28k...

[Tanner Stauss]

I am just curious what some of the practical applications would be for a 28K sensor. I can think of a couple uses, but I want to hear what's really in store. I think that there could be a significant rise in opportunities for cinematography and visual story telling in general as a result of such a large amount of detail and output format.

Thoughts?

[Dan Hudgins]

I am just curious what some of the practical applications would be for a 28K sensor.

How much of the frame will be 28K at the compression ratio is one issue, but a bigger one is can any lens at more than one f/ stop resolve 28K at 80% MTF?

If no lens can resolve 28K at 80% MTF, or just at f/4 or something like that, then the extra pixels may reduce aliasing and chroma moire, but they will not make much of the out of focus parts of the frame "sharper" than they would be shot at 5K or something...

Once you get a so called 28K image, how can you project it on a screen at 80% MTF so people can see what detail their might be, it is very hard to make a f/2 projection lens that can resolve that level, and any thermal shift in the projection lens's focus would reduce the resolution down from its peak. Movie 35mm projectors cannot even hold 2K resolution at 80% MTF much of the time, so building a true 28K projector at 80% MTF is maybe not something we will see in the next 10 years (if ever)?

The best use of a 28K sensor would be to down sample (using 100% of the sensor data) to 4K or 2K to get lower noise in low light and such. When you downsample you can interpolate extra brightness data bits, so you can reduce tone banding and such, in this case it depends on what artifacts the compression has added in the first place...

[Wesley Scoggins]

Things I have heard are... projection on massive screens (like in Time Square, or an HD Jumbotron), having a bigger image to work with so you could easily filmout to whatever (IMAX, whatever, doesn't matter), having a massive area to put two lenses side by side, so you can do 3D on the same sensor, and what Dan said.

[michael zaletel]

One important use for a 28K sensor will be security surveillance in crowds, at airports, casinos, city streets for facial recognition and crime scene forensics. You know how we all laugh when they zoom in on an SD video frame of a wide crowd shot to determine that the criminal had two gold teeth and the victim was holding a credit card that expired yesterday? Well, one day they'll be able to zoom in on the tiniest details even from a wide crowd shot. Storage and processing power will probably catch up by the time this is a reality. I also believe higher pixel counts on moving imagery will open the doors to improved VR. Using something like the VR-360 one-shot mirror, you could record more than the equivalent of (8) 4K quadrants surrounding the camera/mirror. Eventually we'll have a projector (or series of them) capable of playing that back in a spherical theatre room.

And finally, If I can ever figure out how to transmit visual data straight to the optic nerve, I'll probably need all the resolution I can get to keep your attention. :)

-michael zaletel
(shooter)

[Nir Shelter]

For generating true 3D environments, higher DR, lower noise and aliasing.

[Gavin Greenwalt]

Ultra Wide Angle panoramics.

Put three of those suckers back to back with fish eyes and stitch them together.

360 background plates.

[J Davis]

Shooting 3D
With two lenses, one sensor.

[J. Eric Camp]

VFX. Plain and simple

[Tanner Stauss]

Makes sense to me. I do like the idea of 3D on one sensor.

[Chad Lancaster]

75 inch fine art prints

I can't wait!

[Deanan]

Where's Waldo, the movie.


:)

[Enrique Blein]

Photograph a stamp from a military satellite

[Chris Mierzwinski]

The 3d possibility sounds intriguing but unless the lenses physically move to be able to converge at different points, it might not work that way on the single chip. Again unless red has something unique at it's sleeve to make 3d work on the static sensor.

[Dan Hudgins]

Again unless red has something unique at it's sleeve to make 3d work on the static sensor.

With so many pixels you can crop and shift the images to get convergence adjustments without a mechanical adjustment to the lens pair.

14K for each 3D image would be good, 3D needs more fine detail than 2D since the spatial position of subjects comes from very small relationships in the stereo pair, otherwise you get "cardboarding".

Compression does more damage to 3D stereo images than in 2D viewing, so using 28K the compression artifacts may be smaller on the end result and have less impact on the stereo viewing. Even if you down size to feed uncompressed 4K signals taken from the 28K compressed data the end images from the two 4K projectors may look better than showing a pair of compressed 4K signals...

[Tanner Stauss]

Where's Waldo, the movie.


:)

wow, knowing the crazy goings ons at RED, I don't doubt this was the original reason for it's creation. :-)

[Pawel Achtel]

Dan has a point, there is no lens in the world that can possibly resolve 28k or anything near that even with 5% MTF. I don't think it is possible at all.

