Too Much is Not Enough
Thursday, November 13, 2008 at 8:57PM 
I was excited about the RED announcements, and they did not disappoint. Watching Jim trickle in his grand vision to the teeming throngs of online fans was good giddy fun and masterful PR. I watch Jobs keynotes the same way, but not with such a close-knit group of online friends. The brilliant Jim Jannard saved himself the expense of the Moscone Center, instead creating the world’s biggest virtual film-nerd sleepover for the mere cost of some extra server bandwidth.
With the usual caveats that their announced specs often change, RED also did not disappoint there—because they announced every spec anyone ever imagined and more. It’s the Forest Gump of technology unveilings—everyone sees what they want, and therefore believes it was made just for them. The renderings are very pretty, and anyone from a fan of the original Scarlet design to a stereo IMAX ride film producer can see their needs borne out in shiny black ray tracings.
And that’s the idea, of course—that anyone should be able to assemble their perfect camera for their needs from Jim’s bag of building blocks.
But I’m afraid that I’m not seeing my perfect camera among the options.
I’ve mentioned before that what I like about RED One’s 4K resolution is that it makes a very nice 2K image. I don’t think films benefit much from 4K, and for my own filmmaking there are a hundred places I’d rather put my time, money and energy than into more pixel count than most audience members can perceive under the best of conditions. But a rock-solid 2K image requires some downsampling from a bayer-pattern sensor. The sweet spot for great-looking 2K RGB is between 3–4K bayer origination.
So 4K is the cap on my resolution interest. But as far as sensor size goes, I do quite like the choices offered, especially right in the middle of the lineup. A cinema-sized Super35 chip is nice (RED’s S35 size), and a full-frame DSLR (or VistaVision for movie folks) sensor is kinda cool too (RED calls this FF35). Both would work very well with lenses I already have, which was one of my hopes.
But the least expensive RED “brain” with a cinema sensor is the Scarlet S35, and it’s $7,000. That’s without LCD, buttons, or lens. Although it’s the cheapset option with a big chip, it shoots 5K, which is way more resolution than I care about.
The Scarlet FF35, which has a sensor the same size as the Canon 5D and 5D MarkII, is $12,000 for just the brain. It shoots 6K, adding more overhead to my workflow that won’t wind up on the screen.
Worse still, these boxes only overcrank up to 30 fps, whereas the $2,500 2/3” Scarlet can run up to 120 fps at a very usable 3K.
If you want more frames-per-second for slow motion, you have to go to Epic, where a S35 sensor costs you $28,000 and a FF35 sensor is $35,000. For that you get 5K and 6K respectively and up to 100 fps. Still not as much overcrank as the cheapest brain with the most sensible resolution, and a ton of dough for an unwieldy data rate.
I look at these options and I feel like I’m paying for pixels, when I’d much rather be paying for dynamic range, frame rate, and sensor size. For a stills camera, these 5K and up resolutions make sense. But for filmmaking (and I’m talking professional filmmaking, not DV Rebel—this is not a DV Rebel camera discussion here!) it’s just not of any real value to me. More pixels means less dynamic range. More pixels means more data being pushed around that doesn’t make my movie better. More pixels means more compression to meet the low data rates of the less expensive brains.
The brain I’d be excited about would be a S35 or FF35 sensor at 3K, with up to 120fps. The pixels would be huge, the dynamic range would be great, and the 3K bayer would downsample to a lovely 2K image, which would be more than sharp enough to show me that I’d miss-focussed, or that my lead actress had an unfortunate allergy to her makeup.
If you absolutely had to make this thing 4K that would be OK, but not if that limited me to 100 fps, and not if it increased the price.
Remember that RED’s raw recording methodology means that you can’t decouple the Redcode Raw recordings from the native sensor resolution. If you buy a 5K brain, you can’t shoot anything other than 5K unless you’re willing to window in on that big chip, eliminating the whole point of the cinema-sized sensor.
By aiming for a piece of the stills market, RED has encumbered the sweet spot of their movie-making product line with unnecessary pixel counts, sacrificing much more important things along the way.
Reader Comments (96)
All these great comments/introspection... I am now interested in how much $$$ it will take for RED's "minimum build"...ie. something that you could shoot with. No one knows how much the rest of the stuff is yet? But it is starting to look like I could settle for the cheapest one and still get crazy good images.
