The FezGuys
Just How Dumb Do They Think We Are?
(A Review of RealAudio's beta G2 Encoder)

[ No. 25 - November 1998 ]

Another Year, Another Tradeshow

The Audio Engineering Society (AES) hosted it's 105th tradeshow, convention and handshake/business-card-exchange boot camp from September 26th - 29th, deep underground in the cavernous Moscone Convention Center in earthquake-prone San Francisco. You could see it all: from 80-channel mixing boards that cost as much as a one-bedroom apartment on the Upper West side of Manhattan, to tiny pieces of metal and plastic guaranteed to improve the quality of your listening experience. There was a moment of confusion when, upon entering the convention center, it was discovered that it is actually two halls, one of which was hosting an Annual Seafood Show. They wouldn't let us within sixty feet of the sushi. Must have been our badges: the color was wrong. The status of attendees is displayed by the color of the stripe at the bottom of the badge. Yellow for press, purple for regular attendees, red for people actually working at the booths, gold for speakers, etc... Ray Dolby was spotted at the counter getting his own badge. A man of the people, his badge had no color stripe. It proves that in a roomful of audio luminaries he's the biggest Rock Star.

It was mostly the usual supects but a couple of Internet-related companies were present. RealNetworks were seen pushing their G2 system from a small booth by the back wall. LiquidAudio, more centrally located and actively staffed, told us they're adding MP3 capability to their suite of tools and making plans to add the new AAC codec (sometimes nicknamed MPEG 4) as it matures. LiquidAudio products have previously been based solely on Dolby Laboratories' AC3 (now called "Dolby Digital") codec. Over in the loud section Waves (the audio software newcomer from Tel Aviv) were sharing info about upgrades of their live multi-stream encoder (soon to support RealAudio's G2), as well as new versions of EQ, limiting and compression software. At the end of an aisle Telos (proud parent of Audioactive MP3-based products) displayed some bitchin' Real Time Audio Encoder hardware. Apparently we no longer need to route through a Windows computer for their encoder to stream audio over the Intenet. This box attaches straight into the 'Net; plug and play all the way. Seems pretty cool...a player can open a connection to it directly, for single, point-to-point streaming and there are several interesting possibilities (other than the still-supported Netshow) to tie it into a multiplex for live broacast to multiple listeners. MP3 is an open standard; basic interface and other software apps could easily be written for it. There was also a new version of their MP3 Encoder software. We're on it, we're on it... we'll do a followup later, we promise. Of course, at the rate this stuff is taking off, we may need to start our own magazine!

This month's review is practically identical to last month's but with rather different results. We auditioned the new RealAudio G2 encoder in a subjective listening test using a Windows NT System (dual 200 mHz Pentium processor), a 13" color monitor (very lo-rez) and common or garden variety, tiny, little, powered computer speakers called Labtec CS-150s. Test tubes, retorts and beakers with mysterious, brightly-colored fluids bubbled late into the night at the secret underground FezLab at Grok House.

As before, we put an analog signal from a portable DAT player (see column #3 on <>) into the computer, used CoolEdit to create the WAV file, and then compressed it, this time into a RealAudio G2 file.

The G2 encoder is found at: <>. Because it's still in beta it's currently free. If they follow suit with their 5.0 release, a stripped-down version will remain freely available. There are still no versions of the encoder or the player available for Macintosh or UNIX users. (Hello? Is anybody at the helm at Real Networks?) Downloading and installing the encoder is simple and to the point. After installation we open up the GUI interface of the encoder and poke around. Our first look brings us up short. Ordinarily there would be a choice of various bitrate settings to facilitate a users' needs based on file size, connectivity and processing power. We can't believe it! There is nowhere to set the bitrate! There is only a menu bar that says: "Target Audience." Apparently we are being asked to make our encoding parameter choices based not on our technical needs but on unknown values set by the manufacturer. This is just wrong. Even after we get over our shock and actually encode a file this interface continues to come up short. It doesn't show us the compression parameters, doesn't offer any information on elapsed time to encode a given file and doesn't tell us the file size.

We sit back from our computer and draw a deep breath. Is this where Real Networks wants to take us? Looks like they've decided what's going to be best. We only have to point and shoot. This must be the Microsoft influence. You know the signature computer behavior: an app doing non-intuitive things you didn't ask for automatically and you can't stop it? Real Networks (the parent company of RealAudio) has been working closely, maybe too closely, with the software Goliath. This beta G2 encoder is offering a letter perfect example of the "dumbed-down interface." Less user options apparently means, to Real Networks, more efficiency. We're stunned. There isn't even a setting to encode for 14.4 modems.

Now maybe this kind of overly-simplified interface might be appropriate for their "Wizard" (or beginner-user) mode but a more accomplished power-user should be able to override it. We define a power-user as someone who has the desire and ability to use an application in a new and creative way beyond the manufacturer defaults. This is what artists do. It's what creativity is all about. If a Stratocaster guitar only allowed a player five notes it would have been quickly forgotten. If a piano only allowed for music in the the C-major scale (the people's key) there would be no Glenn Gould Bach interpretations. Have we made our point? We'd love to override the GUI on this app but that option is not available. This isn't what we'd call "user-friendly."