A hypothetical lens, given the CoC of the sensor, would need to operate at relative apertures of about f/4 or wider to avoid diffraction limit. At the sensor size, the physical aperture would be huge - a lot of fine glass we are talking about! I can't see how it can be much cheaper or smaller than the Hubble telescope.

3D? Well, why not place two sensors close together, it would be MUCH cheaper due to (much) better yield from the wafer. And, it is not the sensor distance between left and right that is a problem, it is the lens size.

Sorry, 15 posts and not a single usage that would make sense to me :confused1:

[Roberto Lequeux]

Sorry, 15 posts and not a single usage that would make sense to me :confused1:

Bragging rights? :sifone:

[Stefan Christou]

At the sensor size, the physical aperture would be huge - a lot of fine glass we are talking about! I can't see how it can be much cheaper or smaller than the Hubble telescope.

Well perhaps Nikon can use some of their photolithography glass. :coolgleam:

[J. Bernard Vallon]

Tracking something that cant reasonably be tracked by panning and tilting. Shoot wide, crop in, no quality loss.

As far as lenses, we'll all be buying old school large format lenses, and living at f8. This is a daytime camera.

[Chris Newman]

Dan has a point, there is no lens in the world that can possibly resolve 28k or anything near that even with 5% MTF. I don't think it is possible at all.

A hypothetical lens, given the CoC of the sensor, would need to operate at relative apertures of about f/4 or wider to avoid diffraction limit.
Actually, the pixel size is the same as the FF35 and 645 Monstro sensors, so the diffraction situation is no worse than with those, so the sharpest images would likely be around f8 or so.

For example, large format photographers manage to get sharp images that utilize most of the resolution of the film.

[Joseph Ward]

Because they didn't want to go further in Ks yet.

[Robert Kurhajetz]

the digital version of what these folks are doing on emulsion....

http://www.gigapxl.org/gallery.htm

[Justin Anderson]

Actually, the pixel size is the same as the FF35 and 645 Monstro sensors, so the diffraction situation is no worse than with those, so the sharpest images would likely be around f8 or so.

For example, large format photographers manage to get sharp images that utilize most of the resolution of the film.

How do you calculate where the sharpest range on a lens is, btw? Why isn't a lens sharpest at "open"? Anyone have a good resource I could read on this?

...don't mean to derail the thread.

EDIT:
Neeeevermind...found it.
http://www.luminous-landscape.com/tutorials/understanding-series/u-diffraction.shtml

[Dan Hudgins]

Why isn't a lens sharpest at "open"?

There are mostly two kinds of lenses, Anastigmat and Petzval.

In a Petzval you can get sagittal and tangential to align but only on a curved focal plane, and since the sensor is flat you cannot photograph things at "equal" distance.

To shoot on flat film or sensors you mostly use an Anastigmat, but its focus is split and only aligns in one circle near the edge of the frame (and the center), because there is ring of soft focus in all Anastigmat lenses, you need to stop down one to three stops to reach the defraction limit in all parts of the frame. If the maximum stop is too small you cannot stop down enough to keep the resolution at the single pixel size since stopping down makes the circle of confusion larger.

All lenses have aberations and do not focus in small points with all rays going to the same place, the smaller the f/ stop the better the aberration issue since there are fewer ray paths to align. But when you stop down defraction makes the rays bend to hit a larger area, so the two opposite things happen, open up it gets soft, or stop down it gets soft.

With small sensors the defraction limit becomes more important at smaller f/ stops so in 8mm you should stay more open than f/5.6, for 16mm f/8, for 35mm f/11 and so on, for contact prints you can stop down to f/64.

As far as lens design goes, it can be easer to balance the issues in lenses of longer focal lengths that work on larger formats to get a better "total" pixel count in the image, so a large sensor can make the lens easer to design for super-resolution images.

The vague area about what f/ stop is best comes in part from the red vs. violet wave lengths giving different resolutions at different f/stops, so when red starts to blur larger than a sensor pixel, the violet light still has enough resolution, this gives about a two or three stop plateau to the lens resolution where details can be made out at about the same size.

Here is a link that talks about astigmatism:

http://toothwalker.org/optics/astigmatism.html

[Tanner Stauss]

There are mostly two kinds of lenses, Anastigmat and Petzval.

In a Petzval you can get sagittal and tangential to align but only on a curved focal plane, and since the sensor is flat you cannot photograph things at "equal" distance.

To shoot on flat film or sensors you mostly use an Anastigmat, but its focus is split and only aligns in one circle near the edge of the frame (and the center), because there is ring of soft focus in all Anastigmat lenses, you need to stop down one to three stops to reach the defraction limit in all parts of the frame. If the maximum stop is too small you cannot stop down enough to keep the resolution at the single pixel size since stopping down makes the circle of confusion larger.