GlennC, all other things being equal, fewer pixels does equate to greater dynamic range. Fewer photosites means larger photosites, larger photosites gather more light, which means they require less gain, which means less noise which means greater dynamic range.
This is well understood by the various photography review sites, all of which have been actively deriding the "megapixel race" that has resulted in some consumer cameras with enormous file sizes and poor low-light performance.
I personally think that 2/3" with S16 and c-mount lenses is kick-ass tool.
You can get very serious class for c-mount for the price you are renting your Motion lenses. They can be more tedious to work with (smaller barrel, deviations from lens to lens and some of them dont have scales on)
But think this, you can use T1.3 on most high rez c-mounts for shooting, no need to stop down. Suddenly low light becomes possibility. The rig is much smaller, the DOF is there, rez is exactly on the sweet sport or bit over.
The price for the set is excellent (Dont even want to recall what I paid for my XL1 back in the day)
I will most definitely will get one extra to these SI's that I own.
Hmmm... I've just received a marketing email from Nikon. Maybe Nikon is listening. This is probably coincidence but it's interesting that Nikon should send out the following email the day after RED's announcement:
"Welcome to the launch edition of In-Frame – Nikon’s informative e-newsletter that we’ll be sending you each month, enabling you to keep up to date with all the latest photographic trends and innovations.
A WORLD FIRST - Nikon D90
The world’s first D-SLR with movie function is now available from Nikon! This new 12.3 megapixel D90 offers stunning image quality, innovative high-performance features direct from the award-winning D300, and truly cinematic results with its unique High Definition movie function.
..."
Go on Nikon, release a compressed-RAW full-frame movie camera for under $5000 and beat RED at their own game. You know you want to ;-)
Stu, I started to leave a comment 3 times yesterday, but then canceled it waiting to see what you said first. I COMPLETELY agree with you.
1) the DSLR part of DSMC is sidetracking RED away from what I want from a video camera. At the same time, the stills aspect of these cameras don't look to be so great. They are huge, heavy, and can't be NEARLY as refined or low noise at high ISO as Canon/Nikons are currently.
2) I don't want more resolution, I want more sensitivity and dynamic range. Can't emphasize this enough.
3) If it's modular, give me a modular compression engine so I can take a low bit rate recorder out for a documentary, and then swap and use the high-end codec recorder for commercial or post work!
Maybe I should start thinking more about 2/3" CCDs and T1.3 C-mount lenses. I swore to myself that my next camera would render my DOF adapter obsolete! Dang... why is new technology always so frustrating!
There are still good things in this announcement- the wireless remote is a sweet idea. Having things so modular has got to be better for ergonomics than brick designs. I can theoretically rent the high-end sensor for those shoots, and own the lower end sensor. (assuming my rental house gets one).
Gaah, at least we have a year or so to fume about this, and hope Red changes some things.
Stu, you pretty much nailed it, but let me break it down even simpler for you:
What I value:
• 1080p resolution
• 24p framerate
• Full size 35 sensor
• Uses affordable SLR lenses
• RAW recording/workflow
• <$4000
Nail that and you have an indie dream camera that will FLY off the shelf.
The Canon 5D MkII actually comes closer to this than the new RED offerings. The problems: Lack of 24P, no RAW recording option.
Higher specs in any area are nice, but these are the fundamental 'bases' we all want covered--ALL IN ONE CAMERA PLEASE!
Well said sir, well said.
I'm with CareyD. But I'd also like to add 120 FPS.
I hear your arguments. Here are some counter arguments. The cost of the sensor mostly comes from it's over all area, not it's pixel count. So if you want to make a larger sensor, you might as well make the pixel count higher. Also, there is usually a "sweet spot" in terms of pixel size vs S/N. What I'm getting at is that at a certain point it just doesn't make sense to produce a large sensor with a low pixel count. The trade offs just aren't worth it. The senor would cost the same as a larger sensor, but wouldn't provide a good enough improvement in S/N to make it worth the cost. Plus, as many have mentioned, it's the pixels that sell. Having to educate ever buyer to explain to them why your pixel counts are lower than everyone else is a tricky proposition. I don't know why any manufacturer would want to go down that path.
