Current Server Time: 02:36 (Central Europe)

#apertus IRC Channel Logs

2018/11/10

Timezone: UTC


05:23
RexOrCine
changed nick to: RexOrCine|away
05:35
Spirit532
joined the channel
05:35
Spirit532
hey, I have a question
05:36
Spirit532
I was googling about integrating spheres since I found myself in need of one, to attempt external calibration of a cinema camera that lost its calibration and can't be repaired, but still functions otherwise
05:36
Spirit532
and I saw a post of you milling a polystyrene sphere in 2017
05:36
Spirit532
did that get anywhere in the end?
07:17
se6astian|away
changed nick to: se6astian
07:27
se6astian
Spirit532: yes, the sphere is finished
07:28
se6astian
"attempt external calibration of a cinema camera" <- what are you trying to calibrate exactly?
07:29
Bertl_zZ
changed nick to: Bertl_oO
07:39
se6astian
changed nick to: se6astian|away
08:02
BAndiT1983|away
changed nick to: BAndiT1983
08:57
niemand
joined the channel
09:13
illwieckz
left the channel
11:34
PhillyCee
joined the channel
11:34
PhillyCee
Hi Folks..
11:36
PhillyCee
Are there any news before christmas planned? Its nice that panasonic starts the discussions new wiith the s1/s1r model if there might
11:36
PhillyCee
Be 10bit 4:2:2 processing in a prosumer cam .. would be nice to use the beta in 2019 instead ;D
11:38
PhillyCee
I'll check the log for answers - have a nice weekend and good luck on your journey!
11:42
PhillyCee
left the channel
13:17
se6astian|away
changed nick to: se6astian
13:17
se6astian
changed nick to: se6astian|away
13:18
se6astian|away
changed nick to: se6astian
13:18
se6astian
changed nick to: se6astian|away
13:19
se6astian|away
changed nick to: se6astian
13:21
se6astian
changed nick to: se6astian|away
13:22
se6astian|away
changed nick to: se6astian
13:22
danieel
Spirit532: i think the lens mount tube will quite interfere with your intent
13:24
se6astian
changed nick to: se6astian|away
13:24
danieel
maybe rather shot a defocused well lit white wall - or even better, get a reference camera to first calibrate the wall / light source, and from same position profile the target device
13:25
se6astian|away
changed nick to: se6astian
13:27
se6astian
changed nick to: se6astian|away
13:27
se6astian|away
changed nick to: se6astian
15:01
RexOrCine|away
changed nick to: RexOrCine
15:12
Spirit532
se6astian, nonuniformity and FPN
15:12
Spirit532
the sensor is just *raw*
15:12
Spirit532
danieel, ^
15:12
Spirit532
I am also having the issue you described
15:13
Spirit532
the optical path is causing some minor, super minor vingetting
15:14
Spirit532
the issue I'm really having is that there's a really ugly magenta transition on the image set at around 2/3rds the absolute brightness
15:15
Spirit532
it follows the non-uniform light(slightly going out in a ring from the middle), but the actual transition itself is parametrically identical throughout the sensor
15:16
Spirit532
I can correct nonuniformity pretty easily by just averaging a hundred dark images and subtracting, FPN goes away at the same time
15:16
Spirit532
no need for dark current adjustment, which is what I feared
15:17
se6astian
changed nick to: se6astian|away
15:18
Spirit532
I can PM anyone who is interested a set of test images
15:18
Spirit532
but it's a very large set and there may be some duplicates in terms of brightness
15:18
Spirit532
(4.8GB, because the sensor is 4608x2592 @ 12b)
15:19
Spirit532
also, trying to do it through a lens gives even more massive light vingetting, which is why I tried it with a highly uniform flat panel stuck against the bayonet, but it wasn't enough
15:22
Spirit532
it's an URSA Mini 4.6K that works absolutely fine, but the thing doesn't remember anything specific
15:22
Spirit532
the hardware ID box in the menu is just "Val." and it doesn't know its own serial number or calibration
15:23
Spirit532
and blackmagic said it's beyond economical repair(which is a lie, it's probably a tiny eeprom that stores the ID), but I want to make it at least somewhat useable
15:41
BAndiT1983
changed nick to: BAndiT1983|away
15:48
danieel
is there a function to execute calibration?