All lenses have aberations and do not focus in small points with all rays going to the same place, the smaller the f/ stop the better the aberration issue since there are fewer ray paths to align. But when you stop down defraction makes the rays bend to hit a larger area, so the two opposite things happen, open up it gets soft, or stop down it gets soft.

With small sensors the defraction limit becomes more important at smaller f/ stops so in 8mm you should stay more open than f/5.6, for 16mm f/8, for 35mm f/11 and so on, for contact prints you can stop down to f/64.

As far as lens design goes, it can be easer to balance the issues in lenses of longer focal lengths that work on larger formats to get a better "total" pixel count in the image, so a large sensor can make the lens easer to design for super-resolution images.

The vague area about what f/ stop is best comes in part from the red vs. violet wave lengths giving different resolutions at different f/stops, so when red starts to blur larger than a sensor pixel, the violet light still has enough resolution, this gives about a two or three stop plateau to the lens resolution where details can be made out at about the same size.

Here is a link that talks about astigmatism:

http://toothwalker.org/optics/astigmatism.html

:blink: ...got it

[Pawel Achtel]

Good explanation, Dan. However, I would like to clarify/disagree on one point you've made...


With small sensors the defraction limit becomes more important at smaller f/ stops so in 8mm you should stay more open than f/5.6, for 16mm f/8, for 35mm f/11 and so on, for contact prints you can stop down to f/64.


The diffraction limit depends on CoC, which in the case of the 28k sensor, is still around 6 microns.

The diffraction limit also depends on physical aperture, but not relative aperture.

Two sensors of equal pixel count that are different physical size would still peak at around the same relative aperture which, at 4k resolution is around f/4 - f/5. If resolution of the sensor is increased, the relative aperture of the lens must increase to take advantage of it.

Clear as mud... :)

[Roberto Lequeux]

Film a city from a hot air balloon... figure out what your movie is about in the cutting room. :thumbsup:

[Petr Dvorak]

just a few:

Military - one quick shot and you get tons of informations and details for later analysis / survey on aircracft cariers, airbases atc. with loop recorders / situation on battlefields
Forensic, FBI, Security - precious details in combination with laser scanners for precious reconstruction of criminal sites after years ago when some omited details can be revealed
Archeology - detailed and complex pictures of sites for later analysis
NASA - STS, ISS at orbit / with HD they started to measure objects outside station - very helpfull / Analyzing of hires footage if something happens during start of STS (measuring of debris)
Archives, Museum, Galleries - detailed images and footage of valuable artpieces
Science, research, sport, renovation, conservation, geology, agriculture, ecology, civil engeneering, harbors, logistic, espionage ... etc.

But it will be mostly limited by available lenses.

[Dan Hudgins]

The diffraction limit also depends on physical aperture, but not relative aperture.

Diffraction depends on the angle of the marginal rays, wide NA microscope lenses have a small physical aperture with small focal length but have a higher resolution per mm than lenses with much larger physical apertures but f/ ratios that give higher numbers.

So I cannot figure out what you imply?

The f/ number tells you the minimum CoC (in the absence of aberrations). It does not matter if the front element is 1mm wide or 1000mm wide.

The photos in this link show higher resolution and less diffraction in the images with the lower f/ stop i.e. with Oil to widen the f/ stop (even though the physical aperture is very small),

http://www.microscope-microscope.org/advanced/numerical-aperture.htm

==

Its easer to make a super-resolution lens for larger formats since the pixels would be larger in a larger format so the best working f/ number for that larger size sensor could be smaller. Lenses with increasing f/ numbers are easer to correct for aberrations caused by glass. Hence the APO lenses used in copy work that shoot at stops like f/8 but give a pixel count over their working area 16"x20" of more than 28K.

==

Smaller sensors require wider f/ stops. But correction of lens aberations is in ratio to the focal length, so at some point as you make the sensor smaller you cannot resolve what you need to since you cannot make a lens faster than f/0.5. For the most part lenses faster than f/2.3 cannot resolve better than 1/2000 their focal length, and lenses faster than f/1.2 better than 1/750 (at 80% MTF). So there is a minimum sensor size you can use to make high resolution images. At the large end you are better off except that the subject distance gets closer to the focal length which reduces the DOF because of bellows extension, so bigger than 8x10" sensor also has issues.

[Daniel Browning]

I am just curious what some of the practical applications would be for a 28K sensor.