So, more pixels (well, smaller pixels really) usually equal more noise. But looking at the Scarlet s35 sensor data, it's clear the pixel size is actually larger than the 2/3". (5.4 vs 3.2 Micron)
In terms of the work flow argument... Yes, the overhead of the extra resolution of going "Full Ap" on the S35 could be problematic. But aren't you forgetting the wavelet aspect of REDCODE? At worst the extra res just means the files are larger on disk. For rendering, you could just decode at half-res and get a 2.5k file to down-sample to your target 2k.
Also, if you had the time for it, you could render from the full 5k wavelet and scale down, gaining the benefit of the extra oversampling, which would provide more dynamic range by averaging the pixels together. The only bottleneck is CPU in that case. And once the native REDCODE support matures you'll only ever need to do that once, when you render your final output. ( And let's not forget third party options like Cineform and their "express" option.)
I understand the issue of the frame rate being stuck at 30p and having to window to get anything higher. But even still, the s35 Scarlet looks like an incredibly versatile camera. My guess is that 3k you'll be able to get the same frame rate as the 2/3" version.
For those looking for a "better HV20" the fixed lens 2/3" Scarlet could be the answer. Red has yet to announce pricing. It might end up being a really great value if you don't really care about having an s35 size sensor.
Even after the Red Centre podcast, I am unsure whether the fixed lens 2/3" is going to be more or less expensive than the interchangeable 2/3". If it is less, then I think that's a pretty kick butt rebel camera, if it is more, i dunno.
With the rest, I am not impressed with the price/capabilities in terms as a stills camera, but as digital cinema cameras they seem to open up some great possibilities. Canon & Nikon have to address their features and namely their compression/codec before I can really compare them as motion cameras. True we say Canon has it except it's compression and 24p. Well those two things are HUGE. Will Canon & Nikon care enough to get their moving images into professional & robust files that can stand up to heavy image manipulation and CC?
Careyd you suggested:
• 1080p resolution
• 24p framerate
• Full size 35 sensor
• Uses affordable SLR lenses
• RAW recording/workflow
• <$4000
Add to this 'good manual control' and I'm there.
(And the price closer to $2500 as I have to pay for this with NZ dollars.)
Wow, Stu, spot on. Some great points in there that I hadn't really thought about. A larger S35 sensor that could at least do 30fps would be simply awesome. Still probably out of my price range, but it would be the first camera I'd reach for if shooting a low-budget feature. Nice. In the meantime, I pine for the lost '3K for less than $3K' Scarlet of old...
Yet another brilliant post and why this blog is my fav of all filmmaking blogs.
Stu - is there anyway you could design a camera? I know you don't have Jim Jannard's pockets, but Red does have competition.
I'm sure you could gather a bunch of support via phone calls, emails blogs to a manufacturer.
And manufacturing companies often hire outside designers who are not necessarily engineers. After the huge market success of the DVX100 and HV20/30 series, you would think some company would get it by now.
Yet, here, Red, in absolute perfect position, did not. I suspect (and yes this is speculation) that because JJ/Red is more about status and ego, making a humble "rebel camera" that can't brag and beat up "the big guys" is at best, not their priority, or at worst , they feel it's beneath them.
Here's hoping for an HV30 with manual lens, controls and XLR audio or a DSLR tweaked for indie DV shooting.
CareyD: You nailed it.
RocketBoy: Regarding your follow up to Carey... not "GOOD" manual control, FULL manual control. Maybe that's what you meant, but when you had a camera 4 years ago with the level of control as the DVX, there is no reason to settle for "good".
Lord Tangent: I could not possibly disagree with you more. When you are setting out to make the best cameras possible, make them. And that means ignore the pixel count and focus on what matters. You say the customer only looks at the pixel count? That hogwash. That's my Aunt Betty when she is buying a $99 Walmart digicam. Red says and THINKS they are making cameras for PROFESSIONAL CINEMA, well, I know one or two ASC cinematographers that understand what DYNAMIC RANGE is and how it is better than "megapixels".
Those that dont understand, like the fanboys at RedUser who get excited about 6k and 9k - EDUCATE them. Dont F*ck up the cameras you make available just because some people need a little 'splaining.
(Caps are for emphasis, not yelling. I'm far to lazy to type in BOLD brackets everytime.)
