do not want, there is always the option to hire some developer/programmer to add/remove the parts you like
20:23
dmj_nova
rexbron: yes, of course!
20:24
dmj_nova
I honestly doubt we'd be able to keep lossless raw support off the axiom even if we tried
20:26
Bertl
the only reason I was asking was because it would make sense to compress the dng data by color channel with knowledge of the row length
20:27
Bertl
(and I would expect something like that to be part of the DNG specification)
20:28
Bertl
(still talking lossless compression here)
20:29
dmj_nova
Bertl: perhaps there's a mode that separates the planes?
20:29
Bertl
I'm pretty sure there is, doesn't help much with external compression though
20:30
dmj_nova
Bertl: what do you mean external compression?
20:31
dmj_nova
I mean that many raw formats compress their data
20:31
Bertl
if there is no lossless compression in dng working on the image data, then I need to compress the dng as binary blob
20:32
Bertl
if we talk about 'our own' raw format, then we can do whatever we like of course
20:34
gcolburn
Here is what the DNG spec says: "For lossless JPEG, the internal width/length/components in the JPEG stream are not required to match the strip or tile's width/length/components. Only the total sample counts need to match. It is common for CFA images to be encoded with a different width, length or component count to allow the JPEG compression predictors to work across like colors."
20:35
gcolburn
CFA=Color Filter Array (Bayer)
20:36
Bertl
okay, so that might be an option after all
20:36
gcolburn
the DNG spec doesn't say much else
20:37
dmj_nova
dng allows for lossless jpeg compression, which should give about 2.5:1 reduction in size
20:38
dmj_nova
hmm, that may mean we need to do two planes of green?
20:40
Bertl
well, it's probably better to keep the two green channels separated anyway
20:41
Bertl
(they will overlap when binning is used)
20:41
FergusL
I'm not quite following everything here !
20:42
Bertl
not that important, mainly speculations and hand waving ... but if you have specific questions, just ask
20:43
dmj_nova
I'd rather avoid having our own raw format unless it provides a clear and tangible benefit
20:44
dmj_nova
like say reducing file size on lossless compression by a factor of 2 or something
20:44
dmj_nova
which I doubt is possible unless somebody designed DNG in a very stupid way
20:45
Bertl
I'd say a factor of 50% would already be a good argument :) but anyways, I agree, if existing standards can cope with what we need/want, then we should strive to use them
20:45
Bertl
(50% = reduction to 2/3rd)
20:47
dmj_nova
yeah
20:52
FergusL
Bertl: so how is it going with the IT8 chart ?
20:52
Bertl
not working on the IT8 ATM, I've done the incandescent snapshot and that's it for me (ATM)
20:53
Bertl
it's up to you folks to make heads and tails out of it now :)
20:53
FergusL
cool !
20:53
FergusL
it's in the same folders as others ?
20:53
FergusL
(haven't checked this channel much for the past few days)
20:53
Bertl
in the raw folder, yes
20:57
rexbron
Bertl, dmj_nova: BMD have implimented the lossless Huffman part of the cDNG spec
20:57
FergusL
Bertl: do you have a color temperature value for the incandescent source ?
20:57
dmj_nova
rexbron: yes, that's likely what we'll do as well
20:57
Bertl
nope, but the guess around 2900K is probably valid
20:58
dmj_nova
only possible deviation would be if there's a way to achieve significantly better compression than lossless huffman
20:59
dmj_nova
which I suspect is only possible if they're blocking data for huffman in a stupid way or in pathological corner cases
21:00
rexbron
dmj_nova: Even if that was the case, adding an extra step in your post workflow until it gains standard status and then having to get vendors to support it stack the deck against you.
21:00
dmj_nova
rexbron: right
21:01
dmj_nova
I'm not saying we intend to switch even if there is a way.
21:01
[1]se6astian
joined the channel
21:01
rexbron
The only reason ML gets away with it is because, as I understand it, they are directly dumping memory to disk and that needs to be converted into something that is recognizable as an image
21:01
dmj_nova
and that could again become an *option* for say shooting longer with the same cartridge
21:02
rexbron
dmj_nova: sure but also consider the context of production. You aren't going to be shooting uncompressed or even lightly compressed raw for documentary.
