View Full Version : Food for thought regarding Digital Cinema vs Film
David Fairbanks
01-28-2007, 09:13 AM
Since our goal with Red it go get a camera that helped up produce products that look more like film without all the detriments of film. We'll now have a camera that shoots higher quality images than film (with the possible exception of 70mm) without the grain (now addable and controlable in post). But one thing still illudes us as far as the look of film, motion. Digital still has a crispness that film doesn't. I don't think this is something that the Red Team has looked at.
On digital, as soon as the sensor is turned on for a frame it is recording information. As far as I know there is none or almost no ramp up time and then it cuts off just as quickly. This produces a rather sharp edged blur on all moving objects.
Film on the other hand takes time for an exposure to even register an image. Plus the shutter itself only exposes a sliver of light at first then gradually opens to full apature. Then at the end of the exposure, the shutter gradually closes to a sliver. This produces a soft gradual start to any motion blur.
I haven't actally tested this, but I would suspect that film would have an exposure level over time of something like a bell curve. While an electronic sensor (either CCD or CMOS) would have an exposure curve much more like a square. These images are just a guess, but I think they're close to what is going on.
http://www.cinemagraphica.com/tech/images/CCDExposure.jpg http://www.cinemagraphica.com/tech/images/FilmExposure.jpg
At this point this would probably have to be a feature of Red Two, but it sure would help make digital cinema look more like film.
Phil Becque
01-28-2007, 09:46 AM
I'm not sure I agreed with your first statement. I'd like RED to better than film! More like reality! But that's all subjective.
Hmmm - I know what you mean about the edge rates but I don't think that is cause of edge breakup in fast motion digital shots. I think the main causes are:
1) Frames per second
2) Codec artifacts
3) Limited data rates of ancilliary equipment i.e. anything your images pass through.
4) Incompatibilities between playback & monitoring systems
The images we get from any sensor are effectively sampled with whatever light is hitting them at the time. Doesn't the same apply to an image on a strip of film? Once it's exposed - it's there - burned in for good.
I should say I don't really know about film - I've never used it. Maybe the way film is displayed is the difference?
David Fairbanks
01-28-2007, 10:08 AM
There's an inherent difference in the way film is exposed. Film does not instantaneously render an exposure. It take time (miniscule it may be) for it to ramp up to enough exposure to produce an image. But the main point is that the shutter does not allow 100% of the light in at the beginning of the exposure. It's a gradual increase because the shutter is wiping accross the gate.
I'm not taking into account the compression or display in any of this. I'm just speaking from years of experience in film and now digital both stills and motion picture.
donatello b
01-28-2007, 10:10 AM
anyway you look at the how the image is recorded it's apple-oranges ..
digital does not record images like film so all the fine details of comparing how film is recorded and how digital is recorded is totally different and depending on if you are pushing your producer to shoot film you use those points - so if in ones mind is that digital must copy the way film does it = it ain't going to happen = well just shoot film ...
bottom line is the IMAGE ... look at RED images - they are not film images they are RED images - i like what i've seen from RED ...
"Digital still has a crispness that film doesn't"
i've seen that in HD and on point & shot digital still camera's ..have not seen it on RED images or the top of line Digital still camera's
next up in the chain is digital projectors - again it delivers the image to the screen nothing like film projectors = it's a changing technology - for those of us that have watched FILM projected for many years with the 72fps light strobes ( 24fps x 3 bladed shutter ) we will miss it on digital projection ... in another 15 years and those that only know digital projection will say whats with the light strobing when they view film projection ( the few times they see it) ...
Tony Lorentzen
01-28-2007, 12:46 PM
In my mind the whole film vs. digital boils down to this:
In 20 years from now people will have gotten used to digital. Sure - for the next years we will feel somewhat "lost" in transition, but after that digital will be what people expect and film will be like vinyl records is today.
SalaTar
01-28-2007, 12:49 PM
The digital transition has not fully been completed….
Some prefer Analog and some like digital…it will stay like that for years
Peter Richardson
01-28-2007, 12:58 PM
I think David makes a really good point here -- I remember seeing the first images from the Sony Cinealta at a workshop and a lot of people were taken aback by the "strobing" of the image -- a much harsher look than film. That being said, by the end of the workshop, no one seemed to notice it, as our eyes had adjusted (to Don's point).
