3-D Cinema and 3-D HDTV resolution and frame rates.

[Cesar Rubio]

First off, I don't want to sell you anything with this topic post (I don't sell anything anymore) . I just want to share all my knowledge and experience that I've acquire working with Digital Images for more than 8 years.

3-D CINEMA:

 

What is the best resolution and frame rate for 3-D Cinema?

 

In my opinion, true 2K at 48 or 60 frames per second.

 

True 1080p resolution would work too as demonstrated by Avatar 3-D and U2-3D. Nobody complained

about the resolution on the BIG screen with such movies.

 

By “true” I mean a 1080p (1920x1080 pxs) capture by a 3 sensors camera or a 2880x 1620 pxs .capture by a single bayer sensor camera down sampled 1.5 times to 1920x1080 pxs. Like the Alexa camera. (Or 35mm film scanned at 2K).

 

Why not 4K?

 

For many reasons, first because unless is for IMAX 3-D screens there is NO need for that. How many IMAX 3-D screens are installed in the world? Not many. What it would be the percentage of those screens in comparison to all other cinema screens worldwide? Probably less than 1%.

 

If that 1% is important for YOU, then by all means shoot true 4K. That would be 6K bayer resolution capture down sampled 1.5 times to 4K. There is not yet a digital camera that can do that. But it seems that the Sony F65 camera will capture at 8K and output to 4K. That would be down sampled 2 times, even better than 1.5. The Epic can do 5K bayer or true 3.3K resolution. I think they will release more resolution cameras in the future for a true 4K output.

 

I saw Avatar 3-D in a HUGE screen almost the size of an IMAX screen and it looked good.

 

3-D since its projecting double frames per second (left and right) doubles the perceived resolution. Also double+ frame rates like 48 fps. or 60 fps. seems to increase the resolution as well.

 

Another important issue would be the ability to record uncompressed RAW or RGB, it doesn't really matter which format you choose in controlled circumstances when you have the time to use a exposure meter.

 

I shoot RAW for event photography where most of the time you don't get perfect exposures due to the fast pace of shooting. Then I can manipulate the Raw file if I underexposed or over exposed a shot. Unless you are shooting documentary style video, there is not a real benefit between RAW or RGB besides the media space needed on the camera/ recorder. Raw files are 1/3 smaller than RGB. But in post it does not matter since all RAW must be converted to RGB anyways.

 

What would I use for 3-D cinematography capture with what is available in the market at present time?

 

Two Alexa-M's recording Uncompressed RAW (Arriraw) at 2880x1620 pxs. (then down converting to 1920x1080 pxs. in post) Or direct Uncompressed RGB 1920x1080 pxs. Into the Codex or S.Two recorders. The Raw format would permit record more data into the recorders since 1080p RGB will take double the space of 2.88K Raw.

 

Either way at 48 or 60 fps.

 

Another extra reason to go for true 2K or 1080p is that there are not 3-D cinema projectors capable of 4K/48p or 60p. The most you can go up to 60 fps is 2k or 1080p 3-D projection.

 

Why I prefer the Alexa-M over the Epic-M/X? Simple, because I prefer Uncompressed Raw over Compressed Raw even with more resolution.

 

I have used Cineform Raw and Uncompressed Raw and compared the results of both. There is NO comparison, when you see Uncompressed RGB against Cineform HD (already debayered) even at the highest 4:4:4 Filmscan 2 quality, Uncompressed RGB will take your breath away. Hands down is the best video I have ever seen in my life, period. Close to film scans. The colors are so rich and saturated that the images almost pop up out of the screen.

 

Even 8 bit Uncompressed Raw/RGB is way superior to Cineform Raw/HD at 12 bit. (The Redcode is similar to Cineform Raw).

 

Working with Uncompressed Raw/RGB there is no visual difference between 8,12 and even 16 bit. Probably because most monitors/projectors are 8 bit anyways. (24 bit in full RGB color)

 

I saw a huge quality difference between Cineform Raw when recording 8 and 12 bit, so using compression higher bit is helpful.

 

And last but not least, the bigger pixels on the Alexa-M camera offer 1 more native stop of Dynamic Range over the smaller pixels on the Epic M/X camera.

 

3-D HDTV:

 

For 3-D HDTV I would use true 720/50 or 60p.

 

What cameras I would use?

 

If you CAN, by all means use the Alexa-M's as well!

 

If not (like me), then look for more affordable solutions.

 

Like the Sony F3's cameras (recording Uncompressed to external recorders).

 

Recording 1080/50 or 60p will give us a true clean 720/50 or 60p output that would work for 3-D networks or for Bluray 3-D delivery.

