Arri Alexa and Mysterium-X...

[Blair S. Paulsen]

Indeed you are individual people. I'm sure I've even seen you together, so unless you're like Clark Kent with an android duplicate for use at trade shows....

Graeme

An android duplicate for use at trade shows - that would be a popular product. Seems like the next domino in Jim's march - Oakley, Red, MitchMichaels, ...:hand:

Cheers - #19

 

[Graeme Nattress]

I'm pretty sure that someone is going to be shooting a movie or show that will have both Epic and Alexa footage on it. RED Users are pretty much "pro raw" as is RED itself, and as is Arri. It makes sense for RED Users to get all the info - pros and cons, on both camera systems so that they can work effectively in the mixed environments that are sure to occur.

It's always a good thing to be able to speak direct to the makers of a camera for the best answers to technical questions. And sometimes those questions will be hard. A lot of times that's because there's no "right way" to do things, just an "educated compromise" and that's often due to mutually contradictory desires for optimization on cost / weight / heat / size.

Graeme

 

[Michael Hastings]

Just a suggestion for a few of the vibrant posters ..., if you want Michael, (it's Michael not Mitch as was stated by one member*** Mitch beat me to the punch) to poke his head out and have an honest conversation, you might wan to put down the pitch forks and torches.

Yes ArriRaw, is Arris version of RAW. Redcode is Red's version. Different parties will argue for different classification requirements for "raw." They each have come up with their own. Lets actually look at them and get some solid answers and comparisons and make informed judgments shall we? I for one would love to hear more about the specs of ArriRAW. What is the accompanying metadata, is the WB baked in like before etc.

I find it fairly obvious by where I am posting this, where my bias lies. However my belief can only be truly validated by honest discussion and comparison.

Nice post Eric. Until we have mathematically lossless REDCODE Red raw is REDRAW with whatever they have chosen to throw out and whatever secret sauce is added - same with ARRIRAW.

I have no interest in buying an Alexa, but I am interested in understanding the state of the art in our industry. I'm concerned that the level of discourse drives away serious people that otherwise could provide valuable insight. Mike has been hanging in there, but I suspect that a lot of people that I would be interested in hearing from never even bother because of the reputation.

As far as Bravin, Jim asked the question at 6:36pm and by 2:30 AM he is already being accused of dodging the question? WTF?

Maybe he's not sure exactly what that means and wants to check..
Maybe Arri isn't sure and is considering offering it either way as an option..
Maybe he's answering questions from actual Arri customers via phone, email, whatever...

Maybe he has a life....

Jim and Jarred dodge questions all the time and this is their forum! Maybe we aren't always entitled to know everything about an unshipped product - or even everything that's coming up for a current product.

I suspect it will be answered in due time.

PS Edit:

And well said Graeme - I missed your post while I was thinking about/writing mine.

 

[Scott Crawley]

That's a valiant defense Michael. I'm with you in principle, we should cut him some slack. Also consider the fact that Jim's post was not the first to pose the question.
He's a grown man and can handle himself. If he were here for kicks I'm certain he would have cut and run by now. He has a serious job, and big bucks are at stake. The amount of shit he's willing to take likely varies in proportion to the amount of money Arri pays him to represent them.
He must have felt like he was being sent on a camping trip to Hades when they suggested he come here. You'ld have to pay me a whole boat load of money to do what he's doing.

Michael Bravin, you have my empathy... Sympathy, not so much.

 

[Steven Caesare]

A RGB stripe pattern colour filter array (CFA) is just a CFA like the Bayer Patterns is - just a different pattern!

Indeed... therefore I thought that they must do something to transform the data... I was getting the impression that the striped array (rather than the alternating pattern) was simply handled by a different domosaic algorithm... one that in your opinion didn't handle co-location as well because of the lack of interleved color samples.

On this forum, and I'm sure someone can dig it up, Panavision told us they did no processing to account for the lack of co-sitedness in the RGB stripe pattern. I was questioning that as I predicted it would lead to chroma moire unless a very strong OLPF was used. I'd heard reports of aliasing on the Genesis which, if true, would indicate not enough OLPF.

At that time, I had not seen proper measurements of a Genesis. However, now I have had conducted measurements of an F35 (which has the same sensor as the Genesis) and it does indeed show quite clearly that the RGB channels are not co-sited and no processing is done to try to bring them into alignment, and it does show the chroma moire, and strong luma aliasing. Indeed, it show pretty much what I predicted upon learning the CFA pattern and that no processing was done.

"Wow" indeed. I wonder if they thought nobody would notice?

Graeme

I can certainly see why that would be. I'm frankly rather astounded that they aren't at least applying something to attempt to "demosaic" the pixel data.

Thanks Graeme.