This seems like your most successful post to date! Could your next blog post be about how these RED announcements affect the DV rebel in particular? How good or bad is this for DvRebels willing to spend around $6,000 for their camera? Thanks.
I got my post over at scarletuser locked after only 7 posts.
http://scarletuser.com/showthread.php?t=1730
Perhaps a little too close to the truth.
Nice, Casey. Not even a warning or a mod comment. Certainly no "defense" from the Red team. Just closure. LOL.
RocketBoy: Regarding your follow up to Carey... not "GOOD" manual control, FULL manual control. Maybe that's what you meant, but when you had a camera 4 years ago with the level of control as the DVX, there is no reason to settle for "good".
Yep, by 'good' I meant full control and well implemented. Like having a focus ring and someway to quickly adjust the exposure without having to dive into a menu.
An update on the frame rate limitations: Although RED's Ted Schilowitz stated that SCalet S35 and FF35 were maxed at 30fps in the Red Centre podcast recorded yesterday, today Jim posted specs that show windowed overcrank modes for the bigger Scarlets:
http://reduser.net/forum/showthread.php?t=22052
60fps max.
Epics are to go up to 200 fps and possibly more as you window in.
Window in .. and higher fps... That is downright nifty. I need to make more money. Stu, is there any money to be made in filmmaking?
;)
Matt, you say "window in" like it's a good thing.
windowing down to 2k for 60fps (for the S35 Scarlet) will use more surface area on the sensor than the 3k Scarlet... Not a bad compromise.
if i need a higher fps or ramping, i'll rent an epic brain for a day.
Dude I have an HV20.... I have windowed in on that on "professional" projects... it's ridiculous.. the window on the FF35 is like 2 RED Ones big
[i]Matt Moses said...
I said that 5k = 72MB a frame... I was wrong. It's 140MB a frame. (6000x4000 @ 16bit depth)[/i]
Sure, but no-one forces you do debayer the RED output to such a format, if you don't want to.
Go for i.e. the industry standard 2K 10 bit log for all the VFX elements that don't need more. Heck, go for 720p if it's enough...
But IF you need to i.e. zoom into a plate, and/or do extremely delicate color correction, you have the [i]option[/i] to bloat your file to 6000x4000.
16 bit depth is nothing shot of an overkill - the original Scarlet files are 12 bit linear, Epic is 13 bits at best. Converting to 12 bit log should give more than enough precision. 10 bits is fine for Scarlet.
But if you need to bloat your file to the specs you described, you have the [i]option[/i] to do so.
How can more options be a bad thing?
Cool.
All I want is the 5D MarkII with no or less compression (REVERIE impressive still), timecode and manual audio controls.
RED is a truly amazing company I have most respect for!!!! However I am also truly amazed that Scarlet is no 5D Mark II-Killer at all - especially not for 2000-3000 Dollars.
S35 frame
boatloads of dynamic range
3k-real, usable pixels
PL-mount & Nikon interchangable lenses
great handheld ergonomics
120-300fps
in other words, the A-Minima of the digital world.
Still waiting...
Brian,
I'm assuming you simply disagree with my assessment of dynamic range vs resolution in terms of weighing priorities. I totally get the importance of dynamic range. It's one of the areas where I think digital still seriously sucks. ( And You don't need to wave around your association with ASC members. Much like Stu, I work professionally in the film industry as an effects artist. Been doing it for over 14 years now. I'm quite experienced with both film and digital. I know what I'm talking about and it isn't second hand knowledge. )
Except for the sensor area lost due to "fill factor" smaller pixels are not necessarily the end of the world in terms of the effects on dynamic range. As I said, it's a trade off and there is a sweet spot. You can always combine the pixels together, which is to say, SCALE THE IMAGE DOWN, from a higher res capture which has the effect of improving S/N (AKA "Dynamic Range") on the smaller output image.
My point is, on paper, the s35 Scarlet is still an excellent design. It's very flexible. You can literally "have your cake and eat it to". Resolution or dynamic range? You choose. Now, it might not be cheap enough to get a "Rebel" excited. But it is an excellent design. Heck, it will even accept SLR lenses. That is huge.