 

If those cameras are still out of the game for you, then see this camera: (note: I don't have affiliation with Basler nor I sell their cameras or get paid for promoting them).

 

Basler acA2000-340km/kc

 

http://www.baslerweb.com/beitraege/unterbeitrag_en_115108.html

 

It offers almost the same quality wise as the Sony F-3 camera.

 

The Basler cameras advantage that I see over the Sony F-3's for 3-D capture (other than the price for a complete 3-D recording system) would be a little extra resolution (2K vs 1080p), the ability to record side by side stereo without a beam splitter rig and the advantage of recording Uncompressed Raw.

 

The Sony F-3 advantages over the Basler camera is that offers a half stop more of native Dynamic Range, and like a half stop more of base ISO/ASA (ISO 800 vs ISO 640), but that is lost with the more than a full stop that you lose with a beamsplitter rig.

 

Cesar Rubio.

[Cesar Rubio]

I just want to add these interesting additional info,

 

In another 3-D forum I was discussing IMAX resolution with a member who said that he thinks that IMAX resolution would be true 12K Digital resolution. And I said this to him:

 

“I don't know if an IMAX 65mm frame really has a true 12k resolution. It depends in what settings you scan it.

 

In my own experience scanning 35mm full frame film (not S35), I would rate it at around 3K resolution. You can scan it at higher resolutions if you want to, but its pointless, the visual resolution won't go higher than that.

 

And actually in my comparisons making large prints from my Nikon D70 camera (3k resolution) it beats anything shot with FF 35mm film.

 

I have made 30x40 inches enlargements with the 6MP file and the results are actually comparable to the prints I get with a Mamiya RB67 film frame!

 

I would personally rate an IMAX 15/70 frame at around true 6K resolution (9K bayer resolution).

 

Some people who have seen Epic 5K footage projected on a big screen say that it is comparable to a 65mm film... “

[Cesar Rubio]

And then I posted this info today:

 

I just saw that Red plans to release a 9K camera, that would camera would give you true IMAX resolution.

Epic 645

 

Sensor size: 56×42mm (2.2×1.7in)

 

Resolution:9,334 × 7,000 (9K)

 

Price (“brain” only) $43,000 US Dlls.

 

But the problem with such large sensor is the same that always existed for IMAX 3-D films, DOF (Depth of Field) and low light shooting.

 

This is extracted from the IMAX Bugs 3D movie website:

 

“The complex motion-controlled camera system weighed three tons, had more lights than it normally takes to light a cathedral, and all this focused on a scene sometimes not bigger than a postage stamp. In studio this meant wearing sunglasses to film and bringing in industrial air conditioning units to keep the ambient temperature down and localized air conditioning units to keep the sets/bugs/crew cool.”

 

http://www.giantscreenbugs.com/HTML/bugs_html.html

 

 

To get the same DOF of f1.4 in a 2/3” optical format (11mm wide approximately) with a 56mm wide optical format, I would need f5.6!

 

How many more lights would I need from f1.4 to f5.6?

 

4 more? WRONG.

 

Lets say I am using 1 candle of light for f1.4, then for:

 

f2 I would need 2 candles

 

f2.8, 4 candles

 

f4, 8 candles

 

f5.6, 16 candles

 

So from f1.4 to f5.6 I would need 15 times more power in lights!

 

Just so you can make calculations as how much light those numbers mean, when shooting in a normal house light (“1 candle”) I would need a ISO/ASA 800 1/60th shutter speed and a f1.4 lens opening.

 

Probably you can say, no problem I can set my ISO/ASA higher on camera....BUT, you will increase the noise levels, and noise looks horrible in 3-D! ( just like if they were "dancing" bacterias all over the frame).

 

So the only professional choice you will have is to bring in more lights!

 

Film is very “forgiving” with artificial lighting, but video it isn't, you have to be very careful in lights placement so the light can feel more “natural”.

 

Also the heat and discomfort to the eyes that powerful lights brings in actors is something that you should avoid when possible, so they can work more comfortable specially in long hours days of shooting.

 

Another important aspect would be the needed recording media speed and capacity for a dual 9K capture, especially working with Uncompressed Raw/RGB for the best Image quality, and the needed hardware in post to work with such large resolution.

 

Such hardware doesn't exist yet in the consumer field. 9K is more than 4 times the resolution than 4K, and many people are having a hard time in post with a 4K workflow going from a reasonable priced 4K monitor/projector to the fastest consumer available workstations....and they work with Compressed Raw (Redcode Raw) and the Red Rocket.

 

In conclusion, I am NOT against more resolution in 3-D Digital Cinema in general, and I love IMAX 3-D films. But to be honest, not all 3-D productions would need such resolutions and hassle to work with...

