Public are groomed to have insatiable appetite for high resolution, vivid colour images and videos. This unending desire in one end helps the consumer durable manufacturers to have market for new model televisions and cameras. In other end the engineers have to burn their mid night oil to find ways and means to solve the challenges posed by high resolution images.
Increase in image resolution has a direct bearing on generated volume of data. Few examples are presented to substantiate the claim. Next, reasons for the inability of conventional CMOS interfaces to transport digital data pertaining to high resolution images are discussed.
Case I: Digital TV
Increase in image resolution has a direct bearing on generated volume of data. Few examples are presented to substantiate the claim. Next, reasons for the inability of conventional CMOS interfaces to transport digital data pertaining to high resolution images are discussed.
Case I: Digital TV
In Television (TV) series of frames (images) are shown in succession to create a illusion of motion or continuity. In American system 30 frames are shown per second and European systems use 25 frames per second. In olden day televisions, picture resolution is measured by number of lines and not by number of pixels. They use interlaced scanning with an aspect ratio of 4:3 (Width Vs. Height) is employed. In a digital (Standard Definition or High Definition) TV picture resolution is measured by pixels . Each colour pixel have red, green and blue channel with 8 bit width. So, each colour pixel takes up 24 bits.
Net data generated by conventional digital TV titled 480i (suffix i denotes interlaced scanning mode) is
480 (V) x 640 (H) x 30 (fps) x 24 (bits) = 0.22 Gbps (Giga bits per second) (1)
where V = Resolution in Vertical direction, H= Horizontal direction, fps = frames per second
1080 (V) x 1920 (H) x 60 (fps) x 30 (b) = 3.73 Gbps. (2)
In High Definition (HD) TV aspect ratio is 16:9, frame rate is 60 and scanning is progressive. In 1080p, the number 1080 represents lines and suffix 'p' denotes progressive scanning. Switching from, conventional 480i to HD television results in 17x increase in data rate. If the refresh rates are increased to 120 /180/ 240 then accordingly data rate increases manifold.
Case II: Digital Still Cameras (DSC)
DSC are used to capture high resolution images. Reflected light from the subject (human being(s) or object(s)) and background is made to fall on the camera's photo sensor array. The array converts the light into electrical pulses and passes it to the image processor chip present in the DSC. The image processor chip compress the data and stores in the Flash memory as well as display the image in the LCD screen which is used as viewfinder. Now, let us look at the quantum of data transferred from the array to image processor chip.
The total pixel count present in compact DSC ranges from 5M (5 Million) to 12M.
5M (pixel) x 10 (bits) x 15 (fps) = 0.75 Gbps. (3)
We all know combination of red, green and blue produces colour. We also know, in images each primary colour is allotted 8 bits. So, 24 bits are required to produce a colour pixel. But in the eqn. 3 only 10 bits is used for calculation. To resolve this mystery one has to know the functioning of colour image acquisition. In a camera over the photo sensor array a 'colour filter cell' array is placed as in Figure 1. It has red (R), green (G) and blue (B) colour filter cells in the following order Blue-Green-Blue-Green..., Green-Red-Green-Red..., and so on.
 |
| Figure 1. Colour filter cell array (Bayer pattern) on the top of photo sensor array. Image Courtesy: Wikipedia |
Human eye is more sensitive to green than red and blue colours. That is why more green cells are there. The pattern is called 'Bayer pattern' to honour the inventor. Each photo sensor will have one colour cell. The other two primary colours are borrowed from the adjacent photo sensors. Human eye is more sensitive to grey scale variation than colour variation. So, employing borrowing strategy will not hamper picture quality. Initially they get 10 resolution pixel value. This is mapped to eight bits later.
The purpose of DSC is to take still pictures. If so, why 15 frames are taken per second? This is required to create a natural viewing experience on the LCD screen of camera. Thus 0.75 Gbps data has to be transferred from the photo sensor array to image processor within the DSC. If the pixel count is increased to 12M with 12 bits per pixel and 30 frames are used then the data rate increases by six folds.
12M (pixel) x 12 (bits) x 30 (fps) = 4.32 Gbps. (4)
Likewise 24M pixels with 16-bits at 60 frames results in 13.44 Gbps.
CASE III: Digital video cameras (DVC)
Let the DVC capture a High Definition Video at 60 fps with a frame size 1080i or 1080p and having 12 bits per pixel. Then generated data is
2M × 12 (bits) × 60 (fps) = 1.44 Gbps (5)
If 4K or 8K definition are employed then correspondingly data generated is increases manifold. As in DSC here also the data has to be transferred from photo sensor array to image processor.
In the DVC, VGA (Video Graphics Array, 640 x 480 pixels) resolution display screens are employed. The data generated is
640 (H) × 480 (V) × 30 (fps) × 10 (bits) × 3 (channels) = 0.27 Gbps (6)
If the display screen resolution is increased to HD then it results in 3.7 Gbps data rate.
CMOS interface
Photo sensor arrays are invariably built by CMOS (Complementary Metal Oxide Semiconductor) transistors. The conventional CMOS ICs cannot exceed a data rate of 100 Mbps. The optimum data rate is 75 Mbps [1]. The conventional 480i generates 0.22 Gbps and to transfer them 3 lines are sufficient (0.22 Gbps / 75 Mbps =3). At the same time to transfer 1080p HD video one needs 51 lines (3.73 Gbps / 75 Mbps). Having 51 pins makes the chip expensive. Next, more the number of pins increases EMI (Electromagnetic Interference) as well as SSO (Simultaneously Switching Output) noise.
Thus moral of this article is in coming years video data rate will be in Gbps range and the conventional CMOS interface is not suitable to handle it. So, new interface technique like Low Voltage Differential Signaling (LVDS) is employed widely.
Reference
[1]. "High-speed Interface Technology for Image Data Transmission," Find, vol. 26, no. 1, pp 1-5, 2008.
