Today's Transistor Radio™
Now the First to Play Adobe® Flash® Audio Streams

HIGH FIDELITY STREAMING AUDIO BASICS
Credibility Department with Geek Information

Our years of experience in the professional broadcast audio field has taught us that there is a serious lack of understanding of analog and digital audio fundamentals and quality among those in the computer programming and information technology fields. To refine streaming technology, audio and computer professionals must learn from each other. Mastering the details of audio streaming requires practitioners to know computer networking as well as analog and digital audio. This includes thorough knowledge of how and why professional audio signal processing and sound cards are used for these applications to achieve professional results. To create quality audio that will attract and hold audiences, as professional broadcasters have done since the beginning of commercial radio, audio and computer experts must collaborate effectively.

Consumers expect higher performance from new technology. Satellite Radio and HD Radio are shaping up to be viewed as examples of how new technology is not necessarily meeting consumer expectations. They might be thought of as bridge technologies, closing the gap between terrestrial and Internet delivered radio, just as Mini-Disc bridged the CD to downloads. To make streaming audio a successful and viable business, content providers, player devices, and player applications must be able to provide audio fidelity that is at least as good as FM radio. This is the minimum acceptable “entertainment grade” audio quality that can compete with high fidelity downloads and be acceptable to consumers. If content providers expect their listeners to take them seriously, content providers need to take their listeners seriously and provide quality streams.

An audio codec is the engine that drives streaming audio. Without a good sounding codec, there is no chance to achieve a good sounding stream. All audio codecs are not created equal. There can even be audible differences between different implementations of the same audio codec specification. A given implementation may play, but it may not be working as well as it can. That is why audio encoders and decoders, players, should be chosen very carefully. To give your audience the best possible listening experience, these decisions should be made on pure technical merit.

Audio codec performance evaluation is tricky. The ultimate criterion is how they sound, which can only be determined by well-controlled subjective tests.

“High-quality” codecs are intended to be audibly transparent and serious researchers usually evaluate their level of transparency via so-called “ABC/HR” (“double-blind triple-stimulus with hidden reference”) subjective tests using panels of human listeners according to ITU-R Recommendation BS.1116-1. This Recommendation is intended for use in the assessment of systems that introduce impairments so small as to be undetectable without rigorous control of the experimental conditions and appropriate statistical analysis.

“Intermediate-quality” codecs are not necessarily intended to be audibly transparent. Instead, they are usually assessed by their ability to deliver sound free from annoying, unnatural-sounding artifacts. A technique called MUSHRA (“MUltiple Stimuli with Hidden Reference and Anchor”), described in ITU-R Recommendation BS.1534-1, is most commonly used for subjective testing.

When assessing providers’ claims, it is important to determine whether claims were based on ITU methodology or on some independent and unverifiable testing technique. Many high profile codec providers have made unverifiable claims that are not based on reputable testing and evaluation procedures.

Objective testing using test instruments can be a useful adjunct to subjective listening tests. As part of a first-pass evaluation, spectrum analysis is useful in qualifying codec performance by displaying the codec’s spectral signature. This analysis is useful for discovering in-band spectrum discontinuities, and/or aliasing, and/or sample rate conversion artifacts. However, spectral analysis alone is insufficient to characterize a codec’s performance completely. Although multiple codecs may have the same or very similar spectral signatures, they may have very different sonic performance.

The following spectrum plots were made from streams provided by various high profile content providers currently available on the Internet. There is even a stream claiming “CD-quality,” which would require at least 1500kbps! However, “Near CD-quality” can be achieved using an optimized AAC codec at 128kbps or higher. (Using the ITU-R Rec. BS.1116-1 procedure, AAC at 128 kbps has tested as “transparent” to CD sources with most program material.) A quick examination indicates why so many streams suffer from “entertainment grade” audio quality—they chop off or otherwise mangle the high frequency content that is required for full-fidelity sound reproduction. These streams don't even pass the first pass spectrum analysis test, much less the sonic test. There is also an FM and CD audio spectrum for reference.

The Apple hardware is a very high quality portable player device. Give your audience what they deserve, and deploy HE-AACv2 streaming. This is truly changing the way Internet streaming is perceived and consumed. Both full featured file and streaming encoders are available from Orban. They support several different streaming server protocols, as well as real time PAD metadata injection. More information can be found here: http://www.orban.com/products/streaming/