Current Server Time: 08:56 (Central Europe)

#apertus IRC Channel Logs

2014/07/17

Timezone: UTC


00:02
fsteinel_
joined the channel
00:05
fsteinel
left the channel
00:58
fsteinel
joined the channel
01:02
fsteinel_
left the channel
01:04
Bertl
off to bed now ... have a good one everyone!
01:19
danieel
joined the channel
03:01
intracube
left the channel
03:33
jucar1
joined the channel
03:36
jucar
left the channel
06:34
Topic
apertus° - open source cinema | www.apertus.org | Currently Hot: AXIOM Beta https://apertus.org/axiom-beta | IRC Logs available at: http://irc.apertus.org
06:34
se6astian
has set the topic
08:03
FergusL
joined the channel
09:53
se6astian|away
changed nick to: se6astian
10:13
se6astian
good day!
11:04
Bertl
morning folks!
12:52
danieel
interesting price on the pocket bm camera...
13:00
se6astian
changed nick to: se6astian|away
13:25
Bertl
danieel: how so?
13:25
danieel
dropped to half till end of sept
13:26
danieel
end of august
13:27
Bertl
so price dumping again, I presume
13:30
Bertl
I'm curious what message this sends to customers in general
13:31
Bertl
for exisiting: look, we charged you twice the amount for the same camera a few months ago :)
13:31
Bertl
for new ones: hey, we can't sell the product, so we half the price now :)
13:32
Bertl
or maybe it is a completely different message
13:33
danieel
they maybe do something better (gs, hfr) and those stocked units wont be sellable
13:34
danieel
but i assume there wont be any big announcement at ibc
18:09
se6astian|away
changed nick to: se6astian
21:04
se6astian
time for bed
21:04
se6astian
night night
21:04
se6astian
changed nick to: se6astian|away
22:00
FergusL
left the channel
22:01
troy_s
Egads.
22:02
troy_s
When did the Black Magick camera drop to 500$?
22:02
troy_s
Erf.
22:10
danieel
few days back
22:10
danieel
july 16th is the news announcement
22:10
troy_s
Crazy.
22:10
danieel
what do you think of it ?
22:11
troy_s
I'm not a fan of the cameras due to the smaller sensor and the lack of offspeed, but at 500$ it's damn tempting.
22:11
troy_s
My axes of evaluation for this camera are:
22:11
troy_s
1) Latitude and format (both a win - raw Bayer and latitude are pretty decent)
22:11
troy_s
2) Sensor size (a loss here)
22:11
troy_s
3) Off speed (a loss here)
22:12
troy_s
But when I look at all three and then compare price... it makes for an actual decision.
22:12
danieel
resolution is poor.. (fhd@bayer)... even the F3 got enough cells to produce fhd 4:4:4
22:12
troy_s
(at 1000$ it is ridiculuous)
22:12
troy_s
?
22:12
troy_s
What's the price on the F3?
22:12
troy_s
(Haven't seen it. Is it a Sony?)
22:12
danieel
not comparable, as for the rest of the features
22:13
danieel
i think that there is no other camera which relies on pure fhd bayer sensor today..
22:18
troy_s
danieel: It's a pretty damn good camera at that price
22:18
troy_s
for what it is. The balance is pretty damn reasonable.
22:19
troy_s
The photosites still resolve to the lens which is always going to be an issue for CMOS.
22:19
danieel
what issue?
22:21
troy_s
Well the lens is going to resolve down across a smudge of G/B/R sites
22:21
troy_s
and we will always be facing the reality of missing data
22:21
troy_s
(Unless we go to a CCD or such)
22:21
troy_s
make sense?
22:22
danieel
does not. the ccd is same bayer pattern as cmos
22:22
troy_s
Some CCDs I thought were stacked
22:22
troy_s
But perhaps I am mistaken
22:22
danieel
for stills only (foveon)
22:22
troy_s
I know there were a few sensors out there based on that model
22:22
troy_s
Ah well, you get my point.
22:23
troy_s
(hopefully)
22:23
danieel
they recently went on to loose some of that stacking to increase luma resolution, colors get subsampled
22:23
danieel
the stacked ones can not yet be used for video.. so that is not a point
22:23
troy_s
Interesting. Those sorts of sensors would introduce a film-like crosstalk.
22:24
danieel
crosstalk is just a shifted color space, can be restored or enforced by the 3x3 matrix..
22:25
danieel
i do not like the bmpcc's rolling shutter... even that latitude is great, the yello efect ruins practical use
22:25
troy_s
danieel: Not really
22:25
troy_s
(agree on shutter)
22:25
troy_s
But crosstalk is a very real concern in VFX land
22:26
danieel
explain more in depth, that intersts me
22:26
troy_s
hence why keyers almost always want to be operating on the raw camera "space" (not that a camera really has a space per se)
