Arri Alexa and Mysterium-X...

[Rob Ruffo]

Whether White Balance is baked in or not has nothing to do with mixing color temperature within a scene.

Not exactly sure what ARRI is or is not doing with Alexa (will know soon enough), but I suspect it is similar to what Vision Research does with the Phantom CINE files. And if so, I'm all for it. Think of White Balace as an offset -- the sensor sees what it sees, and White Balance is a shifting of the color ratios. Uncompressed, compressed, wavelet compressed, etc., in the end the information will get funneled down in some way due to bit depth. There are not infinite options. So if a basic white balance offset is used to generally clue the choise of what's most valuable to pack into your available bit depth of file then it can mean a better picture as you will have more of the information that you actually want.

We can talk the theory of this all we want, but it is meaningless until we actually see what comes out of the camera. Not all CMOS sensors are the same, not all color dyes on Bayer masks are the same. Graeme is a smart fellow, but he knows wjat he doesn't know. And that is what is going on inside Alexa and why.

I'm hoping Alexa will be fabulous. That said, mixed white balance can be better fixed with masks in post if you can shift white-balance further (regardless of whole image or just part) without degradation penalties.

 

[Nik Skjøth]

The Red camera does not eliminate the color temp differences between sources, it merely captures them relative to its native color temp balance around 5000K. I could load daylight-balanced film stock into a camera, scan the negative, and get similar results. Or set the ARRI to daylight-balance. After that, the process of correcting the image in post for another color temp is the same. So I don't get this "first time in history you can have mixed color temp in the same scene" sort of line of thinking.

Ehm.. Put your R3D footage on a timeline, copy it, and mask out the window from the top one.. Then whitebalance both files accordingly... TADAA!!!!!

 

[Rob Ruffo]

The Red camera does not eliminate the color temp differences between sources, it merely captures them relative to its native color temp balance around 5000K. I could load daylight-balanced film stock into a camera, scan the negative, and get similar results. Or set the ARRI to daylight-balance. After that, the process of correcting the image in post for another color temp is the same. So I don't get this "first time in history you can have mixed color temp in the same scene" sort of line of thinking.

In a video camera - its' pretty new, at least in a manner that works so well. In film, you are right, of course, it is not. Also quite new are affordable masking tools that would let you create a layer for just a window, track it, and have just that layer (the window) be balanced for daylight and the rest of the room for tungsten. I realize this can be risky (spill, etc.) , and it's better to gel that window or use HMI practicals, - but most likely you do not have to deal (at least, not all the time) with the kind of budgets we roll with here in Quebec.

 

[Rick Burnett]

Except that in the case of digital capture noise is usually pretty linear through the entire range of the image. So while you could record a 5 stop image to a 32 bit file you are just really accurately measuring noise. The confidence in each pixel is lower than the precision of your measurement.

If you want to manipulate your linear image though and make it non-linear then you need increased precision to avoid rounding errors and be able to reconstruct the original linear image without introducing banding.

Graeme is assuming that the analog gain is being applied linearly and clipping. I'm saying the point of the dual-gain system might be to apply a non-linear curve in order to preserve all the data but bake in rolled off highlights for exposure and WB.

RED's philosophy is to capture linear and process out non-linear. FLUT exposure, FLUT WB etc.
If I'm right then ARRI's philosophy is to bake everything in as non-linear but stores it in enough precision to be able to go back to linear if needed.

Personally RED's philosophy on paper sounds better to me. It requires less storage.

If I'm right about ARRI's system then you can think of it like having a LUT baked into the image, but including the reverse LUT in the metadata. If you had a 32bit EXR and you stored an image of 13 useful stops of information then along with the reverse LUT you could effectively reproduce the original exactly, as long as the LUT didn't clip any information. The bandwidth would be greater but the processing demands would be lower.

This seems to be what Michael Bravin is saying. 'We bake it in, But you can un-bake it. It's non-destructively baked in.'

Here is another analogy:
Log vs Linear.

If you stored a RED image in a 16 bit log file you could reproduce the 12 bit linear image without problem. The LOG gamma would be "baked in". But that wouldn't stop you from undoing the lin->log operation.

