Professional webcasting (Designing Web Audio)

6.3. Professional webcasting

While simple HTTP pseudo-streaming is adequate for small-scale broadcasting purposes, it is not a professional solution for large-scale broadcasting. If audio is a mission-critical component of your web site or simply needs to be 100% reliable, use a dedicated RealServer. This section provides a step-by-step guide on how to prepare and serve professional-quality RealAudio.

The first step to professional RealAudio broadcasting is to capture the highest-quality RealAudio source material. If you are using RealAudio in a professional situation, you need to maintain broadcast-quality production techniques.

To achieve the best fidelity:

Start with a high-quality digital audio tape or CD when transferring audio to your computer.

Set the input levels into your sound card as high as possible without distorting.
Remove all unwanted noises and sound artifacts in your original source files with a sound editor.
Make sure to start with a stereo 44.1 kHz 16-bit audio file when encoding to lower-bandwidth audio codecs.

To create the best quality RealAudio clips:

Normalize your sound files to 95% of the maximum dB level.

If your clips do not have enough bass or treble, try boosting the low-end or high-end frequencies in your sound file before encoding to a particular codec. For bass, use EQ to increase the low-end frequencies in the 100 to 200 Hz range. For treble, boost the high-end frequencies from 5,000 to 8,000 Hz. Remember, anything higher than 8,000 Hz will get cut when you compress the file using a low-bandwidth setting. For voice, the sweet spots to boost are 500 Hz and 3,500 Hz. Boosting the bass frequencies for voice below 250 Hz may have little effect or may cause distortion of the voice.
If you are encoding for low-bandwidth codecs (14.4 and 28.8), use a low-pass equalization filter to cut frequencies above 10,000 Hz. (For a comprehensive overview on how to capture and edit high-quality audio, see Chapter 3, "Capturing Original Source Material" and Chapter 4, "Optimizing Your Sound Files".)
Avoid flat or dull-sounding RealAudio clips by adjusting the amplitude of your signal with a sound editor instead of recording directly into the RealEncoder. Recording into a sound editor maximizes the available dynamic range.
Prevent signal clipping by making sure that loud sounds in your original audio file signal do not exceed the acceptable amplitude levels. Clipping or "going in the red" can give rise to harmonic distortion in the form of clicks or crackling sounds on playback due to the introduction of odd harmonics. If your source file contains a clipped signal, your final RealAudio or RealVideo file will have high-frequency background noise or static. Lowering the input volume will help reduce clipping.
Limit excessive low-frequency noise by eliminating any DC offset. (DC offset occurs during the analog to digital conversion process when the 0-point in the waveform isn't actually at but offset either above or below. AnalogX (http://www.analogx.com) offers a Direct X plug-in that corrects DC offset errors. The AnalogX DC Offset plug-in is compatible with any Direct X application such as CoolEdit, Cakewalk, WaveLab, Paris, etc. DC offset is commonly produced in cheap PC sound cards. Several editing applications have built-in settings for fixing DC offset.
Ensure proper grounding in audio equipment to prevent AC 60 Hz line noise. If AC line noise is present in your audio files, you can partially remove it with EQ by lowering the frequency range that contains the noise. You can also try using special noise reduction software to perform this task.

In low-bitrate encoding, every bit of bandwidth is important. Every bit wasted on noise reduces the number of bits allocated to the desired content, thus reducing the quality. Noise includes any audio information other than the desired content such as background noise, distortion, quantization noise.

6.3.1. Selecting the right RealAudio codecs

Once you have captured or imported a high-fidelity sound file into the RealEncoder, you need to select which codecs to use when exporting duplicates of the file. Choosing codecs can be a confusing process. With the release of RealSystem G2, there are six RealPlayers and dozens of RealAudio codecs in use. Compounding the problem, connection speeds vary greatly across the Internet.

Although RealSystem G2 simplifies the encoding process with its "encode once, stream anywhere" SureStream technology, it will take a considerable amount of time for the millions of people still using older players to upgrade to the new G2 player.