-sc

 

[M Most]

If you're interpolating to go from 3k to 3k, then going from 3k to 2k (or likewise 2880 to 1920) is also the same as your chroma is "interpolated" up from 1440x810 to 1920x1080. For green you're interpolating from 1440x1620.

I didn't say anything to the contrary. Everything in the debayer process from a Bayer pattern sensor is an interpolation by definition, whether it's Arri, Red, or anyone else. You're creating pixels that did not exist in the first place.

Prores 422HQ is YCbCr.
Prores 4444 is YCbCr or RGB with an alpha channel (or not)

I was using the term RGB to differentiate it from a RAW sensor record. I meant "color image that can be viewed and used directly for further image processing," not technically RGB as opposed to YCbCr, YUV, XYZ, or anything else.

Why no built in raw recorder?

Probably because Arri has made a technical, business, and strategic decision to make the RAW output uncompressed, and there is currently no technology small, light, fast enough, or compact enough to be directly integrated into the camera to do that. Red's technical, business, and strategic decision was to adopt compression to allow that to happen. Both are currently valid decisions dependent on the intended audience for the product, even though they're different.

 

[Tim Morten]

Indeed... therefore I thought that they must do something to transform the data... I was getting the impression that the striped array (rather than the alternating pattern) was simply handled by a different domosaic algorithm... one that in your opinion didn't handle co-location as well because of the lack of interleved color samples.



I can certainly see why that would be. I'm frankly rather astounded that they aren't at least applying something to attempt to "demosaic" the pixel data.

Thanks Graeme.

-sc

This underscores a point made by some others previously: digital imaging is a whole new ballgame. The previously established players do not automatically have an edge up by nature of having existed for decades longer than Red. This isn't to say that Genesis or Alexa are junk, only that no one should assume they are superior strictly on the basis of their brand name.

Pixel design matters (as Graeme has illustrated). So does baking in white blance (if indeed Arri does so), camera size, weight, resolution, frame-rate, modularity... the list goes on. And then there's the small matter of price.

Since I know someone will jump on me if I don't say it: of course, workflow matters too. We can argue endlessly about the merits of R3D versus ProRes, but the bottom line is that software support for Red's format is wide and growing, and high quality real-time transcoding exists if you really need another format.

The jury will be out until side-by-side image comparison happens, but if ProRes-out-of-camera really winds up being the only thing that Alexa has to offer over Epic, all of Epic's other advantages make Epic the clear winner for my dollars. Again, just my opinion...

- Tim

 

[Grady Wilson]

Nobody has actually answered the question about white balance yet.


I am still hesitant to believe that ARRI has chosen to make the first camera RAW format in the history of RAW camera formats to burn in white balance.

Maybe they are not answering because that was the plan, they realized they screwed up and are scrambling now to fix it?

 

[Albert Armes]

Nobody has actually answered the question about white balance yet.


I am still hesitant to believe that ARRI has chosen to make the first camera RAW format in the history of RAW camera formats to burn in white balance.

Maybe they are not answering because that was the plan, they realized they screwed up and are scrambling now to fix it?

Well, after reading all this, and searching out the D21 manual and reading through that, it does sound that ARRI have already made the first "burn in white balance" RAW format camera. And if they thought it a good idea for the D21, maybe they think it's a good idea for the Alexa. Maybe it is a good idea??

Albert

 

[Rick Burnett]

I see a lot of mention of Bayer pattern and demosaicing with the RGB columns, but I don't think that these terms are being used properly. A bayer pattern is specific to using 50% green, 25% red and 25% blue just as in the patent that this method comes from. Likewise, demosaicing is due to the mosaic of pixels and the interpolation technique, which is not the case with the stripes.

 

[M Most]

Sounds like a really dumb idea. Can someone, anyone, explain to me how that could possibly be a wise decision?

Not speaking for Arri, and not knowing exactly how the camera is architected, one possible reason - if indeed this is the case at all - is that any correction can be applied in the analog domain, before the A to D conversion, allowing more possible levels and therefore cleaner correction than doing it after conversion. This has long been the case on video cameras, where doing any adjustments to the color channels up front allows access to a lot more information than you have after formatting for 10 bit digital video.

 

[Graeme Nattress]

I see a lot of mention of Bayer pattern and demosaicing with the RGB columns, but I don't think that these terms are being used properly. A bayer pattern is specific to using 50% green, 25% red and 25% blue just as in the patent that this method comes from. Likewise, demosaicing is due to the mosaic of pixels and the interpolation technique, which is not the case with the stripes.

A Bayer pattern is a colour filter array pattern. An RGB Strip is a colour filter array pattern. They are both mosaic patterns, but different mosaic patterns.

You can rapid extract 2k from 4k on a Bayer CFA pattern, but it's a sub-optimal way to do things. You can do a similar in concept "rapid extraction" on RGB stripe CFA too, and just like it's sub-optimal on the Bayer, it leads to artifacts on the Stripe too.

