Speeds and Feeds

I got an e-mail from the folks over at O'Reilly Media mentioning that keynotes and other presentations from the company's ETech 2009 conference, held earlier this month, were now online at the ETech 2009 site. I missed that show, but I was interested in one of the keynotes, so I surfed on over to take a look.

The keynote I was looking for was indeed online: Mary Lou Jepsen, CEO of Pixel Qi and formerly CTO of the One Laptop Per Child organization, talking about "Low-Cost, Low-Power Computing." You can watch a video of the presentation on Blip.tv or download the PowerPoint slides direct from O'Reilly.

The talk is well worth watching, but it's flawed in many respects. Jepsen still has the over-the-top attitude displayed all too commonly by participants in the OLPC initiative, who often act as if they had actually saved the world instead of merely doing something good for it.

She started out with a series of very squishy claims that overstated the importance of her work, including references to the "digital divide" (a term of propaganda) and statements about how "97 percent of adolescents live in the developing world." Both claims rely on entirely arbitrary definitions.

Jepsen's 97 percent figure in particular assumes that all populations outside "Europe, the U.S., Canada, Australia, New Zealand, Japan, Korea, and Taiwan" are part of the developing world, but of course that isn't true. If I lived in Hong Kong, Riyadh, Shanghai, Tel Aviv, or various other places, I would probably resent Jepsen's implication.

Jepsen singled out power consumption as the most critical requirement for success in laptops for the developing world, but I think an equally strong case can be made for cost, ruggedness, or ease of use. All of these elements must be in place, or the machines aren't worth the effort of deploying them...as many countries approached by OLPC have decided over the last few years.

Jepsen also said that "Bill Gates, Steve Jobs, Craig Barrett, and Michael Dell said it was impossible" for the OLPC project to deliver on its early promises.

I don't know that they did, and she didn't exactly cite references. But after all, it was impossible. And, in fact, OLPC did not deliver what it promised: a $100 laptop. It still hasn't done that and shows no sign of doing so anytime soon.

But most of my problems with Jepsen's presentation stem from her claims about technical matters.

For example, she said that the OLPC XO-1 laptop--for which she was the lead designer--is a "1 watt laptop," "more than 10 times lower (power) than the next laptop up."

But the XO-1 certainly is not a 1 watt laptop; on an apples-to-apples basis, it's more like a 4 watt laptop, and the best traditional PC laptops consume only about 50 percent more power--in spite of larger screens, faster processors, and a complete array of external interfaces. The Thinkpad X series (originally from IBM, now from Lenovo) has included several models down in this power range, according to my own laboratory testing.

Jepsen wanted her audience to believe that the key breakthrough in the XO-1 design was the screen, but the fact is that low-resolution, low-quality, low-power displays have been widely available for years. They aren't used in PCs because they don't meet the expectations of PC users. And frankly, the display in the XO-1 wouldn't meet most PC users' expectations either.

Speaking about the component consuming the most power in PC laptops, Jepsen said this: "It's the screen, not the CPU, not the motherboard." But this is not true in most laptops, especially in the size range of the OLPC XO-1, or in most usage conditions. A 12-inch LCD may consume only 2 watts to 3 watts, usually less than the motherboard in the same system. While an x86 laptop processor may consume less than a watt when idle, it will generally consume over 10 watts when busy.

The purpose of Jepsen's remarks was to create the impression that the display has become more important than any other component in a laptop computer, but I don't think that conclusion is supported by the facts.

Why would she want people to believe the display is so important? As she said in the presentation, she's still working to raise funding for her display start-up, Pixel Qi.

It's traditional for CEOs to overstate the significance of their start-ups, but that doesn't mean they ought to get away with twisting facts in the process.

And it's just not necessary, either. As critical as I've been about Jepsen's hyperbole here, I still think that the work Pixel Qi is doing will be valuable to the industry. Pixel Qi does seem to have a very narrow focus, but I certainly appreciate the fact that start-ups have to start somewhere.

From what I can see on the company's site, Pixel Qi's first product will be a 10-inch LCD that appears to bridge the gap between the E Ink display used in Sony's Reader and Amazon's Kindle e-book readers and the LCDs commonly used in PCs and TVs. With both a high-quality reflective monochrome e-book mode and a "fully saturated color" mode, such a display could be a good choice for many small notebook and tablets. In her speech, Jepsen said the company is also working on a low-power (under 10 watts) HDTV screen, which is also likely to be popular.

