#apertus IRC Channel Logs

america :)

mmm

it plays 192 kHz/24 bit files

probably converted from lower khz

and you dont want too good sounding headphones. at least not at work. stopped bringing my shure IEMs because they distracted me too much with good-sounding music.

well, FLAC is the most sensible choice

96/24 would be enough...

48kHz would be enough for most

intracube, that makes sense. i meant it is not just flac. it supports 192/24 too

and other strange stuiff

No you can't. What you probably hear is a problem in your equipment.

gear for 192k is not so common (its more like Alexas)

i have a demanding content... my wife does classical/opera :)

but might sound just as well on 44.1 ... if mastered well

about speakers, ill stick with my linkwitz orions :

http://www.linkwitzlab.com/orion_us.htm

"axiom alpha felt like nothing before!"

so with 96k/24 i am perfectly fine. It is as reality. but if you do not have any match how the content should sound like, you can be fine with a mp3 :)

good ole nyquist

the frequency range barely is, at 22khz... but i dont hear up that high anyway. and nothing interesting there either

aah monitors, nice \ufffdD

got some B%W as my computer speakers.... love them too :)

Q_: Agree. The old argument is just misplaced.

troy_s: Exactly.

Q_: Same sort of statements. Almost identical. Same misinformation.

I had a fellow try to insist that there was a difference in the signal (digital) going across silver.

Not to mention that most recent music doesn't even have a decent dynamic range, it's all been compressed.

Q_: Why stop there? Aesthetics are emergent.

i also had an argument with a guy about speaker cables and interconnects ( my point : if cables sound differently, then the cable is badly designed)

was a bit ugly

troy_s: The only good argument I've ever heard for quality of digital cables is the bit error rate.

It isn't like a subtle timbre. LOL

blind a/b testing is always the answer

It isn't magic though. You can hear digital oopsies as they are glaring. Not some unnoticeable qualia.

Hence the D to A part.

i'm with Sabre :)

ESS32

Q_, thats why my DAC also does preamp duties, and then feeds active speakers

(anedio DAC2) ... cute little thing

doesnt sound that way, troy_s :)

seku: I am. I have some older school decade old stuff. Holds up fine. Newer consumer receiver.

I have toyed with bumping up to Paradigms but alas... Hard to justify.

at audible frequencies?

I dont see the point to be argued though

Digital transmits just fine down a fricking coat hanger.

You can measure it with an oscilloscope if you want

me? im using well-shielded copper

LOL.

wait, let me get you a link

Doesn't using silver and gold actually cause more noise because of their difference in electrode potentional then the copper?

They do, and gold is worse than copper

for thousands

Interesting.

But obviously it is sub Nyquist or whatever in digital land?

(case in point : mine are soldered)

troy_s: Yes.

Genius.

Oh oh... Nerd alerts.

the cable is an inductor, with resistive properties going along it, then a capacitive effect in between the wires

Q_: all the neutrik XLRs have gold contacts afaik

comradekingu: Looking at their website, none of their XLR connectors have it.

Oh, some do.

well, if one day i win the lottery, i will get me these : http://www.htguide.com/forum/attachment.php?attachmentid=12520&stc=1&d=1211369490

acoustic room treatment is much more important on the sound you hear than those technical matters

and it is almost never being done properly

Nyquist levels yes?

Nyquist levels determine a frequency, so then you are using a cable dimensioned which has to be non-broken

Dark alchemy.

Its not alchemy at all, its plain math

Alchemy. It is lower than the floor of my understanding. ;)

like vinyl!

seku: Again, aesthetics are emergent. :)

To be fair to vinyl, the vinylistas have a valid point on one vantage; most vinyl is mastered very differently to generic MP3s

Perhaps more so these days, where vinyls are limited / exclusive releases

comradekingu: ?

i was in err, and returned to have my offset corrected

The joke is you need to align the offset of your cd-rom otherwise you cant rip CDs with matching filesums

brb sometime later

forward

what do you mean with 'hit'?

was at 120k :)

night!

ah, yeah, well, we'll see :)

(I know... Discussed previously.)