15:49
Spirit532
only black reference
15:49
Spirit532
which destroys the image
15:50
Spirit532
(since the camera doesn't know the rest of its parameters)
16:33
alexML
Spirit532: not familiar with Ursa; if I understand it well, the issue is that camera cannot remember its calibration parameters or reference images?
16:33
Spirit532
completely, yes
16:33
Spirit532
it's a bizarre failure mode
16:34
Spirit532
I'm not going to take it apart enough to try to find the eeprom that probably holds the settings
16:34
Spirit532
the thing is a flex-rigid PCB nightmare
16:34
Spirit532
origami construction
17:12
niteesh
joined the channel
17:17
niteesh
left the channel
18:09
illwieckz
joined the channel
18:34
danieel
I thought only the pocket cam is made that way, due size... but at ursa? does not make much sense to have it on single pcb :/
18:39
Spirit532
it's a single flex-rigid assembly for the main processor and IO
18:39
Spirit532
there's a few daughterboards, one for the sensor, one for the CFast slot, and I believe one for power/XLR audio
18:39
Spirit532
though they're mostly structural
18:39
Spirit532
they're super high layer count too, from what I can understand
18:39
Spirit532
there's basically no traces on the top/bottom layers
19:04
parasew[m]
hey will someone from apertus be at the 35c3 this year?
19:09
comradekingu
I will be there, but I am only community
19:13
parasew[m]
we will be setting up a stage and i was wondering if we could have an apertus talk there
19:49
BAndiT1983|away
changed nick to: BAndiT1983
20:17
se6astian|away
changed nick to: se6astian
21:34
BAndiT1983
changed nick to: BAndiT1983|away
22:15
BAndiT1983|away
changed nick to: BAndiT1983
22:56
BAndiT1983
changed nick to: BAndiT1983|away
23:25
Bertl_oO
Spirit532: so what is the plan on getting it working again?
23:25
Spirit532
Bertl_oO, calibrating it in post
23:26
Spirit532
it shoots raw or prores 4444 hq, that's plenty of data
23:26
Bertl_oO
ah, I see, so the images you get are 'reasonably' good even without any of the calibration data
23:26
Spirit532
they're absolutely awful because of the magenta transition
23:27
Spirit532
and I don't know where it's coming from
23:27
Spirit532
the FPN and nonuniformity can be fixed in one go though
23:27
Spirit532
so that's nice
23:27
Spirit532
well I suppose they're the same thing, more or less
23:27
Bertl_oO
so with magenta you actually mean a red + blue transition from full on to full off (or the other way round)?
23:28
Spirit532
the other way around
23:28
Spirit532
I can PM you the samples if you want
23:28
Spirit532
I have a full gradient step from almost dark to almost saturated
23:28
Spirit532
and a few dark frames
23:28
Bertl_oO
got a single frame to look at (preferably at an image hosting service)?
23:28
Spirit532
yep
23:28
Spirit532
https://i.spirit.re/Etu1F.dng
23:28
Spirit532
here
23:29
Spirit532
you can see the magenta at around the 3/4 of the way up
23:29
niemand
left the channel
23:29
Spirit532
this is without FPN removal or dead pixel correction by the way
23:29
Bertl_oO
yeah, but I also see all kinds of colors right above the knurled know thingy
23:30
Bertl_oO
it is the back of a flashlight, right?
23:30
Spirit532
top
23:30
Spirit532
and yeah
23:30
Spirit532
that's the magenta stacked with FPN
23:30
Spirit532
the FPN just makes it a lot more ugly
23:30
Bertl_oO
and this is YCrCb or RGB raw?
23:30
Spirit532
this is raw from the camera
23:30
Spirit532
not debayered
23:30
Bertl_oO
(and with raw I autally mean cookes which they claim to be raw :)
23:31
Bertl_oO
*cooked
23:31
Spirit532
nope, this is sensor raw
23:31
Spirit532
here's a dark frame:
23:31
Spirit532
https://i.spirit.re/c0dtD.dng
23:31
Spirit532
the camera literally forgot how to calibrate
23:31
Spirit532
it just spits out the raw images into the codec
23:31
Spirit532
which is either raw, 2:1, 3:1, 4:1, or flavors of prores
23:32
Bertl_oO
yeah, you might be right and this is the first of its kind which really provides raw sensor data :)
23:32
Spirit532
yup
23:33
Bertl_oO
but the pink tint might actually be the result of some misconfigured correction mechanism
23:33
Spirit532
maybe some silly default state?