But more generally, I believe there's still room for plenty of improvement in CPUs, chipsets, and other laptop components, not just displays. I see no reason we can't eventually get $100 laptops, but it'll take improvements in all these areas to get there.

Samsung Electronics, an arm of the giant Korean company (second only to General Electric in annual revenue among conglomerates), held a press event in San Francisco last week to show off its products for the coming holiday season.

Samsung's new Series 9 LCD HDTVs use LED backlights to improve contrast and color quality

(Credit: Samsung Electronics)

I'd been looking for an excuse to go up to the city, so off I went-- taking Caltrain rather than driving. Conveniently, the Samsung event was just a few blocks from the train station in San Francisco.

... Read more

I just saw an interesting piece over on the EE Times website (here) written by David Carey, president of Portelligent, an analyst firm well-known for doing teardowns of popular electronic products.

The XO-1 laptop from the OLPC Foundation

(Credit: OLPC Foundation)

Here are some of the key points I learned from the article:

According to Portelligent, the LCD is pretty similar to previous transflective LCDs (that is, LCDs that can work from transmitted backlight or reflected ambient light) except for the arrangement of the color subpixels. I've seen nothing particularly remarkable about mine. It offers poor off-axis image quality, as I described in my first look (here), and doesn't even work as well as the transflective LCD on my Sony DSC-T1 digital camera, which displays color in reflective mode. The XO-1's display may be more power-efficient than previous transflective displays, but Carey doesn't report information on that point in this article.

Portelligent did report whole-system power figures, though, seeing 6W to 7W of power consumption during "full-tilt operation," whatever that means, exactly. In my own brief tests I saw power consumption in the range of 4W to 5W when the machine was doing almost nothing. Any machine with such a narrow range of power consumption isn't exhibiting good power management.

Carey's article (and a great illustration, here) goes on to provide some good information about the specific chips found in the unit they took apart. The article is also an advertisement for a full report Portelligent sells with all the details of the teardown, but I don't mind that at all in cases like this where the free article stands on its own.

I'm a little late to the party with this unboxing of my new OLPC XO-1 laptop, but the machine arrived while I was out of town visiting my family for Christmas. In fact, there's a story there.

Before I left, I started hearing that people were receiving their XO-1's, and I realized that if mine didn't show up before I left, it would almost certainly arrive while I was gone. The OLPC people sent out no shipment notifications and didn't reply to several emails, so I had no way to delay the shipment or contact the carrier.

I left a note on my front doorstep: "PLEASE DO NOT LEAVE PACKAGES HERE. HOLD FOR PICKUP. THANKS."

But on Dec. 21, a FedEx delivery person left the XO-1 box right next to the note, and they were both still there six days later when I got home. All that time, the package was in clear view of the street. Never mind New York-- I love Cupertino.

If you get an XO-1, don't throw away the box! You'll need it for the free year of Internet access through T-Mobile WiFi hot spots. The box has the reference number for account activation.

In keeping with the low-cost nature of the XO-1, its packaging is minimal but adequate.

The OLPC XO-1 comes with only a few sheets of basic “Getting Started” documentation.

(Credit: Peter N. Glaskowsky)

The XO-1 comes with no manual, just two sheets of paper: one showing the hardware and software features of the unit plus some warning icons, and one with a thank-you note from OLPC founder Nicholas Negroponte.

There's also no warranty booklet. The XO-1 comes with a 30-day limited warranty, but that's it, and it isn't written down anywhere.

I was somewhat surprised-- and pleased-- to see that OLPC provided a toll-free support phone number. As I'll describe in my forthcoming review, that might prove to be an expensive decision; the XO-1 is not yet very well documented, and some aspects of its operation are difficult to understand.

Of course, there's some XO-1 documentation online. Negroponte's letter points buyers to the laptopgiving.com website, which in turn points to the main laptop.org site, and from there a diligent search will reveal more detailed information on the OLPC Wiki.

But many aspects of laptop operation that are familiar to Windows, Mac, or Linux users aren't documented anywhere, as far as I can tell, probably because they aren't even supported. I can't find any way to control power-management features, for example.

