Bayer Pattern

[Jannard]

One of "the BIG Studios" spent a lot of time and energy testing film and film scans to really see what was up (so as to see what was the best quality and not to waste money). They wanted to know if there was and advantage to scanning S35 film to 4K or if 2K "was good enough". The results were as I already stated. And it was certainly better to scan film at 4K, just as a 4K Bayer pattern camera has much more resolution than a 1080P camera... with or without 3 sensors. This is because the "native resolution" of film is greater than 2K... and less than 4K. You always put your letters and junk in a drawer large enough to hold everything. But putting the same letters and junk in a much bigger drawer doesn't mean you have more stuff.

Jim

 

[Dan Hudgins]

What you see...

What you see...

Our 35mm movie film scanner captures enough resolution to see that the gunk on the film is sharper than the film grains.

The size of the grains image dye clouds is about 3 to 5 pixels at near 4K pixel resolution, so 35mm movie film at 80%MTF should be concidered about 720 video lines, I get that this way,

(3800*0.75)/4=712.5

3800 is the pixels per image width as scaned
0.75 is the inverse of the 4:3 ratio
4 is the pixel width of the grain

The value of 1280x720 is what projection tests show is what you can see off a typical 35mm movie print.

Moving the film tends to look twice as sharp since your eye connects frames together and sees details that are made up of grains from more than one frame, but with the projector out of focus as it is most of the time now you get half or less so you end up back at about 1280x720.

Its good to scan film at 4K to reduce aliasing and to increase the tonal values from sample averaging.

It is possable to shoot fine grain film like pan-sound recording film with special processing to get near 10x higher resolution that color film, and do that in the Technicolor camera, but you cannot project the images in a way that would let you see the extra detail from film.

We shot some sound recording film and the negatives seemed to have better than IMAX detail on 35mm, but when projected in a profesional screening room it was imposable to hold the focus for more than a few seconds, as the lens was focused you could see there was detail on the print, but the lens and projector could not hold the focus close enough to be able to view it for any length of time.

It would be nice to make a larger sensor three chip camera, you do not need a prism behind the lens like 2/3" cameras use it could be made using plate beam slpitters and relay optics. It would be as large as the old TV cameras, or the old Technicolor camera, and for the most part it does not matter since you can now do "keys" on any color in the frame to replace the color shot with any other color, and so the concept of "shooting" images is passe.

Even if you could build a 3 chip camera using large sensors, the noise levels would require Bayer like chroma filtering. Even film scans get processed with heavy software filters now, there is no such thing as a RGB image in the way old "one shot" 4x5 color cameras made today, its all processed and re-worked to the point that it is in some way computer graphics rather than photography.

For the most part people will like a software processed image BETTER than one that is direct off the film or sensor, raw film scans have all kinds of grain and dust problems, and raw sensor data has noise issues that need to be dealt with.

Like it or not the days of film cameras are ending for the most part. All anyone can do now is to remember what was good about film and try to make Digital better in every way that can...

(It would seem that higher numbers for film resolution people get are from the way that the scanner uses a tall narrow slit to average grains, see Kodaks RMS granularity instrument and the readings are at very low MTF%, and taken off the negative with a microscope like device, or that the scanner has had some video or digital sharpen used. 80%MTF off a 4th generation print is the useful value and taken off the screen in a normal theatre, because that is what you see and remember as movie film.)

Try looking at a 2K single pixel checker board pattern on your 2K monitor, then make a filmout of that uncompressed file and project the print (a print off the film recorder negative is like an answer print, but normal prints are made off a dupe negative made off a master positive) and tell me you can see the checker board at 80%MTF, you will see a gray screen for the most part. Then try a twoxtwo pixel checker board to see if you can see 1K, which you might in the front row. Then try a fourxfour pixel checker board and you will begin to see the black and white spots at 512 per screen height. Now make the checker board yellow and white and see what you can see, not too much because the yellow layer in film is the most soft, maybe you can see an 8x8 pixel checker board in the yellow and white, that 64 pixels. Try sitting at various distance from the screen and see which of the grids you can see as dots. Notice that the focus in the projector may make the dots not visable over the whole screen at the same time and that the focus weaves in and out. Try projecting it with both scope and flat lenses to see how that changes the resolution you can see on a screen from a normal distance.