I'm especially not really feeling Stus argument regarding handling of the higher res output. It's just not strong enough. Especially considering "Moores Law" and the fact that by the time these things actually ship, the "problem" of dealing with the extra res of a 5k Bayer image will be a non-issue. And even today, the waveletty goodness of REDCODE lets you work at smaller resolutions as required.
Lord Tangent, I'm sorry you took that personally. I dont question your credentials at all, and congrats on the job, but I was not suggesting that I know people better than YOU, I was challenging RED and their seemingly "megapixels are better" attitude due to ignorance when they are trying to market to the PROFESSIONAL film industry, hence ASC DPs. (And DPs from any other organization or no organization at all.)
So don't worry, I'm sure you know things, I just don't agree that as long as you are making the sensor bigger that you should add more pixels.
I also understand that there is a sweet spot, but just "pack them in 'cause we have more sensor" isnt really the best way to find it. Also, from my understanding, while DR and S/N are RELATED, they certainly aren't the same thing. I understand that shooting at 4K and downrezzing to 2K will minimize the appearance of noise, but I don't understand how that gives me more exposure latitude.
Here's some interesting reading on sensor sizes, noise, etc.:
http://www.cambridgeincolour.com/tutorials/digital-camera-sensor-size.htm
Let's take a quick look at Scarlet's competition. All the HD camcorders under $10,000 can only shoot a maximum resolution of 1920x1080, and they are only able to shoot 1080 at a maximum framerate of 30fps. Also, the largest sensor size of all those camcorders is 1/2 inch (Sony EX1). And all of those camcorders need a 35mm adapter to take advantage of 35mm lenses and their DOF (along with a significant loss of light!). When you take all of that into account, Scarlet is an incredible achievement.
It was great to learn that the s35 and ff35 will be able to shoot 60fps at windowed 3K, but I'm hoping to see the spec change to allow 96fps at windowed 2K (I prefer to shoot higher framrates in multiples of 24!) Hopefully the hardware "brain" would allow for this, and IMHO it would turn the Scarlet s35 and ff35 into the ultimate indy filmmaker cameras.
Clearly, it's a little painful to see that 120fps feature on the Scarlet 2/3", knowing that it will never be the camera for serious filmmaking. Increasing the maximum framrate on the s35 and ff35 are really the only thing about RED's new lineup that I would like to see change.
Paul, you are comparing the announced RED specs with the cameras that are available right now. The RED cameras are at least 6 months to a year away.
Who knows what Sony, Canon etc are planning? I am sure they will not just sit back and let RED make all the running.
Stu, excellent article. The section on lens diffraction is also good reference for people not familiar with the phenomena. That plays into MTF and makes another argument for larger pixels.
Brian, S/N is essentially the technical factor of a linear digital sensor that determines how much "dynamic range" one can hope to usefully extract from it.
Think of exposure on a digital sensor this way: Successfully capturing the entire dynamic range of a scene isn't about the MAXIMUM value of light the sensor can record, it's about the MINIMUM value it can record with acceptable S/N. That is because to control exposure (the MAX value in the image), you can always just stop down to let less light in to the camera. What limits how much you can do that before the bulk of the image becomes un-usable is the S/N on the sensor. Am I making any sense? Remember, we are talking about linear sensors here.
I'm an advocate of programmable LOG CMOS for just that reason. It is more like film. Actually superior in dynamic range if set up correctly. LINLOG sensors are a great way to get around the S/N issue at an analog level on the sensor. Amazingly for some reason no major manufacturer (other than scientific imaging folks) have chosen to go with them.
And, just so I don't look like I'm being a jerk, or inflexible on my opinion on this I want to mention... a good buddy of mine reminded me of a huge portion of Stus original argument for the 3k s35 Scarlet. Frame rate. Stus hypothetical 3k s35 Scarlet would be able record frame rates similar to the 2/3" Scarlet without the more expensive compression circuits the Epic requires.
Holistically, it is a very good blend of features. And in that respect, I think it is a great idea. But it would not cost any less. It would cost around the same as the 5k Scarlet.
Personally, I would rather have the higher res, for the superior over-sampling of detail in the scene, removing more of the "curse" of shooting with a single chip cam. As I mentioned, higher dynamic range can also come from over sampling. The higher frame rate isn't as important to me, personally. But now that Jim has announced the s35 Scarlet will shoot 60fps 2k windowed, we know there will be a fast frame rate available for when we really needed slo-mo. Sure, 60fps isn't as good as 120. But it's good enough for me.
