Streaming media products: Moving from software to hardware

It is clear from the compression ratios that must be achieved that every opportunity to make the job easier ought to be taken. In the early 1970s I remember
Author:
Publish date:

It is clear from the compression ratios that must be achieved that every opportunity to make the job easier ought to be taken.

In the early 1970s I remember trying to understand the papers given at the SMPTE Winter Television Conference on “Bit Rate Reduction.” The ideas seemed foreign — taking perfectly good video and squashing it into digital form and then wringing out the excess content to make it smaller for storage and digital transmission. Of course today we would find nothing out of the ordinary. BRR — video compression — is used in consumer camcorders, hard disk personal video recorders, and perhaps most ubiquitously in little windows (and big windows) on our PC screens. The basic technology stems from research into removing redundancy in complex data.

It can be shown that mathematically loss-less compression can achieve something slightly less than a 3:1 compression ratio. Typical professional video recorders generally operate at closer to 10:1 compression. Satellite transmission for program backhaul is done at about 25:1 reduction. DBS signals are at almost 50:1 compression.

Then there is streaming media, which operates at a tiny fraction of the data rate of other uses. Try a 700:1 compression ratio for 300Kb/s. If you have only a dial-up connection, the ratio becomes an astounding 6000:1. It is against this backdrop that we consider the issues related to hardware and software solutions for streaming media.

Compression for streaming

When the concept of streaming video to the Web was created a few years ago, it sufficed for most purposes to simply show a “changing” picture, for moving images would seem to dictate a frame rate in positive numbers. The quality of compression for streaming has improved from both major vendors (Microsoft Windows Media and Real Networks RealVideo), and the ability of the Internet and private networks to handle high bandwidth media has improved at the same time, multiplying the net effect to consumers. As both the technology and bandwidth continue to improve the quality of the delivered application should become quite good.

The process of streaming involves two distinctly independent processes: encoding and delivery. Encoding requires an input card for audio and video, the compression software and storage for the output file, unless it is passed immediately to the delivery system in a live broadcast. Delivery usually involves passing the file to a server from which it is accessed by the consumer. This server could be a single point source for content not expected to require wide distribution or a considerable number of concurrent users, or a distributed network of servers which cache the stream nearer to the users' location, a scenario illustrated in Figure 1. This method enables delivery to multiple users, while avoiding often-clogged segments of the Internet.

The media to be streamed must be converted from analog (or digital) video to the appropriate compressed format. This process is not at all different from ingesting for a broadcast playout server. Until recently this process was generally done in a computer with an appropriate input card, loaded with software that completed the appropriate compression. It is clear from the compression ratios that must be achieved that every opportunity to make the job easier ought to be taken. Feed lousy, noisy video in, and get a “less than stellar” result. It is also important to make sure that the intended target bit rate is one that can be delivered end-to-end to the consumer. Poor video, made worse by dropped frames and constant recovery from network errors, will make what might have been a successful experience one the consumer will reject as a failure. Anecdotally, I marvel that users who would not tolerate a VHS tape stopping periodically will actually watch video on a computer that seems to suffer from too much cholesterol.

Products for streaming compression

In the last year several manufacturers have introduced products intended solely to compress and either store or forward to the delivery server video from professional formats. Indeed a recent issue of a consumer computer industry magazine reviewed such products. They tested quality and features and pronounced their judgments.

These products include simple systems intended to perform only one type of compression at a time (usually selectable between Windows Media Player and Real Video formats). But the marketplace has moved considerably beyond making the ingest computer more user friendly. Some products allow both formats to be created at the same time. Others provide the ability to combine the compressed stream with other elements, automating the assembly of a coherent and locally branded webpage. One carries it much further, with control of cameras, switching and effects, character generation, web advertisements, and other elements in a very sophisticated presentation. This might be viewed as TV studio in a box, with the output dedicated to Web access.

At the extreme edge is a system from one manufacturer that combines all of the above plus user interactivity that allows users to assemble a newscast from their choice of stories. While one might question the journalistic value of some people's judgment, it is a natural extension of the freedom of choice on the Web. Such streaming on demand playlists may well demonstrate a major change in the way users interact with streaming media in the future, allowing a user to assemble a virtual broadcast channel for seamless viewing, provided bandwidth was not an issue, which of course it still is.

Various factors need to be considered when picking streaming media products for installation into a broadcast station. I encourage customers to look back down the wire from the consumer's viewpoint. If the customer can reasonably be expected to accept all streaming formats, it is clear that support for a single format would be a limitation you should avoid. If you are placing your newscast on the Web you may be able to select a single format, for your viewers will likely come back regularly and therefore take the time to load the appropriate codec. On the other hand, if your product has wide use, or is one which viewers are less likely to visit regularly, you should consider support for all likely target formats. If the time sensitivity of your content is such that you cannot take the time to sequentially encode in multiple formats you should consider either a single solution which does both simultaneously, or perhaps redundant systems, which will allow you to create two streams at the same time from separate encoders.

The input side of the encoder is very important to the engineer who must implement the system. Some systems are built with unbalanced audio inputs and poor quality NTSC decoders. Others have component digital and AES inputs. Streaming encoders work very much like more mainstream compression products, and perform significantly better if you can deliver either component digital video or at least NTSC that has been decoded well (assuming a component input on the streaming encoder). Some encoders allow you to control the source tape deck when encoding from a VTR, so you might look for an RS-422 port. A few permit 1394 inputs (DV streams), which would allow a simple and very effective interface.

It is important to recognize the issues that affect the ease of use and ultimate quality of the product at the output of the encoder. The size (pixel map) of the encoded image, bit depth, encoded bit rate and encoded frame rate are all linked in complex ways to allow the end user the most effective image possible. Some encoders using lower quality decoders suffer from being unnecessarily soft, or do not adequately remove NTSC artifacts from the signal prior to encoding. Others do not maximize the potential quality of the encoding process. Some allow two-pass encoding, analyzing the complexity of the material and then tweaking the encoding to apply the largest number of bits from the target bit rate to the most challenging portions of the content.

As with many things in our industry, buying by brand name has become difficult. Some of the manufacturers of streaming products come from the computing industry and have added video-based products only recently. Others who have roots in video graphics have leveraged their television expertise to move towards Webcentric products. The best advice possible with these products is to be sure the company you pick understands how you run your business, can support you well and has the financial stability to remain in business as the .com bust proceeds.

John Luff is vice president of business development for AZCAR USA.