Bottom line: the OLPC developers have a lot of work to do. These early systems don't even qualify as beta-test devices; they're just an alpha release, not feature-complete.

But they do work, and I still believe the XO-1's primitive state of development could actually be a positive benefit for bright children, who will be challenged to learn about these machines in ways they'd never have to do with a mainstream laptop PC.

Other than the documentation, the box contains only three items: the XO-1, the battery, and the AC adapter.

(Credit: Peter N. Glaskowsky)

The XO-1's limited hardware budget isn't wasted on unnecessary doo-dads. It arrives with the bare minimum of accessories: a battery and an AC adapter.

The battery is rated at 6.5V, 3.1AH (20.15 watt-hours); the AC adapter is rated at 12V, 1.42A (17.04W).

(Credit: Peter N. Glaskowsky)

Both of these items are in keeping with the low-power design of the XO-1. Most laptops today come with larger batteries, often in the 50 watt-hour range; the XO-1's battery provides only 40% as much capacity. The AC adapters for full-size notebook PCs usually provide over 65W of power; this one is about a quarter as powerful.

But these are advantages, not disadvantages. A low-power laptop is like a lightweight car. A lighter car can use a smaller engine, brakes, and suspension without compromising performance. If the car gets heavier, the other components have to bulk up too. Similarly, reducing a laptop's power consumption saves weight in the machine itself and in its battery and power adapter.

The XO-1's battery compartment is well integrated into the bottom of the unit.

(Credit: Peter N. Glaskowsky)

You can see here that the whole surface of the XO-1's hard plastic case is covered by a pattern of nubbly dots that make it easier to grip without making it any more difficult to clean-- a wise decision by the developers. There's also a bit of whimsy around the handle section, where the openings are ringed by little "X" shapes that form the XO-1 logo.

There are four soft narrow feet at the corners of the unit.

(Credit: Peter N. Glaskowsky)

Since the hard plastic would still be too slippery on a desk, the XO-1 has molded-in feet made of some non-skid rubbery material. They aren't very tall; since the XO-1 consumes so little power, there's no need to create airspace under the case.

Each XO-1 gets a logo with a distinctive color combination.

(Credit: Peter N. Glaskowsky)

My XO-1 came with a nice blue/green logo color combination. I don't know how many combinations there are, but I gather it's a large number, reducing the odds that two students in the same class will have the same colors.

Open, the XO-1 shows its most distinctive feature: the antenna “ears”.

(Credit: Peter N. Glaskowsky)

The XO-1's ears contain 2.4 GHz antennas shared between the WiFi and proprietary mesh networks. They're also the locks that hold the machine closed. They engage with spring-loaded pins so the top will snap closed even if the ears are stowed first.

The left side of the XO-1 provides microphone, headphone, and USB jacks.

(Credit: Peter N. Glaskowsky)

There aren't a lot of I/O options on the XO-1, just the basic requirements. The microphone jack can also be used as a generic analog input; the XO-1 comes with an application that works like a simple oscilloscope. Neat.

(Actually, applications are called "activities" on the XO-1. Sometimes it seems like the developers are thinking too differently.)

Two more USB jacks are located on the right side.

(Credit: Peter N. Glaskowsky)

Another clever design feature on the other side of the unit: two USB jacks are positioned at different angles to make it more likely that awkwardly-shaped USB devices can be accommodated.

The XO-1 display is flanked by more buttons including a D-pad and a cluster of buttons like a game controller.

(Credit: Peter N. Glaskowsky)

The XO-1's display is about what I expected. Resolution is good, but colors aren't as vibrant as on traditional LCDs. As I should have predicted, color saturation is related to the ratio of backlighting to ambient light. Outdoors or under a strong indoor light, colors are very washed out even with the backlight cranked up all the way. In sunlight, color disappears entirely, and you might as well turn off the backlight since it doesn't help.

The LCD viewing angle, unfortunately, is very poor. At little as 30 degrees off-axis, contrast begins to drop sharply. Two children sitting side-by-side would have trouble viewing the screen together. For ebook reading, the XO-1's display can't match those of the Sony PRS-505 Reader and the Amazon Kindle.

Not shown here is the Secure Digital (SD-card) slot, which is under the lower edge of the right side of the display unit. The positioning helps protect the slot, but there's no way to get clear access to it, which may limit the range of SD-card peripherals that can be used with the XO-1. I'm not sure this was so clever.

