How music got free

CHAPTER 4

Their state funding running out, the Fraunhofer team traveled to industry trade shows across Europe and America to promote the mp3 standard. They had a customized booth, with brochures and demonstrations of the technology, but there wasn’t much interest. Struggling to attract potential customers, they kept hearing the same thing: the mp3 was “too complicated.” Meanwhile, across the trade show floor, the mp2 booth was three times the size of their own, and mobbed. Philips had done its job well, dumping promotional money into its own product while undermining the competition.

In head-to-head listening tests the mp3 remained superior. Only Fraunhofer couldn’t get anyone to participate in such tests anymore—MPEG had run those competitions, and everyone knew the results. Standardization of computer hardware had made team member Harald Popp’s expertise less relevant, so Brandenburg reassigned him to sales. In his pitch, Popp told potential customers about the mythology of the “complexity problem” and about the “political” nature of the MPEG decision, but some of his explanations sounded more like excuses.

They were saved in the end by a guy named Steve Church. Grill had first met him at a trade show in Las Vegas the previous year. The CEO of a start-up called Telos Systems, Church was a former radio talk show host and studio engineer who saw a market for improving the quality of audio broadcasting. Like Brandenburg and Grill, he didn’t trust MPEG, as he had seen these “impartial” standards committees make biased decisions before. He agreed to an independently refereed head-to-head listening test between the mp2 and mp3, and was startled by the results.

The mp3 was way better! Shortly after the demonstration, Church called back to the home office in Cleveland and arranged to repeat the experiment over a newly installed digital telephone line. The demonstration material was an encoding of Steely Dan, a band as beloved in Ohio as it was in Bavaria. Telos became the mp3’s first—and for some time, only—enterprise-scale customer. Church commissioned several hundred mp3 conversion boxes called Zephyrs, the size of VCRs, capable of streaming mp3 audio in real time. He then turned around and licensed these to his biggest customer: the National Hockey League.

Here, finally, was a stroke of good fortune. One of the key reference materials in Bernhard Grill’s menagerie of exotic sounds was a recording of a German-league professional hockey game. The sound of scattered clapping had always been a challenge for the encoder, particularly when set against a dynamic soundscape of scraping skates and brutal, bone-crushing checks. The sample was a small snippet of on-ice action, followed by a few seconds of indifferent applause. Grill had listened to it hundreds of times, isolating the encoding errors and working with Brandenburg to implement fixes. The NHL was the perfect customer: the mp3 had been specifically calibrated to the sound of the game.

But the league had certain technical requirements, and these took months to meet. By the time the units finally shipped in late 1994, the hockey players had gone on strike. That year’s shortened season didn’t officially begin until January 20, 1995—the official start date of the mp3 revolution in North America. The fastest game on ice was not widely understood to be a pioneer in digital acoustics, but as the first puck dropped on center ice that year, fans of the Blackhawks and the Red Wings were an unwitting audience on the cutting edge.

It wasn’t until after the 1995 decision in Erlangen that income from the sales finally began making its way to Fraunhofer, arriving just in time to save the mp3 team. The Zephyr racks allowed radio broadcasters to save thousands of dollars an hour on satellite transmission costs, and were installed in every pro ice arena in North America. Telos’ revenues quadrupled, and Steve Church became a zealous advocate for the technology. Soon he was in talks with every major North American sports league. But Fraunhofer received only a small cut. The licensing agreement they’d negotiated with Church charged on a per-unit basis, and there were only a few hundred stadiums to sell to. The mp3 was alive, but on life support; to earn substantial profits, the technology would need many more licensees.

For Brandenburg, that meant a continued push for the home consumer. Earlier in the year, he had directed Grill to write a PC application that could encode and play back mp3 files. Finished within a few months, Grill dubbed it the “Level 3 encoder,” or “L3Enc” for short. The program fit on a single 3.5-inch floppy disk. L3Enc represented a new paradigm of distribution, one in which consumers would create their own mp3 files, then play them from their home PCs. For the home audio enthusiast, the requisite technology was just arriving. Introduced in late 1993, Intel’s powerful new Pentium chips were the first processors capable of playing back an mp3 without stalling. Plus, the new generation of hard drives was enormous: with storage capacity of nearly a gigabyte, they could store almost 200 songs. The biggest limitation was still the encoding process. Due to MPEG’s forced inclusion of the cumbersome MUSICAM filter bank, even a top-of-the-line Pentium processor would take about six hours to rip an album from a compact disc.

