Panasonic going towards 8K to the masses...

[Christoffer Glans]

Im not sure what film look is as it´s very subjective.

Nope, if Arri focused on matching up the color science with how a certain Kodak stock behaves in DR and colors, that is enough to please the eyes of many cinematographers and it's not subjective but a matchup against a certain source.

Just a minor side note: "Celluloid" ?

Celluloid, chemically is cellulose-nitrate... "gun cotton"/ Parkesine.

Explosive and dangerous. Amateur use stopped around the 1930's. Kodak continued its use until about 1951 (for professional/cinematic use).

Well, a "Red Celluloid" goes well with Red naming cameras in military terminology then :tongue_smilie:

We don't know the specs of Gemini's sensor.

Yes we do, Jarred pointed it out in his post.

 

[Björn Benckert]

Nope, if Arri focused on matching up the color science with how a certain Kodak stock behaves in DR and colors, that is enough to please the eyes of many cinematographers and it's not subjective but a matchup against a certain source.
.

Thats a surplice, So what is the Kodak stock number?


As if developing arri raw to linear for example then its absolutely not trying to look like any filmstock. Thats what we do and then comp LIN CG onto it and grade the comps with Lin2REC709 luts and grading. The "film look" is very much in the grading process then and has very little to do with the look provided from the Arri sensor.

 

[Eric Lange]

I contend that we are rapidly approaching the practical limits of motion capture using CMOS Bayer pattern sensors. I expect the next truly significant breakthrough to be sensors the can read both photon levels and wavelength at every photo site.

Cheers - #19

I kinda like that idea.

A spectrophotometer at every photo site.

 

[Wayne Morellini]

You mean Sigma Foveon x3, that Canon 7 version two camera with five frequency regions, the old HTC camera phones, and other manufacturers like Sony, working in the technology in the past?

 

[Steven Abrams]

Nope, if Arri focused on matching up the color science with how a certain Kodak stock behaves in DR and colors, that is enough to please the eyes of many cinematographers and it's not subjective but a matchup against a certain source.

Where are you getting that from? To match up to film stock would require non-linear processing. For Arri to fit into ACES workflows that are scene referred would mean reversing these non-linear transforms to get to Linear XYZ, but you can look at their IDTs they supplied and they don't. In fact, they do very standard things that everyone does.

https://github.com/ampas/aces-dev/b...exa/v3/EI800/IDT.ARRI.Alexa-v3-logC-EI800.ctl

Even look at how they calculate these IDTs here - https://github.com/ampas/aces-dev/b...idt/vendorSupplied/arri/alexa/v3_IDT_maker.py

Here all they do is have a conversion from their LogC curve into linear, and 3x3 matrix to convert from their wide gamut into XYZ. All that is going on here is that their primaries for their wide gamut are well chosen (but still nothing to do with film) and they use a log curve that was similar to cineon so that colorists used to grading scanned film felt comfortable. This is what Arri has said they did, and it's been reinterpreted by many to be they matched their color science to film. No. Their pixels are well designed and the metameric effect from their CFA quite pleasing compared to others. They also prioritized skin tones when matching their sensor response to their wide gamut. But there's nothing else really special, just the basics done well. Don't believe art adams about purposely changing saturation with luminance either. Again it would break ACES unless there was the inverse of this function in their IDT's, but there isn't and their images transform well using the standard linear approaches.

It's worth noting Arri supplied a 3x3 matrix for transforming LogC to more closely match scanned film when using a FPE LUT (or you can select the option on camera) but basically no one uses it. But why would they add this if it was already matched up to film? Answer = It wasn't.

http://archiv.arri.de/AlexaColorProcessing-23Feb2011/html/filmstyle.html

 

[Hrvoje Simic]

Here all they do is have a conversion from their LogC curve into linear, and 3x3 matrix to convert from their wide gamut into XYZ. All that is going on here is that their primaries for their wide gamut are well chosen (but still nothing to do with film) and they use a log curve that was similar to cineon so that colorists used to grading scanned film felt comfortable.

