OLPC Origins: US and Taiwan's Hardware LovechildS

In November of 2005, Nicholas Negroponte and his OLPC CTO Mary Lou Jepsen traveled to Tunisia for the UN-sponsored World Summit on the Information Society, where they were able to present a "working" $100 laptop concept to Kofi Annan, UN secretary general. No longer did the machine rely on that pop-up rear-projection display; it was smaller, made of green plastic, and had a crank for the kids to work—for 10 straight minutes per hour of use—when they had no other access to electricity. It was a vast improvement over that January's pup-tent rear-projection laptop, hampered only by the fact that it was an absolute fake.

One attendee described it as a "balsa model with a keyboard and an LCD with a thick cable attached to a box under the counter." Others noted that the screen froze up during Negroponte's demo. Worst of all, when Annan himself approached the device with the sole intention of making Negroponte and his mission look good, the secretary general broke the crank handle. Clean off. Overnight, the broken handle story became an internet vote of no confidence. It was time to make this concept into something that would work. And that would take help.

OLPC Origins: US and Taiwan's Hardware Lovechild

Make It So?

As we covered in Part 1, whether Quanta founder Barry Lam's deal was a charitable donation or a crafty decision to get some fresh MIT thinking into his own laboratories, the dreamers from Cambridge started in earnest with the best laptop engineers in the world. Jepsen and others spent two years shuttling back and forth from Cambridge, MA to Taipei, Taiwan. (She actually filed her income tax return as a Taiwanese resident in 2006.) The American scientists would bring ideas to Quanta's engineers, who would either approve them, or disapprove.

"Oftentimes I'd be presenting my ideas in this large boardroom with five EVPs on one side and five EVPs on the other side, and they would point out reasons it wouldn't work, things I couldn't have known, that are somewhat trade secret. I would take these down, saying 'These are really good points—when I have answers, can I come back and present them to you?' In three months, I'd come back for more."

As engaging a process it was for the Americans, it seems that Quanta's people were a little annoyed by the game, at least according to what Quanta's OLPC overseer Dandy Hsu told a reporter from Taiwan Review. For one thing, OLPC didn't order up its laptop from Quanta's speedy-delivery menu, like the big brands do. And for another, OLPC's unfinished OS made motherboard testing a huge pain. "This made the process long and complicated," Hsu told Taiwan Review. "We'd make assumptions in the design, but later, when we got the software, we would need to make changes to the hardware."

What Hsu didn't tell the Taiwan Review was that throughout this process, any motherboard changes made by the OLPC team—that is, some of MIT's brightest minds— were added to Quanta's reference design, and therefore its intellectual property. Quanta had 100 staffers on OLPC on a regular basis, and perhaps 200 at peak times, but it got some good R&D in return. Quanta built a consumer-friendly prototype of the XO to add to its menu—one that has become a standard netbook reference design—and OLPC can't challenge the company to a patent dispute.

As much as it sounds like a Terminator movie, it's true: Quanta could use the XO's own blueprints to build something that could destroy it.

Big Differences

The point of OLPC's hardware endeavor was to create a laptop that didn't resemble all the others out there. In spite of the ambitious $100 target price tag, the largest constraint on product design was not cost but power consumption. Because CPUs tend to draw as little as 1-2 watts of power, while screens draw 7 watts at the minimum, the radical idea was to build the rest of the system around the screen. Cheap, safe battery design and energy efficient wireless connectivity added to the challenge.

Green Inside and Out

In the early days, when Negroponte and crew were talking to manufacturers, a big laptop maker—"who shall remain nameless," says Jepsen—told them that it would cost $25 to $30 extra to make them green. "We made the difficult decision that we couldn't afford it, but also vowed to do the right thing at every step of the line to make them environmentally friendly." Because they were choosing parts from scratch, and because they adhered to the "reduction of hazardous substances" (ROHS) standard, the XO became the greenest laptop ever made. Though many colors were tested out, the "playful" green and white design that was chosen was perhaps the most fitting. (Though I still kinda like the red limited-run ones I saw at OLPC headquarters.)

Battery

"Normal laptops use lithium ion, but every six months or so, there's a large recall because sometimes, in bad batches, they explode," says Jepsen, arguing that they wanted to "go the extra mile on safety." So, if they couldn't use batteries that occasionally blew up and burned at 1000º C—and nickel metal hydride was too expensive thanks to a precious metals boom—what was left?

The answer, one that Quanta brought to the table, was lithium ferrophosphate. It rarely explodes, and burns at a slightly less painful 100º C in a fire, and it has a very long life, 2000 charge/recharge cycles. Most laptop makers aren't into it, since its charge density is light—it's currently more often considered for electric cars. Still, because the OLPC team believed they only needed a low-power battery, they opted for it, ordering up a battery that produces just 20-watt-hours, versus a typical laptop battery's 80-watt-hours, and has no circuitry of its own so the replacement cost is very low. Those perceived benefits, particularly the lack of circuitry, have been called into question, though, because the XO doesn't get the battery life it's supposed to. NiMH may be a tad greener overall, but according to Jepsen, "there's some suggestion that [the lithium ferrophosphate] decomposes into fertilizer."