22:26
troy_s
Well a non-crosstalked set of data has some nuances to it that are smudged apparently when you take it into a different transform.
22:26
danieel
isnt it confused then with a color subsampling? (usually you get 422 out of cameras)
22:26
troy_s
(for example, you and I might be able to pull a key using the edge information strictly from a luminance pass across the raw bayer)
22:27
danieel
that is unusual
22:27
troy_s
It makes sense though
22:27
troy_s
The lowest to the steel data you get of the lens resolving to some data
22:27
troy_s
is in that raw bayered image
22:27
danieel
(recently saw some pro bw photographer who pulls bw shots from raw stills)
22:27
danieel
ah
22:27
troy_s
And as soon as you debayer you are 'smudging' that
22:27
danieel
i get it where the issue comes from
22:28
troy_s
the difference between what is captured (critical for keying) and what color it represents is blurred
22:28
troy_s
(literally I suppose :) ()
22:28
danieel
when you apply 3x3 matrix / arbitrary "crosstalk", then you are corrupting the physical position of pixels
22:28
troy_s
Bingo
22:28
danieel
as the input elements are sampled at different pos
22:28
troy_s
Exactly.
22:28
troy_s
So your citing goes to shit
22:28
danieel
good, finally what we agree both on :)
22:28
troy_s
(I thought we agreed often. Communication sucks.)
22:28
troy_s
(Little did I know)
22:29
danieel
:)
22:29
troy_s
(Citing is actually a huge issue in subsampling anyways)
22:29
danieel
both have our set of truth
22:29
danieel
citing?
22:29
troy_s
(IIRC Greg's work on "Toward Better Subsampling" covered that a bit)
22:29
troy_s
Citing
22:29
danieel
whats that?
22:29
troy_s
where the given sample lay in spatial sense
22:29
troy_s
If you think of a standard lerp
22:29
troy_s
the interpolated value rests precisely at 0.5 distance
22:29
danieel
lot of new words today.. lerp?
22:30
troy_s
LERP == Linear Interpolation
22:30
troy_s
Hrm. How to describe.
22:30
troy_s
Think of it this way
22:30
troy_s
Think of three pixels
22:30
troy_s
expressed as a bar graph
22:30
troy_s
looking at the side
22:30
troy_s
when we draw a line from the middle of the first bar graph to the third
22:30
troy_s
the line precisely crosses the middle of the middle bar graph at the midway point
22:31
troy_s
A cubic scale
22:31
troy_s
will create a curve that doesn't quite end up precisely at that mid point - the citing is off.
22:31
troy_s
(As for tents, box, Lanczos, Sincs etc.)
22:32
danieel
so this is upsampling of the subsampled color channels, where they do not match? got a little lost
22:32
troy_s
Yes
22:32
troy_s
upsampling and how that RGB set is generated
22:32
danieel
and that error multiplies by consecutive color transformations
22:33
troy_s
http://www.glennchan.info/articles/technical/chroma/chroma1.htm
22:33
troy_s
Exactly.
22:33
troy_s
Glen's work has pictures!
22:33
troy_s
It's great to look at as some issues that are obvious become clear, despite rarely entering one's world view.
22:33
danieel
(that is why i preferred to keep my stuff in DNG... it keeps bayer, unless other formats)
22:33
troy_s
Sure.
22:34
troy_s
But at some point you have to futz with crap in RGB
22:34
troy_s
or a tri-color world
22:36
danieel
btw how would you deal with moare? strong sw - adaptive processing, or just proper OLPF?
22:36
troy_s
God that's a huge question.
22:36
troy_s
I know that the typical method goes as follows that I have seen:
22:36
troy_s
1) Filters on lens. Smudge it. Likely not preferred by some.
22:36
danieel
have noticed that f5 or f55 has a set of olpf for different uses
22:36
troy_s
2) Change the goddamn thing.
22:36
troy_s
(as in if it is wardrobe, change it.)
22:37
troy_s
that's the reality I have seen
22:37
troy_s
Yes. There's filters.
22:52
darthrake
left the channel
22:53
darthrake
joined the channel
22:54
jucar1
left the channel
22:54
PipeManMusic
left the channel
22:54
darthrake
left the channel
22:54
troy_s
left the channel
22:54
Bertl
left the channel
22:54
skinkie
left the channel
22:54
danieel
left the channel
22:54
mars_
left the channel
22:55
darthrake
joined the channel
22:55
jucar1
joined the channel
22:55
danieel
joined the channel
22:55
troy_s
joined the channel
22:55
skinkie
joined the channel
22:55
Bertl
joined the channel
22:55
mars_
joined the channel
22:55
PipeManMusic
joined the channel
22:59
FergusL
joined the channel