To make sure your audio is available to the widest possible audience, include several duplicate audio clips encoded at different bitrates. Generally, three or four codecs -- such as 20 Kbps mono, 56 Kbps ISDN, and stereo ISDN -- are adequate. Bandwidth negotiation is a RealServer feature that automatically detects the bandwidth of a listener and then streams the appropriate RealAudio clip. To use bandwidth negotiation, you need to select several RealAudio codecs.

To determine which codecs to use, remember that the cumulative bitrate of all files in your RealMedia presentation should equal no more than 75% of your target bandwidth. For example, if your target audience has standard 28.8 Kbps modems, your presentation bitrate should be 20 Kbps. Table 6-1 shows the optimum streaming rate for each target connection speed.

Table 6-1. Suggested maximum bitrate for different connection speeds

Target connection speed	Suggested maximum bitrate for streaming files
14.4 Kbps	10 Kbps
28.8 Kbps	20 Kbps
56 Kbps modem	34 Kbps
56 Kbps ISDN	45 Kbps
112 Kbps ISDN	80 Kbps

When choosing a RealAudio codec, keep in mind that different types of sound require higher-bandwidth codecs due to their frequency range. As discussed in Chapter 2, "The Science of Sound and Digital Audio", higher-pitched sounds need a higher sampling rate to be accurately reproduced. Most popular music recordings have a frequency range between 60 Hz to 16,000 Hz, whereas speech primarily resides in the 200 Hz to 5,000 Hz frequency range. Thus, 28.8 Kbps RealAudio clips can reproduce frequencies only under 4,000 Hz to 5,500 Hz. Accordingly, a 28.8 codec can reproduce speech quite well but not music.

You will need a higher bitrate codec for high-quality music broadcasting than for simple speech. Table 6-2 shows the low- to medium-bandwidth encoding options and lists which codecs the various RealPlayers support. Make sure that some of the codecs you use support older 2.0 player versions.

The table also lists each codec's frequency response. The higher the frequency response, the better the audio quality. Notice that the newest G2 codecs produce a much higher frequency response resulting in greatly improved music playback than older codecs. To view the full table, go to the RealAudio 5.0 Content Creation Guide.

Table 6-2. RealAudio's bandwidth codec chart (from low to medium)

Audio Codecs Supported	Players Supported	Frequency Response	Comments on Codecs
5 Kbps Voice	G2, 5	4 kHz
6.5 Kbps Voice	G2, 5, 4	4 kHz
8 Kbps Voice	G2, 5, 4, 3, 2,	4 kHz	Superceded by 8.5 Kbps Voice Codec
8 Kbps Music	G2, 6	4 kHz	Generation 2 Codec
8 Kbps Music	G2, 5, 4	4 kHz
8.5 Kbps Voice	G2, 5, 4	4 kHz	DolbyNet Codec
11 Kbps Music	G2	5 kHz	Generation 2 Codec
12 Kbps Music	G2, 5, 4	4 kHz	DolbyNet Codec
15.2 Kbps Voice-Mono	G2, 5, 4, 3, 2	4 kHz	Superceded by 16 Kbps Voice Codec
16 Kbps Voice-Mono	G2, 5	8 kHz	Highest bitrate Codec for voice
16 Kbps Music-G2 Mono	G2	8 kHz	Generation 2 Codec
16 Kbps Music-G2 Low Resp.	G2, 5, 4	4 kHz	DolbyNet Codec
16 Kbps Music Mono Medium Resp.	G2, 5, 4	4.7 kHz	Suitable for Pop/Rock music DolbyNet Codec
16 Kbps Music Mono High Resp.	G2, 5, 4	5.5 kHz	Classical Music -- DolbyNet Codec
20 Kbps Music-G2 Mono	G2	10 kHz	Generation 2 Codec
20 Kbps Music Stereo	G2, 5, 4, 3	4 kHz	DolbyNet Codec