21:03
rexbron
The feature I shot was 10Tb of camera originals by the end of the 8 day shooting schedule
21:03
dmj_nova
*cough* russian ark
21:03
rexbron
dmj_nova: different beast entirely
21:03
dmj_nova
though that's a corner case again
21:03
rexbron
that's more factual than documentary, semantic difference but important
21:03
se6astian
left the channel
21:03
[1]se6astian
changed nick to: se6astian
21:03
dmj_nova
most features, you're quite correct don't need that
21:04
dmj_nova
right now the "custom raw format" thing is just a thought experiment
21:04
rexbron
I'm not entirely up to speed but has the team defined exactly who the audience is for Axiom?
21:04
rexbron
dmj_nova: sure, experiment away :)
21:05
dmj_nova
probably the best test is just break a few images down in different ways and compare size to what the cinema dng allows
21:05
Bertl
I don't think we need to 'define' the audience, we will simply see who shows interest
21:05
rexbron
troy_s_ could talk at length on context in design
21:06
rexbron
Bertl: unfortunately you do. You can't change camera ergonomics easily or cheaply. Pros don't scream loudly, amatures and hobiests do. Blackmagic's forums are proof enough of that.
21:07
rexbron
Pros find tools that work for them. If it doesn't present it self that way, it gets passed on.
21:08
dmj_nova
rexbron: Bertl: it really does need to be a highly useable tool for filmmaking foremost
21:08
dmj_nova
good ergonomics and workflow
21:08
rexbron
Here is a link to my BMCC rig, http://imgur.com/Rp7Z9bt
21:08
rexbron
The camera is one of the least expensive parts there
21:08
rexbron
the actual body I mean
21:09
Bertl
rexbron: well, I guess then the pros have to start speaking up if they want to shape the axiom :)
21:09
rexbron
Bertl: that's why i'm here ;)
21:09
Bertl
I thought so :)
21:10
rexbron
http://www.aaton.com/images/delta-700c.jpg
21:10
rexbron
probably one of the most ergonomic cameras ever designed
21:10
rexbron
too bad it bankrupted Aaton
21:10
dmj_nova
rexbron: what did you feel made it ergonomic?
21:11
rexbron
dmj_nova: Notice how it's curved, meant to put the center of balance below the shoulder, not above it
21:11
rexbron
ergonomic for handheld I should clarify
21:12
rexbron
notice how the viewfind is integrated into the body but ajustable
21:12
rexbron
notice how all the controls for the camera are on the opposite side (camera right) to the oporater, the camera is meant to have an assistant
21:13
Bertl
maybe it might be a good thing to put the various aspects together in a document (via email to the mailing list or probably even better on the wiki?)
21:13
rexbron
Bertl: sure
21:13
dmj_nova
rexbron: so it's designed to be ergonomic for a 2-3 person team to use?
21:13
rexbron
dmj_nova: yes
21:13
Bertl
btw, are there 'left handed' camera(wo)men?
21:14
rexbron
Bertl: there are left eyed camera ops, but the camera is always on the right shoulder. Totally convention, no other reason other than right-handed domanance
21:14
dmj_nova
personally, I would like to see the axiom as having a "brain" and a "shell"
21:14
dmj_nova
so you have a very small unit that houses sensor
21:14
rexbron
dmj_nova: that sounds like an epic and epics are terrible to work with in production
21:15
dmj_nova
and a shell that the user interfaces with
21:15
rexbron
the image is great
21:15
dmj_nova
I think that may be a result of a terrible shell
21:15
rexbron
but all the ergonomics of the camera were sacrificed for 3D
21:15
dmj_nova
explain that last bit?
21:15
rexbron
the Epic was designed to be the smallest it could be to shrink the size of 3D mirror rigs
21:15
Sasha_C
left the channel
21:16
rexbron
as a result, the camera has no mounting points for accesories
21:16
rexbron
or anything, not even a battery without extras
21:16
rexbron
great for modularity, terrible when the director wants to go from studio to handheld and it takes 5 minutes,
21:17
rexbron
and they are complaining about how their assistants aren't working fast enough or hard enough
21:17
rexbron
I'm not saying that there isn't a place for that form
21:17
dmj_nova
rexbron: you mean because you have to take everything apart and put it all back together?
21:17
rexbron
dmj_nova: yes
21:17
rexbron
in the photo of the rig I posted, my BMCC
21:18
rexbron
also consider cost, do you want to put that into a body that just works or do you want to create a secondary ecosystem for camera accessories to make your body work with the exisiting equipment standards
21:18
Bertl
looks futuristic!