No, that is not all they do.

That is the easier part.

But there's nothing else really special, just the basics done well.

Having tested a bunch of cameras and their signal behaviour, Arri ones included, and working on Arri transformations for the last year I'd say you are mistaken.

http://archiv.arri.de/AlexaColorProcessing-23Feb2011/html/filmstyle.html

That tells you little about camera performance.

 

[Christoffer Glans]

Thats a surplice, So what is the Kodak stock number?


As if developing arri raw to linear for example then its absolutely not trying to look like any filmstock. Thats what we do and then comp LIN CG onto it and grade the comps with Lin2REC709 luts and grading. The "film look" is very much in the grading process then and has very little to do with the look provided from the Arri sensor.

I'm talking sensor level handling. It's not offical, but most people gather it's base resembles Vision3 5219 after a normal curve is applied. Of course it's done in grading, but if Alexa gives you a base look after basic normal curves applied that resembles 5219 both in how it handles DR and colors, you don't need to grade your image into oblivion to match a film like look. I've not seen the same accuracy to the film look on Red, on any of the sensors and when pushing for it, there's so much in the image that starts falling apart. Fringes and aberrations and so on. Bottom line is that Arri matches up a baseline that's closer to how the kodak stock handles things, then leaving room for grading on top of it. The closest I've got to film-like colors on Red is with the filmconvert LUT tools, but it's crushing the image quite a lot. All grading is destroying the image in some ways so if Arri gives you the look without pushing the image too far compared to Red, it creates a much more pleasing look with minimal tampering.

 

[Steven Abrams]

No, that is not all they do.

Didn't say it was. But that's the bulk of it aside from say sensor calibration in terms of processing. The rest is in the design of the sensor/pixel/CFAs. It doesn't take much to reverse engineer whats going on especially when you have friends that work there.

Having tested a bunch of cameras and their signal behaviour, Arri ones included, and working on Arri transformations for the last year I'd say you are mistaken.

Not according to my friends there. And they're very open about what they do, even publicly. Because that's the real 'secret'. It's doing the basics well, which is extremely difficult. But simple. They don't have to hide anything because it's all standard stuff. They just do it very well. Better than most it seems.

If I'm mistaken, then clearly explain how their IDTs which transform their footage into scene referred Linear XYZ don't require any non-linear or "exotic" transforms? Because it really is simple. Just difficult to execute. I mean even they have not been able to top themselves yet after 10 years. Their sensor has a high full well capacity (likely around 100ke-) with low read noise (around 1.2e- or so for around 16 stops and what they say is 14.5 useable), 12bit dual gain readout , good CFA design, and well chosen primaries for their wide gamut and a log curve people were familiar with. If you think they're doing something special, lets hear it. Why do you think that? What is it you've seen?

People talk about Helium having higher DR than Alexa. Prove it. We looked inside again recently and it appears they're still using 14bit ADCs (I've worked on sensors and calibration/etc before so know what I'm looking at), store it as log and pad it out to 16bit linear after decode. You can easily tell if you analyze the raw files too. Then you look at this chart from phil and it's clear they're not getting more than 14 stops without HDRx on helium which makes total sense given the hardware inside.

That tells you little about camera performance.

Maybe there's a language barrier but I didn't say it did nor imply it. Just that if their intention was to get their color science in line with kodak film stock as Christoffer claims, why did then they add a film matrix to match scanned film? Just think it over and you'll get there.

