new digital camera better than 35mm film?!

[Nicholas Kovats]

Great post once again, Carl. Are you aware what Aaton was attempting with their prototype Digital Super 35mm sensor in their Penelope Delta before they went belly up? The sensor was attached to a moving assembly whereby a half pixel offset was introduced PER frame! You are right on regarding a major "non-linear" rethink is required of conservative grid sensor manufacturers. Source = http://blog.abelcine.com/2012/09/27/aat ... pe-camera/

[Nicholas Kovats]

FiIm is THE reference model.

[PyrodsTechnology]

couldn't it be got in post production ? I mean, let us imagine a full pixel frame capture. Then in post we delete randomly some pixels per frame. Then when in motion we will get an average of random pixels like real motion film picture does (correct me if i am wrong)

[carllooper]

couldn't it be got in post production ? I mean, let us imagine a full pixel frame capture. Then in post we delete randomly some pixels per frame. Then when in motion we will get an average of random pixels like real motion film picture does (correct me if i am wrong)

Any processing done after capture, such as adding noise (dithering), can help transfer a signal to a lower bandwidth channel and avoid issues such as contouring or any other aliasing that might otherwise occur during such resampling. But it won't, in any way, fix any aliasing issues that occured in the original capture (in the original sampling). That aliasing will propogate no matter how much noise you add. Because you can't in any simple way, remove it. It's baked in. And that's the problem being addressed: the aliasing that occurs during capture. To solve it means intervening prior to capture. Perturb the signal before its quantised by the sensor. Or perturb the sensor, ie. in the way that film does. Same result.

Basically you want to neutralise any systematic correlations between signal and response. Noise does that. But it can't neutralise such if it's already baked into the signal. At least not in any simple way. There may be some complex way of doing so, using machine vision/intelligence. But the better solution is to fix it up front. There's an adage which goes: don't leave until tomorrow what can be done today. Another one is: rather than leave something to be done in post, do it now.

Light itself is dithered! Strangely. It's a bit of a mystery in quantum mechanics. I imagine in the future there will be some interesting sorts of digital sensor technology that emerges when the scale at which QM becomes significant, is encountered. The dividing line between sensor and sensed becomes a lot more 'entangled'. Or rather, the assumptions which produce the notion of a diving line in the first place, become no longer sustainable. Alternative models are required.

C

[carllooper]

Great post once again, Carl. Are you aware what Aaton was attempting with their prototype Digital Super 35mm sensor in their Penelope Delta before they went belly up? The sensor was attached to a moving assembly whereby a half pixel offset was introduced PER frame! You are right on regarding a major "non-linear" rethink is required of conservative grid sensor manufacturers. Source = http://blog.abelcine.com/2012/09/27/aat ... pe-camera/

Ah. Yes you mentioned that before. For some reason I thought it was Arri that developed that, but it was Aaton. Ok.

In any case a very good idea, but obviously the economic climate for such ideas is not quite ready for it.

Just like 4K capture of Super8 was a few years ago. But now it's becoming acceptable.

C

[Nicholas Kovats]

Apparently it was delays on the part of the off shore sensor manufacture which killed the Aaton project and subsequently the company went bankrupt for the 2nd time in it's history. Apparently sensor delivery and market prices are an issue for some camera developers.

[Nicholas Kovats]

Aaton was not Arri nor Canon but a meticulous handcrafted "boutique" camera shop. Such is the price of "artful" engineering.

[wahiba]

First time I have read an explanation of the difference between digital and film light gathering. I have not appreciated before the obvious fact that the digital sensors are always in the same position while the chemical sensors are randomly distributed.

To emulate film then the sensors on a digital camera need to reposition randomly for each frame. Not sure how it would be done but it makes sense.

Whatever else one says about digital projection it does not seem to get hairs in the gate.This was Interstellar 0n 70mm!! 2001 digitally projected was superb. Managed to see them within a couple of weeks of each other. Kubrick won.

[Nicholas Kovats]

I recently saw in the last 2 months both Interstellar (15 perf 70mm film print - 92 ft wide) and 2001 (5 perf 70mm film print - 40 ft wide). Film won. Interstellar did happen to have a few frames with random back projected dust particles. However, as a perfectionist this did not distract me too much and both films nailed me to the wall. Emotionally, visually and intellectually. 15 perf IMAX is a wonderfully immersive and excellent medium. I could not take in the complete vertical space with my eyes. No sense of top or bottom edge of screen.

Now back to my super 8 camera systems.... The micro formats are just as fascinating. Doug Trumbull is on to something with his huge inflatable micro cinemas that seat a maximum 12-25 people. Huge semi wrap around curved screens. I see a place for our beloved tiny Super 8 format.

