Arri Alexa and Mysterium-X...

[Rick Burnett]

Which is why in my later post I opted for the term "demosiac", as opposed to debayer... or "transform", as I also used.

Point being, is that they are colored-pattern subsamples of full RGB space (weather Bayer or striped patterns), and thus should have some mathematical transform applied to them... and NOT simply be treated as RGB samples of the same pixel... they aren't.

-sc

You don't know that, and from Graeme's analysis of the F35 as he posted earlier, they did just that! Now what they've done on this new camera, no one has tested it like this, so no one can say for sure, but looking historically at what they've offered before, they aren't offering any details that things are being done differently.

 

[Jannard]

We need to keep this discussion on a more professional level. Reasonable questions are allowed in a respectful package.

However, calling a file RAW when white balance is burned in does not seem at all reasonable to me.

All any of us want is to have the question answered.

Jim

 

[Mitch Gross]

And, respectfully, there are those who stated quite clearly when RED was first introduced that a codec that compresses the data should not reasonably be called RAW.

I guess it depends on how we define our terms.

 

[Ian Laurie]

Seemed like a straight forward question... not even looking for reasons why, one way or the other. Is the WB baked into arriRAW?

 

[Graeme Nattress]

And, respectfully, there are those who stated quite clearly when RED was first introduced that a codec that compresses the data should not reasonably be called RAW.

I guess it depends on how we define our terms.

They did say that, but because we've always stated it was compressed it's quite reasonable for us to describe REDCODE RAW as a compressed raw format, the word RAW applying to the lack of image processing and lack of demosaicing rather than the use of compression.

Graeme

 

[David Mullen ASC]

Almost every manufacturer processes their "RAW" signal to some degree, whether its analog gain, compression, whatever. I don't see why we have to get into some argument over whose version of "RAW" is more pure than others.

All that matters are the results, how it looks, how it goes through post, etc.

 

[Graeme Nattress]

It's not about who's raw is rawest - but having information to make informed decisions about camera setup.

With the RED, there is no need to worry in any way about setting white balance critically on set. It's purely metadata.

With the Alexa we don't know if it's critical to get it right on set or not. That impacts shooting practicalities, workflows, who's job is it, etc. etc.

Graeme

 

[jonnycom]

You gotta give Arri credit, the Alexa is going out on some huge shows this summer.... including a bunch of 3d shows like the Scorsese 3d show.... were getting a lot of calls for it.

Hurry up with the EPIC please!

 

[Rick Burnett]

Almost every manufacturer processes their "RAW" signal to some degree, whether its analog gain, compression, whatever. I don't see why we have to get into some argument over whose version of "RAW" is more pure than others.

All that matters are the results, how it looks, how it goes through post, etc.

Which I agree with as well. I'd say the issue being circled about here is when a manufacturer tells you "RAW" what should you expect. There isn't really a definitive standard as far as I am aware, and so what that tells me is that when making a camera purchase, one must really push the manufacturer to explain what and why they've made any modifications to the sensor data so you understand what they were thinking it terms of workflow, efficiency and expectations of working with their product.

I don't think this is unreasonable, and if I were a potential Alexa customer, I'd want to know what the reasons for baking in the white balance are for. Obviously a decision was made to do this (if the documentation is correct), and I think people are genuinely curious as to why. Whatever their belief in advantage to this is. I am sure given their expertise in camera making it just didn't happen on its own.

 

[Steven Caesare]

You don't know that, and from Graeme's analysis of the F35 as he posted earlier, they did just that! Now what they've done on this new camera, no one has tested it like this, so no one can say for sure, but looking historically at what they've offered before, they aren't offering any details that things are being done differently.

I think you may be misinterpreting my post... I said "...they aren't" as in "Don't treat them as RGB images... they aren't".

I am making no suppositions as it weather the Alexa does this or not.

-sc

 

[Rick Burnett]

I think you may be misinterpreting my post... I said "...they aren't" as in "Don't treat them as RGB images... they aren't".

I am making no suppositions as it weather the Alexa does this or not.

-sc

I might have misunderstood then what your point was, I apologize. From what I understand, and you as well, the Bayer pattern requires a demosaic(interpolation) of the CFA pattern to build the full color image. The RGB stripe does not. They can build each final pixel value from three adjacent color values directly without having to calculate anything. I think this was the point of the difference, and the results seen by Graeme.

Whether the Alexa does anything different in this regard we don't know. That said, given their arrangement of pixels, it still means the horizontal and vertical resolution capabilities are not equal.

I'm still curious to know if the actual pixels are taller than they are wide.

 

[Andrew M.]

Almost every manufacturer processes their "RAW" signal to some degree, whether its analog gain, compression, whatever. I don't see why we have to get into some argument over whose version of "RAW" is more pure than others.

All that matters are the results, how it looks, how it goes through post, etc.

Applying WB on analog side of the sensor or even by installing lens filter saves on available DR at the end. Maybe Alexa WB is not so much baked in the RAW but it is applied on analog side, before digital RAW data is actually acquired from analog to digital conversion process.

 

[Graeme Nattress]

I might have misunderstood then what your point was, I apologize. From what I understand, and you as well, the Bayer pattern requires a demosaic(interpolation) of the CFA pattern to build the full color image. The RGB stripe does not. They can build each final pixel value from three adjacent color values directly without having to calculate anything. I think this was the point of the difference, and the results seen by Graeme.

Whether the Alexa does anything different in this regard we don't know. That said, given their arrangement of pixels, it still means the horizontal and vertical resolution capabilities are not equal.

I'm still curious to know if the actual pixels are taller than they are wide.

