Current Server Time: 17:40 (Central Europe)

#apertus IRC Channel Logs

2014/03/14

Timezone: UTC


00:00
eadmund
because they are very peaky, and so stuff like integrating the observer function becomes hairy.
00:00
Bertl
yeah, I can imagine
00:01
eadmund
even *if* you manage to measure the led accurately, which consumer-grade equipment may not necessarily be able to do.
00:01
eadmund
eg I would not bet on an i1Pro or equivalent, maybe a minolta.
00:01
eadmund
might work
00:02
Bertl
well, if you just want to know the spectral profile, measuring the LED shouldn't be too hard, you can assume a gaussian distribution I guess (centered at the major wavelength)
00:03
Bertl
and the center wavelength is known for most LEDs
00:04
eadmund
the problem is finding the center accurately.
00:04
Bertl
if it is given as e.g. 560nm (from the manufacturer) does it really matter if it is 558nm or 562nm?
00:05
Bertl
I mean, would you be able to tell the difference?
00:05
eadmund
yes. as I was saying all the power is delivered on one spectral locus or near one locus.
00:06
eadmund
so if the wavelength is slightly off in a region where the observer functions overlap
00:07
eadmund
your observer functions will differ quite significantly, I believe. Also the observer function computations in the instruments are not really designed to be precise in such a situation.
00:07
Bertl
okay, but that would be a very bad choice of wavelength for the LED :)
00:07
eadmund
so different instruments will compute xyz differently if they're a bit old.
00:08
eadmund
unfortunately this is what happens for rgb-le displays (not white led displays).
00:09
Bertl
really? interesting
00:09
eadmund
maybe we can talk about the latest operas instead of this idiocy?
00:09
Bertl
what operas have you seen recently?
00:10
eadmund
luckily, none :)
00:11
Bertl
so you're not even prepared to talk about operas :)
00:11
eadmund
any new images form the camera?
00:11
Bertl
tons
00:12
eadmund
I have a nice reading list of wavelet transform books, if you're interested. I have discovered booza.org :)
00:12
eadmund
bookza.org :)
00:13
eadmund
so - show a pic!
00:13
Bertl
give me a minute to upload one
00:18
eadmund
I will talk to Trenz tomorrow. Maybe they are willing to publish the docs for their board.
00:18
sb0
left the channel
00:18
sb0
joined the channel
00:18
Bertl
that would be nice
00:19
eadmund
http://www.trenz-electronic.de/de/download/d0/Trenz_Electronic/d1/TE0701/d2/HW_design.html
00:20
eadmund
maybe not the right one?
00:20
eadmund
got it! http://www.trenz-electronic.de/de/download/d0/Trenz_Electronic/d1/TE0720-GigaZee.html
00:21
eadmund
looks fairly well documented to me :)
00:21
eadmund
open is a different story of course.
00:21
Bertl
no, the problem is, there is just a library to integrate the module
00:21
Bertl
i.e. no schematic or layout for the module itself
00:22
Bertl
and the library was contributed by a caring third party :)
00:22
eadmund
yes, I wonder who?
00:23
eadmund
I will ask. what level of hardware documentation d'you expect?
00:23
Bertl
steffen.mauch
00:23
Bertl
(it's in the library)
00:24
Bertl
well, first a complete schematic of the board (that is basically the essential documentation)
00:25
eadmund
yeah. They're supplying the Xilinx ref files, but I guess you still want schematics
00:25
Bertl
then some kind of layout/mechanical drawing so that the board could be built by a third party
00:25
Bertl
I do not care about software examples and demo code
00:25
eadmund
that I wonder whether they will supply :)
00:26
eadmund
I mean enough to reproduce the board.
00:26
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/RAW/rcn.raw12.xz
00:26
Bertl
here is your image
00:26
eadmund
yuck. what is .xz? can you uncompress it to something else?
00:27
Bertl
you can uncompress it with xz :)
00:27
Bertl
but if it helps, I can uncompress it for you
00:27
eadmund
please. . you cna recompress it as tiff lzw
00:28
Bertl
it is raw data, doesn't make much sense as tiff
00:28
eadmund
wait. I think I have xz.