As I mentioned earlier in this thread, I don't buy the dual gain benefits from ANYTHING I have read about them or any of the diagrams pertaining to their implementation. When capturing pixel data, there are a variety of noise sources, but the two main ones we are concerned with are CMOS sensor noise and ADC noise. When designing an ADC, one only has to look at the maximum value of the pixel storage amount and the magnitude of the ADC noise floor. One can simply gain up the output of the CMOS sensor to be larger than the noise of the ADC, select an appropriate ADC that can then handle the range of values from the amplified sensor output at a level where the sensor noise is at the threshold of the bottom that is desirable and there you have it. ADC technology and usage is nothing new.

You DO NOT need variable amplification after the CMOS sensor because it's not helping you. Once you blow out the sensor, that's it. No negative gain is going to recover the lost data. Using variable gains disturbs your noise characteristic in your image. I'd take consistent noise than differing pictures when having to use something that is noisy due to uncontrollable shooting conditions.

You are correct in the sense that if you use a larger bit depth for an image and just spread out the data into that extra precision that you aren't gaining anything in the image (nor are you losing anything though), but, a smart engineer isn't going to do that. If you have an image that is 14 bits of precision due to the sensor, the smartest thing to do is place that 14-bits in the center of the 32-bit data structure IF you allow destructive manipulation in camera (or baked in settings if you will) because you've added numerical headroom around the value to UNDO what you've done to the RAW data.

If you do not have any headroom, as in, your bit depth is too shallow, then depending on how much you've changed your image, you can NEVER get that information back.

The same is true with your LOG->LIN, LIN->LOG example. The ability to reproduce the LIN image is ABSOLUTELY dependent on the strength of the LOG conversion. If the curves are TOO sharp, then the values in the highlights or shadows will quantize to the closest value and when you undo the LOG, it will NOT represent what was there. Those quantized steps will be spread out relative to the multiple of scale between the curve of the LOG at that point and the LIN of the location they came from. (Basically, this can be seen by looking at the quantization steps of the data relative to the tangent line of the LOG curve at those positions to see how the values squash down in the LOG curve).

 

[Mitch Gross]

I'm hoping Alexa will be fabulous. That said, mixed white balance can be better fixed with masks in post if you can shift white-balance further (regardless of whole image or just part) without degradation penalties.

And that asserts an assumption that it would be easier on a REDCODE file compared to an Alexa ARRIRAW file. Since you do not know the how or why of what is occurring in the Alexa or in the ARRIRAW file then I would call this assumption premature.

 

[Rob Ruffo]

How? With extensive roto work to track windows and then mask out all foreground movement that passes them? Using color gels seems like a much simpler option and if someone can't afford photofloods ($5-8 each) or CFLs (less) or a roll of orange gel ($120) I can't imagine they can afford costly post work. And besides...generally, the window light would be sufficiently powerful to negate the practicals' influence entirely--and (depending on window placement and blocking) you'd have huge issues of contrast in this scenario, too, especially since you're pushing against like 500ISO wide open if the foreground (somehow) truly is lit solely by average indoor lighting against like sunny 16 (at 50ISO!) outside.

For the record, I once (just once!) shot a scene at 3200K instead of 5600K and color correcting this will be...a pain. And there is some value to being able to adjust from like 2900K to 3200K totally losslessly in post. But I'm with David, who's right as almost always: if the Arri really does bake in white balance, it's not a big deal at all and I'm certain they have their reasons. While red's method is preferable, if this is a deal breaker for you I hope you never have to shoot emulsion (or even traditional video) of any kind...

Roto is not that expensive when the window is away from characters, like a skylight. It can take 5 minute with a good post op. This is exactly when it is hard to gel. Gelling takes time - on lower budgets, that might mean short coverage or overtime or.... All these little things add up.

 

[Rob Ruffo]

And that asserts an assumption that it would be easier on a REDCODE file compared to an Alexa ARRIRAW file. Since you do not know the how or why of what is occurring in the Alexa or in the ARRIRAW file then I would call this assumption premature.

I made no such assumption - I'm just saying that there is value in not baked in, whatever camera does or does not offer this to whatever degree. My assumption is that ARRI will read this post, and that feedback from all of us is useful on all aspects, as the camera is not yet final.