6.3.2. Using the RealEncoder

There are several methods and tools for converting standard WAV, AIFF, or SND audio source files into the appropriate RealMedia .ra and .rm formats. The cheapest and simplest way to encode files is either to use the free RealEncoder available on the RealNetwork's web site or to export your files from an editing and recording application, such as CoolEdit, that supports .ra and .rm file extensions. The better alternative for .ra and .rm file conversion, if you can afford an extra few hundred bucks, is to usea batch conversion application such as Wave Convert Pro by Waves or BarbaBatch by Audio Ease. See Appendix A, "Creating the Ultimate Web Sound Studio: Buyers Guide and Web Resources" for more information about these applications. These batch conversion tools feature superior processing and file optimization capabilities and will make mastering your source files into high-quality RealAudio clips a much easier process. Here is a description of the options in the RealEncoder 5.0:

File Information: In the Properties section of the RealEncoder window, you can enter title, author, and copyright information for your output file. This information is displayed in the player.
Mobile Playback

Setting the proper timeout values

Before batch-encoding, make sure to change the default timeout value from 60 seconds to a greater value -- such as 9,000 seconds -- to ensure adequate time for encoding all your files.

6.3.2.1. Encoding RealAudio with a batch processor

Running files through the RealAudio encoding algorithm is a time-consuming process. If you have to process dozens or hundreds of sound files at a time, a batch processor application can save hours and days of work. Wave Convert Pro for Mac and PC and BarbaBatch for Mac are great batch conversion and mastering tools. Wave Convert Pro and BarbaBatch will import an entire folder of audio files, normalize them, perform EQ enhancements, and convert them to RealAudio clips on the fly.

Video capture cards

Looking for the best video capture card? The Osprey 100 ($199) for Windows NT is the most highly recommended one on the market. In fact, it is the card RealNetworks uses for all their live video capture needs. If you have a reasonably good sound card installed, the audio sync is as good or better than more expensive cards with on-board audio, according to RealNetwork engineers. To capture video from a VHS tape, first capture the footage to an uncompressed AVI file, then compress it in non-real-time. (This takes large amounts of disk space, however.) You can find the Osprey 100 at http://www.real.com/products/tools/.

Most video capture cards have a 2 GB AVI file size limitation. Some video card manufacturers purport to have solutions to break this barrier. However, the SGI is the only platform that truly supports 2 GB+ video file formats. RealNetworks has an encoder for the SGI IRIX OS, which in conjunction with the video file format, offers the best solution for long format video clips.

Advanced users can use the command-line RealAudio utilities for batch processing of multiple files. Within the DOS command line, you can automatically encode several files sequentially by using the following syntax:

RVBatch RVEncode.exe options

Table 6-3 lists the command-line batch encoding options in the Rvencode directory.

Table 6-3. Command-line batch encoding options

Option	Description (default in parentheses)
`/I`	Use this option to specify an Input File.
`/O`	Use this option to specify an outfile or directory -- Output File Name or Directory (infile.rm or dir\YYYYMMDDHHMMSS.rm).
`/L`	Use this option to specify Use Live Input.
`/S`	Use this option to specify `server[:port]/file` -- Server Name, Port, and File (Port 7070).
`/W`	Use this option to specify password -- Server Password.
`/D`	Use this option to specify hhh:mm:ss -- Maximum Encoding Duration (continuous).
`/A`	Use this option to specify an Audio Codec (0).
`/V`	Use this option to specify a Video Codec (0).
`/F`	Use this option to specify a frame rate -- Frame Rate (Optimal).
`/B`	Use this option to specify total Kbps for clip (100).
`/N`	Encoding Speed range 1 to 5, where 1 = normal, 5 = fastest (1).
`/M`	Optimal Framerate Bias, where 1 = sharpest image, 3 = smoothest motion (2).
`/Q`	Use this option to specify Quality, 1 to 100 (100).
`/T`	Use this option to specify a Clip Title.
`/U`	Use this option to specify a Clip Author.
`/C`	Use this option to specify a Clip Copyright.
`/R`	Use this option to enable Selective Record. Valid options are Disabled and 1 Enabled (1).
`/K`	Use this option to enable Mobile Playback. Valid options are Disabled and 1 Enabled (0).