00:28
Bertl
you can display it like this:
00:29
Bertl
xzcat rcn.raw12.xz | display -depth 12 -size 4096x3072 gray:-
00:30
Bertl
note that it is an image from an engineering sample (i.e. known defective sensor)
00:30
eadmund
I don't have "display".
00:30
Bertl
part of ImageMagick
00:31
eadmund
gimme a minute, I will see if pd can deal with it. what does depth 12 mean?
00:31
eadmund
gimme a minute, I will see if PS can deal with it. what does depth 12 mean?
00:31
Bertl
it is raw data with 12bit values
00:32
Bertl
i.e. 4096x3072x12bit
00:33
eadmund
hmm. I can read bytes, not sure about nibbles
00:38
eadmund
I cannot find the trick to read this into Photoshop, maybe you can turn it into some other format so I don't have to write a parser.
00:39
Bertl
I'm completely convinced now that PS sucks :)
00:39
eadmund
yeah. strangely enough I managed to read in CUPS raster files
00:40
eadmund
I'm the color geek for Gutenprint so I was forced to ...
00:40
Bertl
who forced you to use PS? :)
00:41
eadmund
well, I found that not a single Linux distribution can print a CMYK tiff correctly.
00:41
eadmund
unless you turn it into a pdf.
00:42
eadmund
the print system on Linux mostly does it's own thing.
00:43
eadmund
and if you feed it an rgb file, good luck with stopping your favorite distribution from color-managing the print path.
00:44
Bertl
I've given up on non-postscript printers
00:44
Bertl
and personally I prefer network printers anyway, so ...
00:44
eadmund
postscript is a proprietary Adobe thing :)
00:45
Bertl
yes, it is, but it has been around for so long and documented so well :)
00:45
eadmund
yeah, I love getting pstscript printers to compute my fractals :)
00:46
eadmund
so, where is that file?
00:46
Bertl
I love drawing flowers which change with the time of day/year :)
00:47
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/RAW/rcn.tiff.xz
00:50
sb0
left the channel
00:51
sb0
joined the channel
00:52
eadmund
I am getting a grey square with some junk in it. Is that your intention?
00:52
eadmund
btw, what do engineering sesnors cost?
00:53
Bertl
that is the FPN corrected image the sensor produces under weak illumination
00:53
Bertl
the junk is the result of having a defective sensor
00:53
Bertl
they roughly cost half of a good one
00:55
eadmund
yes. ugh. btw are you familoar with rawdigger?
00:55
eadmund
http://www.rawdigger.com/
00:56
Bertl
I know libraw, but I haven't heard of rawdigger before
00:56
Bertl
not sure what I would use it for?
00:56
eadmund
maybe you don't want to use the software, but you might look at the type of output it generates.
00:57
Bertl
like?
00:58
eadmund
hehe. well you might wish to generate raw histos on the fly for the user.
00:58
Bertl
we are already generating live histograms
00:58
eadmund
and for evaluating and calibrating sesnors it has a bunch of useful visualisations.
00:58
Bertl
like?
00:59
eadmund
talk with Iliah :)
00:59
Bertl
okay :)
00:59
Bertl
but for now, my work is done, and I'm off to bed ...
01:00
Bertl
have a good one everyone! cya!
01:00
eadmund
as I use mainly "semi-cooked" data and have my own toolchains I don't need it.
01:00
eadmund
Good night.
01:01
eadmund
left the channel
01:12
jerknextdoor
joined the channel
01:23
jerknextdoor
left the channel
01:30
sb0
left the channel
01:31
sb0
joined the channel
01:41
intracube
left the channel
01:42
sb0
left the channel
01:42
sb0
joined the channel
02:07
sb0
left the channel
02:07
sb0
joined the channel
02:53
troy_s
Bertl: Agree on shapers and matrix. if you can unpooch the data and get a consistent response, shaper + matrix is likely pretty decent.
03:30
sb0
left the channel
03:31
sb0
joined the channel
04:04
sasha_c
left the channel
04:34
sasha_c
joined the channel
06:38
sasha_c
Hey Bertl, is this to your liking: https://wiki.apertus.org/index.php?title=CMV12000_Register_Blocks
06:45
niemand
joined the channel
06:47
sasha_c
left the channel
06:53
niemand
left the channel
07:29
sb0
left the channel
07:32
sb0
joined the channel
07:54
se6astian
joined the channel
07:55
se6astian
good morning
08:08
Bertl
morning folks!