21:18
rexbron
Bertl: the epic, Aaton or BMCC?
21:18
Bertl
(at least to me :)
21:19
Bertl
your rig
21:19
rexbron
hah
21:19
rexbron
it's a pain in the ass that cost me more than the body
21:19
rexbron
and it's a decent one lol
21:20
dmj_nova
rexbron: my own thoughts for a "shell" system is that the shell should be a complete and singular thing
21:21
Bertl
rexbron: do I see that right that it has 3 displays?
21:21
rexbron
another example of Aaton's style of design
21:21
dmj_nova
and the "brain" just plugs in and is interchangeable with separate bodies, even if a given user never wants that.
21:21
rexbron
Bertl: two, the built in touch screen and SmallHD AC7
21:22
rexbron
dmj_nova: that could be great but there in lies many design challenges
21:22
Bertl
rexbron: what's the thing on the right side?
21:22
rexbron
Bertl: hinged V lock battery plate by viewfactor
21:23
Bertl
ah, so that's the battery ..
21:23
rexbron
Yes
21:23
dmj_nova
rexbron: have you seen the axiom modules concept?
21:23
rexbron
dmj_nova: I have from the website, my concern is mechanical robustness
21:23
dmj_nova
Yes, I have a feeling we'll be learning more about that quite soon
21:24
dmj_nova
I believe se6astian has just ordered some housing pieces.
21:24
dmj_nova
am I right on that se6astian?
21:24
rexbron
dmj_nova: Production is hard on cameras. What material are you looking to have them made out of?
21:24
dmj_nova
not sure what the final plan is
21:25
Bertl
for now, we are still working on the prototype
21:25
Bertl
which will get a shoe box :)
21:25
Bertl
(a fancy, transparent shoe box, but nevertheless a shoe box :)
21:26
dmj_nova
rexbron: I'd imagine that a shell would fit around that assembly in the concept
21:27
rexbron
Bertl: the first engineering prototype of the Red One was in a wooden box :P
21:27
Bertl
so we are already far ahead of that then :)
21:27
dmj_nova
rexbron: you mentioned it's hard to go from handheld to studio mount with the RED, what is the difference in setup between the two (other than I assume having to exchange shoulder mount + handles for a tripod)
21:28
rexbron
dmj_nova: depending on the production, studio mode may consist of just a tripod, so having an arri standard dovetail and shoulder brace makes it fairly quick
21:29
rexbron
but say your on a production zoom like an Angineaux 24-290 or 28-340. Those lenses weigh 25-30 lbs by themselves.
21:30
dmj_nova
note to self: never put a production zoom on my moco rig
21:30
rexbron
so you need to switch matte boxes (from say a 6x6 to a 4x5.65 clipon), lenses and a few other tings
21:31
rexbron
Take my rig for example
21:32
rexbron
Because of how small the body is, how small the lenses are, and where the monitor needs to be for me to see it, all the weight is over my hands
21:32
rexbron
it's not balanced at all
21:32
rexbron
It makes operating a more taxing experience that it needs to be
21:32
rexbron
but since that rig is nearly all alluminum, it already weighs 25lbs
21:33
rexbron
I can look at remounting the battery to the back on longer rods to compensate a bit
21:34
rexbron
but you get into rediculus dslr style rigs
21:34
dmj_nov
is a dslr shooter and avoids stupidly ridiculous rigs
21:35
dmj_nova
So you're saying that weight distribution is key
21:35
rexbron
yes
21:35
rexbron
also by lowering the center of gravity to below the shoulder, you get less shake
21:36
rexbron
more fluid larger movements and less judder
21:36
rexbron
http://www.ocon.com/inspiration/labs/rod-standards-explained/
21:36
rexbron
^ Side note, clearest explaination of the different rod standards in cinema
21:37
rexbron
In North America, the east cost uses 15mm studio offset and the west cost uses 19mm. troy_s_ does Vancouver use 15mm or 19mm?
21:38
rexbron
19mm is nomially stronger but means you need to carry two sets of rods for studio and lightweight
21:38
rexbron
Axiom, be it with a shell or not, should adhere to those standards to accessories just work
21:38
rexbron
s/to/so
21:43
Bertl
not sure what standards there are (to accessories)
21:44
Bertl
but I don't see any reason not to follow standards if they are open
21:45
dmj_nova
yes, we should support standard accessories
21:45
dmj_nova
rexbron: why is the RED bad with the rod system and accessories?