 

[David Mullen ASC]

Sure you can put gells on all the bulbs in a particular house, or you can put a filter over the window, or have a location scout look for a different house to shoot in which by chance has daylight bulbs installed. Heck you can keep spending time and money anyway you want, or you can shoot with a camera which dosnt care.. You decide.

The Red camera does care. It sees the difference between tungsten practicals and daylight windows and it records it balanced around 5000K, closer to the windows' color temp than the tungsten practical. It sees the green spike in fluorescents as well. What we are talking about is how that image gets white-balanced, not that white-balancing has now been rendered unnecessary.

 

[Gunleik Groven]

The Red camera does not eliminate the color temp differences between sources, it merely captures them relative to its native color temp balance around 5000K. I could load daylight-balanced film stock into a camera, scan the negative, and get similar results. Or set the ARRI to daylight-balance. After that, the process of correcting the image in post for another color temp is the same. So I don't get this "first time in history you can have mixed color temp in the same scene" sort of line of thinking.

If anyone interpreted me that way, I am sorry... -

But you have to have a pretty short memory not to remember (and this is not for you David) how badly RED ONE could perform if the WB was too far off from "native" when you started to correct. IMHO this is one of the major pieces that finally fell into place with build 30 (started to get there with B21)...

Up untill B30, there were significant trade-offs setting the WB low when you started correction when it came to noise and thus resolution.

From B30 the WB settings works like you would hope and expect.

To me FLUT isn't the biggest thing about B30, but this is. Without that improvement, this whole WB issue vs ARRI would be purely academic... Now it's more interesting.

If - as Michael implies - there are no trade-offs in ARRIs approach for RAW WB, fine.

The one trade-off I DO see vs RED, for ARRI (forgetting resolution here...), is the lack of a compressed RAW, which will exclude the RAW options many situations. Many will probably just shoot prores, and argue that it CAN do RAW, but still use it mostly as any tape based RGB cam. Because that's where you keep the compact form-factor. But prores is "good enough". Hey... It's an ARRI....


But seeing how DP's and ACs really love D21 "because it works just like filmcameras allways have done", I think there might be a bit more than a name to ARRIs solutions...

 

[Nik Skjøth]

The Red camera does care. It sees the difference between tungsten practicals and daylight windows and it records it balanced around 5000K, closer to the windows' color temp than the tungsten practical. It sees the green spike in fluorescents as well. What we are talking about is how that image gets white-balanced, not that white-balancing has now been rendered unnecessary.

I seriously hope that you are wrong.. to me is understood as red raw is raw like DSLR raw.. No whitebalance untill post. What you are saying is like, that the camera does bake the whitebalance at 5000K, and so you cant change it in post. Say what?

 

[Tim Whitcomb]

I am not dismissing Alexa - I am only dismissing the reasons for buying/renting expressed so far by any and all parties - I am commenting on info quality - I realize I cannot yet comment on the camera itself, and I am not doing so.. I specifically stated that I hoped more info would be presented in the future, and that it would be a successful camera.

Rob, I am curious why you make up 7/8ths of every comment on this thread lately? You often sound like you pull stuff out of your ass only
to take 4 or 5 more posts to then explain "what you meant" and its nothing at all like what you first wrote.

Dont take it personally, but can you please ignore the impulse to respond to every single post... several times over?

I cant speak for others, but I thought it was tolerable if you WERE a teenager... well apparently you arent a teen? OK/.

with all dude respect. can you at least let other people have a say and quit the ARRI trashing and misquoting, it makes us all look bad.

@ Graeme, Michael said they are still tweaking Alexaraw and that info isnt public yet. Hence my persistence stance lately to please please close this thread for a month or two...

@ Jim - please keep those sexy specs coming!

 

[Dj Joofa]

I seriously hope that you are wrong.. to me is understood as red raw is raw like DSLR raw.. No whitebalance untill post. What you are saying is like, that the camera does bake the whitebalance at 5000K, and so you cant change it in post. Say what?

I think David Mullen is right. The difference between Arri and Red seems to be that once the signal comes out of the sensor then it is recorded as it is, more or less, by Red, where as Arri scales it before recording, but not necessarily color-matrix it, just a simple color-wise scaling. This scaling will be done by Red in software later.

 

[Matt W.]

