How to stream
Jan 1, 2001 12:00 PM, By Chuck Leavens
As the Internet continues to play a larger and larger role in our lives, the role of the radio station must change to suit the times. With the introduction of audio streaming, listeners can access stations around the world. For a minimal investment, we can grab those listeners. Radio can take advantage of this new medium with two types of streams: static and live.
Static streaming takes a pre-recorded sound file and delivers it on demand. The file is recorded, then encoded and archived on the server hard drive. Good examples are demos, feature length programs or archived programs. This type of streaming does not require the use of a dedicated audio stream server. It can be streamed right off of your Web server to the client player. Some players begin playing the file before it is done downloading (quickstart), which is helpful for large sound files. Static streaming is supported by a variety of common audio file formats including .WAV, RealAudio, Quicktime, .WMP and .MP3.
Live streaming is a continuous audio feed that is always being encoded and delivered. The transmission side of a live streaming setup requires two main pieces of equipment: the encoder, which converts the audio signal to the delivery format, and the server, which accepts the single stream from the encoder and distributes it to the end user's client software.
Choosing a system
There are a growing number of streaming systems competing for dominance in the marketplace. Choose a system that compliments your audio and has players for every possible combination of hardware. Any system you choose will sound good at high bandwidth, but the trick is to choose a system that sounds good for low-speed users. It may pay to optimize your audio stream to the lowest common denominator.
Once a system has been chosen, decide how your stream will get onto the Net. There are a variety of ways in which to stream live audio.
- 1) Encode the audio at your facility and send the output of the encoder to a streaming provider. This is usually done with an encoding computer at your facility and a fixed Net connection � either a DSL or fractional T-1 to the provider. If you send your stream to a provider via DSL, remember that Net congestion and outages can affect streaming to all of your users. Some streaming providers insert advertising or picture clips in the player before the audio plays. Read your agreement carefully to be sure any ad insertions are appropriate for your station.
- 2) Encode the audio at your facility and send the output of the encoder to an in-house server. This requires at least a full T-1 access to the Internet. Top level access, providers that give direct access to the Internet backbone, is available through carriers such as UUnet, Verio, Digex, AT&T CERFnet and Exodus.
- 3) The third choice is to send your audio to a facility or ISP and have them handle the whole thing. This was the most common way to stream in the beginning. In many cases, a tuner was placed at the ISP and streamed directly to the Net. This method is the simplest, but a tuner feed of your air signal is usually a disaster unless you have a classical station with little to no processing on-air. Limiter by-products that are of little consequence to the air signal wreak havoc with the encoder. The extra harmonics and information cause distortion and noise in the decoded audio stream.
Nuts and bolts
Let's go through an on-site streaming system with Apple's Quicktime, and RealNetworks' Real Audio 5.0 and G2. These are the three systems we serve at WDUQ.
Prepare the audio with proper processing and equalization. Processing for the Net is different than processing for on air. Avoid traditional air-chain limiting and processing because heavy limiting does not work well with the encoder algorithm. Equalization ahead of the AGC is important, not to change frequency response, but rather to remove the frequencies that the encoder cannot pass. By placing the EQ ahead of the AGC, the AGC will not react to the frequencies that will not be heard anyway.
Listen on the player that your listeners will most commonly use. Adjust the processing while listening to a decoded signal. You want to hear the effects of the algorithm path. Hear what the user hears, just like adjusting air chain processing while listening carefully off air. There can be a considerable delay between the encoder input and audio decoder output, so take your time with adjustments. Many new processors are designed just for Internet encoding and have the settings and processing power needed for clean streaming.
Setting up an encoder
PCs and Macs work well as encoders. PCs should have a fast, modern CPU (Pentium class and at least 128MB RAM), a good sound card and network interface card (NIC). Older PowerPC Macs work well (128MB RAM and 100MHz PowerPC minimum). We encode on leftover Macintosh 8100s. Since Macs have very good on-board audio, a separate sound card is not needed. We use RealProducer software for our RealAudio streams and Sorenson Broadcaster with a Q Design PRO codec for our Quicktime streams.
The bit-rate speed is set at the encoder. Setting the speed too high will prevent modem users from receiving your stream. Setting it too low degrades audio quality. While cable modems and DSL are becoming more popular, a typical radio station will serve most of its users by POTS modem. Most of these users are still connecting at 28.8- to 33kb/s despite the availability of 56k modems. Some encoders, such as RealAudio's G2, are able to provide both higher speeds to users who can handle it and lower speeds to modem users. This is a great advance for streaming and really helps give you better choices.
Setting up the server
You have a choice of several server platforms. We chose UNIX because it is a mature, reliable and well-established operating system. I obtained a refurbished Sun Microsystems workstation for $50 and went to work. I was amazed to see how the older, smaller and slower 50MHz UNIX machine could handle the Real 5 server without being overtaxed. We later upgraded to a larger multiple processor, a Sun SparcServer1000E. We also added the Quicktime server and Realaudio G2. One workstation runs all three by setting the input and output control to different ports for each service.
I recommend downloading the free versions of the server software to determine your needs and get a sense of how many users you will have. RealNetworks offers a free 25-user G2 server, and Apple offers a free 1,000-user Quicktime server. Buying a streaming server license based on concurrent users may not be wise because it may be overkill. It does not take much traffic to saturate a T-1.
Chuck Leavens is director of engineering of WDUQ-FM, Duquesne University, Pittsburgh.