21:46
mikkael
joined the channel
21:46
Bertl
welcome mikkael!
21:53
rexbron
dmj_nova: Mostly my complaints come as an assitant. You rent a Red from an owner operator and it's missing half the things you need to make it functional.
21:53
rexbron
when getting a camera from a rental house, they have all that stuff
21:54
rexbron
is Axiom including rental houses in their target audience?
21:55
Bertl
we are not excluding anything atm
21:58
dmj_nova
so it's got rod compatibility, but it doesn't come as a functional unit?
21:59
dmj_nova
why wouldn't the operator have all the needed things?
21:59
rexbron
huh, I can't save an edit to a discussion page on the wiki without getting into a captcha loop
21:59
rexbron
dmj_nova: because they bought a ferarri but don't have any money left for tires
21:59
dmj_nova
ah
22:00
rexbron
it's happened so often in Toronto
22:00
dmj_nova
then how do they use it, just making do with minimal stuff?
22:00
rexbron
like a big dslr
22:00
rexbron
Canon EF mounts on the epic is the biggest culprate
22:00
dmj_nova
ah, yeah I could see people doing that
22:01
dmj_nova
so like a big dslr with worse ergonomics than the Canons
22:01
rexbron
Bertl: re: rod standards, they are just physical mesurements
22:01
rexbron
dmj_nova: yup but the operators try to get full rate for the rental
22:03
dmj_nova
so they're renting out the body-only as a complete-package
22:04
rexbron
as little as they can get away with
22:05
rexbron
http://mythofcapture.com/the-cat-and-the-box-part-1-of-3-an-essay-on-cameras-form/
22:05
rexbron
great essay on camera egonomics
22:05
rexbron
much more on the art aspect of design
22:07
dmj_nova
I will say that DSLRs have brilliant ergonomics for a box camera
22:07
rexbron
Does a wiki account need to be activated?
22:07
rexbron
dmj_nova: for stills, not for motion
22:07
rexbron
if they had the same optical viewfinder in motion, I'd be more likely to agree with you
22:08
rexbron
and that's not to say it can't be done
22:08
rexbron
but you just can't replicate the feel of a 35lbs camera balanced on a shoulder. That's what cinema handheld looks like
22:08
dmj_nova
the hardest part is stabilization and follow focus with them
22:09
dmj_nova
Oh, not saying DSLRs are good at replicating that
22:09
dmj_nova
they're much better at the things that a cinema camera on the shoulder isn't good at and vice versa
22:10
se6astian
rexbron, did you manage to save the wiki page?
22:10
rexbron
nope, got a message saying the link I posted was blacklisted
22:11
se6astian
the captcha is normal, hmm thats mollom then
22:11
se6astian
let me check
22:11
rexbron
se6astian: it's the link I pasted above
22:13
rexbron
With future capture power arriving in smaller camera bodies we increasingly have to choose between two types of camera form and two distinct operating possibilities: Box or Balance. Rigid stabilization, tripod or otherwise, will limit some of the differences between these two forms. However, if we’re departing from stabilization we’ll have to make the choice between capturing with a Box Camera in our hands, or a Balance Camera on our shoulder. A
22:13
rexbron
form cannot be both things at once, in capture we will have to choose one.
22:13
se6astian
can you save the page just without the link?
22:13
rexbron
To pull a quote from the intro
22:13
se6astian
I moved you into the "emailconfirmed" group
22:13
se6astian
maybe that lightens the limits
22:13
rexbron
ues
22:13
rexbron
yes, I can post now
22:14
se6astian
great
22:15
rexbron
se6astian: Does the captcha happen every edit?
22:16
rexbron
ok, well I need to head out to dinner with some friends. It was great chatting with you all!
22:17
dmj_nova
rexbron: chat again, I learned quite a bit
22:19
se6astian
yes the captcha is always enabled
22:19
se6astian
we have zero spam since implementing that ;)
22:20
dmj_nova
se6astian: So I think we may need to design separate "shells" for Box and Balance
22:26
se6astian
sorry I was busy and have not follow the discussion, can you make a summary/proposal to the mailing list?
22:27
se6astian
time for me to go to bed
22:27
se6astian
good night!
22:27
se6astian
left the channel
22:28
gcolburn
left the channel