The only piece I might add here is that, while the Red images shown at the LA screening were stunning (from accounts I have read -- wasn't there, unfortunately) I seem to remember some people mentioning this "strobing" effect of the images and that it was particularly pronounced in the REd. It would be interesting to hear Graeme's thoughts on this (if he has time) and whether this factor has been taken into account with the codec or whether there is really anything that can be done to address it. Thanks for bringing this up David -- very important and definitely overlooked aspect.
Peter
David Fairbanks
01-28-2007, 02:52 PM
Ultimately it will come down to the end product. While film has been the gold standard, digital has the protential to be even better, and I think Red is the first leap in that direction.
I agree with tlorenzo, that film will go the way of vinyl records.
Kjetil Haugen
01-28-2007, 03:05 PM
This being a suggestion box kinda forum for the Red guys, I think this is probably the best suggestion I've read here in a while... I've also been stuggling with accepting the "strobing" part. And I agree that over time things will be forgotten, but if something infact CAN be done, then why not try? Nice comment, David!
Cail Young
01-28-2007, 07:52 PM
Would this be why ARRI put a mechanical shutter in the D20?
Nook Kim
01-28-2007, 08:02 PM
Personally, I like the idea of having Red's own look, like people have
mentioned, before. It's like a new film stock coming out.
Some people don't like the digital filmmaking because they think it's
everywhere with so many different formats/looks to consider/learn.
They say digital filmmaking needs a "standard."
However, I really enjoy having all these different digital cameras with
the looks that each offers differently. Depending on the project, I
can choose the one that I want.
Why bother to come out with a new technology when there's film that
has its look? I don't think it would be fun for the Red team to do all
this development if there were to imitate something.
Also, about video looking image mentioned before, I really don't think
more than 1% of regular audiences notice that Superman Returns
or Apocalypto was shot digitally.
Sorry that my post doesn't really have a point. Just a thought.
Nook
Andreas Fernbrant
01-29-2007, 02:35 AM
Interesting thread.
I think this is VERY important and valid thought!
From a personal view I always look at the picture as a whole. Resolution charts dosn't do anything for me.
Latitude, movement/blur and texture are the things that turn me on.
Like many others I can't wait to see some new shots!
Direct sunlight and fast movements. Go shoot some motorcross guys:)
Zach Hilton
01-29-2007, 10:55 AM
Personally, I like the idea of having Red's own look, like people have
mentioned, before. It's like a new film stock coming out.
Some people don't like the digital filmmaking because they think it's
everywhere with so many different formats/looks to consider/learn.
They say digital filmmaking needs a "standard."
However, I really enjoy having all these different digital cameras with
the looks that each offers differently. Depending on the project, I
can choose the one that I want.
Why bother to come out with a new technology when there's film that
has its look? I don't think it would be fun for the Red team to do all
this development if there were to imitate something.
Also, about video looking image mentioned before, I really don't think
more than 1% of regular audiences notice that Superman Returns
or Apocalypto was shot digitally.
Sorry that my post doesn't really have a point. Just a thought.
Nook
That is a strong assumption. Superman Returns and Apocalypto had a completely different look. Which is mostly due to the fact that a large percentage of it and the actors were digital. I think the percentage is a lot more than 1%. That's besides the point. I think the issue of this whole topic is that of having the OPTION to make something look like film or having the OPTION to make it look like crisp HD. Options. Of course those have been readily available in film, HD, etc... But what about having those different looks all in one device. Namely, the RED. We (independent filmmakers and others alike) are investing in a piece of equipment that will provide us ultimate control over the look of our film (in this respect, movie). So instead of having to shoot on film, or shoot on HD to achieve those looks, wouldn't it be grand if we could choose the look we want in pre-production and such, and actually attain that with one camera? That would seriously lower the cost or even the thought of "well lets just shoot it on film." Even though RED does have its unique and special look, wouldn't it be great if it were fully programmable, and customizable to achieve whatever look the project calls for. Because in the end, that is what really matters.
__________________________________________________ ______________
RED ONE #1108
kraemer
01-29-2007, 12:24 PM
I haven't seen mention of frame rate here. I think most significantly, "temporal resolution" will be vastly improved by usage of the 60fps frame rate. Then when you are panning right or left, you'll actually be able to read words and signs, as well as see movement very clearly.
Also, when special effects are being created in post, they'll spend a lot less time trying to make them blurry like 24fps film.