No one at Fraunhofer quite knew what to do with L3Enc. It was a miraculous piece of software, the culmination of a decade of research, capable of taking 12 compact discs and shrinking them to the size of one, unencumbered by any digital rights management. On the other hand, the speed limitations of encoding made it cumbersome. After some internal discussion, Brandenburg made an executive decision: to promote the mp3 standard, Fraunhofer would simply give L3Enc away. Thousands of floppy disks were made, and these were distributed at trade shows through late 1994 and early 1995. Brandenburg encouraged his team members to distribute the disks to friends, family, colleagues, and even competitors.

Meanwhile, Popp continued to make scattered sales of the encoding racks, mostly to curious academics and broadcasting professionals. But the door was open to anyone who called, and that summer they met with another struggling entrepreneur, a former fiber-optic cable technician turned music impresario named Ricky Adar. Like Seitzer, Adar had hit on the idea for a “digital jukebox.”

Adar believed that in a few years you’d be able to download music directly over the Internet and dispense with the compact disc entirely. The hitch was that audio files were large, and would have to be compressed considerably for the approach to scale. Fraunhofer, of course, had spent years working on exactly this problem. Even so, when Adar arrived at their offices, he wasn’t hoping for much. Given his past experience with audio compression, he expected the mp3 to be a tinny and unusable bust.

Instead, it reproduced CD music with near perfect fidelity at one-twelfth the size. Adar was astonished. The mp3 seemed a marvel beyond technical comprehension. An entire album at only 40 megabytes! Forget planning for the future—you could implement the digital jukebox right now!

“Do you realize what you’ve done?” Adar asked Brandenburg after their first meeting. “You’ve killed the music industry!”

Brandenburg didn’t think so. He thought the mp3 was a natural fit for the music business. It was just a question of getting them to understand their economic incentives. Adar, however, knew better. His digital jukebox idea was struggling, mostly because he couldn’t get the licenses. The music industry feared that Adar’s digital jukebox would cannibalize physical music sales, and he’d spent the last two years being told no. He explained to Brandenburg the mindset of the record companies: the splendid profit margins of the compact disc, the covetous attitude toward intellectual property, the indifference—indeed willful ignorance—toward both the Internet generally and the future of recording technology specifically. Adar had spent a lot of time trying to get these guys to sign on to his digital jukebox scheme. He’d gotten nowhere. The music industry wasn’t interested in streaming. It was married to the compact disc, in sickness and in health.

The Fraunhofer team already had some notion of the industry’s resistance to change. In October 1994, shortly after being re-assigned to sales, Popp had finagled a meeting with Bertelsmann Music Group, one of the Big Six music labels. It was the first time that Fraunhofer had approached the recording industry directly. Popp made his pitch, and the BMG executives listened. Then they smiled and nodded politely, and reminded him to return his visitor credentials to building security on the way out.

Popp was in some ways a natural choice to manage sales. Of the Fraunhofer team he was certainly the best fit for polite company. He was dark, bearded, and unusually handsome. He wore glasses, dressed sharply, and spoke with a deep and sonorous voice. But he was still an engineer, and not naturally predisposed to the art of the deal. What Fraunhofer really needed was a closer—and then, almost on cue, one arrived.

His name was Henri Linde, and he worked as a licensing manager for the French conglomerate Thomson SA, where he had spent his career negotiating. Along with AT&T, Thomson had acted as the corporate sponsors of the mp3, and by late 1995 the company had sunk more than a million dollars into the project. In fact, researchers at Thomson had independently secured basic patents on the technology, and they had an outsized stake in its future revenues. But no one at Thomson’s headquarters in Paris had the faintest concept of what they’d actually invested in. Linde was dispatched to Erlangen to author a situation report.

He approached the mp3, in his own words, “unburdened by knowledge.” He had no engineering background. He did not understand the math. He did not build his own loudspeakers. His sole qualification for the job was that he spoke German. When the Fraunhofer team tried to explain how the technology worked, with talk of reference frames and bit reservoirs and polyphase quadrature filter banks, it was his turn to nod and smile politely. Yet, unlike the BMG executives, he could see at once that the engineers had achieved something remarkable: they’d obsolesced the compact disc.

