Thursday, September 08, 2011

Sony 4K Digital Cinema Sensor Better Than Film

Sony F65 camera brochure (or web version of it) mainly talks about its sensor innovations: "The F65 exceeds the resolution of any previous digital motion picture camera (as of August 2011), the result of a remarkable Sony Super 35 image sensor".

Sony states: "Through all the decades, our design goal has always been to match the photographic quality of 35mm film. But now we’re setting our sights even higher: to surpass the limits of human vision. The F65 image sensor is the first of this new breed".

"The 20 million photosites of the F65 deliver a powerful imaging advantage. The difference is “supersampling”:

Regarding the extended 14-stops DR: "Thanks to Sony CMOS advances, the F65 empowers the cinematographer with prodigious exposure latitude, high sensitivity and extremely low noise":

About color gamut: "Many HD cameras are limited to the Rec.709 color space. While this produces broadcast-legal color for television productions, it still falls short of motion picture film. The Sony F65 transcends this limitation. The camera features a new Color Filter Array on the sensor, along with a 3D lookup table (LUT) and proprietary color management systems. In this way, the F65 can shoot in either Rec.709 or F65 Gamut, which actually exceeds the SMPTE specification for color negative film in every direction":

And the last: "the F65 provides over- and undercranking at frame rates of 1 to 60 frames per second (4Kx2K resolution), and up to 120 frames per second (4Kx1K resolution)".


  1. Funny how rotating a traditional Bayer pattern 45 degrees can be presented as such a new thing. I also wonder if that does not create strange artefacts in an image.

  2. Because Sony have publicly trashed use of bayer patterns in large single sensor cinema cameras and they think that rot45 makes them immune to criticism that they're hypocrites.

    Next BS is their "8K" which would need 37mp, whereas they only have 20.

  3. A 45deg line viewed by a Bayer pattern sensor is the same as a vertical line viewed by a rot45 sensor. Concentric rings are exactly the same to both. What's the "new" part?

    I also don't understand their gamut diagram. Why the rounded corners on the F65 plot? Three color filters makes three points on the chromaticity diagram. Three points makes a triangle, not a rounded blob. Where exactly are the three points for the F65?

  4. @ What's the "new" part?

    The new part is the over-sampling.
    They create a 4Kx2K (8mp) image from a 20mp sensor.

    1. Bit too late, but what the heck.

      I don't see any over-sampling here. On the websites it sais 1 green photosite dedicated to 1 pixel (which truly i 4K), and 2 blue pixels from 1 photosite (undersampling), and 2 red pixels per 1 photosite (again - undersampling).

      Green is 4K, blue and red are only 2K.

      Aspect ratio is 1,9:1: How they calculated 20MP from that is beyond me. If I assume the CMOS chip is a square, then yes, you have oversampling on the vertical (but not on the horizontal), but then their claim (1 green photosite per green pixel) is false.

  5. The fact to rotate the Bayer's pattern looks equivalent to original Bayer's pattern. But the tricky thing is that our world is very artificial and most of the objects have rectilinear forms.

    So the 45° rotated Bayer's pattern will give you subjectively better image quality.

    I've proposed this idea in a seminar talk in 1997 at Taipei....

    -yang ni

  6. Human vision is quite sensitive to horizontal resolution loss, less so to vertical one and even less to diagonal. So, it looks to me like a good approach to invent such a pattern that trades diagonal resolution for horizontal.

    It would be even better to trade some vertical resolution for a better horizontal one. BTW, it is only possible in video cameras where there is no such a thing as portrait orientation.

    @ "I also don't understand their gamut diagram. Why the rounded corners on the F65 plot? Three color filters makes three points on the chromaticity diagram. Three points makes a triangle, not a rounded blob. Where exactly are the three points for the F65?"

    I'd guess Sony makes some color space transformation here. Thus we do not see the triangle corners in green and blue. Ideally, they should be located at at the crossing points of the "blob" sides, if you continue them as straight lines.

  7. rot45 sensors have been used by Fuji DSLRs for many years.

    There is a preponderance of horizontal and vertical structures in the natural world as well. This is because of gravity. You either fight gravity or you don't. If you don't then you are (usually) horizontal. If you do, then the most efficient way is to be exactly vertical. This is because the compressive strength of most materials is greater than their tensile or tortional strengths.

    See "Statistics of natural images: Scaling in the woods",

  8. Sony's goal is to win over paying customers, not win over the technical community.

    With this blitz, despite the technical arguments one might make, Sony's competitors will need to go on the defensive.

    Sony's sensors are pretty good, maybe even excellent. You have to respect your competition when they are making money and a leader in the pack. Even if it is OVT some of the time.

  9. What is not mentioned though (unless I missed it) is a quantification of the lattitude. They only describe it as being great...but what does that mean?

  10. disregard the previous post...just saw that I indeed missed the 14 stop remark.

  11. Previously Sony announced that the new sensor will be with Global Shutter but actually they did come with Rotating shutter and Rolling shutter. Maybe they could not achieve the noise level that they wanted?

  12. Any ideas if this sensor technology is patented by Sony?

    I would love to see a similar sensor with 3.5K resolution in a 2/3" size for 3-D Cinema.

    Cesar Rubio
    Wisconsin & L.A.


All comments are moderated to avoid spam and personal attacks.