CPU

The team figured out that they could save an awful lot of energy by having the CPU flick off in tenth of a second when it wasn't needed, and power up again when something demanded its attention. Jepsen says this would be deemed heresy at Intel, but they wouldn't hesitate to shut the screen off. The complexity, in Bletsas' mind, was that there are three different sets of firmware all interrelating: the main one that runs the show, the embedded controller chip for the keyboard, and the networking chip, which has its own CPU. In the end, the on-and-off flickering power-saving innovation hasn't fully worked on the Linux OS, but it apparently does work when the system is running Windows XP, so the issues are considered software problems.

OLPC Origins: US and Taiwan's Hardware LovechildS



Networking

The Wi-Fi radio works independently of the CPU because it has to form the XO mesh network. A big part of Negroponte's dream, this network is famously intended to blanket a town so children's data can easily be shared, even when given computers are in a state of rest. This caused a problem for the component, a USB device, and required tricking the radio to stay awake when its master went to sleep. "It took many tries to get that right," says Bletsas. Also, because of the USB interface, the flicking on and off of the CPU played havoc.

Those Little Antennae

The XO wouldn't be the XO without its hallmark green antennae, an attribute that gives it a ridiculous Wi-Fi advantage over, say, a MacBook Pro. The MBP's aluminum body is actually especially bad for radio transmission, says Bletsas, OLPC's networking guru. It was he who argued strenuously for the antennae, as you can see in his none-too-artful sketch:

OLPC Origins: US and Taiwan's Hardware LovechildS

Though Bletsas proved that such a placement would double the wireless reach of the XO without demanding a milliwatt of additional power, the request didn't make designer Behar terribly happy: "Two ugly antennae to place on top of a laptop?" he recalls. "It was a huge challenge." Luckily the team turned those pointy green lemons into lemonade. "We used them to give the laptop personality, make it look friendly and fun," says Behar. "It became an icon for the laptop." As you can see, they were almost little stubby things, less rabbit, more cat:

OLPC Origins: US and Taiwan's Hardware LovechildS

Keyboard

One of the most underrated developments on the XO is the keyboard. Admittedly it's a little tricky to type on it with grown-up hands, but it's the only keyboard that is shows Ethiopian characters and layout, for instance, and it can be easily configured for other languages and dialects as well. The process is achieved through screen printing. "With one single printing pass, we can very quickly localize the keyboard," says Behar. It can be modified in low quantities and reasonably quickly too, with only a month or so leeway to set it up.

Screen

Beside the ears and the impressive mesh network, the screen was a signature trait for the XO, and one of the only innovations that merited some solid patent applications. For starters, the flickering CPU meant something had to be done to keep the screen from strobing as well. The team designed a controller chip with onboard memory which kept the images on the screen alive.

More importantly, the LCD panel itself, though it came straight from a plain-vanilla LCD fab, had some never-before-seen twists: In basic use, the $35 component displays color at a resolution of 800x600 pixels, with a backlight so that it's easy to see indoors. However, when outside, a switch can change it to a completely different screen, one that is monochrome, uses reflectivity to conserve power and improve visibility, and has a far tighter resolution of 1200x900 pixels.

Though she was OLPC's chief technology officer, this is Jepsen's claim to fame—some people on the internet refer to it simply as "Mary Lou's screen." Even critics who don't wish to give Jepsen undue credit for the whole XO laptop say that without her LCD breakthrough, the screen would have cost at least $20 more, would have had far fewer pixels, and probably would've tied up the development for six extra months.

Triple Trouble

It didn't take long, though, before OLPC got totally hosed. It started in March 2006, when Quanta was about to mass produce the first run of XO motherboards. OLPC has surveyed a poll of potential buyers in various developing countries, who overwhelmingly asked for a faster processor. That isn't an easy fix, especially when you're about ready to go into mass production. You can almost hear the guys at Quanta saying, "Something that would have been nice to know last year!!"

Then, Quanta screwed over OLPC by selling off its display division. As we mentioned in Part 1, the screen is the most expensive component, and the OLPC had designed a really special one to keep prices down. Quanta's owner Barry Lam may have loved the mission, but business was business, and when he sold the LCD plant, the company who bought it told OLPC to screw off. As Jepsen was returning home, she was not just exhausted and disappointed, but fell ill with adrenal failure. Her flight to Boston was forced to make an emergency landing.

OLPC Origins: US and Taiwan's Hardware Lovechild

With Jepsen barely on the mend, and a whole new stack of problems to solve for everybody, OLPC took a third blow square in the jaw: In May 2006, Intel announced it would go into the business of building laptops for teachers and students in third-world countries. Sound familiar? CEO Paul Otellini even showed off a prototype of the more powerful "Eduwise" laptop, soon to be (wisely) renamed Classmate PC. It was the first of many spiritual clones of the XO, and the beginning of even tougher times ahead for Negroponte and his team who hoped to ship millions of their own laptops.

Come back tomorrow for Part 3, the final installment of OLPC Secret Origins. If you're just jumping in, hop back to Part 1, where Negroponte announces his dream and pulls his team together.