I worked as an electronics engineer for Disc Manufacturing, Inc. from September of 1997 to February of 2000. My main responsibility was project management for a new DVD-RAM production line.
DVD-RAM Replication line:
At the start of this project, we believed that the DVD-RAM disc had the potential to replace the videocassette. As DVDs became increasingly popular there would be a demand for a method of making copies of them, and we wanted to be the main supplier of discs. DMI was already in the business of manufacturing CDs, CD and DVDs, so DVD-RAMs were an obvious extension of that business.
DVD-RAM drives use phase-change technology in which a laser heats an inner layer of the disk to cause reversible crystallographic changes, allowing the data to be rewritten. An injection molded wobble groove provides clocking data, with marks written on both the grooves and the lands between the grooves. The grooves and sector headers are molded into the disc during manufacturing. The discs consist of a layer or layers of a reflective metal alloy (we used a mixture of Germanium, Antimony, and Tellurium) film embedded in polycarbonate. A high power laser writes and erases data, while a low power laser reads data. To write data, the laser heats a small spot on the alloy layer to around 600 degrees Celsius (roughly 1100 degrees Fahrenheit). If the light is instantly shut off, there is not enough time for nucleation crystal growth and the alloy freezes as many small amorphous crystals, resulting in low reflectivity. If the light is ramped down over 100 nanoseconds or so, the alloy freezes as one large crystal, resulting in high reflectivity. This change in reflectivity can be read by a laser, much as CDs and DVDs are read.
The disc manufacturing business has very low profit margins, very high production rates, and extreme precision requirements. These factors make for some interesting engineering tradeoffs. Let's look at each of these factors:
Low profit margins: In the year 2000 the gross revenue for a CD manufacturer was around thirty cents per disc with fierce competition from plants in Asia. A plant that already owned all of the equipment needed could make a modest profit with good management, but it is was and still is impossible to open a new U.S. CD plant and make enough profit to pay for the machinery.
The nature of the disc manufacturing business is such that the first plant to make a new kind of media makes large profits at first, then the profits approach zero as more and more competitors start making the same media. The margins are so low on mature products such as CDs and CD-ROMs that foreign competitors cannot take advantage their traditionally low labor costs - only a pure robotic production line can make a profit. I had to design a robotic production line and get it working very quickly to reap the large profits that come from being first, and I had to do it in such a way that the per-disc cost was as small as possible so that the production line would remain profitable as prices dropped.
High production rates: In the year 2000, there were roughly ten billion CDs and half a billion DVDs produced worldwide, with three billion of the CDs and a quarter billion of the DVDs manufactured in the U. S. We were hoping to position ourselves as the first to market with large quantities of DVD-RAM disks. To meet this goal, I designed a fully automated production line that would produce one DVD-RAM disc every three seconds. This would give us a capacity of roughly ten million discs per year for the one production line. Because of the uncertainty of the market, we planned on building up to three additional lines, depending on demand.
Three seconds per disc is a daunting specification to meet. At the time, the only DVD drives available were 1X, which means that it would take two hours to format a two-hour DVD-RAM disc. This was an ongoing challenge; would 2X or 4X drives be available by the time we started production? Could we modify a 1X drive for 8X formatting? Would we be able to speed up our formatting speed when we started making four-hour or eight-hour discs? Any way it turned out, we were looking at racks full of drives and a robotic disc changer.
Another challenge was the fact that our existing operating system (Windows NT) was marginal for controlling such a complex automated production line. I considered Embedded NT, but the way Microsoft was pushing Windows CE made me worry about the future of Embedded NT. The various flavors of Embedded Linux weren't quite ready at the time (they would become a good solution by the start of 2001), so we chose QNX, which worked out very well for us.
Extreme precision: Another challenge was the fact that a DVD pit can be as small as 400 nanometers across. We had to be able to melt sections of the metallic layer at those scales, along with the usual injection molding challenges. In addition, we had to bond together layers with the thickness of the adhesive layer controlled to within a fraction of a wavelength of the light from a red laser. All in less than 3 seconds per disc.
In the end, it was not technical issues that doomed this project, but rather corporate / financial issues, ending with most of the engineering department being laid off and the manufacturing plant we were working out of being shut down.