Ugh

which sw supports mlv? i do not recall any grading suites to open these...

Where is the spec?

Going to piss off an entire audience base overnight

afaik mlv was born out of the misery of Canons firmware : the format needed to index a lot of out-of-order data and save it correctly to the limited speed of a 100mbyte/sec CF card

hi g3gg0 , you can explain that way better than i

Where is the spec though? And a sample?

afaik grading suites are mostly fine with CinemaDNG. compressed cinemadng would be even nicer :)

g3gg0: Spec and sample?

LOL

its already late and it takes a few seconds

https://bitbucket.org/hudson/magic-lantern/src/4d5cf113a68128c866eb788cd7ea3245e81fcc34/modules/mlv_rec/mlv.h?at=mlv_rec_lj92

https://docs.google.com/spreadsheet/ccc?key=0AgQ2MOkAZTFHdHJraTVTOEpmNEIwTVlKd0dHVi1ULUE&usp=drive_web#gid=1

thats some more detailed explanation

So RTC is a time code chunk?

RTCI is the real time clock information

every (meta)data block has a µsec timestamp

sure

[MLVI] [RAWI] [VIDF]

yep

Third...

Yep

VIDF= video frame

So a metadata.

Is the meta variable length?

(I have an OIIO stub already for the raw CMV12000 shots that loads fine in my lab thingy.)

all blocks basically support variable length. for most blocks it wont change though. only INFO blocks that contain ascii text are variable iirc

g3gg0: So is this a V1 or pre-v1 implementation?

v1?

I expect there is a 1.1 or something likely for Axiom

g3gg0: In your estimation then, how much would potentially (guess) shift to support Axiom?

g3gg0: the structures are LE or BE ?

Temporal stripes?

So unordered?

iirc bertl explained that the sensor data isnt read out from left to right, but its sectioned and interlaced

the idea is to allow for recording sensor data without reordering it

this format isnt explicitely supported by the format's metadata yet. i.e. the writer cant tell the reader how the data is ordered.

yes, we do it in the FPGA already on the Alpha

It is like having the old guys shouting from the muppets.

Ah. Good point.

Not that batteries are a tremendous issue given the contexts. But still valid.

or reduce heat :)

g3gg0: is there any lab processing software out there right now for the MLV format?

you mean for debugging the format?

yeah thats one reason for mlv.

mlv also converts fine to cinemadng.

Probably a tiff hack.

danieel: can you note that somewhere (wiki, forum) for discussion?

me? using chmee's raw2cdng

Who is the author of MLRawViewer?

g3gg0: is there any indexing for quick seeking?

troy_s: baldand, andrew baldwin

danieel: .idx files containing a header and only mlv_xref entries

Interesting. He wrote the tag parser.

Yes

g3gg0: It's a decent enough goal, but the overlap in the design is questionable to skip.

many tools are very picky about CDNG.

g3gg0: I've sort of been plucking away on a lab software because I'm interested in getting footage into an offline state, so dealing with the demosaic etc. is useful.

g3gg0: A few things that can be imrpoved almost immediately is tacking on OCIO and OIIO support. Gains you offline dumping to EXR / DPX immediately and full LUT / color support.

*sent

I haven't looked at amaze much. I probably should. I need to start doing RMSEs on generated bayers and see which ones hold up ideally.

g3gg0: OCIO and OIIO are reference standards really.

I'd be skeptical of the shader code

as they are already 4 byte names, consider them being perfect MLV tags already ;)

and OCIO opens up CDL for free.

but its already late, 4h left for sleep

g3gg0: Ping me on the morrow.

Wish my GL were better.

Wholly god does the shader code suck in ellipses.

There has got to be a better way.

the shader code is usually ugly, depends what version of the compiler you can use

opencl is better in that respect, but does not run everywhere

you mean the computational format?

should be 32bit (SP)... the DP rating for gpus is 64bit float

danieel: Just a standard fragment shader. I'm just curious how to enum / query that.

No. I won't. There has got to be a query.

Egads.

well... in the gpu terminology we can start complaining about dimensions being specified in floats

I'm cool with that... I just want a query.

