Digital Video Basics

By Amir Majidimehr

At the risk of stating the obvious, video in the real world is analog. To convert it to digital, we need to decide on how many samples to take per second and the resolution of each. The latter is rather simple in this space. The most common resolution is 8 bits with 10 bits reserved for broadcast/effects/professional space. To put this in perspective, we use a minimum of 16 bits for audio in CD and go up from there.

8 bits represents 48 db of "signal to noise" ratio by the way. Not a very big number and hence the reason even clean video has some noise in it. But we digress.

CCIR 601/BT601 is a SMPTE specification which standardized how many samples are taken per second for standard definition video. The convention is 13.5 MHz (millions per second) which translates into 720 pixels horizontally. And not by accident, this is the same number of pixels in DVD format.

Vertical resolution for standard definition video is fixed by the broadcast standard (NTSC in US is 525, and PAL used in the rest of the world, 625). The actual visible resolution is lower due to some reserved areas and resulting in 720x486 and 720x576 for NTSC/PAL standards respectively. Some people round the vertical resolution a bit so you might see slightly different numbers. Putting it altogether, the DVD resolution therefore is 720x480 in US for a total of roughly 346,000 pixels. Put in the context of typical digital camera, this is 0.3 megapixels per frame of video. Imagine taking a picture at that resolution and blowing it up bigger than poster size to 40 and 60 inches of your flat panel TV. No wonder then that the high definition video standard was created to up the resolution (by up to 6 times).

In video, we do not operate in RGB (red/green/blue pixels) as PCs do. Instead, we separate the color from black and white information. The latter is called Luminance and the former Chrominance. Luma and Chroma for short. Color is separated into two “color difference” values giving us three values per pixel. In other words, each video sample has one Luma value, and two Chroma values. As you may know, in RGB we also have three samples one for each color indicated by the letters.

Depending on who you are talking to, and whether we are talking about analog or digital signals, you will see notations such as YCrCb, YPrPb, YUV, etc. They are all the same and indicate color difference mode as opposed to RGB. As the theme indicates, “Y” is always luminance and the other two are the color difference signals.

RGB is for computer use and has no notion of separate luminance. Using simple math, you can go from YUV to RGB and back. But not every color can be represented in each space so the conversion may be lossy causing some colors to be lost.

In order to reduce the burden of storing and managing so much video data, color is sub-sampled meaning it is allowed to have lower bandwidth. A notation is used to indicate this. Without subsampling the video is said to be 4:4:4. Next step down is 4:2:2 which translates into half the bandwidth for color. If you halve the color yet again you arrive at 4:2:0. Using 8 bits, 4:4:4 translates into 24 bits as it would in RGB world. The same resolution translates into 12 bits when dealing with 4:2:0.

DVD, Blu-ray Disc, and all forms of broadcast television use 4:2:0 (8-bits). This means that the color is substantially reduced in bandwidth as compared to what is available in a professional setting. Fortunately your eye is not very sensitive to color frequency (due to the difference between rods and cones in your eye). So in real life, you don’t see that softness. But put up a color bar from a test disc and you can easily see that the edges are quite soft as one color switches to another. Compare that to the very sharp lines in the black and white boxes.

For HD there is also a difference in color space. What is color space? In a nutshell, the color space tells you what numbers to use to represent a specific color. As you can imagine, there a lot of variations in the color red. But we must pick one to be the absolute red color so that we can then reproduce it at the receiving end. CCIR 601 defines a certain color space used for SD. Recommendation 709 does the same for HD. It has a slightly expanded color gamut. Your TV always expects 709 if it sees a signal higher in resolution than SD. So when a DVD player upsamples the video, it must also change the color space to 709 or the colors would look (slightly) wrong.

As for levels, the TV world uses 16 to 235 in an 8-bit word even though the range of values is 0 to 255. This is to allow signals to go above and below the min and max and not get lost. In computer world/RGB, we use the full values. If your display is set up for PC use/0-255, feeding it video signals results in washed out video since it thinks “16” means something well above absolute black. Likewise, if your display thinks you are in video mode and you feed if computer graphics, the levels can be crushed into 235. Something has to do the right conversion between the two. If not done right, you get banding and crushed video.

Finally, if your source is 8 bits as is the case with everything you buy today, having higher resolution does you nothing. Even though a display may advertize 10-bits, 16 or whatever, you are not able to experience that (assuming the display can actually show those levels as opposed to having that input resolution only which is commonly the case). So unless you are doing graphics/effects work on the display, you can safely ignore these numbers.

This should get you started at least in learning more about this subject. If you have subscription to WideScreen Review magazine, I also wrote an article in there with a lot more detail on this topic.

Back to Articles