 

[Björn Benckert]

Color fringing is a helium thing. Dragon has a noisecfloor issue when moving from the first olpf that had the red orb issue. Both of these things are quite easy to stay clear from. Enough fill and dont overexpose, and that means enough fill in all chanels and a stop of head room in all chanels. Monstro does not have any of these problems and looks great both when clipping and dipping into black. As in looks better than anything else out there Arri included. To me thats not a taste thing its a fact, but sure it will most likly pass mostly un noticed by most DPs as they got burned on red before, did not understand the cameras and simply felt more comfy shooting arri. Its been easier to shoot arri in the past. But now with the minis and all the itterations of arri cams out there I would say thats not the case any more.
Try rig a mini, in comparision weapon is third iteration, and much better thought through. Then all the different options of arri, from only prores 16:9, external raw recorder, alexa m, alexa studio 4:3 with spinning mirror, Alexa XT recording Open gate, not able to record open gate at certain framrates, then only 422 at some framerates. ACs got into arri as they used their film cameras and it was smart of arri to make first a simple digital camera and then slowly add things to it. They could do so as they allready had the clients. Red needed to be more disruptive to get their share of users, both with price and workflow.

 

[Karim D. Ghantous]

I think that ARRI can be happy that people talk about their sensor so much. On a RED forum. Also, it's lucky for ARRI that nobody seems to care that some Alexa films are either no different to other digital shoots, or are simply not that nice.

It's worth noting that RED is not insecure about what they make. If you don't like it, use something else. There's room for every camera. If you do like it, great! Jim Jannard seems to know that you can't be successful in business if you're reacting to your competition.

 

[Sabovic Adis]

(...)but most people gather it's base resembles Vision3 5219 after a normal curve is applied. (...)

That's a color negative film. What normal curve are You talking about?

 

[Christoffer Glans]

That's a color negative film. What normal curve are You talking about?

Got that wrong, I meant at it's sensor level it's like Vison3 5219 and then all curves applied happens just like it does after scan, when color profiles are attached before grade.

 

[Sabovic Adis]

Got that wrong, I meant at it's sensor level it's like Vison3 5219 and then all curves applied happens just like it does after scan, when color profiles are attached before grade.

Hm...after scan of Vison3 5219 You're in "Cineon space", which, when not converted to color positive and transformed to 709, looks awfull. Like this:

This actually is scanned off a film positive (print film, toe and shoulder inclusive!) and untransformed it looks like this:

Then, after ADX10 to 709 transform, it looks like this:

And that's all there is to it. So, ALEXAs LogC Wide Gammut resembles - what exactly?

 

[Eric Lange]

I think that ARRI can be happy that people talk about their sensor so much. On a RED forum. Also, it's lucky for ARRI that nobody seems to care that some Alexa films are either no different to other digital shoots, or are simply not that nice.

It's worth noting that RED is not insecure about what they make. If you don't like it, use something else. There's room for every camera. If you do like it, great! Jim Jannard seems to know that you can't be successful in business if you're reacting to your competition.

+1

From my spidey senses and being out of the loop vis a vie RED-Strategy I think SONY is the problem not ARRI. [This has been mentioned recently in few places here an' there.].

Also just from a marketing perspective, product differentiation is very important. So trying to desperately prove that a RED Montro or other can be just like an ARRI is kind of a loosing proposition from a psychological and sales perspective. Monstro obviously has its own "Brilliant corners". It's a terrible idea to make (what used to be called "Me too" products). Usually companies in a broad market space figure out what they do differently and better than their competitors rather than try to be like the "Other guys".

Red is not going to Out-Arri ARRI nor is ARRI going to try and Out-RED RED YKWIM? Totally different culture, locations and history and different goals, purpose and objectives with each company.

Meanwhile what is SONY getting up to "Big-picture" ?

 

[Hrvoje Simic]

Didn't say it was. But that's the bulk of it aside from say sensor calibration in terms of processing.
The rest is in the design of the sensor/pixel/CFAs. It doesn't take much to reverse engineer whats going on especially when you have friends that work there.

So that's mostly it. And oh yeah just the processing.

Good luck with reverse engineering a camera in post.

Not according to my friends there. And they're very open about what they do, even publicly.

Your friends told you everything about Arri's color science and signal processing ?