But I'd like to know what depth I'm at.

danieel: Any links?

section 4.7.2 at https://www.opengl.org/registry/doc/GLSLangSpec.4.50.pdf

it causes rounding problems...

is anyone here?

developers wouldn't mind something less. I would just think that the added expense is still much less than any aftermarket solution and prolongs the camera's life. And since it's designed as a box anyhow...

Hey, are the irc logs broken?

indeed the sunday one is empty

And I was thinking you finally took some rest.

"a thief copied large amounts of highly scandalous internal communication from the high security server"

need to sell the logs on the black market!

its funny how one spends $100M's on a movie, and at the end it sucks because all the VFX is global shutter like while real scenes got heavy rolling shutter artefacts

newest transformers

danieel: I couldn't see any obvious signs in the transformer 4 trailer

hi, looks like irc archive isn't showing today's log

indeed, let me check

nah, the logger ran out of space, is already fixed

ahh okey thaks bertl..... the log was looking strange :-)

test3

okay, logs should be back and working

d20 or that 1990's military one?

most of the people working on arri came from frauenhofer research..

do you know if there is a change with the access from the documents from truesense because on semi has bought them?

probably not, they finished the transfer like a week or two ago

i hope i can go it's for me just 2h drive so not far away....

public, but you could figure that out if you've been actually looking :)

the DS is on the product page :)

good evening

danieel: I can check, but pretty sure it was shot on a global shutter camera.

*if it was shot whit rolling shutter equipment

Domestic: $245,435,927   22.7%+ Foreign: $835,500,000   77.3%= Worldwide: $1,080,935,927

That puts it at slot 11 of all time worldwide grosses.

danieel: are there any clips online that show the issue?

danieel: May have been. You also have to remember that Epics are used when cameras are going to get crunched.

it is most visible in nonvfx action shots (people figting, camera panning quickly), let me find it

wasnt it supposed to be shot on imax/3d ?

the seccond transformer had compositing problems too

global can be simulated for <180deg with an liquid crystal shutter

danieel: Rare to shoot under the 180. Is that what they were doing?

the new red sensor does that thing with liquid crystal shutter

Tapered pillars.

danieel: that is from what film?

once you start making a camera, you cant watch movies... RS besides the poor dynamic range is most noticable :)

speaking of dynamic range and highlights: http://images.static-bluray.com/reviews/6812_20_1080p.jpg

White and black are creative decisions.

prometheus was quite good.. was not distracting much

at least that's what it looks like

you can adapt your range level to the darks... and then they surprise you with a sun or light...

danieel: I think you're missing my point

then they've been brought back to non peak levels where you'd expect to see detail

its intentional bring down as i wrote.. but yes, there is lack of detail :)

so it really depends on the shots leading up to it

Just saying'.

If at any point someone felt that was a mistake, I can state beyond a shadow of a doubt that particular shot would have been reshot. Nothing to it.

I will say it again for folks that perhaps didn't grasp / embrace / comprehend:

Could Bertl craft a nice PLE that gathers up a much larger DR? Absolutely. Should a DP throw out all but one stop of it? Absolutely.

Aesthetics go through the typical cycles of “out of place with purpose� “fresh new bleeding edge� “mainstream� “cliched vernacular�

See Cross Cut.

very interesting what I'm saying here?

ah, nevermind, window update problem

which in film is, among other things, chemistry to math (the color correction aspect of it)

So I'll call into question the trope that many professionals “don't know what they are doing�

Everyone from the DP, to assistants, to other HoDs have varying experiences that insulate.

research doesnt always reward us with results

aombk: I am offering my opinion purely as anecdotal from experience. In my personal experience, one can have a gripe with aesthetic choices etc, but I have met very few that “don't know what they are doing�.

(EG a DP may be thinner on color knowledge, but has a generally experienced DIT on their side.)

troy_s: who has said that many professionals don't know what they're doing?

aombk: Exactly. Privileged rubbish terms.

intracube, i said something like that

troy_s, well in the end, i dont agree with you on that

There is also the very real fact that seasoned veterans are often doing things in mainstream movie making that well... Haven't been done before. So of course they don't know what they are doing.