 

[Steven Caesare]

I belive one of the tradeoffs that makes using three 35mm sensors (as opposed to a sinlge bayer/stripe pattern sensor), is that not only does the required prism block introduce optical/chromatic aberations, but that it's physically impossible to place the block such that you maintain the requred distance to the sensor backplane and still accomodate a 35mm lens. the required dimensions of the prism are just too big.

Thus, while you could construct a tri-35mm sensor camera, it would in effect be orphaned in the huge ecosystem of existing standard 35mm PL mount lenses and optical accessoriess...

-sc

 

[Graeme Nattress]

Good point Steven - I think you're right.

Graeme

 

[James Brundige]

There is a direct analogy here with audio, which is about ten years ahead of digital imaging, because digital audio is so much easier.

When the CD came out, it was called "cold". People were used to the low frequency rumble of LPs, and the lack of high frequency resolution felt "warm" The truth is, CD's give more high frequencies, and clearer, more accurate sound.

Film is like the LP. The inherent noise is grain, which many people like. High Rez digital has a measurably higher resolution, which I prefer. I don't like grain - it's just distortion to me. I like looking through the medium to the image itself, which is clearer without the grain. But there were about four years where some producers I worked with added a grain filter to video footage to make it look like 16mm, because it smoothed out the color transitions in betacam.

So the bottom line is individual taste in what the image looks like. The real revolution in Red, from my POV, is they have eliminated the edgy look of video. Colors now flow into one another more subtly (some call it film-like), but without the distortion of grain.

 

[James Brundige]

I've been looking around on Kodak's website a lot, and here's where my original inquiry in this post stemmed from:

I'm not a total idiot. I know that resolution is certainly not the only thing to consider when assessing the "supremacy" of an image format, but being able to extract the kind of 'raw' data that you can from film is going to subsequently effect bit-depth, color space, and the resilience of the footage as far as post-processing/vis effects are concerned, yes? Considering that no digital cinema camera can preserve the same kind of information that celluloid can, this gives it a quantifiable advantage over d-cinema... or at least a uniqueness to it (regardless of whether or not it's perceptible to the human eye).

Critical, verifiable tests are all over this forum demonstrate that digital can preserve more information than celluloid. Perhaps there is an agenda on the Kodak website. Do they include any actual resolution charts?

 

[Dan Hudgins]

Cubes

Cubes

Thus, while you could construct a tri-35mm sensor camera, it would in effect be orphaned in the huge ecosystem of existing standard 35mm PL mount lenses and optical accessoriess...

-sc

The way 2/3" cameras use the prism is the wrong way to build a 3 chip camera.

The Japanese use the prism in the non parallel beam, and that requires that the lenses be desigined to incorporate the prism into their perscription.

With relay optics like were used in the first color TV cameras you can use normal PL mount lenses in front of the system, they focus on a field lens and then the mirrors and additional relay optics are behind them.

Today you can register the images using digital processing after the three channels of true RAW data are recorded.

The reason that things that are "better" are not made or used is that there is no return on the box-office for a better looking image, that why most films are finished at 2K and 65/70mm is not used. 65/70mm only costs about 2.5x the film and lab cost to make and that is a small part of the film production and the results do LOOK better on the screen (when shot 65mm).

Why did 65mm die? It did not make enough return in the multi-plex small screens to turn an extra profit.

Kodak could make a fine grain film at EI/ISO 8 like the first Technicolor camera that had great color. But even Technicolor went to higher speed stocks for their later films and reduced the quality of the results, why, because the power bill was too high to fry the actors.