I think you're getting mixed up. The thread drifted off to Sony F35 / Genesis HD cameras which use RGB Stripe CFAs. The Arri cameras use a standard Bayer pattern CFA, no doubt for similar reasons to RED, in that it produces the most perceptually relevant image via an efficient use of photosites.

Graeme

 

[David Battistella]

We need to keep this discussion on a more professional level. Reasonable questions are allowed in a respectful package.

However, calling a file RAW when white balance is burned in does not seem at all reasonable to me.

All any of us want is to have the question answered.

Jim

I agree. Asking a question is fair, having it answered is fair.

Flogging and name calling gets us nowhere.

David

 

[Rick Burnett]

I think you're getting mixed up. The thread drifted off to Sony F35 / Genesis HD cameras which use RGB Stripe CFAs. The Arri cameras use a standard Bayer pattern CFA, no doubt for similar reasons to RED, in that it produces the most perceptually relevant image via an efficient use of photosites.

Graeme

I think you are right! I stand corrected. I finally found a discussion on arridigital that talks about what they used. (http://www.arridigital.com/technical/bayermask)

Thanks for setting me straight!

Pertaining to the white balance, or also analog gain in general, there is no way not to have that baked in. It exists before the signal becomes digital in the first place.

It is this sensor that uses the two analog gain paths per pixel that it combines?

 

[Ian Laurie]

I see this from 2 different perspectives,

1. as a DIT I need to know whether or not the white balance is backed into the image or not... this may have a MAJOR outcome of the images. is this something i need to sit down and talk to the camera crew about to make sure it happens in each setup? In this situation i am not choosing the camera, But i DEFINITELY need to know.

2. as a R1 owner and operator, this is a curious development. I want to know their reasoning for this. not because i want to tear down their company or badmouth them, simply because i want to know the logic and find it interesting.

One of these situations is critical to know whether or not WB is baked in. I am sure ARRI would much prefer a little trash talk on a user forum for a competing manufacturer to having productions publicly slam the camera for not being good because some DIT or DOP didn't understand that they needed set the white balance before hand... I have nothing against ARRI. I would love to be able to afford some of their products. But whether or not WB is baked in is essential to the work flow of the camera! This is a very important professional question. Is the WB baked in?

 

[Sanjin Jukic]

Can somebody with more sort of engineered knowledge try to describe and explain main differences between Alexa and D-21.

I see it like a both Alexa and D-21 still use the same recording scheme:

ARRIRAW S35
23, 24, 25, 29, 30 (compressed)

and uncompressed RAW
(Dual link HD-SDI 444 / 4444
1080P / PsF / i -
23, 24, 25, 29, 30, 50, 59, 60

also ARRI M-SCOPE (anamorphic).

Alexa has new and better sensor than D-21, 800 ASA native, 13.5 stops of DR,...

Alexa added "SxS slots that records ProRes for Direct To Edit functionality, that works with FCP and AVID" (Michael Bravin).

Also Alexa has a smaller and lighter body than D-21, new redesigned camera control system, new cooling system, optional and kind of universal lens mount (!?),...

Alexa EV cost much less than D-21 with optical VF.

More or less that's all.

So in the case of new recording option actually nothing is new with RAW (compressed and uncompressed) recording

between Alexa and D-21 except addition of ProRes recording integrated inside of Alexa's compact body.

 

[Graeme Nattress]

I think you are right! I stand corrected. I finally found a discussion on arridigital that talks about what they used. (http://www.arridigital.com/technical/bayermask)

Thanks for setting me straight!

Pertaining to the white balance, or also analog gain in general, there is no way not to have that baked in. It exists before the signal becomes digital in the first place.

It is this sensor that uses the two analog gain paths per pixel that it combines?

I think the magic gain tricks Arri do are somewhat independent of using differential analogue gains for white balance pre A-to-D, although the use of the magic gain tricks may complexify any use of analogue gain for other purposes.

As far as I can think, there are three places to do white balance (in RAW camera situations)

1) as analogue gain pre-a-to-d.
2) as digital gain post a-to-d before recording.
3) as digital gain in raw development.

We chose 3 so that we could make accurate white balance something which you don't need to determine on set and something which is not critically set on set - which is, for the very most part, how DSLRs do it. It made sense to us to do that way too.

Graeme

 

[Chad Clemans]

It's not about who's raw is rawest - but having information to make informed decisions about camera setup.

With the RED, there is no need to worry in any way about setting white balance critically on set. It's purely metadata.

With the Alexa we don't know if it's critical to get it right on set or not. That impacts shooting practicalities, workflows, who's job is it, etc. etc. Graeme

Thanks for not baking in white balance! That's always been part of how I choose to define RAW.

 

[Rick Burnett]

As far as I can think, there are three places to do white balance (in RAW camera situations)

1) as analogue gain pre-a-to-d.
2) as digital gain post a-to-d before recording.
3) as digital gain in raw development.

We chose 3 so that we could make accurate white balance something which you don't need to determine on set and something which is not critically set on set - which is, for the very most part, how DSLRs do it. It made sense to us to do that way too.

Graeme

I know your comment is specific to white balance, but pertaining to any sort of analog gain at all, I've never fully understood the technical reasons (taking cost and speed out of the equation) for variable gain pre A/D in terms of image quality.

If a particular sensor has a particular noise floor, and the A/D also has a noise floor, wouldn't you want to use an analog gain that is fixed that brings up the levels high enough to overcome the noise of the A/D, but use an A/D that has a larger bit depth so you can push the signal back down in digital to push out the A/D noise?

I've never understood the inherent benefit to adjusting that gain as all you are doing is pushing the noise further up above the A/D noise, but that is not a variable issue if heat/power issues are held constant in my opinion.

In other words, other than speed and cost (possibly) I just don't see an advantage that analog would provide over digital in this case from a quality of results perspective.