Sure. But it is also ideological I believe.

If we say “That movie sucked� is is basically a statement of constancy (a myth) and a checksum of extremely complex things.

And over time, that constancy is more assaultable. See DW Griffith's earlier cinema for example, which at once is celebrated as monumental feats and also simultaneously grotesquely racist.

aombk: Sense?

are you a dp?

What core?

hi troy_s

That professionals do or do not know?

at the core of what you are saying i think lies relativism

postmodernist relativism

Indeed.

Or in actuality, a vastly more workable model for sociological manifestations? ;)

troy_s: if i may ask, whats your background in film making?

g3gg0: Studied film /photography in art school, shot / produced blah blah crap, and also exists as a minion to the machine.

just curious

so both

On some projects, such as album covers, sometimes none.

Still trying to learn.

so colorspace conversion stuff is one of your competences?

the things behind the scenes. really good to know. then things make a lot more sense

welcome PRN!

g3gg0: Yes sadly.

g3gg0: Take chart. Develop 3D LUT, transform to XYZ. Transform to output +/- a Bradford or CAT02 in there. ;)

CAT02? ok looking up. the other things i did myself too

g3gg0: CAT02 is just one of the newest breeds of CA.

anyone from apertus team going to Broadcast India, Mumbai?

ok is it the one related to XYZ, just with correct chromatic "distance" between colors?

g3gg0: Just a different model. Like a Bradford++.

http://web.stanford.edu/~sujason/ColorBalancing/adaptation.html

This isn't about standards. It is more about attempting to simulate something. So there is a little more directionality.

Bradford superseded Von Kries, as VK superceded XYZ scaling.

it is also mostly used for white point adaption?

Chromatic adaptation, yes.

Although CAT02 is part of a larger thing.

Color appearance models.

(As in many of the color phenomenon illusiony things)

Problem is, some / many may be cultural.

(See Herskovits et al regarding Influence of Culture on Visual Perception)

g3gg0: Explain?

Same goes for display profiling.

Not exactly

it is effective, yes. but there are colors you cannot reproduce with passive lighting

yep

yeah and every channel gets light that is not filtered ideal

Yes. Non-narrow band

So a camera can never really reflect what it sees

Blah blah

Bottom line is that A) 3D LUTs are vastly more accurate B) you need one for each varying of contexts (ISO, illuminant, etc.)

g3gg0: Nope. Hence post-spectral

i tried to find the matrices various raw developing tools use.

Matrices suck

i tried to reconstruct a LUT from that. i somewhen stopped because i couldnt explain what ive seen there :)

here it is https://www.youtube.com/watch?v=9Fz6V5aJSbk

The problem is that some shapes may cover a decent volume of the gamut, but when you get to the edges, your DE goes through the roof

yeah absolutely

g3gg0: quit watching that crap

i made that ;)

The ICC tools can be leveraged to your whim, but you sadly have to know their graphic arts hacks

(aside of RT, dcraw, DPP)

g3gg0: We need to chat more. ;)

OCIO was developed to supplant the limitations of ICCs

In particular:

B) Discarding the D50 rubbish

unfortunately no direct DNG support. didnt check if the TIFF module is capable of writing DNG-alike TIFFS

D) Discarding the display referred-centric nature of ICCs

DPXs are pretty universal. And as I said, if someone added the tags for the TIFF side, quite sure Mr. Gritz would merge them for CinemaDNG.

g3gg0: But I think your transforms would greatly bump up a notch. You could end up with archival stored EXRs without breaking a sweat.

g3gg0: I forced OIIO to load a CinemaDNG, after all, it is a TIFF.

(if i understood it right)

well, lets then call it standardized ones :)

g3gg0: So for example, one might want to process raw to DPX and archive the DPXs.

troy_s: what i dont get is - why would we ever care about any color space / LUT when writing data into an universal format like EXR. to my understanding all the data should match the scene as-shot. applying some LUT that looks the best on cinema screen or RGB monitors or CCTV is one of the last steps?