If people WANTED higher quality film images that needed 4K scans, they would demand EI/ISO 8 film stocks, but they shoot T500. They could shoot T500 in 65mm but they shoot so called "Super35" which has half the film area of Anamorphic 35mm and less resolution because it is easer for FX and shooting 3perf cheeper.

I would enjoy using a good large sensor 3 chip camera because of its more original Technicolor look of color PHOTOGRAPHY like a Carbo-print. But if people want T500 they are not going to get it with a 3 chip camera using relay optics, the T stop loss would be several stops. You might get EI/ISO 32 out of it with the sensors at minimum gain and minimum noise processing.

So you make a highend 3 chip camera that is EI/ISO 32, who is the end user, who wants to shoot like the great films of the past made in 3 strip Technicolor and Dye transfer prints?

Everyone is going in the other direction, they want smaller, faster, cheeper, just good enough to get by and hope the check clears....

 

[Jason Murphy]

Man, I would love to see the rebirth of the Tech IB process for dye transfer prints. Most beautiful color film prints anywhere; there's just nothing remotely like them nowadays; not even Vision Premier. And no digital projector that I've seen comes remotely close to either the colors or the blacks of an IB print.

As and aside, I just recently rewatched the restored version of The Red Shoes. Three strip Technicolor, shot with a camera the size of a small shed, and on film that was super slow. Still some of the most beautiful colors I've ever seen on film.

But you're right, Dan. Everything's headed in the other direction, with the Image quality being just good enough.

 

[Milan Nikolic]

People you know so much! I am learning, - thanks!

 

[Dan Hudgins]

The Garden of Allah

The Garden of Allah

I got to see a NEW IB print, fresh out of the mordant solution, the guy who has the Gateway Theatre here in San Francisco purchased one of the IB print machines, another of their machines went to China and may still be working there?

He reprinted several of the films from the original(?) matrix but The Garden of Allah was the best I think, the slower film stock was soo smooth in color tone, you can see that even in these internet stills,

http://www.widescreenmuseum.com/oldcolor/technicolor10.htm

http://www.widescreenmuseum.com/oldcolor/technicolor-gardenofallah-1.jpg

http://www.widescreenmuseum.com/oldcolor/technicolor-gardenofallah-2.jpg

When Technicolor changed to Eastman color for the negatives all that was lost forever... The green tones were outside the gambit that tri-pack print stock can produce I think, I have been putting extra dichroic + UV/IR filters in our B&H dichroic light house, plus gels Wratten 70, 98, 99 and 1B, the newer ECP print stocks can be developed for 5 minutes rather than 3.5minutes and the results are better that what you see from normal lab work maybe, but not quite IB gambit.... (you can't reporduce them on a normal monitor, maybe a OLED monitor? Maybe someday there will be 30meter wide OLED screens and no projectors with their out-of-focus lenses...)

 

[Steven Caesare]

The way 2/3" cameras use the prism is the wrong way to build a 3 chip camera.

The Japanese use the prism in the non parallel beam, and that requires that the lenses be desigined to incorporate the prism into their perscription.

..SNIP...

Well, if you are saying the way to do it requires a specifically engineered lens to support it, that's my point. You lose compatablity with standard 35mm PL mount cine optics... which is a major point of the format.

-sc

 

[Dan Hudgins]

PL

PL

You lose compatablity with standard 35mm PL mount cine optics... which is a major point of the format.

-sc

What I was saying was that there is more than one way to build a 3 chip camera and it is possable to use standard lenses like a PL mount movie lens on a 3 chip camera if it uses the right relay optics, you just have a large T stop loss but you can get imporvements in dynamic range and color quality because the chroma noise in Bayer filters is higher than using three monochrome sensors, that is do to the red and blue area being 1/4 the sensor area and the un-balance in the red, green, and blue, and the low saturation of the Bayer filter (made that way to reduce other problems, but with a 3 chip camera you can use full saturation filters).

Three and four tube color tv cameras made before the type of prism used in the current 2/3" cameras used an optical system where the light rays were parallel, and the prime lens on the front of the camera focused on a field lens and did not need the same type of correction that lenses for todays 2/3" 3 chip cameras need.