Macintosh users can use AppleScript to batch-encode audio files. Table 6-4 describes the encoding parameters for Macintosh. These can also be obtained by opening the RVBatch Dictionary installed with the RealEncoder.

Table 6-4. Macintosh encoding parameters

Parameter	Description
encode	Use this option to specify an Input File.
output	Use this option to specify an FSSpec or full pathname of the output file.
audio	Enable audio encoding.
using audio codec	Audio codec name.
video	Enable video encoding.
using video codec	Video codec name.
at frame rate	Frame rate.
optimal bias	Determines the behavior of optimized encoding and can be one of the following options: sharpest image, normal, or smoothest motion.
cropping	Sets the cropping rectangle and follows this form: top, left, bottom, right.
at bitrate	Total bitrate.
at quality	Video quality setting (1-100).
encoding speed	Speed of encoding -- the faster the speed, the less the quality of the encoding. Can be one of the following options: normal, medium, fast, faster, or fastest.
title	Title string.
author	Author string.
copyright	Copyright string.
Mobile Playback	Enable Mobile Playback.
Selective Record	Enable Selective Record.

Preparing and encoding audio for RealVideo broadcasting

Audio quality is the biggest tradeoff when preparing content for low-bandwidth RealVideo. Because video is so data-heavy, only a small percentage of the bandwidth can be allocated to audio. For most RealVideo broadcasts, better picture quality takes priority over audio quality.

At Fox News, RealVideo clips generally contain shots of newscasters talking. Stationary video footage of "talking heads" are ideally suited for low-bandwidth RealVideo encoding. If there is little motion in the video feed, you can set the ratio settings in favor of video quality while still preserving speech intelligibility.

According to David Dadekian, technical director for Fox News, the RealVideo 20 Kbps stream allocates 6.5 Kbps to audio and 13.5 Kbps to video. The 50 Kbps Fox News stream allocates 8.5 Kbps to audio and 41.5 Kbps for the six frames per second video. Devoting a higher ratio of bandwidth to video preserves picture quality. Raising the audio portion of the audio-to-video data ratio degrades the respectable 176-by-144 pixel video quality.

For some content, increasing the ratio of video bandwidth will not help picture quality. High-motion video, such as sports footage with rapid action or music videos with dancing, are not suited for low-bandwidth codecs. For instance, video clips of regular season NBA basketball highlights encoded with the same settings used for Fox News footage resulted in indistinguishable blurry pictures. Reducing the frame rate of the basketball reel produced a higher-quality picture.

The best approach for encoding sporting events is to create a slide show with the video set to one image per second and the rest of the bandwidth allocated to audio. You do not get true video action, but at least you get high quality stills.

Encoding video content with music requires a higher ratio of bandwidth dedicated to audio. Music content is generally unacceptable at the audio rates of 6.5 Kbps and 8.5 Kbps used by Fox News. When David encodes RealVideo clips with music content for TV Guide Entertainment Network, he sets the audio at a minimum of 15 Kbps. For a 50 Kbps stream, he sets the audio at 32 Kbps and the video at 18 Kbps. Setting the bandwidth ratio at 32 Kbps/audio and 18 Kbps/video produces an engaging slideshow of one image per second with good music quality.

Once you have encoded your source material using the appropriate RealAudio codecs, you are ready to stream your RealMedia files with the RealServer.

6.3.3. Broadcasting audio with the RealServer

To broadcast audio with the RealServer, you can either install your own RealServer or contract with an outside service provider to host your RealMedia content.

If audio is a small component of your web site, you probably want to consider installing a 25-stream Basic RealServer available for free at the RealNetworks web site. If you need more streams or advanced features, you will need to consider one of RealNetwork's commercial server options, which range from $695 to $30,000, depending on the range of simultaneous streams you need and extra features you want, such as live broadcasting, RealFlash broadcasting, and commerce functionality. Visit RealNetworks' Servers and Solutions page at http://www.real.com/solutions/servers/index.html.