Perhaps the only thing more remarkable than Fraunhofer’s accomplishment was their total failure to capitalize on it. Although he liked the Fraunhofer group personally, to Linde it seemed they weren’t really businessmen. They were scientists. They didn’t understand the marketplace, they didn’t understand sales, and they definitely didn’t understand how to profit from intellectual property. Looking over the paperwork, Linde realized that even the few licensing agreements the team had so far signed would have to be renegotiated.

Linde reported back to Thomson headquarters with the startling news: on an overlooked line item in the corporate R&D budget, six German nerds were sitting on a gold mine. The response from corporate was skeptical. If the mp3 was so great, how come no one was using it? Perhaps Linde should try selling laser discs as well. But Linde kept pushing, and finally his corporate overseers conceded that, in the unlikely event that he ever found a customer for the mp3, he was authorized to license the tech. They also made clear that this was a side project, and that the work was not to interfere with his day job.

Linde, a born competitor, didn’t believe in MPEG’s profits-by-committee approach, and pushed the Fraunhofer team to innovate. And so they did. Late in 1994, Harald Popp had commissioned a manufacturing run of dedicated mp3-decoding chips. Now, combining one of the chips with a power source, a soldered-on headphone jack, a primitive flash memory card, and a circuit board, he commissioned an engineer to jerry-rig a prototype of the world’s first handheld mp3 player.

The device was about the size of a brick and could store one minute of music. Which minute? The Scorpions were the obvious choice—or perhaps Suzanne Vega—but the team worried that encoding “Wind of Change” or “Tom’s Diner” to mp3 format might infringe on the artists’ copyrights. Fearing pushback from an industry that already disliked them, Fraunhofer eventually settled on an original composition, from team member Jürgen Herre: “funky.mp3.”

Another long discussion grew out of a second question: did Popp’s mp3 player constitute a distinct invention of its own, or was it merely an implementation of an already patented technology? Linde pushed the team to apply for a patent on the device, but ultimately the Fraunhofer group decided that an mp3 player was nothing more than a storage device.

A visitor to the Fraunhofer booth at the Audio Engineering Society’s Paris trade show in February 1995 would have been presented with a compelling vision for the future of music distribution: the encoder on a floppy disk for creating the files, the home computer for playback, and the handheld player for portable listening. The economics of the arrangement were compelling. The encoder was free, the PC market was exploding, and the handheld player could be built by any consumer electronics firm for a minimal per-unit licensing fee. The whole suite was scored to a finely calibrated strain of synthetic German funk.

But there was little interest. Without cooperation from the music industry, what good was any of it? The industry had decided on the mp2, and that was that. To their competitors, Fraunhofer’s attempts to push this complex piece of technology into the hands of the home computing enthusiast looked deranged. At the same Paris trade show, a Philips executive broke it to Grill point-blank: “There will never be a commercial mp3 player.”

The mp3 was caught in a bind. The music industry wouldn’t license the technology without a critical mass of mp3 players, and the electronics industry wouldn’t manufacture the players without a critical mass of mp3 users. Fraunhofer was beginning to realize that, while scrappy independence was fun, it was also unprofitable. They needed corporate support. Despite Linde’s involvement, Thomson didn’t really seem interested in the project anymore, and AT&T had walked away after the final disappointment in Erlangen. So they began to discuss a new idea: they’d replace the mp3 with a second-generation psychoacoustic encoder, one that would be faster to run, one that would be easier to use, and one that would not use MUSICAM’s goddamned filter bank.

Brandenburg originally called it “NBC”: Not Backward Compatible. The name was a distinct rebuke to MPEG, a signal that they would not enter any more beauty contests. With time, though, this chippy attitude faded, and Fraunhofer eventually gave the project a less combative name: Advanced Audio Coding (AAC).

Brandenburg enlisted corporate stakeholders into the AAC project from the very start. Sony, AT&T, and Dolby were all given large shares, with the understanding that they would fight as hard for AAC as Philips was fighting for the mp2. Politics or no, the next thing Brandenburg made was going to get used. He directed his team to wrap up their work on the mp3 and focus on AAC instead. A new crop of graduate students was assigned to the project, and once again James Johnston provided support. Meanwhile, Grill was given one final piece of mp3-related work: Brandenburg directed him to make an mp3 player for Windows 95.