It makes for a larger camera, but the quality can be high with digital image registration and you can use any movie lens on it.

You could also use relay optics on the 2/3" camera type prism to be able to shoot with PL mount lenses, bit the f/ stop would be limited. If you use parallel rays only through cube prisms in the relay optics system, you can increase the f/ ratio to work with f/1.2 lenses if you want to spend the extra money.

There is no technical reason that one cannot build a fine 3 chip camera that uses PL and other movie optics.

The issue is that there are few people that would pay the extra money for such a camera which is why RED ONE has done well, it would cost maybe 20 times as much on the same pricing ratio, and with people getting by with what is out there, where is the market?

If you could make a lower cost 3 chip camera there might be a market if you can do head-to-head image quality tests, but even if it looks better, would the extra weight stop people from using it?

The other way to go is to do what RED is doing, increase the number of pixels so that you can average several in groups and still have more groups than the needed 1280x720 RGB resolution on the movie screen. As you make the pixels smaller the four pixel groups ack more like a foveon because you have 16 or more within the lens circle of confusion much of the time, then the Bayer artifacts become less important. That seems to be Panavisions view with their small pixel sensor, it has more pixels (photo sites) than you
need for filmmaking so you can group them then the chroma moire gets blured out.

To get 1280x720 well from a Bayer sensor you need about 3x the pixel or photo-site count, so a sensor with 3840x2160 is enough to work, which is why the RED ONE footage works. If you go to 4x or 5x then maybe with a 4K digital projector or IMAX print you can see more, otherwise you are pushing the chroma moire down even more, and that can only improve the end results.

I replied because there seemed to be the implication that there was some problem using PL and other lenses on a 3 chip camera, and there is not, such 3 tube cameras with image areas about 50mm wide have been made before and with modern CNC tools could be made well enough to get up to the needed 1280x720 alignment or better with digital registration. Some of those types of color camera use a turret to change lenses before large zoom lenses where introduced, and those prime lenses were based on 35mm movie camera lens designs.

 

[Xander Fischer]

I replied because there seemed to be the implication that there was some problem using PL and other lenses on a 3 chip camera, and there is not, such 3 tube cameras with image areas about 50mm wide have been made before and with modern CNC tools could be made well enough to get up to the needed 1280x720 alignment or better with digital registration. Some of those types of color camera use a turret to change lenses before large zoom lenses where introduced, and those prime lenses were based on 35mm movie camera lens designs.

Wait, are you saying that three sensors, 50mm each have been put into a camera, and this beast could utilize PL lenses...? By all means, point me in the direction of a site where I can see this beast, but I don't believe it. I might just be misreading what you're saying, but that sounds absurd.

 

[Jason Murphy]

I got to see a NEW IB print, fresh out of the mordant solution, the guy who has the Gateway Theatre here in San Francisco purchased one of the IB print machines, another of their machines went to China and may still be working there?

Well, color me envious. I'm pretty sure the IB print machine in China stopped printing in the early 1990's - I think its last film was Zhang Yimou's 'Ju Dou.'

 

[Dan Hudgins]

TK-41

TK-41

Wait, are you saying that three sensors, 50mm each have been put into a camera, and this beast could utilize PL lenses...? By all means, point me in the direction of a site where I can see this beast, but I don't believe it. I might just be misreading what you're saying, but that sounds absurd.

http://www.ev1.pair.com/colorTV/myTK41_1.html

PL mount lenses do not cover 50mm wide sensors, my point was that "sensors" of an area larger than is needed have been used in that sort of relay optics system, we have some of those, f/0.7 Cooke relay optics we picked up for making such a camera perhaps.