I think the real unspoken potential here will be in camera HDR when you have 90 fps data rates and some sort of digital aperture. (Red two?) You could then have a 30fps clip with three exposure levels for each frame being composited in-camera.
http://en.wikipedia.org/wiki/High_dynamic_range_imaging
Júlio Taubkin
01-29-2007, 12:34 PM
I don't know. We're talking about minor diferences, here. Grain is different, as you said even motion can be a little different. But When you print a RED on 35mm and show it, It will look darn close.
Billy Summers
01-29-2007, 12:52 PM
Wow, I think what David and Peter are saying goes hand in hand. The Film rolls through the frame being exposed by light from the bottom to the top of the frame while moving (bluring) a little. But when the CMOS/CCD records a frame the entire frame refreshes and exposes the image without going through the same subtleties that the film is subjected to. Hence the strobing, effect and differences in "visual feel" betweeen film and digital.
Very interesting...
"2cents"
donatello b
01-29-2007, 01:07 PM
"Film rolls through the frame being exposed by light from the bottom to the top of the frame while moving (bluring) a little""
there are film camera's that have horizontal shutters ( left to right or is it right to left ) .. from my experience camera with horizontal shutters have slightly smoother pans but more strobing on tilts vs. vertical shutters have more strobing on pans and slightly smoother tilts ...
Stephen Williams
01-29-2007, 01:11 PM
"Film rolls through the frame being exposed by light from the bottom to the top of the frame while moving (bluring) a little""
there are film camera's that have horizontal shutters ( left to right or is it right to left ) .. from my experience camera with horizontal shutters have slightly smoother pans but more strobing on tilts vs. vertical shutters have more strobing on pans and slightly smoother tilts ...
Hi,
Modern Arri cameras will have horizontal shutters, so the lens mount can be recentered for S35 without hitting the mirror. The Arri II & III series have vertical shutters.
Stephen
Graeme Nattress
01-29-2007, 01:22 PM
Judder is more to do with edge sharpness (and is made worse by HD and it's awful ringing sharpness edges). Shallow DOF will also help with judder too.
http://www.bbc.co.uk/rd/pubs/whp/whp...les/WHP034.pdf
Explains a fair bit.
Graeme
Stephen Williams
01-29-2007, 01:34 PM
Judder is more to do with edge sharpness (and is made worse by HD and it's awful ringing sharpness edges). Shallow DOF will also help with judder too.
http://www.bbc.co.uk/rd/pubs/whp/whp...les/WHP034.pdf
Explains a fair bit.
Graeme
Hi Graeme,
I get page not found!
Stephen
Jeff Kilgroe
01-29-2007, 01:42 PM
Also, when special effects are being created in post, they'll spend a lot less time trying to make them blurry like 24fps film.
The blur is often your friend... It can cover lots of nasties. Not to mention that 60fps would require processing and/or rendering of 2.5 X as many frames. ;)
I'm all for 60fps and the extra clarity though... I'm anxious to work with it and see how cinematic it can be made to look. 720p60 from the HVX200 almost always looks like video no matter what you do with it.
Nick Shaw
01-29-2007, 01:42 PM
http://www.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP034.pdf
Thomas Mathai
01-29-2007, 01:58 PM
I haven't seen mention of frame rate here. I think most significantly, "temporal resolution" will be vastly improved by usage of the 60fps frame rate. Then when you are panning right or left, you'll actually be able to read words and signs, as well as see movement very clearly.
Also, when special effects are being created in post, they'll spend a lot less time trying to make them blurry like 24fps film.
I think the real unspoken potential here will be in camera HDR when you have 90 fps data rates and some sort of digital aperture. (Red two?) You could then have a 30fps clip with three exposure levels for each frame being composited in-camera.
http://en.wikipedia.org/wiki/High_dynamic_range_imaging
They may spend less time trying to make blurry images, but they will spend more time just doing even basic effects work because there's be 2.5 times more frames to work with for the same time duration.
This in turn raises the costs because no fx house wants to do more work for less, though they may be forced to.
kraemer
01-30-2007, 09:15 AM
Not to mention that 60fps would require processing and/or rendering of 2.5 X as many frames.
spend more time just doing even basic effects work because there's be 2.5 times more frames to work with for the same time duration.
Thats technically not correct. A majority of the time spent is getting an animation to look right at 24Hz and then making it look like the existing piece of film. The lower the frame rate the harder it is to integrate, not easier. A really significant amount of computing power goes into the process of taking the animation through an accumulation buffer to "filmize" it.
Faster frame rates = easier. Ask around.