He was done within a month. Dubbed WinPlay3, this player also fit on a 3.5-inch floppy disk. As Grill tended to write software for other engineers, his sense of design was poor. WinPlay3 was an ugly, uncustomizable blue-on-gray box, with no ability to make playlists or edit the names of songs, and its user interface unnecessarily mimicked the appearance of a monochrome LCD screen.

The finishing touch was the filenames. Microsoft required all files on Windows 95 to have a three-letter filename appended to them. This had led to some rather strange naming conventions, like “.jpg”—Joint Photographic Experts Group—and “.gif”—Graphics Interchange Format. Grill, at this point, pushed for the technology to be rebranded. The name “Moving Picture Experts Group, Audio Layer III” could certainly be improved upon, and doing so would also allow Fraunhofer to distance themselves from the politically biased standards committees. But after some discussion the team decided to embrace their heritage and use the filename extension “.mp3”. Steve Church’s promotional work in the United States meant they already had something resembling a brand. Fraunhofer also realized that MUSICAM was encoding files in Windows. They were using the filename “.mp2,” and that meant MPEG had given them an unexpected gift. While the two technologies were bitter rivals that had been developed in parallel, the naming scheme implied that the mp3 was somehow the mp2’s successor—a misconception that worked in Fraunhofer’s favor.

Grill finished the program in July and began distributing it on floppy disk as “crippleware.” WinPlay3 had the capacity to play twenty songs, and then, like a message from the Impossible Missions Force, it self-destructed. If you wanted to continue using it past that point, you were required to send a registration fee to Fraunhofer and wait for them to send you back a serial number. WinPlay3 debuted in August, and Grill waited for the sales to trickle in.

Nothing. After some discussions with Linde, Brandenburg and Grill came to understand the problem. Why would anyone purchase a music player if there was no music to be played on it? Before they could sell an mp3 player, they’d have to generate a critical mass of mp3 files. And, to do that, they’d first have to sell a bunch of encoders. And to do that, you’d have to have an mp3 player, which no one was going to make without a bunch of files.

It was a classic catch-22, but Brandenburg wasn’t giving up. What did you do with a locked-out technology? You lowered the price. In the first, unsuccessful attempts, Brandenburg had tried to charge users $125 per encoding license. By the middle of 1995, in consultation with Linde, this had been lowered to $12.50. By late 1995 it was down to $5. Frozen out by MPEG, Brandenburg still worked relentlessly to push the technology into as many hands as he could.

Watching Brandenburg hustle, Linde began to revise his initial impressions. Brandenburg wasn’t just a scientist. Underneath the geeky exterior beat the heart of a cunning business strategist. He was a terrible salesman, to be sure. He generated no excitement in potential customers, and his idea of effective marketing material was a binder full of single-spaced engineering data. But he could think strategically, was comfortable with delegation, and had an excellent understanding of his position in the marketplace. He worked relentlessly, and he had terrific instincts about where new opportunities might be found.

The mp3’s first website went live in late 1995. In the top left corner, a spiky red starburst shouted, NEU! Below this were a half dozen blue download links on a plain white background in hand-coded hypertext. The links offered versions of the L3Enc mp3 encoder for DOS, Windows, and Linux. Apple was not included—Bernhard Grill found the company’s programming environment cumbersome and their user interface patronizing. With such a small share of the home computer market, he didn’t think building an encoder for Macintosh was worth his time.

The download links on the Fraunhofer page offered L3Enc for sale at the new price point: zero. L3Enc was “shareware,” a freely distributable demonstration program that permitted users unlimited access to the software. Accompanying the application was a small text file from Fraunhofer, encouraging users to share the program with others, and if they liked it, to please send 85 deutsche marks to Fraunhofer IIS in Erlangen, Germany, payable by mail or fax.

Brandenburg figured this would expose the technology to a wider audience and maybe make some money on the side. Linde was unconvinced but willing to follow Brandenburg’s lead. Grill was skeptical—practically offended—and felt they’d been reduced to begging. And, initially at least, his skepticism was justified. The shareware encoder was a flop, and very few users faxed in their deutsche marks. Over its lifetime, the downloadable L3Enc demo earned less than five hundred dollars.