Graeme

 

[Rick Burnett]

A Bayer pattern is a colour filter array pattern. An RGB Strip is a colour filter array pattern. They are both mosaic patterns, but different mosaic patterns.

You can rapid extract 2k from 4k on a Bayer CFA pattern, but it's a sub-optimal way to do things. You can do a similar in concept "rapid extraction" on RGB stripe CFA too, and just like it's sub-optimal on the Bayer, it leads to artifacts on the Stripe too.

Graeme

Well that is certainly contradictory to my understanding. The patent of a Bayer pattern is specific about the arrangement of the CFA. While they both are in fact CFA, looking at the original patent indicates the specific arrangement (US03971065) as part of his initial filing.

In fact, EVERY reference to Bayer patterns I have ever read have indicated that the specific pattern is as been implemented by the RED sensors. If someone were to tell you that a camera used a "Bayer" pattern, it would be instantly assumed to be alternating rows of GRGR and BGBG, not a stripe nor any of the other patterns that use additional colors in the CFA.

I can understand if the "industry" is attempted to change the definition, but this has not been the case and is going to confuse people about what they are getting (in terms of Arri using it for their sensor).

 

[Graeme Nattress]

Rick - we use a standard Bayer pattern. Genesis / F35 use a stripe pattern, not a bayer pattern. Both are colour filter array pattern sensors though. On one hand, we hear from Panavision that Bayer CFA patterns are "bad" (chroma moire being pointed out as an issue), but somehow, the RGB stripe CFA pattern is ok, even though when we test it, it produces significant chroma moire (and luma aliasing vertically).

Graeme

 

[Stephen Williams]

Sounds like a really dumb idea. Can someone, anyone, explain to me how that could possibly be a wise decision?

I think it is official, ARRI has definitely lost its mind, and lost it's soul.

Hi,

I am more interested in how the pictures look, they had the best Bayer pictures last year, it was only film that looked better.

 

[Tom.Wong]

Sounds like a really dumb idea. Can someone, anyone, explain to me how that could possibly be a wise decision?

I think it is official, ARRI has definitely lost its mind, and lost it's soul.

I think we all need to stay objective. There are reasons for everything, whether we agree with them or not. Jumping to such conclusions is what starts up paths to really BAD "discussions" Once we get the answer to the question does ARRIRAW burn in WB, then we will get explanations that go with it, and I think we can discuss pro's and cons from there.

 

[Steven Caesare]

I see a lot of mention of Bayer pattern and demosaicing with the RGB columns, but I don't think that these terms are being used properly. A bayer pattern is specific to using 50% green, 25% red and 25% blue just as in the patent that this method comes from. Likewise, demosaicing is due to the mosaic of pixels and the interpolation technique, which is not the case with the stripes.

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

 

[David Battistella]

Maybe they are not answering because that was the plan, they realized they screwed up and are scrambling now to fix it?

I have to echo some of what others are mentioning in this thread. Comments like this tend to make the REDUSER crowd look like a bunch of dorks, rather than professionals engaging in a professional discussion.

Try to keep your comments aimed toward positive flow to a conversation. If you are making a joke use a smile or an LOL to illustrate that you are trying to be funny.

If people sit on a RED high horse and start trying to bash ARRI (when they have decided to engage in a welcome conversation in the REDUSER forum), then you can expect that these types of constructive open dialog will cease.

Would you have said this in a group of people or a room full of people at a presentation? I really doubt it.

David

 

[Tom.Wong]

+1,2,3,4,5

+50

 

[Rick Burnett]

Rick - we use a standard Bayer pattern. Genesis / F35 use a stripe pattern, not a bayer pattern. Both are colour filter array pattern sensors though. On one hand, we hear from Panavision that Bayer CFA patterns are "bad" (chroma moire being pointed out as an issue), but somehow, the RGB stripe CFA pattern is ok, even though when we test it, it produces significant chroma moire (and luma aliasing vertically).

Graeme

Oh, okay, I think I was unclear on your response, it seemed to imply their RGB was also a Bayer pattern. Understood.

Pertaining to your sensor design, I just don't see how they can make that determination as their chroma moire on a horizontal line of pixels would be worse given their photosites have 2 photosites between like colors and yours only has one. Granted, in a row of pixels in the vertical direction, they have no space inbetween, which would be better, this mismatch in vertical/horizontal resolution means in one direction, your chroma moire issues are going to be worse and why wouldn't you pick a design that minimizes chroma moire in both directions?

The obvious tradeoff is that green samples higher than red and blue, but given my understanding of the eye and it's sensitivity curves, this has always made sense to me.

I understand the speed difference in terms of processing power to generate a frame from the sensor between both methods, however, if you take that out of the equation, pixels being equal, I don't see how the bayer pattern isn't going to give you an overall better images in all circumstances.