Current Server Time: 23:17 (Central Europe)

#apertus IRC Channel Logs

2022/10/05

Timezone: UTC


08:23
se6astian
polyrhythm: we decided to not take any venture capital so VC means video call or video conference :)
08:24
polyrhythm
yeah, that's what I thought!
08:24
polyrhythm
i've worked in california for too long obviously
08:25
se6astian
digitalbolex, great, I also have access to some academic equipment, though probably most of it not useful, I spectrally measured our UV/IR cut off filters
08:26
se6astian
https://www.apertus.org/axiom-team-talk-15-4-optical-filters-firmware-2-article-june-2020
08:26
se6astian
I am not sure though this particular device can measure non transparent objects though
08:27
se6astian
as its a separate emitter and measurement unit with a small gap in between
08:27
se6astian
Max does not have any high CRI led lights currently and would start with tungsten lights
08:28
se6astian
would that be a good first step or will the measurements only make sense when we also have a daylight set?
08:28
se6astian
for which I would need to organise the daylight leds for a day or so, but I cant hand them over to max as they are used in projects every now and then
08:45
polyrhythm
I think we can start with tungsten lights...
08:45
polyrhythm
at the end of the day we want two different matrices, one under tungsten and one under daylight, but one is better than none
08:45
se6astian
LEDs would be aputure LS 300X with CRI ≥ 96
08:45
se6astian
great!
08:47
polyrhythm
Aputure LS 300X seems a-okay :)
08:55
polyrhythm
regarding spectral profiling of a sensor, it does not look cheap...haha.
08:55
polyrhythm
https://www.image-engineering.de/products/technology/spectral-sensitivity
08:55
polyrhythm
this company seems to specialize in it.
08:55
polyrhythm
https://www.image-engineering.de/products/equipment/measurement-devices/588-camspecs-express
08:55
polyrhythm
16,500 euros!
08:55
polyrhythm
Weta Digital built their own spectral measurement device dubbed "Lightsaber" which always profiles sensors.
08:55
polyrhythm
basically you point a light of a known wavelength at the sensor and record the output
08:56
polyrhythm
I don't think the thing you used will work in this case, but if you have access to anything remotely like what I linked to above (or anything with a name like "monochrometer") let me know
08:57
polyrhythm
the problem this solves is essentially figuring out, in a very accurate way, the absolute limits of what the sensor can read.
08:57
polyrhythm
test charts can get us a decent sample set for typical images but not really the edge cases unless we start devising a ton of clevery-designed setups like certain hues of neon lights and so forth
08:58
polyrhythm
in reality it's not necessary to go this far for a cinema camera. weta's use case is that they try to recreate the camera in CG so they need to go this distance to get the result they need
09:00
polyrhythm
i'm sure Arri, etc., do this kind of thing for their color science though as well
09:00
vup
I mean building something like a monochrometer doesn't seem too hard, right? Using a prism with a "white" light and blocking most of the output (and maybe calibrating with a spectrometer or two different color lasers) doesn't seem too hard
09:01
vup
the main problem is producing the same intensity for every wavelength I guess
09:03
se6astian
I keep grazing ebay for an old spectrometer or sekonic color temperature measurement device...
09:04
se6astian
vup, I think we have different definitions of "too hard" :)
09:14
polyrhythm
yeah I think it's one of those things where the devil is in the details. I think you could probably cobble something together but when accuracy is the name of the game, that's when it gets difficult
09:15
polyrhythm
and that's why I spend $200 for a flimsy piece of paper with dumb little color patches on it. because a manufacturer is assuring me I can trust it, versus just printing one out
09:39
illwieckz
left the channel
09:59
illwieckz
joined the channel
10:04
vup
maybe there are a lot of details, but I think you can get something reasonable with not too much pain
10:04
vup
accuracy is pretty relative, I don't think we need a lot more than 8 to 12 bit, right?