[MovieStuff]

couldn't it be got in post production ? I mean, let us imagine a full pixel frame capture. Then in post we delete randomly some pixels per frame. Then when in motion we will get an average of random pixels like real motion film picture does (correct me if i am wrong)

Any processing done after capture, such as adding noise (dithering), can help transfer a signal to a lower bandwidth channel and avoid issues such as contouring or any other aliasing that might otherwise occur during such resampling. But it won't, in any way, fix any aliasing issues that occured in the original capture (in the original sampling).

That is true IF you have aliasing in the original image which, of course, you would notice while you were shooting and could attenuate the problem to avoid aliasing. So while technically true now, it's hardly an issue that holds up production and I think the issue of aliasing is going to be a total non issue as we move into 4k and beyond. At some point, the pixel count is so dense that, for all practical purposes, you have nothing but pure image and aliasing and moire patterns would be a very rare exception. But the point that Pyrods makes is spot on and something I have always felt would eventually be true. When you get to a point where it takes clusters of pixels to make up a single grain of film, then obviously all you need to do to emulate the random grain pattern of film is to break up the fixed pixel pattern into constantly changing, random clusters of pixels, which would be super easy in post.

Roger

[carllooper]

That is true IF you have aliasing in the original image which, of course, you would notice while you were shooting and could attenuate the problem to avoid aliasing.

Well that's what is being discussed: aliasing in the original image. And yes, the solution is to deal with it during the original sampling, rather than assuming it can be solved in post.

So while technically true now, it's hardly an issue that holds up production and I think the issue of aliasing is going to be a total non issue as we move into 4k and beyond.

It's always been true that it won't hold up production. After all there's plenty of aliasing going on in the early days of television which didn't stop any production. The show went on regardless. But that also didn't stop someone suggesting that the news presenter not wear a finely striped shirt. And the articulation of this advice didn't imply the nightly news not otherwise go to air.

At some point, the pixel count is so dense that, for all practical purposes, you have nothing but pure image and aliasing and moire patterns would be a very rare exception.

Yes, that helps. However the proposition is that it's the not-so-obvious-aliasing responsible for the difference in look between film and digital. We're talking about how digital might be re-engineered to look like film, and if the proposition holds, then it could address this not-so-obvious-aliasing in the ways suggested or by some other means.

But the point that Pyrods makes is spot on and something I have always felt would eventually be true. When you get to a point where it takes clusters of pixels to make up a single grain of film, then obviously all you need to do to emulate the random grain pattern of film is to break up the fixed pixel pattern into constantly changing, random clusters of pixels, which would be super easy in post.

It only helps maintain signals being transferred to a lower bandwidth (eg. from 13 bits to 10 bits). Otherwise you can just leave it as is. There's nothing to be gained by adding noise. It doesn't help resolve any difference, be it obvious as a moire pattern, or otherwise felt as a difference in "warmth", already baked into the signal.

C

[MovieStuff]

There's nothing to be gained by adding noise. It doesn't help any difference (be it obvious as moire, or otherwise felt as a difference in "warmth") which is already baked into the signal.

C

But, if the pixel count is high enough then there is no moire and there is no aliasing that needs to be addressed. So the point of creating pixels in groups the size of grain that are randomly placed (i.e. "digital grain") isn't to avoid aliasing or moire patterns but, rather, to simply emulate the look of film grain as it dances across our retinas from frame to frame. This is a solution to an aesthetic problem, not a technical problem.

Roger

[carllooper]

But, if the pixel count is high enough then there is no moire and there is no aliasing that needs to be addressed. So the point of creating pixels in groups the size of grain that are randomly placed (i.e. "digital grain") isn't to avoid aliasing or moire patterns but, rather, to simply emulate the look of film grain as it dances across our retinas from frame to frame. This is a solution to an aesthetic problem, not a technical problem.

That's not correct. Certainly many believe this is the only purpose of dithering (otherwise called 'fake grain') - which is to emulate grain effects in film. But one needs to ask why the effect doesn't work in terms of this. Why it still doesn't look like film. Certainly if one likes the look of fake grain then sure, there's no reason not to use it. But if one is interested in why it doesn't actually look like film, then a different answer emerges.

Adding noise doesn't alter the signal. You can still see the original signal through the grain. If the original signal is one captured by conventional digital means then that is the signal you will see through the grain. The proposition is that there is a an observable difference between a film signal and a conventional digital signal, sometimes expressed in terms such as "warmth". And that the reason for this difference has nothing to do with "dancing grain effects", for the difference in warmth would still be there even if you degrained the film signal.

C

[MovieStuff]

But, if the pixel count is high enough then there is no moire and there is no aliasing that needs to be addressed. So the point of creating pixels in groups the size of grain that are randomly placed (i.e. "digital grain") isn't to avoid aliasing or moire patterns but, rather, to simply emulate the look of film grain as it dances across our retinas from frame to frame. This is a solution to an aesthetic problem, not a technical problem.

That's not correct.

Actually, it is.

Roger

[carllooper]

Actually, it is.

And what is your argument in defense of this proposition?

Or are we simply to have faith.

C