 

And don't get me started on the non sense 4K and 8K resolutions for HDTV!

 

I have a Sony SD TV (720x480 pxs) and people think that is a “HDTV”...also I have a Samsung 720p HDTV at home and everything looks good. I watch Bluray Discs with a 720p projector and I use a more than 100 inches screen size and everything looks superb.

 

What we need for home HD viewing is not more resolution, but LESS compression!

 

HDTV transmission bandwidth at the moment can not do more than 720p (or 1080i) and for 3-D we need double that bandwidth.

 

I don't think that anytime soon TV Broadcasters will invest billions and billions of dollars in infrastructure worldwide so they can start transmissions at 4K, not even at 1080p that requires double the bandwidth than 720p. 4K is more than 8 times 720p!

 

And all that just because TV manufactures now can do 4K TV panels more affordable...

 

TV manufactures are going the way still digital cameras went at the beginning, more and more resolution to be “ahead” of the competition...

 

Now still digital camera manufacturers have reached a wall, especially in those tiny sensors packing millions of extremely diminutive pixels in point and shoot consumer cameras.

 

I can get better pictures with a DSLR camera with 6 MP (Mega Pixels) than with a 14 MP consumer camera, that proves that “more” is not always better. Why? Because the pixels are larger in DSLR's and offer better low light performance, dynamic range, color reproduction and yes better visual resolution!

 

CR.

[Cesar Rubio]

Somebody posted this link in another forum:

 

http://carltonbale.com/1080p-does-matter

 

That's why in Cinemas when you seat in the middle of the theater, even true 1080p or 2K material looks fantastic. But if you seat in the front row it will look like crap...(even 4K).

 

CR.

[Cesar Rubio]

At first sight, one thing that concerned me about the Sony F65 camera is that in order to pack 8K resolution into a S35mm format they would have to make the pixels size smaller, that would be the case if they used a typical Bayer pattern sensor. But they don't!:

 

pro.sony.com/bbsccms/assets/files/.../F65_Camera_CameraPDF.pdf

 

 

That sensor array reminds me a little bit of what the Fuji S3/S2 DSLR cameras offered in their sensors:

 

http://www.luminous-landscape.com/reviews/cameras/s3-pro.shtml

 

 

They also include a clever “rotary shutter option” on the F65, to get rid of the rolling shutter artifacts of traditional CMOS sensors (see their explanation in what those do in a 3-D beam splitter rig).

 

The Sony F65 camera being 8K Bayer will be close to an IMAX 15/70 (65mm) frame. And for 3-D work, having a S35mm sensor size will be better than the larger sensor on the Epic 645 camera (9K).

 

But only if they could make the Sony F65 smaller, like the Epic-M or Alexa-M cameras for 3-D work...

 

And I couldn't see anywhere on the Sony PDF brochure if they offer Uncompressed Raw recording on the F65, does anybody knows if they do?

 

Thanks,

 

CR.

[Cesar Rubio]

I am not sure if the Sony F65 offers TRUE 16 bit Raw output...

 

According to the info here:

 

•”60 Minutes of 16 bit Linear RAW file recording onto a 1TB SRMemory card at 24FPS”

 

http://www.engadget.com/2011/09/07/sonys-f65-cinealta-4k-camera-now-available-for-a-paltry-65-000/

 

Let's do some math...I love it ;-)

 

For 1080/24p I get a file size of 3 GB per minute with 8 bit Uncompressed Raw.

 

For 2K/24p that would be 3.2 GB/minute.

 

For 4K/24p would be 12.8 GB/min.

 

For 4K/24p 16 bit that would double it at 25.6 GB/min.

 

Now, 25.6 GB x 60 min=1536 GB or 1.5 TB.

 

What those numbers tell me?

 

That either the Sony F65 is outputting 12 bit Uncompressed Raw, or it's using 1.5:1 Compression for 16 bit Raw...

 

CR.

[Cesar Rubio]

And those numbers also show that the F65 camera does NOT output direct 8K Raw resolution like their PDF brochure suggests:

 

-“Your choice of resolution: gloriously supersampled HD, supersampled 2K, true 4K or EVEN 8K”

 

It's “super-sampled” 4K resolution...

 

CR.

[Cesar Rubio]

This whole discussion might sound like if I am an “anti” Red guy, I am not. I like the Epic Camera form factor very much. 8 years ago I fell in love with the Panasonic DVX-100 camera the first time I saw it because of the same reason, small handheld form factor. I never liked “full size” (shoulder mount cameras).They are horrible and a pain in the butt to work with. I think the DVX-100 is one of the best camera designs ever, a classic and I will never sell it nor my DVC-80 (same camera but with only interlaced recording).