11:26
danhanes
joined the channel
11:46
danhanes
left the channel
13:15
Bertl
off for a nap .. bbl
13:31
intracube
joined the channel
14:53
Bertl
back now ...
14:59
rainer_vie_
joined the channel
15:45
danieel
left the channel
15:57
se6astian
time to leave the university
15:57
se6astian
left the channel
17:24
rainer_vie_
left the channel
18:09
se6astian
joined the channel
18:17
intracube
changed nick to: intracube_afk
18:27
se6astian
left the channel
18:31
se6astian
joined the channel
18:44
intracube_afk
changed nick to: intracube
19:51
intracube
left the channel
19:54
niemand
joined the channel
20:47
niemand
just found this when browsing through the talks of the last year: https://cfp.linuxwochen.at/de/lww2013/public/events/35
20:48
Bertl
thanks!
20:49
niemand
:)
20:50
niemand
would be great to have him also at the lww
20:50
niemand
we will invite him
20:51
niemand
and something else Bertl: one major topic of the conference will be 3d printing, concerning the casing of the cam :)
20:52
Bertl
okay, good to know ... do you know what technologies will be discussed?
20:55
niemand
well the cfp is running. but the people talking are very experience, so some of them are working on improvements of printers etc. E.g. we have Bernhard Kubicek who talked at fabcon
20:56
Bertl
ah yes, I know him well
20:57
Bertl
anything DLP related?
21:03
niemand
Nothing now. is your target to find people with experience in this field?
21:06
Bertl
more personal interest (i.e. just curious :)
21:47
troy_s
left the channel
21:48
troy_s
joined the channel
22:05
niemand
left the channel
22:31
se6astian
time for bed
22:31
se6astian
good night
22:31
Bertl
cya
22:31
se6astian
left the channel
22:35
eadmund
joined the channel
22:36
eadmund
so I had a chat with Trenz
22:36
Bertl
okay?
22:36
eadmund
you there?
22:36
Bertl
looks like :)
22:37
eadmund
their module is fully documented, except for the actual traces (gerber files)
22:37
eadmund
they have a carrier support board for development
22:37
eadmund
actually two of them. https://wiki.trenz-electronic.de/display/4X5B/TE0701+Carrier+Board+for+TE07xx+Series+User+Manual
22:38
eadmund
and of course they do custom designs, already did a CMOSIS chip once.
22:39
eadmund
actually their TE0701 looks interesting on paper because of the cameralink interface it has. I wonder if this can be ised as output ...
22:41
eadmund
anyway, they would be interested in doing a chip-carrier board. I guess such a board would be open source if one pays for it :)
22:43
eadmund
Any new images? :)
22:44
Bertl
well, I haven't found any schematic (yet) maybe you can point me to them
22:45
eadmund
I'll ask them for a link. They have stuff on github
22:45
Bertl
okay, please keep me updated
22:46
eadmund
I spoke to Mr. Trenz - so if he says the schematics are public, I'll believe him :)
22:49
eadmund
I still don't completely understand why you want to dump the exiting carrier board, it seems it would work with the Trenz module and their VITA carrier
22:50
eadmund
????
22:51
Bertl
I doubt it will work out of the box
22:52
Bertl
but we'll see once I have a look at the documentation
22:53
Bertl
also having a schematic is only the first step, as you well know ... i.e. what price will be get for which module, what availability is there, etc
22:57
eadmund
if it did, would that be useful?
22:57
eadmund
go ahead :) btw did you consider an APtina sesnor?
22:57
eadmund
how many bits/pixel d'you get out of the CMOSIS sensor?
22:57
eadmund
is it just 10 or 12?
22:57
eadmund
or d'you need to do a trick?
22:58
eadmund
I don't know. They have a price list and in fact stock for various qualities (temperature ratings) and can manufacture and test.
22:58
Bertl
the cmosis supports 8,10 and 12 bit, not sure the 12bit mode is really significant though
22:58
eadmund
for me the electronics price is much lower than sesnor price so I don't rally care.