It's not just the resolution, but the sensor size. RED ONE has 11.3 stops of dynamic range. Let's say the Monstro in 617 has two additional stops at 28K. When scaled for spatial frequency (e.g. downsampled), the dynamic range increases another *three* stops. That's 15.3 stops of dynamic range, and everyone one of those stops has much less photon shot noise, due to the much higher quantity of light. It's the same reason that DSLR have more dynamic range than digicams.

Size matters.


Dan has a point, there is no lens in the world that can possibly resolve 28k or anything near that even with 5% MTF. I don't think it is possible at all.

Remember that it's much, much easier to build a 6x17 lens that resolves 28K than a 28K Super 35 lens. Google turns up 6x17 film scans at 13.7K (http://www.stockholmviews.com/canon_8600f_review/8600fpage5.html) that have good MTF.



At the sensor size, the physical aperture would be huge - a lot of fine glass we are talking about! I can't see how it can be much cheaper or smaller than the Hubble telescope.


Modern Schneider 6x17 lenses seem to be around 5,000 USD.



Two sensors of equal pixel count that are different physical size would still peak at around the same relative aperture which, at 4k resolution is around f/4 - f/5.


I think there is a mistake in that statement. A 28K 6x17 shown at 4K will not have any visible diffraction even if shot at f/16. Whereas a Super 35 sensor shown at 4K and shot at f/16 will show visible diffraction.



A hypothetical lens, given the CoC of the sensor, would need to operate at relative apertures of about f/4 or wider to avoid diffraction limit.


I'm just going to say that I agree to disagree with you here. We've had this discussion before (http://reduser.net/forum/showthread.php?t=20711), so there's no sense in doing it again. :)



The diffraction limit depends on CoC, which in the case of the 28k sensor, is still around 6 microns.


Agreed. Another factor is display size. If the 28K is shown on a 4K display, then that will make the acceptably sharp CoC much larger.

[Roberto Lequeux]

Surveillance for the Red store front? :)

[Pawel Achtel]

I'm just going to say that I agree to disagree with you here. We've had this discussion before (http://reduser.net/forum/showthread.php?t=20711), so there's no sense in doing it again. :)


Thanks, Daniel. I was looking for that thread. It was a good discussion.

After chewing over what you said, I think I agree to agree with you :)

[Tanner Stauss]

This has been a great discussion,. I have learned much, thank you all. But to be honest I am a little overloaded by some of the terms and heavy diffraction, micron, spatial frequency stuff. What is MTF? and CoC? Is there any way in hell of a very brief, simplified version of all that? Thanks again.

[Petr Dvorak]

CoC - Circle of confusion
http://en.wikipedia.org/wiki/Circle_of_confusion

[Chris Newman]

Circle of Confusion is the size of the circle made on the sensor by a point of light being photographed. If the image is out of focus, the point makes a larger circle as seen here:
http://farm4.static.flickr.com/3575/3293740317_6e71e177e2.jpg

Even if it is in focus, points of light still make a small circle that is determined by diffraction, aberration, etc.

Wikipedia gives more details:
http://en.wikipedia.org/wiki/Circle_of_confusion


MTF is a measure of the contrast of a lens when photographing a given spatial frequency (black/white line pairs per mm). MTF charts are a way to show how sharp and/or contrasty a lens is.

Chris

[Darren Orange]

I think it will reduce camera movement. Setup three cameras and record a scene, crop out the shot you want and there you go. Focus might be the only thing to be adjusted rather than whole camera movement. Fight sequences will be fantastic and ability to do easier Green screen keying. I agree that VFX will drool over it to work in, just not want to deliver it lol. :head_explode:

[Dan Hudgins]

What is MTF? and CoC?

Another way of thinking about MTF is that the CoC is not just a nice spot, it can be a brightish point with dimmer blobs of light reaching out maybe 10 to 20 times as far, usualy away from the center of the frame.

The parts of the CoC that shoot out are what people talk about as "contrast" since they pile up and rase the overall light on the sensor like lens flare, but only near image features.

So with that in mind, put a circle of given size over the irregular blob of light that the CoC is, some parts of the actual CoC will be inside the circle of given size and some parts will be outside, you can then count the rays that are inside vs. the rays that are outside and get a % value.

The 100% of the rays inside the circle of given size would require a very large circle of given size but would read a value of 100% MTF, for a 35mm movie camera lens that might be a video resolution of 250 lines per image height in the center of the frame and 100 lines per image height in the corners. Because the 100% MTF rating gives very low resolution values lens manfactures usualy rate lenses at maybe 35% MTF meaning that only 35% of the light from any given point falls withing the circle that defines the resolution stated and 65% is falling all over the place fogging the details around that point.