If I'm mistaken, then clearly explain how their IDTs which transform their footage into scene referred Linear XYZ don't require any non-linear or "exotic" transforms?

Because signal behavior and signal mapping are two different things.

I wouldn't use IDTs and ACES as a basis for camera interpretation. Ever.

Their sensor has a high full well capacity (likely around 100ke-) with low read noise (around 1.2e- or so for around 16 stops and what they say is 14.5 useable), 12bit dual gain readout , good CFA design, and well chosen primaries for their wide gamut and a log curve people were familiar with.

And specific processing. Which cannot be copied by a matrix.

Log curve is just a gamma and tells little about particular color science.

If you think they're doing something special, lets hear it. Why do you think that?

Because the signal does not behave like a signal coming from a digital camera. In short.

Maybe there's a language barrier but I didn't say it did nor imply it. Just that if their intention was to get their color science in line with kodak film stock as Christoffer claims, why did then they add a film matrix to match scanned film?

Perhaps you should ask Christoffer that question.

Generally, shooters notice the "film like" feel of Arri image.
It is not visible in math and it is there even without that film stock matrix.

 

[Steven Abrams]

So that's mostly it. And oh yeah just the processing.

I'm wondering how much experience you have compared to me in doing image sensor processing and color science. Actually I'm not because it's quite obvious you have very little, probably none. Calibration is not about imparting a look into a sensor. Its getting the sensor response to actually be linear, managing noise, etc. So yeah, the "processing" isn't creating a film response. It's attempting to achieve a very good linear response.

Good luck with reverse engineering a camera in post.

Who said reverse engineering in post? You make some crazy assumptions. About everything.

Because signal behavior and signal mapping are two different things.

You don't say. Well you just blew my mind. Genius.

And specific processing. Which cannot be copied by a matrix.

Again, who said a matrix can copy any other process than a matrix? What are you even talking about and who is it directed at?

Log curve is just a gamma and tells little about particular color science.

Are you just saying basic stuff for the benefit of other readers? Hey guys guess what, a lens is not a camera! No kidding.

Because the signal does not behave like a signal coming from a digital camera. In short.

The linear signal actually behaves how a well calibrated linear sensor/camera should respond. Got evidence to the contrary? Show us then. But talk to Arri engineers and they will tell you that the idea that their camera is specifically calibrated and processed to look like film is baloney. The sales guys may be okay with you believing that junk though.

But I get it, you need people to believe that "special processing" is needed so that you can sell your wares after all the time you invested. And for me to say a good image like Arri is just the basics and a simple approach done really well makes that difficult for you.

Perhaps you should ask Christoffer that question.

I did. And then you came along and replied with unrelated comments.

Generally, shooters notice the "film like" feel of Arri image.

So why did Arri add a film matrix (that basically no one uses and many dislike) to match scanned film? Just think it over and you'll get there.

 

[Hrvoje Simic]

I'm wondering how much experience you have compared to me in doing image sensor processing and color science. Actually I'm not because it's quite obvious you have very little, probably none. Calibration is not about imparting a look into a sensor. Its getting the sensor response to actually be linear, managing noise, etc. So yeah, the "processing" isn't creating a film response. It's attempting to achieve a very good linear response.

Self-fluffery to force authority.

Impressions depend not just on the input but in logic and character which define its processing.
And wondering tends to lead mind astray.

I wasn't talking about sensor calibration nor imparting a look into a sensor.

Who said reverse engineering in post?

Logic did.

With your matrix, friends, reverse transforms and simplification used as premises.

I'm waiting for your "pretty easy reverse engineering" of Alexa.

Again, who said a matrix can copy any other process than a matrix? What are you even talking about and who is it directed at?

Your rhetoric suggested oversimplification.

Are you just saying basic stuff for the benefit of other readers? Hey guys guess what, a lens is not a camera! No kidding.

Interesting reaction after glancing over log gamma mapping from a high horse and it's there "because colorists got used to it".