Notice the lens turret, if you get one of those type of camera you can replace the large IMAGE ORTHICON TUBE (yes we have some of those as well so I know how large the image circle is) with those 65mm size Dalsa CCD sensors (monochrome ones), and you can adjust the front relay optics to use the smaller image circle of the PL lens, or use 65mm camera lenses on it. If you use three Bayer sensors you can get HDR since the other two colors act as ND filters once you filter out all the light of the other two colors, so when you interpolate you get a strong signal from the matching colors, you use that for the shadow detail, then you use the other two colors for the highlight detail. It would be better to make chips WITH ND filters rather than the Bayer patten, but it will do to get started in making a three chip HDR camera.

To build a real 3 chip camera you would make it about half as big, and use optics optimized for larger f/ stops and higher resolutions, but what has already been done 50 years ago is close to all that is needed.

If you look at the bottom of the page at this link you can see the size ratio of the sensor end of the 2/3" tube and the size used in broadcast three and four tube cameras,

http://members.chello.nl/~h.dijkstra19/page4.html

So optical systems of larger than is needed have been made, and if you look at the tapes of some shows shot with those cameras they have a special look more like Technicolor due to the bandpass of the dichroic which is much like the dichroic mirrors used in the B&H Dichroic light house used for fine motion picture color printing.

BTW, did you know that the first electronic television image was produced in San Francisco, there is a palque on a rock in front of the building downtown marking the spot. What Zworykin had at that time was is seems a "monoscope" not a camera, although he would not say so... Farnsworth was making "images" of moving light projected on the sensor surface of his device.

 

[Steven Caesare]

What I was saying was that there is more than one way to build a 3 chip camera and it is possable to use standard lenses like a PL mount movie lens on a 3 chip camera if it uses the right relay optics, you just have a large T stop loss but you can get imporvements in dynamic range and color quality because the chroma noise in Bayer filters is higher than using three monochrome sensors, that is do to the red and blue area being 1/4 the sensor area and the un-balance in the red, green, and blue, and the low saturation of the Bayer filter (made that way to reduce other problems, but with a 3 chip camera you can use full saturation filters).

...SNIP...

I'm not an expert in "relay optics", but so far none of the folks much smarter than I have demonstrated that such a thing would work acceptably:

http://en.wikibooks.org/wiki/Movie_Making_Manual/Cinematography/Cameras_and_Formats/Genesis:

Unfortunately, most lenses designed for film cameras cannot be adapted to work on 3-chip (or 3-tube) video cameras, because of the need to allow room for a dichroic color separation prism. On a typical 2/3" 3-CCD video camera the clearance between the rear lens element and the focal plane needs to be at least 50mm (2"). Most cine lenses have far less clearance than this, as they have only ever needed to allow room for the shutter, so on a video camera, the image will want to focus somewhere in the middle of the prism block. Panavision originally tried to overcome this problem with optical adaptors that fitted between the cine lens and the video camera but these have all produced an unacceptable drop in image quality.

-sc

 

[Dan Hudgins]

More than one way,

More than one way,

I'm not an expert in "relay optics", but so far none of the folks much smarter than I have demonstrated that such a thing would work acceptably:

http://en.wikibooks.org/wiki/Movie_Making_Manual/Cinematography/Cameras_and_Formats/Genesis:


-sc

The type of prism used in newer 2/3" cameras is NOT the same as the type that can be used.

The relay optics make the rays parallel through the cube prisms or beam splitters of plate or Pellicle Beamsplitters.

Cameras have been made with those type of optics, just see the links above for 3 chip tube cameras, people have not made CMOS type cameras like that maybe because it is a new art, but based on what was done before there is no reason to think that it would not work. Talking about current 2/3" type cameras is not relevant.

==

There is also the option of making a two chip camera with one sensor all red pixels and the other a closer checker board of green and blue, that using a Pellicle Beamsplitter beam splitter can be used with PL mount movie lenses without relay optics and would me much lower cost.

I posted a drawing of that here a few years ago. Using RED for the full sensor gives advantages since the other sensor is just blue/green so it is more like the 3 strip Technicolor camera where there was just one beamsplitter and two colors were in bi-pack.

You can do it with a monochrome and Bayer sensor as well, using the Bayer sensor for chroma, but the special blue/green checkerboard with an all red sensor should produce better results, maybe.