On to the test portion of our program. As we stated: instead of setting bitrates and sampling rates we are forced to select a "Target Audience" (choices are: 28k, 56k, single ISDN, dual ISDN, LAN/T1 [low and high]). Each "Target Audience" selection then offers a choice of "Audio Formats" based on the content of our source material. The choices are: Voice, Voice with Background Music, Music With Vocals, Instrumental Music and No Audio (for encoding video only). For our purposes we'll stick to files encoded for streaming over standard phone modems, skipping the higher-bitrate compression settings for ISDN and LAN/T1. We'll compare the elapsed encoding time and sound quality of identical WAV files using both the G2 beta encoder and RealAudio's earlier 3.1 version. The source WAV file is a 3 min. 35 sec. song that uses 36.2MB of disk space.

RealAudio 3.1 (with "Show Audio Signal" option turned off for more speed) -

  • At a 4.7kHz sampling rate, 16kbps file, mono (for streaming over 28.8k modems): Encoding takes 53 seconds. Audio quality: sounds barely acceptable but sonically legible.

  • At an 8kHz sampling rate, 32kbps file, mono (for streaming over 56k modems): Encoding takes 55 seconds. Audio quality: sounds a little better but still merely ok.

RealAudio G2 (using the "Music with Vocals" setting in the "Audio Format" menu) -

  • At an unknown sampling rate, 16kbps file, mono (the "28k" setting): Encoding takes 1 min. 10 sec. Audio quality: sounds good and actually better then the 3.1 (the G2 16kHz sounds identical to the 32kHz 3.1 encoded file.)

  • At an unknown sampling rate, 32kbps file, mono (the "56k" setting): encoding takes 1 min. 25 sec. Audio quality: sounds very good.

Based on our observations; a 32kbps RealAudio G2 encoded file sounds better than a 3.1 encoded file at the same bitrate. But we can't tell whether the G2 is playing back in stereo or mono. An accurate comparison of the quality of the two applications is difficult because there's no way to set identical parameters on the G2 system. We had to read the kbps value of a G2-encoded file by looking at the streaming info on the players' GUI while listening back to the song.

Overall the RealAudio G2 encoder sounds pretty darn good, though there are some serious problems with the interface design. We found ourselves visiting the Help menu frequently which, in the arena of audio encoding software, is typically unnecessary. We're aware that this is a beta version but it seems that in Real Network's desire to simplify and streamline their encoding product they've done the opposite. It feels overly complex, somewhat top-heavy (many irrelevant options) and rather totalitarian (many useful and common-sense choices are glaringly absent).

This new G2 codec is based on a "Cook" compression algorythm, a name we admit we're not familiar with. It sounds better then the MP3 encoded files from last month's Xing Streamworks encoder review (see FezGuys column #24 at: <>). Our listening tests showed that the 32kbps encoded Xing MP3 file sounded only slightly clearer than the 16kbps encoded G2 file. The RealAudio 3.1 16kbps encoded file sounds clearer than the Xing MP3 16kbps encoded file. Interestingly, a Xing MP3 32kbps encoded file sounds clearer then the 3.1 32kbps file. And the G2 32kbps encoded file sounds better (good clarity, approaching FM-quality) then the Xing MP3 32kbps encoded file (somewhat thin with slight high-end swooshiness). Each encoder has a unique sonic footprint. On a purely subjective note (is there any other kind?): the Xing MP3 has a more natural, warm sound. The RealAudio G2, however, has a more "digital," or colder, sound to it. The G2 sounds oddly enhanced; as if a virtual surround element has been introduced into the encoding process.

Bottom Line: Internet audio products are evolving quickly and the quality of compressed audio streamed at low bitrates is improving dramatically. It's clear, however, that interface designers would benefit by involving people who actually *use* the product in their developement process. This might provide perspective into how musicians or music-lovers actually think and interact with computers and software. It might improve the way tools are created. All things are possible.

(Note: as we went to press Real Networks introduced a "second beta" [gamma?] release of the G2 encoder. Go to: <>. Of course it's still Windows-only... Natually we will put it through the wringer as soon as we can.)

Letters To The FezGuys

Dear FezGuys - Do you have any experience with Shockwave audio (SWA) streaming? While RealMedia is the defacto standard for audio streaming, there are some attractive media synching features within the SWA model that we may wish to use. But I am concerned about the performance of SWA streaming of audio files (FezGuys note: SWA files are encoded using the MP3 codec and have an extra file header) longer than a few seconds or minutes in length. Thanks - Glen J. Stephan

Dear Glen - ShockWave Audio's (SWA) ability to sync different events within a larger ShockWave interactive piece is a very nice capability. RealNetworks does have some ability in this as well, and in fact there are ties into it from with the ShockWave architecture. Nonetheless, using SWA for streaming long clips is no different than streaming short clips. Encode the audio at the bitrate that suits your need (i.e.: 16kbps encoded stream for 28.8k dialup users, etc...). ShockWave also features the ability to include sound clips in their entirety for downloading. Keep these non-streaming files as small as possible in order to keep the download to the user short as well. - Happy Encoding! The FezGuys

The FezGuys welcome your comments.



About the authors:

Jon Luini is a working technophile, a musician (bass player/singer) with full-blown facility and extensive experience on the Web and no free time. He is a co-founder of IUMA and MediaCast, co-creator of Addicted To Noise, and runs an Internet and music consulting and technology company, Chime Interactive (formerly Evolve Internet Solutions). <>

Allen Whitman is a working musician (bass player/singer/producer) with a keen, real-world interest in the practical use of the Web. Music credits include: The Mermen, "Brine-The Antisurf Soundtrack, biL, Deep Field South, Doormouse, Delectric and Drizzoletto. He has written for the San Francisco Examiner, Wired, EQ, Revolution, Yahoo Internet Life, Prosound News, Surround Professional, Replication News and others. <>

©1996-2003 The FezGuys™