Ask Sony about their SLog2 SGamut trainwreck and orange people.

The linear signal actually behaves how a well calibrated linear sensor/camera should respond. Got evidence to the contrary?

I wasn't talking about how linear signal behaves.

Which can be easily deduced by a categorization before the term "signal".

But talk to Arri engineers and they will tell you that the idea that their camera is specifically calibrated and processed to look like film is baloney.

Please forward this comment to Christoffer.

Meanwhile, I will consider copying Strawman fallacy wiki page link in the clipboard when I go on the forum.

But I get it, you need people to believe that "special processing" is needed so that you can sell your wares after all the time you invested.

And here we have the application of wishful thinking and projection of state of character and system of values, existing as an option in one's own mindset. With attempt of disqualifying a person in a discussion.

That's a second "ad hominem" fallacy in your post.

I'm saving this one for later, thanks.

And for me to say a good image like Arri is just the basics and a simple approach done really well makes that difficult for you.

I'm sorry.

My oversimplification alarm got activated.

So why did Arri add a film matrix (that basically no one uses and many dislike) to match scanned film? Just think it over and you'll get there.

Arrogance with repetition won't add sense in the rhetoric.

Let's make a distinction first between a generalized "film look" and "film-like image behaviour" so the third attempt and cognitive dissonance can be avoided.

 

[L. Langer]

+1

From my spidey senses and being out of the loop vis a vie RED-Strategy I think SONY is the problem not ARRI. [This has been mentioned recently in few places here an' there.].

Also just from a marketing perspective, product differentiation is very important. So trying to desperately prove that a RED Montro or other can be just like an ARRI is kind of a loosing proposition from a psychological and sales perspective. Monstro obviously has its own "Brilliant corners". It's a terrible idea to make (what used to be called "Me too" products). Usually companies in a broad market space figure out what they do differently and better than their competitors rather than try to be like the "Other guys".

Red is not going to Out-Arri ARRI nor is ARRI going to try and Out-RED RED YKWIM? Totally different culture, locations and history and different goals, purpose and objectives with each company.

Meanwhile what is SONY getting up to "Big-picture" ?

Sony is primarily focused on the cinema side of digital cinema - projectors and systems that they lease out - and their non-CineAlta camcorders, which are pretty much the only cameras with their logo on it that anyone is using these days, outside of their mirrorless stuff. At this point, CineAlta is floundering because ARRI took over the bulk of TV and film production in key markets where Sony's cameras were once the standard. They are working feverishly in their R&D facilities to generate new sensor tech for their mirrorless cameras that can be carried over to their professional broadcast and cinema products, but it's likely that Sony kills the CineAlta line and refocuses solely on low-end cameras, if Venice is a failure.

 

[Steven Abrams]

I wasn't talking about how linear signal behaves.

Then you're talking about what exactly? Because from linear there's just a 3x3 and a curve to LogC. So where's the magic?

Impressions depend not just on the input but in logic and character which define its processing.
And wondering tends to lead mind astray.

No, you convinced me with your overwhelming proof. I now believe in fairies and unicorns. ENJOY!

 

[Blair S. Paulsen]

Sony is primarily focused on the cinema side of digital cinema - projectors and systems that they lease out - and their non-CineAlta camcorders, which are pretty much the only cameras with their logo on it that anyone is using these days, outside of their mirrorless stuff. At this point, CineAlta is floundering because ARRI took over the bulk of TV and film production in key markets where Sony's cameras were once the standard. They are working feverishly in their R&D facilities to generate new sensor tech for their mirrorless cameras that can be carried over to their professional broadcast and cinema products, but it's likely that Sony kills the CineAlta line and refocuses solely on low-end cameras, if Venice is a failure.

If Sony had done things differently... But they didn't, and despite making mostly excellent imaging devices, still struggle to get out of their own way. Just sayin'

Cheers - #19