==

What will work "acceptably" is not a objective measurement perhaps, maybe some people would like the look of a large 3 chip camera, some would like the look of a 2 chip camera, and others the look of a Bayer chip camera.

My point was there is no obstacle technically to all three using standard motion picture optics like PL or BNCR lenses. The fact that it is not on the market to rent or purchase is an economic one and perhaps the skill of the camera makers and the fact that it would be larger and weigh more, up to now there has not been much reason to build once because people were shooting movie film, when film production stops then maybe there will be large sensor 3 chip cameras, or just Bayer sensor cameras with smaller sensors and more oversample or foveon may get something fast enough to shoot with.

If the argument is that Bayer sensor cameras are smaller and cheeper to make, then sure, but that does not stop anyone from making a 3 chip large sensor camera that takes PL mount movie lenses if they want to spend the money to do so.

 

[Steven Caesare]

I don't want to derail this thread any further with technical minutiae...

Do you know of any 3-chip 35mm cine cams that can use standard PL-mount lenses?

If you do know of one, then we can discuss it's technical merits in light of the OP's question...

If not, I assume there's good reason why they haven't been made, and in any case it renders the issue moot for the purposes of this thread.



-sc

 

[Michael McVicar]

The difference seems to be with the unknown optical manufacturing paths (ever hand grind a lens) or the the known path of the bayer pattern. Math on known quantities is absolute. "The best of all possible worlds." Voltaire

 

[Marc Wielage]

I'm doing a research project on Film vs. Digital Cinema, and it's been a really good excuse to delve into very in-depth research. As with many people, I have always favored film, but the digital revolution has got me questioning my loyalty...

The (very controversial) February issue of American Cinematographer got into this in some detail. I would also recommend John Galt's The Future of Pixels article, bearing in mind that the article is written from a very pro-Panavision point of view.

The recent ASC open house discussion panel debated the pros and cons of film vs. digital for a good 90 minutes, back in early February. I believe this video will be up on the web shortly on the American Cinematographer website. The top DPs on this panel (including Oscar-winning Dean Cundey) seemed to express a preference for film, but at the same time readily admitting that digital can make good pictures, and that it's all about the lighting. They were very open to new tools and technology, including the Red, the Genesis, the Arri, various Sonys, the Canon 5D, even 16mm, and stressed that it was an exciting time for cinematographers because of the many choices available.

I think it's important for more people to understand that the meaningful differences between film and digital are more than just pixels and pickups. From my experience as a digital colorist for 30 years, I think resolution is only one of many factors that go into what makes a good picture. I think exposure, data compression, and motion artifacts are far greater issues, particularly for D.I. theatrical releases. I'm very glad that the new Mysterium-X chip and the Epic addresses all these problems and advances the Red platform.

If you're doing a research project, get some back issues of Videography and track down Mark Schubin's past columns, many of which have delved into dissecting the digital vs. film conundrum. I think Schubin has a very objective viewpoint as to Bayer filtering, multiple chips vs. single-chips, CMOS vs. CCD, what makes up the "film look," the real-world differences between 2K and 4K, and so on. Mark's also a multiple Emmy-award winning video engineer who knows digital cameras backwards and forth, and he has a good grasp of the realities and limitations of the subject.

It's fair to say this is a controversial, religious topic on which many disagree. Speaking objectively, I think you can make good pictures and bad pictures on many devices (even film), and I'm not convinced the camera alone is the most important factor. One of my colleagues is fond of citing the old chestnut, "it's not the arrows... it's the archer." That's never been more true when applied to the art of cinematography. A great camera in the hands of an amateur will not necessarily make great pictures; I'd put my money on a so-so camera in the hands of an experienced artist.

Having said that: no question, the Red has helped democratize the filmmaking process more than a lot of developments over the past five years. But this is a very fluid, fast-changing business, and we're going to see many more options happen in the next few months from many manufacturers. I'm bracing myself for some blockbuster news at NAB in six weeks.