Like I said, on paper the road map of Red cams is very impressive. If they added something like Stus Scarlet to the line up, it would add more choice. Assuming BOTH were options, I wonder what most people would go for if forced to choose between a 5k s35 Scarlet and a 3k s35 Scarlet with bigger pixels and better over cranking. You already know what I would choose.
Great post, and I totally agree.
If anything, the $2500 Scarlet would be my pick. I'm not familiar with C mount lenses though. As for resolution, 3k would be perfect, and with a frame rate of up to 120fps at that resolution, it would be brilliant. Plus, I'm sure the all the 35mm adapter guys will find a way of enabling their adapters to work with a 2/3" sensor. Thankfully at least I can't see these Red cameras putting them out of business anytime soon.
I agree with your comments and I was puzzled, disappointed really, with the Scarlet specs that seem favor raw resolution numbers for seemingly, gratuitous reasons.
Just to be specific, I'm looking for the following: 2K resolution, 35 full frame imager, durable codec with 4:4:4 color, multiple frame rates, ability to use my Canon DSLR lenses, all in a system that will give me a complete machine for under $4K (minus lenses). The best bet right now, looks to be that a DSLR manufacturer will get it right. Maybe Panasonic?
I wish RED would have hired you as a consultant for Scarlet, or at least read your blog.
This may be a stupid question, but I hope techies out there can answer it.
If a 1000 res sensor with fatter pixels yields more dynamic range that a 2000 res sensor - the logic being that the first sensor's pixels has twice the size to pick up light; then why not have the 2000 rez sensor double up each pixels. So two pixels work as one. Why would wouldn't that have the same functionality of a native 1000 res sensor... and the option of being a 2000 res sensor.
Enquiring minds like to know...
I've been researching c-mount lenses. I'm thinking those could be the way to go with the 3k Scarlet. I found these at B&H - http://www.bhphotovideo.com/bnh/controller/home?ci=0&shs=schneider+c+mount&sb=ps&pn=1&sq=desc&InitialSearch=yes&O=jsp%2Fproduct.jsp&A=search&Q=*&bhs=t&Go.x=0&Go.y=0&Go=submit
Now look at that 17mm lens, it's an f/0.95! According to my calculations that would be roughly the equivalent to a 66mm f1.8 to f2.0 35mm SLR lens. Which I think would be fast enough for most people.
Opinions anyone, Stu?
I agree with Stu. I have to chime in on this, in hopes that RED/Jannard reads this stuff and will be swayed:
the PR work on these is great; but one gets the feeling it's all for the sake of gear porn (and the comparison with Steve Jobs is apt in that respect too).
A lot of bloggers etc. have repeated the RED press releases, mentioning the 5/6/28k resolution, saying "yeah well more resolution is better, right?" but of course it isn't.
How about a great pocket sized camera with a fullframe 35mm sensor and beautiful lenses that even shoots *just* 2k? It would be a bigger revolution for cinematography than any of the models they announced.
I think RED's assumption that they have to compete with digital stills cameras is a huge misstep. I think they should have let the Canon 5D mk2 and the Nikon D90 blow by. Nobody is going to shoot a feature on these things if RED can offer them a *real* cine-camera that shoots 2 or 3k images with a bigger sensor and better dynamic range and a post-production workflow that is rock-solid for the same money.
Proof-of-concept and PR are not what counts at the end of the day. Filmmaking is still a different challenge and a different process than still photography and lumping them all together in to some kind of hybrid is a mistake. How about thinking about what the special challenges of filmmaking are, that make it different from other digital image capture, and go with that? Then do the same for photographers? Both will appreciate you much more in the end.
The whole FPS vs Canon Lenses really has me PISSED! I live in Jackson Hole so no suprise that I have at arms reach access to nearly every Canon lense made givent heamount of still photographers around. I also made a consious effort to get into Canon still knowing the glass would provide a platform one day. That day is here and it looks like slo mo went out the window!!!
