[This post is written by Ariel David, Video Technologies Expert in RADVISION's Networking Business Unit. Ariel has close to a decade of experience in video technologies and has embedded systems for video and image coding, video enhancement technologies and the management and development of integrated solutions. He can be reached at firstname.lastname@example.org ]
High definition (HD) is everywhere. If you haven’t read about it in one of the posts here, you probably saw it featured in a commercial. If you haven’t seen any commercial, you probably heard about it from one of your friends, who bought a new TV set. HD is the next big thing. The question is not if HD will spread globally, but rather if we have really stopped to consider all the implications of HD?
When it comes to video conferencing, there are now enough endpoints out there that support HD to admit that the market has completely gone to high def. I’m not talking about outrageously expensive conference rooms. I’m talking about commoditized HD endpoints, from vendors like LifeSize or Aethra.
Most users would definitely enjoy the HD experience. In ideal conditions it is a definite “WOW”, especially when compared to the small, not to say blurred SD experience (and HD sure makes CP look better than ever). It’s bigger, it’s sharper, and you can see details you’ve never seen before.
That’s why the video conference world has “jumped” to HD (actually skipping middle resolutions such as 4CIF). Modern endpoints and MCUs support up to 720p. Some even support 1080p. But a very crucial part of the puzzle was somewhat neglected. Bandwidth, which is necessary for the transmission of video over IP, hasn’t progressed as fast as the endpoints. Actually, it’s not only a matter of progress, but also of price. In other words, bandwidth is expensive.
If bandwidth is expensive, compression becomes a necessity. The video encoder is the unit responsible for compressing the video to fit the bandwidth requirement. In an ideal world this encoder would have unlimited compression ability, with no implementation problems or any visual artifacts. Therefore, a video in any resolution can be sent over any given bandwidth with no effort at all.
Unfortunately, back in the real world, even the most advanced state-of-the-art encoder used today – H.264 – has many limitations, in terms of compression, complexity and visual artifacts. The reason it is considered the most advanced is due to its ability to compress the same video at the same visual quality to half the bandwidth required by previous encoders (such as MPEG2, H.263, MPEG4, etc.).
Still, for a good quality (or, God forbid, superb quality) 720p resolution, 30 frames per second (fps) video conference scene (known as “talking head”) compressed with H.264 you probably need at least 1.5-2 Mbps (that is Megabits per second). This means that an HD endpoint that sends a 720p resolution video stream would require a guaranteed 1.5-2Mbps bandwidth. I must warn you that this is a ballpark number, as it depends very much on numerous factors (such as the quality of the camera or lens, the quality of the encoder, the complexity of the scene itself), so take it with a grain of salt. Nevertheless, the graph below shows the visual quality that the end user can expect with a given bandwidth using H.264 encoder.
Visual quality resolution vs. bandwidth using a H.264 encoder for SD
It is very clear why CIF/240p and 4CIF/480p resolution are still the most common resolutions today. For example, a good quality video at CIF/240p would require a 512Kbps link. This is something most users can afford. This would definitely be the norm, if video conference becomes very common in the enterprise and there will be many video conferences running in parallel in the enterprise network. Even a good quality 4CIF/480p video is still affordable in most cases. Around 1Mbps is also not that much to “pay” for good quality video.
Frame sizes comparison for CIF, 4CIF, 720p and 1080p
Jumping to HD introduces a very “nice” jump in bandwidth requirement. I suppose this should be quite obvious as there is more than twice as much information in a 720p picture than in a 4CIF/480p picture and 5 times more information in a 1080p picture. For something a bit more staggering, 1080p has more than 20 times the information in CIF. All this can be seen very clearly in the graph above. Still people usually tend to forget this and expect good quality HD to consume somewhere around 2Mbps. But even with 2Mbps per user, assuming there are many video users and conferences in parallel, the enterprise will have to invest a lot in its IP network.
Visual quality resolution vs. bandwidth using a H.264 encoder for HD
Now let’s make the final and painful leap to 1080p. If you invest in a 1080p endpoint, you probably expect superb quality (and not only a “good” one). This minor experience will cost you somewhere around 6Mbps.
I am guessing that seeing the required bandwidth for 720p or 1080p makes it crystal clear why the HD video conferencing is not spreading that fast and is still more of an executive toy rather than a mass communication tool. Most enterprises, even if they can afford an HD endpoint and the associated connection, can most probably not afford this in high volume. This means that at best there will be only a few HD endpoints in some of the conference rooms in the enterprise, maybe a few for the executive board.
There is also another issue. If we look at the larger scale, the video conference market today is still filled with SD (CIF or 4CIF) endpoints and the common links between sites range from 384Kbps to 2Mbps. An enterprise and/or user that has invested in HD, expects to get “the HD experience”, but is strongly affected by his peers – other users that attend the same conference and other users in the same site that share the same bandwidth capacity. For instance, a participant connecting with a low-end endpoint, using low resolution or low bandwidth will send poor quality video and will look bad even on an HD endpoint. Actually it will look even worse on the large HD screen. Eventually, the HD users will be reduced to poor quality, not because of his endpoint or connection, but simply because of other participants in the conference.
But still, HD is next. If it is not here now, it’s definitely on the way. HD-enabled MCU’s and endpoints are being released by all the leading vendors and prices are dropping. Everyone agrees that soon HD will become a commodity. But bandwidth is a problem and it seems that it will remain a problem in the future. So, for all those who want and are ready to invest in the HD experience, PLEASE don’t forget to invest in bandwidth. It’s as important.