22:59
eadmund
this is very problematic, because 10 bits is not enough for still photography, as channel imbalance costs you a stop
23:00
eadmund
I mean light temperature correction
23:00
Bertl
regarding Aptina: I havent seen any 'suitable' sensors there yet, but I haven't looked either
23:00
Bertl
so if you have specific sensors (with datasheets) you consider suitable, just drop me a link
23:00
eadmund
maybe you are using some integration tricks?
23:01
Bertl
you can definitely increase bit depth with sampling more than once
23:02
eadmund
strangely enough, no one wants to make datasheets *public*
23:02
Bertl
kodak doesn't seem to have a problem with that
23:03
eadmund
I see. And where are their CMOS area sensors?
23:03
eadmund
btw google "AR1411HS"
23:03
Bertl
http://www.truesenseimaging.com/
23:04
eadmund
neat. stops at 12MP
23:05
Bertl
@AR1411HS, HiSPi(TM) interface
23:05
eadmund
btw, if you manage to get the Aptina full datasheet, please email it to me
23:05
Bertl
I'm not going to register, so probably no :)
23:07
eadmund
anyway, if you find a large CMOS sensor, please tell me, the problem with CMOSIS is that they will be bound to Leica
23:08
Bertl
bound to Leica?
23:09
eadmund
yes. if you want to buy the 20K the "customer" needs to assent, and I guess the same will be true of a larger chip. I thought you knew this?
23:09
eadmund
I mean you are very insistent on "open source"
23:10
Bertl
not looking at the 20K :)
23:10
Bertl
for our purpose the 12k is more than enough for now
23:11
eadmund
maybe there are people here who want to do a still camera :)
23:12
Bertl
and we are not fixed on the cmosis sensors, although I doubt that we will adopt the Aptiva ones soon because of the HiSPi interface
23:15
eadmund
My problem is that with the large sensors the sensor costs become huge, so if you don't have volume discounts the open source solution is not attractive. It's impossible to compete eg. with SonyA7
23:16
Bertl
you're free to adapt a proprietary solution to interface with the core axiom system
23:16
eadmund
I don't understand.
23:16
Bertl
well, you are saying, that open source is 'not attractive' for your purpose
23:17
eadmund
it's not the open source - it's user who won't want to pay 3x the price of a Sony for an open source solution
23:17
Bertl
then he should get the sony solution instead
23:18
eadmund
yes. you know the reason we managed to get Linux into the universities and firms was tat it could run on obsolete hardware.
23:18
Bertl
there are a number of companies out there which do heavy price dumping
23:18
Bertl
we are not trying to compete with them on the price :)
23:19
eadmund
they are not doing price dumping. they are doing industrial production with a reasonable profit.
23:19
Bertl
I'm sure about that :)
23:20
eadmund
if they were price dumping it would be ok, because it would stop at some point.
23:20
eadmund
these guys have low dev costs because their development is incremental.
23:20
eadmund
sensor cells are reused, camera signal processing is reused etc
23:21
Bertl
good for them
23:21
eadmund
it takes only 2 or 3 guys to design a RAM or a sensor once are set up. Processors are people intensive.
23:22
eadmund
of course for images you also need color science etc, but basically we should not be angry at these guys for doing their job properly.
23:23
Bertl
I couldn't care less
23:23
Bertl
i.e. I'm definitely not angry at any company.
23:24
eadmund
anyway, I would like to find a reasonably cheap and large sensor, without having to do what I once did with flash RAM, go and buy devices on the open market and rip them apart :)
23:25
eadmund
my friends are now suggesting to buy used Sony cameras on ebay, and I think they may have a point :(
23:26
Bertl
maybe an appropriate solution for your purpose
23:26
eadmund
Bertl, I wish you a very good night. I'm off to learn wavelets - and my offer of a bibliography :) still stands
23:27
eadmund
also I need to look at the latest Imatest, I have a customer who wants me to do MTF for "cheap" and Norman doesn't feel philantropic.
23:28
Bertl
a good night to you as well, and good luck!
23:28
eadmund
left the channel
23:41
sb0_
joined the channel
23:44
sb0
left the channel
23:44
troy_s
left the channel