 

I also like the Red Camera Company policy of not sending to build their cameras overseas...that's what we need now here in the USA to improve the economy. 400 families have a descent life style because of the decision of one man. If more CEO's had that in mind, instead of more profits at all cost the economy wouldn't be as bad as it is now. And the whole world economy would be improved as well because Americans are expenders by nature and everybody benefits if they have jobs and money in their pockets to spend.

 

That reminds me of the movie “300” that is of my favorite movies of all time. Jim you have awakened “Xerxes” congratulations! Now you have to do what David did against Goliath, select very careful the stones you are going to throw...

 

This is a free “consulting” if you will, my 2 pesos.

 

First, OPEN the UNCOMPRESSED RAW GATES! It's time, you CAN do it. I am not saying to drop the Redcode Raw compression, some people might like that for certain projects. But give them FREEDOM to choose what they want, Uncompressed or Compressed workflows.

 

Second, The movie industry (and by default you) should move two steps forward and 1 step "backwards" from the “standard” S35mm/24fps format...

 

For 2-D productions, (and given the fact that the Sony F65 camera is already here) bring in Epic FF 35mm 6K ASAP!

 

Offer Uncompressed Raw 8 bit at the moment for that camera, 6K 8 bit requires the same bandwidth than 4K 12 bit.

 

Those thousands of Canon 5D shooters aren't going back to S35 after the beautiful Bokeh or out of focus backgrounds they can get with the FF 35mm format. Drop Epic-X and offer an upgrade path for current Epic-M owners to Epic FF35. Leave the Red One-MX camera as it is with 4K resolution in the S35 sensor size (or replace it with an Epic size design but with the same specs).

 

That's one step forward, the other is double frame rates, 48p or 60p. You already can do that so it's not a big deal.

 

For 3-D go one step “backwards”...to the 16mm format (or 4/3” that is like 18mm wide).

 

Offer a 3K or even better, a 3.3K camera sensor in such format to shoot parallel stereo and have spare resolution to crop in post. If people wants more resolution than that for 3-D, then a Foveon X3 sensor approach or even a traditional 3 sensor video camera design will give them that without increasing sensor size (the Sony F65 sensor approach is a clever idea too, but I would guess that they have that sensor design already patented and they wont sell it no anyone else any time soon).

 

Thousands of people probably will hate me for this but oh well business is business, drop the 2/3” 3K Scarlet...leave it at 2K resolution to compete with the prosumer market like Ikososkop dll and Sony F3 cameras. Those at Scarletuser forum will be angry at the beginning for sure, but it will pass and they will get over it sooner or later...or move on to the next “level”.

 

I think that is all for now.

 

All this might sound crazy, but hardcore 3-D people like me are kind of “nuts” anyways, we have stick to an almost “impossible throw” (quote from Hector in the movie Troy-2004) with S3-D for years...

 

Take it from a 7 year old boy who fell in love with 3-D the very moment he saw the first Viewmaster reel back in 1978. And in 2011, 33 years later he is still in love with it like the first time.

 

Cesar Rubio.

[Cesar Rubio]

Someone mentioned to me that the Alexa camera when recording 1080p (RGB) outputs less resolution than the 2.88K Raw full sensor resolution.

 

And I know that 1080p is less resolution than 2.88K, but I thought that internally the whole sensor resolution was read and then downsampled to an 1080p output.

 

But it seems that is not the case. I already brought this issue on the Alexa Digital forum on the Arri website, but so far no comments about this issue.

 

Probably what the Alexa camera does is what DSLR cameras do when recording video, skipping lines of pixels to record less resolution without “windowing” the sensor...

 

Also it seems that when recording directly RGB color with a Bayer sensor you lose light sensitivity: (the person previously mentioned says that as well of the Alexa when recording RGB):

 

http://info.adimec.com/blogposts/bid/54091/The-Pros-and-Cons-of-Color-Processing-with-Machine-Vision-Cameras

 

So for the time being, for 3D Cinema I will stick with the Alexa-M camera recording Uncompressed Raw 2.88K (Arriraw) at 48 or 60 fps. Then downconvert in post to a final 1080p delivery.

 

And for 3DH-DTV productions after reviewing all the pros and cons more carefully, I think I will stick with 2K bayer resolution in a 2/3” sensor camera like the Basler camera recording Uncompressed Raw 2k at 50 or 60 fps. Then downconvert in post to a final 720p delivery.

 

Thanks,

 

Cesar Rubio

Wisconsin & LA.

http://dna-rubio-3d.blogspot.com/

http://dnarubio3d.wordpress.com/