Is it possible that Red is making a creative statement by limiting the fps on some of the cams, because there's already way to much slo-mo and speed ramping in a lot of editing these days? ;P
Me,
Combining pixels is exactly the strategy that Panavision uses on the Genesis. The way they do it is a little strange and IMHO, sub-optimal though. They do it right on chip which leads to some aliasing problems.
I would rather have access to the full res of the chip with a traditional Bayer pattern (rather than the funny RGB filter of the Genesis) and the ability to scale in software (unlike the Genesis). That would give a lot more flexibility in how to use the camera.
And for those who just can't let go of this "Gimme a low res camera" stuff. Remember, to get a good MTF with a single chip camera you MUST over-sample. A 2k Bayer patterned sensor doesn't provide 2k of usable resolution.
So, my argument is that by oversampling you gain
1. Resolution.
2. Dynamic range.
I don't care to use the 5k sensor to get 5k. That is worthless to me. I would use it to get 2k with +2 stops of dynamic range (over it's native dynamic range) and better oversampling on fine details in the scene.
Lord Tangent, you're making a convincing argument to me on how you can get improved dynamic range when you oversample. So what is the path in software to take that 5K footage and create 2K w/ extended dynamic range?
Lunelson: I agree completely on the gear porn factor.
All: For documentary/event workflows, what would be the viability of using the 1080 HDSDI outputs on the RED and send them into one of those Convergent Design Flash XDR recorders? http://www.convergent-design.com/downloads/Flash%20XDR%20Brochure.pdf
Can you set up gamma curves, knee, etc in RED camera enough to give you a good image in this path? I realize it wouldn't be nearly as nice as REDCode, and you wouldn't get any real benefit of the oversampling that we talked about, but it would be a shortcut to longform recording off a S35 sensor with full manual controls.
Finally, does anyone know if the Nikon DX or Canon EFS lenses will cover a S35 sensor completely? Those lenses are smaller, lighter, and sometimes more cost effective. Also, with the S35 sensor you'd be missing out on the wide side of things with your DSLR lenses that many of us plan on shooting with this kind of camera.
I did some digging to answer my own question- Red One sensor is a tad bigger than APS-C (canon's EFS size) and S35 is quite a bit larger. So you'll be using FF35mm lenses on both the S35 and FF35 scarlets. Not sure what the crop factor of FF35 to S35 is, but it's certainly not as severe as 1.6 for APS-C.
Tom,
To get the extra DNR you would set the camera ASA two stops faster than its "rated" or optimal ASA. On output you would debayer to 5k and then scale the image down. It's that simple. Sort of. Really, exposure on a RAW shooting camera is more complicated than that. You should only ever shoot that way if you are hell bent on shooting like it's a film camera.
Ideally, exposure-wise, your goal should be to expose for the whites. The brightest things in the scene determine your exposure. You should always strive to use the entire DNR of the sensor if you want a clean image.
By averaging 4 or more pixels into one you get a superior S/N (averaging 4 pixels should be equal to about 12 db improvement in S/N which = 2 stops) The results you got would depend on the scaling filter used. But any good software should be able to do what amounts to basically a 2:1 scale. The main thing is you want to do the image scale in linear space BEFORE the grade if possible. This is where using a node based image processing package would be good. Though I guess the he same thing could be achieved in After Effects using nested comps.
The specifics of the work flow depend on a million factors. What are you editing in? Etc.
Here is an idea for a simple cheap solution: Cut in Premiere (using the native Redcode plug-in, which should totally be mature by the time Scarlet ships) and do the final render out of After Effects. It loads up Premiere edits directly. AE functions in Floating Point so you have everything you need right at your fingertips, for cheap.
As an engineer all I can say is WOW. From the perspective of making sure everything works together in every possible combination, I'll have to wait this one out. Given it's price, I'll have to stick to the 5D and the D300 for the time being.
Can't add anything much at this point, except another amen. Thanks for the blog, Stu -- I hope that if RED isn't listening, the other manufactures are. I want low light sensitivity over super high resolution. 1920x1080 is plenty, although 3K sounds nice for cropping and image stabilization in post, and I think RED Raw format is brilliant (if you don't want to use it in post, just transcode to ProRes or DXnHD). I do not know enough about sensors to know if this is possible, but what would a 3K sensor that's actually 35mm wide look like? The photosensors would be huge, and the low light sensitivity/dynamic range would be off the charts.