Even though MTF can fall to small % of the light, if the CoC has a bright point the images when viewed can show "low contrast" details in the image, so some lenses can have high resolution ratings but still look soft because only a small amount of the light is within the stated circle of given size. The best lenses will be able to get more light into smaller circles of given size, so that will have both high contrast and high resolution, and therefore higher MTF at higher frequencies.

In the case of the about 3.2K or so rating for the RED ONE, the MTF % was not stated clearly but was taken to be "just visable" or maybe 5% MTF, to get a 100% MTF resolution for the RED ONE you might only read 1K given the OLPF and lens effects at wide stops etc.

So same with the so called 28K camera, the OLPF will blur the image so each CoC spreads over maybe 9 pixels area making the 100% MTF value maybe 8K or so.

If you sharpen digital images you can increase the MTF % values in high frequencies, but you add noise, aliasing, "video look", chroma moire, and fixed pattern noise. And in the case of compressed images like REDCODE and H.264 you increase wavelet and blocking related artifacts when you sharpen the images. That said, you can even out the MTF of good and bad lenses quite a bit with some sharpen of the images to the point that the bad lens may look sharper than the good one depending on the settings.

[Tanner Stauss]

I think it will reduce camera movement. Setup three cameras and record a scene, crop out the shot you want and there you go. Focus might be the only thing to be adjusted rather than whole camera movement.

See, I guess a certain style could be achieved by doing this, but then you really don't have the aesthetic of a moving camera. The changing perspectives of a dolly shot, or the life like feel of a hand-held shot can never be achieved by cropping out of a locked off shot.

Agreed, this could be a great tool in some situations, but you really lose some dynamic in that sense. It sounds more like a back-up.

[Tanner Stauss]

Dan and everybody, this is all certainly a mouthful, to say the least. I will have to read these over a few times and let it marinate. I really appreciate the time you have taken to explain this stuff to me. Trust me, I will do my best to try and understand it all. Please continue to add any other ideas or info that you may have. Thanks again.

[Cory Mitchell]

Red isn't the only one making 6x17 sized cameras (or digital backs). http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f934.cfm, and they use theirs for landscape shots (at least that's what they're showing). And since Red's 617 has more resolution, those landscapes could (potentially) be all the more sharper.

Also, if Adobe could adapt this lens for the 617, it would be awesome (since it has so much real estate):
http://www.wired.com/gadgetlab/2007/10/adobe-shows-off/
(Check out the video link).

I, personally, would like to use it for panoramic landscapes and VFX, and if they had an Adobe lens, then I may not even have to use a green screen (assuming software could auto-roto out the background).

[Petr Dvorak]

Red isn't the only one making 6x17 sized cameras (or digital backs). http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f934.cfm, and they use theirs for landscape shots (at least that's what they're showing). And since Red's 617 has more resolution, those landscapes could (potentially) be all the more sharper.

...


I think that Seitz is only scan back camera (just portable scanner with lens), with linear sensor which has only 7500 pixels and you have to wait 1 secon to cover whole picture. Red 28K has one huge sensor with matrix of zillions pixels. That's It.

[Cory Mitchell]

Yeah, the Red 617 has a 'few' advantages (like 30fps, 276+ MP (correct me if I'm wrong), etc...), but either way, the Seitz 617 pictures are pretty amazing.

[Casey Green]

The 28K RED will also be great for use as a still Camera for huge print ads (think sides of buildings).

Also, it will be great for generating extremely high-res Translight prints for use as TV and Film backdrops.

With HD viewing now common place, how many times have you watched a show and caught yourself thinking, "Wow, that background looks fake".

28K could help suspend disbelief. :-)

[Tanner Stauss]

Red isn't the only one making 6x17 sized cameras (or digital backs). http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f934.cfm, and they use theirs for landscape shots (at least that's what they're showing). And since Red's 617 has more resolution, those landscapes could (potentially) be all the more sharper.

Also, if Adobe could adapt this lens for the 617, it would be awesome (since it has so much real estate):
http://www.wired.com/gadgetlab/2007/10/adobe-shows-off/
(Check out the video link).

I, personally, would like to use it for panoramic landscapes and VFX, and if they had an Adobe lens, then I may not even have to use a green screen (assuming software could auto-roto out the background).


That's some pretty interesting stuff. It's amazing that even a scanner back 617 camera like that can get it's image from such a small lens. I guess it's just the design of the mechanics and glass that makes that possible, yeah? What would some of the other lenses for a 28K/617 camera be like/look like. Zoom lenses?