23:33
Spirit532
I don't know, I haven't looked at the camera's firmware
23:33
se6astian
changed nick to: se6astian|away
23:34
Spirit532
which might be a good idea, but it doesn't come as a separate file, it's packed deep into the updater sw and encrypted
23:34
Bertl_oO
yeah well, I think chances are good that there are a number of mechanisms to compensate for all kind of noise during calibration
23:34
Spirit532
I think there's really only 2
23:34
Spirit532
static nonuniformity and dark current
23:34
Spirit532
the images *are* color accurate after all
23:35
Spirit532
they're not inverted or anything
23:35
Bertl_oO
those are the obvious ones and also the ones quite easy to detect
23:35
Spirit532
there just seems to be some weird knee where two of the color channels(red & blue) drift up and down
23:35
Spirit532
I'd be completely fine with the image it produced if it wasn't for the nonuniformity and magenta
23:35
Bertl_oO
but let's say they have some kind of noise caused by readout or similar and they add some offset or something like this 'under certain conditions'
23:36
Spirit532
but if it's zero
23:36
Bertl_oO
if that offset is encoded in such way that an empty eeprom doesn't give a zero value
23:36
Spirit532
and let's assume it's zero, since the camera thinks its version is "Ver." instead of a hex value
23:36
Bertl_oO
then you end up with weird color changes and even weirder places
23:36
RexOrCine
(20:04:05) parasew[m]: Not to my knowledge. I won't certainly.
23:36
Spirit532
it's an odd distribution though
23:37
Spirit532
there's a point until the image is perfect color-wise
23:37
Spirit532
the response is linear(more or less) to the brightness
23:37
Bertl_oO
but I think it is rather unlikely that the eeprom has lost its content over night
23:37
Spirit532
but then there's a point at which the image just /shifts/ a few hundred counts
23:37
Bertl_oO
can you describe what led to the current condition?
23:37
Spirit532
and sloooooooowly drifts apart - not clipping, but drifts apart until saturation
23:37
Spirit532
I have no idea, I bought the camera this way hoping it'd be an easy fix
23:37
Spirit532
(or sending it to BM would work)
23:38
Bertl_oO
I see, well it might still be an easy fix
23:38
Spirit532
blackmagic said it's beyond economical repair
23:38
Spirit532
but they followed up with "we don't do component level repair" so that explains why
23:38
Bertl_oO
yeah, well, they obviously want to sell a new one :)
23:39
Spirit532
they did offer to replace the entire left side of the camera with the slightly damaged LCD cable for only $150
23:39
Bertl_oO
got a newly developed bad pixel, well, nothing we can do about that, buy a new one :)
23:39
Spirit532
they actually can
23:39
Bertl_oO
for sure they can, but they are not interested in doing so
23:39
Spirit532
they said "we tried to pass this through our image optimization process, but since there's no hardware IDs anywhere, the software doesn't know anything about the camera and can't proceed"
23:39
Spirit532
which is lazy
23:39
Spirit532
it's a single commented line to skip ID checks
23:40
Spirit532
though I assume even if they did calibrate it, they wouldn't be able to write it to the camera
23:40
Bertl_oO
anyway, what I would check first is if all the eeprom pins have a proper connection
23:40
Spirit532
I don't even know where it is
23:40
Spirit532
and I'm not really keen on ripping the camera apart again
23:41
Bertl_oO
because if one of them has a bad solder joint, you basically get a all ones or all zero result on each and every read
23:41
Spirit532
yeah it might be something like that
23:41
Bertl_oO
while the actual calibration data might still be tehre
23:41
Bertl_oO
*there
23:41
Spirit532
I... doubt that one
23:41
Spirit532
they said they flashed it with the factory bootloader before undoing it back to production firmware
23:41
Spirit532
(to attempt restoring the IDs)
23:42
Bertl_oO
and it didn't work, did it?
23:42
Spirit532
so you're thinking that it's the command lines, not the data lines of the eeprom that are gone?
23:42
Spirit532
MOSI, if it's SPI?