Roto is not that expensive when the window is away from characters, like a skylight. It can take 5 minute with a good post op. This is exactly when it is hard to gel. Gelling takes time - on lower budgets, that might mean short coverage or overtime or.... All these little things add up.

I suppose it's a mild advantage in like this one situation (which, again, would not exist if you used kinos, jokers and photofloods, none of which break the bank)--but skylights generally blow out to the extent that the image detail can be sufficiently corrected on film, high-end video, or .dpx files because you just don't have that much detail and a near-white patch isn't the focus of the frame--and at a certain point the advantage is more academic than anything. And besides, a skylight is rarely in frame and, when it isn't, it's usually easily fixed with trash bags and duvetyne, still assuming you're lighting with tunsgten for some reason... And if it is in frame you still have to account for the mixed light on the talent and foreground and balancing that out and then timing the skylight to a believable in-between. It's not an ideal solution with either camera and it's only slightly better with the red than even the D21, assuming you don't saturate color channels.

Yes, red's white balance system is better than the D21's. But we don't even know the details of the Alexa yet. It seems to me like a lot of people are looking for and then playing up flaws in Arri's product that may not actually exist. You being happy with red doesn't exclude others being happy that some affordable competition is finally being made available. In terms of price and features, competition only benefits shooters and even if Arri sticks with a fixed white balance, it in no way precludes them from being competitive.

I seriously hope that you are wrong.. to me is understood as red raw is raw like DSLR raw.. No whitebalance untill post. What you are saying is like, that the camera does bake the whitebalance at 5000K, and so you cant change it in post. Say what?

Red's raw files work (roughly) like those in a dslr. It's just that, like a dslr, if white balance in different parts of your scene aren't matched, you have to render multiple files and do additional post work to get them to match. Which is sometimes easy and sometimes hard.

 

[Martin Stevens]

"The White Balance Dialectic"
is Rated R and is coming to theatres soon.

What I love about Red RAW, or for that matter any film or digital cinema
medium that is very easy to grade, is that it allows me to easily re-purpose
a shot or scene during editing.

I once took a daytime shot of a clock on a wall and graded it to look like night.
A night time shot of the clock was not in the script, but we saved an
editing moment in the film because of the shots gradability.
So for me the more gradable a raw file is the better.

 

[Roberto Lequeux]

All this uproar over WB is a bit much -- in all my years, I've never picked the wrong white balance for a scene, whether I was shooting film or video. The odds are low that you would pick 3200K for a scene and then decide in post that it should have been balanced to 5600K. People who need the options to completely change their mind in post about color don't sound very confident in their choices, or maybe the whole concept of having an artistic concept for how the scene should look -- i.e. previsualization and then execution -- eludes them. Or maybe I'm just old-fashioned and don't get the point of shooting footage in such a vague, non-descript and non-committed approach that it can easily be altered in post in order to "find" a visual concept for the shot after-the-fact. Where's the fun in that?

You are totally right David, but don't confuse some of us with hunger for understanding with people who don't know what they want. I admire you so I felt the need to clarify this because it is very important to me, even if it was unnecessary.

On a side note, I really enjoy control over WB. I do minor adjustments to many of my shots. Since daylight varies through a very broad range, if someone dialed in a custom WB, it is always easier to judge off the field in a proper viewing environment.

In the end it is just more options, and that doesn't hurt.

Did I miss something? Instant dailies would be you when you can simply pull out the memory card and hand it to whoever needs it - like the Alexa prores. AFAIK Red Rocket gives you realtime or better than realtime output from your recorded media - good, but not the same. And again you are throwing out the term true RAW, when AFAIK it is substantially compressed RAW. And isn't Alexa native resolution even without the lookaround something like 30 or 40 percent higher than their intended 2K/1080P delivery resolution?

The ability to play back to a proper monitor -or even a 4k projector- via HD-SDI, off the RAW, with full debayer... THAT sounds like an instant daily if there ever was one. Also, if you get full debayer 1080 compressed to h.264 streamed wirelessly straight off the camera, THAT is the other end of the spectrum... BEST quality, and NOW, or both! Then add a ProRes or NHxHD module if you want, but that is a different deal.