If audio is an integral part of your web site and you broadcast to a large audience of more than 1,000 simultaneous listeners, consider using a third-party hosting service that guarantees superior reliability, fault tolerance, load balancing, and a number of simultaneous streams. For example, the RealNetworks Broadcast Network can support up to 50,000 simultaneous unicast streams. You can find a list of RealMedia-qualified ISPs at http://www.real.com/solutions/partners/isp/index.html.

6.3.3.1. RealServer installation

Installing and configuring a RealAudio Server is a four-step process that entails:

Downloading the appropriate platform-specific server package from RealNetworks
Running the automatic installation process
Launching the server
Testing the server to ensure audio playback

One of the strong points of the RealSystem is easy automatic installation for a variety of popular platforms including Unix, NT, Windows 98/2000, and Macintosh. Once you have purchased and downloaded the appropriate RealServer package, run the automatic installer application to configure the settings for your platform. The RealServer installer asks a series of simple questions regarding personal information, directory names, and default port setting options, then automatically installs itself on your system with the appropriate settings.

To install the RealServer on a Unix machine, create a separate directory to hold all of your RealAudio clips before launching the application. Do not place the RealServer under the root directory; instead create a separate directory with its own user ID to prevent access to your machine (Unix). Once you have placed the RealServer and all of its corresponding files into a separate directory, launch the server from the command line.

Make sure to type in a command in the boot script so that your machine will launch the RealServer whenever the computer is re-booted.

Finally, make sure you test your RealServer by connecting a web browser via HTTP to the RealAudio port 7070 (http://yourservername:7070). This port has a link to pre-installed stock audio samples for testing, so you don't have to encode your own test file. The RealServer diagnostic test detects that you are making an HTTP request rather than a standard RealAudio request and generates system reports on the fly. If you hear the audio playing back, you should be ready to stream RealAudio.

6.3.3.2. RealServer system requirements

To run your own RealServer, you need the following items. (For more detailed information, refer to the RealAudio Server documentation located at the RealNetwork's Developer Zone.)

A computer running one of the supported operating systems: Macintosh System 8 or greater, Digital Unix v3.2, Microsoft Windows NT 4.0, Solaris 2.5, Linux 2.03, Sun OS 4.1.x, Silicon Graphics IRIX 6.2 and 6.3.

Space on the computer's hard disk for the RealServer software and RealAudio clips. You need upwards of 1 GB or more of hard disk space if you are serving high-bandwidth streams with bandwidth negotiation. One 6- to 10-minute audio file encoded with a range of high- and low-bandwidth codecs for bandwidth negotiation requires 15 MB to 20 MB of disk space. If you have disk space constraints, stick with low-bandwidth codecs.

A network connection with sufficient bandwidth to serve your users. For each listener connected to the Internet backbone, the RealServer requires at least 10 Kbps for 14.4 format and 20 Kbps for 28.8 format. A T3 line, for example, could stream 4,500 simultaneous streams. See Table 6-5 for the number of approximate streams for each bandwidth connection.

Table 6-5. Number of streams at different connection speeds

Network connection

Number of 14.4 Kbps streams

Number of 28.8 Kbps streams

ISDN (128 Kbps)

12

8

T1 (1.5 Mbps)

150

90

T3 (45 Mbps)

4,500

2,700

A web server that supports configurable MIME types. RealServer works with the following web servers:

Website for Windows NT

Apache 1.1.1 and higher

Mac HTTP

HTTPD4Mac

Netscape Netsite and Netscape Enterprise Server

Microsoft Internet Information Server (IIS)

NCSA HTTPD (v 1.3, 1.4, 1.5)

CERN HTTPD (v3.0)

EMWAC HTTPS 0.96

Spinner 1.0b12-1.0b15 / Roxen 1.0

Webstar and Webstar PS

Sufficient processor and memory capacity. RAM, not processor speed, is the gating factor for the computer running RealServer. A 90 MHz Pentium processor, with enough available bandwidth, can simultaneously deliver at least 500 28.8 streams. For example, Real Audio Server 3.0 requires 20 KB RAM for each simultaneous stream. A stream of 60 simultaneous audio files requires approximately 4 MB of available RAM. 1,000 streams requires 70 MB of RAM, and so on.

