The purpose of this EBU recommendation is to give technical aid to broadcasters who intend to use current (or future) 2D HDTV infrastructures to produce 3DTV programmes.
NHK is developing an 8K image sensor for its Super Hi-Vision Ultra HD system capable of 120 frames per second, which it plans to unveil in Tokyo on 23 April and should be ready for use at the London Olympics.
The new 33-megapixel (7680x4320 pixel) CMOS sensor will use an advanced two-stage (4-bit then 8-bit) cyclic analogue-to-digital converter to deliver a 12-bit image at the higher frame rate (twice that of the current SHV cameras). This architecture also allows it to reduce power consumption, with the ADC drawing 800mW (out of a total drive power of 2.5W). To deal with the high number of pixels involved, the sensor outputs alternating rows of pixels to ADCs on either side, each of which has 48 parallel outputs.
The sensors will probably still get rather warm at these frame rates, so heat management will be a significant issue. Each pixel measures 2.8 x 2.8 microns, and the 26.5 x 21.2mm chip (which is about the width of a Super35mm sensor, but taller) will use a 0.18-micron manufacturing process. The chip is being developed with Shizuoka University, and engineers revealed details of the sensor at the recent IEEE International Solid-State Circuit Conference in San Francisco.
Two SHV cameras will be used to capture parts of the 2012 Olympics, with transmission to three large screens around the UK (plus one in the International Broadcast Centre) and three in Japan. Having the 120fps sensor will reduce motion blur and allow for much better slo-mo replay.
However, as the resolution increases (and 8K is 16 times the resolution of HD), motion defects become much more noticeable. “300 frames per second might just be acceptable, but 600fps would be better,” said colour scientist and camera consultant, Alan Roberts. At 300fps, the material would also be easily compatible with both 50 and 60Hz display.
Using Long GoP compression, which typically combines a group of pictures in half a second, would lead to a GoP of 150 or 300 frames, but this wouldn’t lead to huge bandwidth requirements. “The motion between frames is very small, so the compression is much easier,” he explained, which means the increase in bit rate needed to convey high framerate material is minimal. “The problem is in the shooting and the editing, because of the monstrous data files you have to deal with,” he added.
By David Fox, TVB Europe