23:43
Bertl_oO
which means that the eeprom (actually it will be some kind of flash memory) didn't get the proper commands
23:43
Spirit532
the camera has an absolutely horrid origami construction
23:43
Spirit532
http://i.imgur.com/2X8disJ.jpg
23:43
Bertl_oO
and without the proper commands (because some pin is shorted or open), nothing will be erased or changed or overwritten
23:43
Spirit532
those flexes feel more like thin PCB than something designed to be bent
23:44
Spirit532
(not my camera btw, just a pic online)
23:44
Bertl_oO
if you find a high resolution image, I probably can point you to the flash memory
23:44
Spirit532
nope, that's the best I found
23:45
Spirit532
it's the only place it can be, so probably something small
23:45
Spirit532
but not one of the switchmode PSUs
23:45
Bertl_oO
well, the one on the top board below the white button, right above the 6 pin unpopulated header (left side)
23:45
Spirit532
and probably not one of the QFNs with a bunch of caps, those are probably switched cap psus or charge pumps to go back up
23:45
Bertl_oO
looke like a hynix flash memory at first glance
23:45
Spirit532
ram
23:46
Spirit532
it's the main and only processor
23:46
Spirit532
but there may be flash too
23:46
Bertl_oO
I would say the stuff on the right side is the RAM
23:46
Bertl_oO
like 6 chips of DDR memory
23:46
Spirit532
it's almost entirely single sided load
23:46
Spirit532
the stuff on the back is just decoupling
23:46
Bertl_oO
but I'm talking about the smaller one right below the white button
23:46
Spirit532
that... might be
23:47
Spirit532
but it's FBGA by the looks of it
23:47
Spirit532
package is floating above the board
23:47
Spirit532
I'll die before I can reflow that at home
23:47
Bertl_oO
yeah, eMMC comes in this kind of package for example
23:47
Spirit532
the little small SOP below the battery looks out of place
23:47
Spirit532
there's obviously two big power supplies
23:47
Spirit532
and a tiny xilinx part
23:48
Bertl_oO
and actually you might get away with simply reheating it with a hot air gun
23:48
Spirit532
that is probably not the best idea
23:48
Bertl_oO
because your hot air gun skills are that bad?
23:48
Bertl_oO
:)
23:48
Spirit532
I mean overheating is an issue yes
23:48
Spirit532
but just generally reheating random parts on a board is the "phone repairman" tactic
23:49
danieel
that flash next to the small xilinx/atmel device is just fpga firmware and that works
23:49
Spirit532
(doesn't work 90% of the time)
23:49
Bertl_oO
not really, you heat up the PCB to something aroun 100-120°C from the bootom
23:49
Spirit532
can't do that
23:49
Spirit532
I have no preheater or oven
23:49
Bertl_oO
and then you use a temperature controlled heat gun set to 250-350°C to reheat the package from the top
23:50
Spirit532
250-350C is vague
23:50
Bertl_oO
well, depends on the solder paste used
23:50
Spirit532
it's going to be rohs
23:50
Spirit532
so probably closer to 300
23:50
Bertl_oO
250 for leaded solder, 350 for unleaded
23:50
Bertl_oO
anyway, preheating can be done with a cheap PTC from china
23:50
Bertl_oO
you can get those for about 10 USD or so
23:51
Spirit532
the issue with this board is that to get it to the state that is in that picture, you have to strip the camera to its last screw
23:51
Spirit532
it's an awful design
23:51
Bertl_oO
but hey, I'm just providing some ideas here, not saying that you should do this or that to your camera :)
23:52
RexOrCine
changed nick to: RexOrCine|away
23:52
Spirit532
http://i.imgur.com/YOtEHSQ.jpg
23:52
Spirit532
this is how it looks in the camera
23:52
Bertl_oO
could be worse
23:52
Spirit532
could be better too
23:52
Bertl_oO
the main problem is the lack of documentation
23:53
Bertl_oO
but that's what you get in the proprietary arena
23:53
Spirit532
blackmagic's business model seems to be a complete lack of technical support beyond what's commercially viable
23:55
Spirit532
what's worse is that the eeprom might be on the sensor board
23:55
Spirit532
which is... http://i.imgur.com/N6ZndDu.jpg
23:55
Spirit532
yeah
23:59
Spirit532
I'm thinking it's a separate eeprom, because if it was some sort of flash device, you'd think the OS would be stored on it too, right?