Ultimately most will want to start native.

“match the scene� is really the end of a 3D LUT, but that isn't ideal either. Ideal for an entry to grading, but not for post always.

Much of post isn't color critical, as long as there are assertions to not mangle the data.

off for now ... bbl

bye Bertl!

assume the LUT wasnt good enough

all your archived data is crap

Hence you only tack that onto the view. (Most imaging apps have a distinct breakdown between the input / working slush / and view)

yeah thats a viewer story then, right?

You set up your views and can toggle between them. So you chain transforms to have a reference to output (say reference to 709) then tack on your 3D LUT display correction

All the while your data is pure.

yes this absolutely does

hey, i shoot MP4 videos. this says it all? ;)

Even then, you can greatly improve your work by following that path.

i am neither a professional videographer nor a photographer. i am reverse engineer and a programmer with hobby photographing :)

Sure. You are a perfect candidate to muck with it then.

Even breaking your MP4s down into stills will increase the flexibility you have.

Everything from controlling the subsampling (a huge impact on your still image) to being able to composite say, fire, to getting to a correct film-like dissolve.

to me this is only 50% of the story

This is because they all rely on ICC tools.

you shoot patches and build a matrix that should match all colors that are physically possible.

Hard to do with reflective charts. You need light boxes etc. But that said, a HCT is a good chart. IT8s are solid too.

yeah

The metal basically takes a few things:

B) A data file that contains spectrophotometer readings of the actual chart print unique to each chart.

especially with pink highlights a1ex found that a bunch of the common imagery tools produce RGB results that are far off the reality.

Matrix!

Less so with 3D LUTs

but then you really need lightboxes that cover a wide spectral range, right?

No... Think of this mistake

(You need to look at the text file outputs to see ideal exposure. It is often 1-2 stops hotter than most folks shoot charts.)

Even if 90% are damn close to the libe

Hence pink or magenta or yellow

There are _many_ places to screw profiling up.

If you would like, I can step through with Argyll and we can cut much of the crap out of the way and you can see the results

still the biggest issue remains the proper lighting of the chart

danieel: That is large. But not largest.

Overexposed is bad too, but so is under.

this is the reason why everyone hopes that making it community-driven will make it successful

That will correct the glaring badness at the upper and low ends, preventing skewing off into grotesque color.

There are a lot of damn sharp minds out there. Collective efforts are tough though.

cant we measure the spectral response for every single bayer filter R, G1, G2, B and built a LUT from that?

look at the QE charts from the sensor makers

Can not work.

So two problems.

And, the sensors misbehave differently with different factors.

So ISO? Different LUT. Color temp? Different LUT.

But the non-narrow band is the big killer there.

Needless to say, decent profiles can be had if the sensor can be tamed.

Color temp of illuminant. ;)

And a bunch of other crap.

nah maybe i expressed myself wrong.

might not be just one bump on the chart

danieel: did all of that already by hand, so basics are known

And mixed illuminants

Really shocked

Little did I know how massaged that was

troy_s: i really thought about measuring every single CFA (wideband) filter for their spectral response. e.g. like those http://www.astrosurf.com/buil/350d/350d.htm (but he made the mistake and didnt use RAW but PP-ed RGB)

or this one: http://www.astrosurf.com/buil/eos40d2/filter.htm

The bottom line is that all glass / plastic adds crosstalk

So the battle is already fruitless

Varying levels of light will mix differently at different zones

Hell, people loved shooting film _because_ of the unique crosstalk

danieel: Do you have the text file with the HCT?

If you get even lighting, you should get a pretty decent set of profiles. I would do ISO and other variables too so you can get a few.

Solid charts.

what i (naively) thought is, getting the spectral response for every CFA filter R, G1, G2 and B to make up a matrix. this matrix is the coordinate system for the colors being measured. e.g. if R/G/G/B is 0/0/0/1 then the real input color can be in range of the spectral response of the blue filter. (e.g. this plot: http://www.astrosurf.com/buil/eos40d2/graph5.png ). i know that recovering a color in XYZ space from the CFA output is as complex a