6.3.3.3. Using bandwidth negotiation

Once you have installed the RealServer or found an ISP running one, use bandwidth negotiation to manage the delivery of RealAudio content. Bandwidth negotiation checks the listeners' modem speed settings then sends the appropriate stream for their bandwidth. If you encode three or four versions of your audio clip, you can reach a wider audience by offering low-bandwidth users a smaller compressed RealAudio clip and high-bandwidth users a CD-quality clip. Note that when bandwidth negotiation checks the connection setting of the client player, the player reports the speed at which it has been set, not the speed of the actual network connection.

In most cases, bandwidth negotiation works fine. On occasion, however, the RealServer overestimates a RealPlayer's "real" connection speed. Many web developers run into problems because the default setting of the RealPlayer is 28.8. This setting may not be optimized for the user's available bandwidth. Frequently, RealPlayer bandwidth preferences are set too high for the actual available data throughput. If a user clicks on a link to a RealMedia clip that streams at too high a data rate for the user's real connection speed, a somewhat cryptic error message will appear indicating that the network connection or CPU speed is insufficient to view the content.

Selecting the right TCP/IP port

If the RealServer is configured on the same machine and IP address as the web server, the web server's TCP port must be changed. The RealServer uses the standard TCP Port 80 for its "cloaking," so the web server has to be relocated to Port 81 or any other available TCP port. A single machine also could have two IP addresses with the web server bound to the first IP address and the RealServer bound to the second.

RealSystem G2's SureStream technology promises to alleviate this problem. The RealPlayer G2 "switches down" to a lower bandwidth without rebuffering the clip when a user's Internet connection degrades or when the user's bandwidth was misjudged. RealEncoder G2 with SureStream generates one master file that includes several different bandwidth settings. In the future, as older players get upgraded, the G2 Encoder will make it easier to reliably stream RealAudio content.

If you are using RealServer 5.0, include a separate file for 28.8 Kbps modems, 56 Kbps modems, and ISDN connections along with one or two files for older players. For example, if you want to encode an important tradeshow presentation so people with 14.4, 28.8, and 56 Kbps modems can listen, encode the live audio to three different files or capture the audio to a .wav first and then encode it three times.

To stream a presentation using bandwidth negotiation, follow these steps:

Encode your audio content into at least three separate files:

14.4: Encode one file at 5 Kbps, 6.5 Kbps or 8.5 Kbps
28.8: Encode a higher-bitrate file with a maximum of 20 Kbps
56: Encode a final audio clip with a 40+ Kbps stereo or mono codec

Create a generic directory (not a file) in the RealServer Content directory: C:\real\server\content\tradeshow.ra.
Place the files generated in step 1 into that directory. For example:

C:\real\server\content\tradeshow.ra\tradeshow_14.ra

C:\real\server\content\tradeshow.ra\tradeshow_28.ra

C:\real\server\content\tradeshow.ra\tradeshow_56.ra

Change the filenames to the following (as listed in the Content Creation Guide):

C:\real\server\content\tradeshow.ra\pnrv.18

C:\real\server\content\tradeshow.ra\pnrv.36

C:\real\server\content\tradeshow.ra\pnrv.43

On your web page, reference the files as pnm://yourservername/tradeshow.ra.

Be sure to add RealAudio 2.0 codecs for people using 2.0 players. Use RealEncoder 3.0 to create RealPlayer 2.0 compatible codecs. Remember, you will have to manually name the files generated by the RealEncoder 3.0 and place them in the proper folder.