Nick Shaw
01-30-2007, 09:23 AM
One of the larger rendering time hits in 3D or compositing is the calculation of motion blur. Just try rendering a simple animation in After Effects as both fields and frames(at eg 25fps). The frames version often takes longer to render, as you can get away with switching motion blur off for a lot of the layers, whereas the frames version needs motion blur on for pretty much any layer that moves.
The relationship between fps and rendering time is certainly not linear, and with careful tweaking of the render settings it is possible to make more fps render in less time.
Nick
Jeff Kilgroe
01-30-2007, 09:50 AM
Hmmm...
I suppose it depends on the level of animation, detail, complexity, etc.. Considering most of the animation work I do (which is mostly what I do these days), motion blur is a small percentage of the render calcs. It also depends on the render system being used. Many newer GI implementations can get various types of motion blur for free in the calcs simply by offsetting the framing or directional camera vectors between render iterations. -- Actually, that is specifically how many render algorithms generate blur these days and much of the motion blur or adaptive sampling is part of the anti-aliasing or iterative render process. Animations in After Effects or Motion is one thing and I can see where higher speed/fps with less blur can be beneficial. But when we're talking rendering of scenes from Mental Ray, F-Prime, Renderman, etc.. And frame render times are measured in hours, not minutes for 10s of millions of polygons with blur effects as rendered or addes as post process are only a fraction of that render time, it's a whole different ballgame. 24p with more blur to add obscurity can be a much faster workflow than trying to do 30p or 60p. Even at 60p, the renderer must still generate motion blur to match with the blur length of any combined footage. Otherwise it just looks like a video game... Oh wait, newer game engines are incorporating real-time motion blur too. Hmmm.
Nick Shaw
01-30-2007, 10:49 AM
Jeff, I'll bow to your superior knowledge on the 3D front! I put that in without really thinking about it, and 3D is not relly my area of expertise. It has however definitely been my experience that it is the case with 2D, in both After Effects and Shake. I'm still running After Effects 6.5, so can't comment on how much GPU acceleration in v7 helps, but certainly Shake (no GPU acceleration) takes a big render speed hit for each element you turn on motion blur for.
Nick
Jeff Kilgroe
01-30-2007, 11:08 AM
Sure Nick.. I agree with you on the 2D ftront. AE, Shake, Fusion, etc.. doing 2D animation work or combining pre-rendered graphical elements where blur must be added manually, is a whole different animal. And yes, that does take more time to add proper motion blur to 24p in that respect. I have to admit, when I originally posted the "rendering 2.5X as many frames" comment that I completely ignored this side of the coin.
Thomas Mathai
01-30-2007, 01:16 PM
Thats technically not correct. A majority of the time spent is getting an animation to look right at 24Hz and then making it look like the existing piece of film. The lower the frame rate the harder it is to integrate, not easier. A really significant amount of computing power goes into the process of taking the animation through an accumulation buffer to "filmize" it.
Faster frame rates = easier. Ask around.
A good amount of bread and butter fx work is fix work, not animation, so paint and roto will take longer for sure.
A lot of animation is rendered in layers, to be comped together later, so motion blur can be finessed on the 2d side. It would still be 2.5x times more data for the same time duration.
By the way I did ask a 3d artist who does a ton of film work and she has never had an issue with creating animation at 24fps, and have it integrate back into the film plates.
Making an animation look good has more to do with the skills of the animator not the frame rate or software.
Software doesn't care what the frame rate is, so I don't see how more computing power gets used up for a lower frame rate.
Since pretty much a ton of great animation has been done at 24fps, I'm not sure it's such a big deal.
There has been experimental formats like ShowScan that use 60fps and has been used mostly in immersive motion simulator rides.
Jeff Kilgroe
01-30-2007, 02:14 PM
Yep, Thomas that sums it up.
You'll also find that when creating animation for broadcast, DVD, etc.. where the animation isn't being matched with video elements, many animators (myself included) will usually still work in 24p at times or will go 25p for PAL or 30p NTSC. It's typically faster to render animation data for 30 progressive frames than it is to render for 60 interlaced fields. While the render calcs themselves take up about the same amount of time for one full res progressive frame as they do for two half-res fields, when considering moving elements and 3D geometry, the computer must re-calculate mesh data and vertex positions for each field just as it would for indidvidual frames.
Also like you said, doing fix work or rotoscope operations, the fewer the frames to work with the better. But maybe not, I've been known to charge by the hour on occasion and I kinda like that sometimes. :D