Rob seems happy to dismiss Alexa without seeing significant samples from the final product and doing true comparisons - as well as pigeonhole their only possible customer base as established DPs that are incapable of judging the "true" quality of the picture because they might not be as digitally savvy as he.

Me? Really??? Are you serious?

:facepalm:

It seems like some people on this thread are fine with throwing out FUD that isn't even based on correct information - which disturbs me because I believe the R1 and upcoming Epic stand up very well on their actual merits and the fanboy-like misinformation only serves to undermine (at least somewhat) the overall reputation of the RED products which I have already invested in and/or ordered.

If you are still referring to me, then you should read my posts before you address my stand.

Please, tell me I am misreading your post.

 

[Nik Skjøth]

Red's raw files work (roughly) like those in a dslr. It's just that, like a dslr, if white balance in different parts of your scene aren't matched, you have to render multiple files and do additional post work to get them to match. Which is sometimes easy and sometimes hard.

I hope so.

Maybe David was speaking about something else.. While Im talking about the difference between shooting raw vs. shooting jpg in a dslr, you miss the white balance you a screwed. in raw you are not.

The red still dosnt like mixed lights in one scene, even tho it dosnt bake the color, sooner or later you have to bake it at SOME color balance, and thats where the problem occures.. But as said earlier, rotoing and masking gives you oportunities you never had before, you can whitebalance the same scene how ever you like in post.

Here is another.. Timelapses.. Sunlight goes from 5400K in direct sunlight and all way up to 7000K in the shade or overcast. Regulate your post whitebalance to follow the sun conditions.

 

[David Mullen ASC]

I seriously hope that you are wrong.. to me is understood as red raw is raw like DSLR raw.. No whitebalance untill post. What you are saying is like, that the camera does bake the whitebalance at 5000K, and so you cant change it in post. Say what?

It doesn't "bake in" the 5000K balance, that's simply the color temp balance of the sensor, what it doesn't "bake in" are any changes to that balance, that change is saved until post. So if you look at the RAW file converted to RGB, it would be balanced for 5000K. Silicon sensors are naturally less sensitive to blue wavelengths so they balance more correctly in a scene with an abundance of blue wavelengths, i.e. daylight.

Just set-up a tungsten-balanced scene on the Red One and switch from 3200K balance, looking with RedColor or whatever, to "RAWVIEW" and you'll see what I mean, the camera naturally records the image balanced for 5000K, so a 3200K scene looks very orange.

The arguments for saving it in post or doing it in camera, for the most part, revolve around noise in each color channel, which method generates the least noise in the recording, and which generates the least noise in post in each color channel as you color-correct the image.

If you don't want to "bake in" a change in color temp, then just shoot the ARRI at near daylight-balance then, which must be close to its native balance, and save any changes until post just like with the Red, if the idea of the camera doing the white balance for you and changing what gets recorded is so onerous. I'm sure with a little testing you can figure out what white balance setting matches the native balance of the sensor the closest if you really feel that strongly about leaving a change, let's say to 3200K, to post.

 

[Nik Skjøth]

Ok David I get it, but the fact is that the native 5000K image I get, can be bended and molded far more than any other camera in the world.. Thats why I call it a multi whitebalance camera.

The only question relevant here is.. Do you have to decide what color temp to shoot with before post or not?

 

[Rick Burnett]

Ok David I get it, but the fact is that the native 5000K image I get, can be bended and molded far more than any other camera in the world.. Thats why I call it a multi whitebalance camera.

That's clearly not true and makes no sense. If you leave the ARRI to record at the native sensor values and do not apply any difference in white balance, then you are left to exactly what you have with RED in post. You are not altering the values. It's a pretty simple concept.

 

[Nik Skjøth]

Then I dont know what RAW is apparently..

Dont tell me that if i shoot redraw in tungsten light Im as much screwed as I would be if I shot still DSLR jpg's balanced to 5000k

I repeat.. the only relevant question here is.. do you have to decide what whitebalance to shoot with on the red prior to post or not? If YES why raw then? what advantage do I get from raw if everything is natively preset.

PS: sorry for hijacking the thread and replying every 3rd comment, but I want to get to the buttom of this now. You missunderstand me or I missunderstand you.. and I suspect this is the very reason for this looooong discussion about that feature.. People who know how it works, cant explain porpperly it to those who dont.