The XO-1 keyboard uses a flimsy rubber membrane over soft springy keyswitches.

(Credit: Peter N. Glaskowsky)

To me, the low point of the XO-1's physical design is the keyboard. The synthetic rubber membrane is very thin and the keyswitches are very soft so there's almost no tactile feedback. Hitting a key feels almost the same as missing one.

Perhaps children's fingertips are sensitive enough to get the feedback they need for good touch-typing. But even if that's true, I fear this keyboard may be too fragile.

The keys are also smaller than necessary, even given the focus on small hands. The keyboard is 15 keys wide, with a double-wide Enter key plus tab, [, and ] keys on the QWERTY row. Although the OLPC developers took a fresh look at pretty much everything else, they slavishly imitated the high key counts of full-size notebooks to their detriment.

Since the XO-1 has multiple modifier keys-- shift, control, alt, fn, "hand", and alt-graph keys-- it would have been better to move more of the punctuation symbols to letter keys, reducing the key count and allowing the keys themselves to be slightly larger, making typing easier.

The keyboard is printed with many international characters, but it isn't as cluttered as it could be. Only one key has four different symbols on it (semicolon, colon, and underlined lower-case a and o characters); most have three, and some have two. G, K, L, Z, X, V, and B are left alone. Oddly, there's a whole extra key just for the "times" and "divide" symbols.

There are also many extra keys for features unique to the XO-1's "Sugar" user interface, which is a good thing. Sugar relies too much on tricks like hot corners and tabs, disappearing borders and drawers, and other features that require a lot of careful cursor motion. Unfortunately, the XO-1's touchpad doesn't operate very smoothly or accurately, at least for me, and there's no apparent way to control its sensitivity or the speed of cursor motion.

Because I was somewhat critical of OLPC in earlier blog posts (here and here) for making strong promises about battery life that weren't supported by the early prototype hardware, the first thing I did with the new machine after charging it for a few hours was to run a couple of simple battery-life benchmark tests.

In the first test, I connected the XO-1 to my home WiFi network (which required falling back from WPA security to the relatively insecure WEP standard), cranked the backlight up to maximum, and opened my favorite webcam page: Ben Lovejoy's auto-refreshing feed for the camera at the public entrance to the Nürburgring racetrack in Germany.

The page didn't load reliably-- sometimes the WiFi connection would drop, provoking Server Not Found errors-- but I kept an eye on it and got it back on track each time it derailed. This wasn't the "heavy use" that OLPC's Walter Bender was describing in his comments on 60 Minutes last May, but at least it was something.

The result? The XO-1 ran for just 45 seconds short of four hours. Not so great.

Well, it's a prototype, and OLPC vice-president Jim Gettys said that "heavy use" could be construed to cover uses as lightweight as reading an ebook outdoors with the backlight off. So I charged the machine overnight and, this morning, from a clean reboot, I started an ebook-reading test with the backlight off. I opened a PDF provided with the XO-1 and pushed the page-down button once every 20 minutes to keep the display from turning off entirely. The machine ran for 4 hours and 59 minutes. (I swear these are the actual numbers.) That's a long way from Bender's promise of "10 to 12 hours... with heavy use."

But still, it's a prototype, and as Gettys explained, there are many opportunities for further power reductions. Similarly, there will undoubtedly be other improvements over time. We'll see.

Today I take my lead from a Reuters article that describes two alternative display technologies that may some day replace (or at least augment) LCDs in mainstream computer systems.

Both are on the market today. OLED (organic light-emitting display) technology, used on some cellphones, creates what amounts to an array of tiny LEDs. This approach is theoretically superior to the way LCDs work, which is to combine a white backlight with colored filters and tiny shutters (the liquid crystals) for each pixel. All the light generated by the OLED is visible to the user, but most of the light in the LCD is stopped in the color filters or the liquid-crystal shutters. If you're looking at a predominantly black image--for example, the main page of Apple.com these days--you may be seeing less than 10% of the light coming from the LCD's backlight; the rest just turns into heat.

The other approach discussed in the Reuters article is what it calls bi-stable displays, after the earliest form of this technology in which the display elements could be set to black or white. Perhaps you've seen these displays in old airports and train stations, where relays would flip dots the size of dimes from one state to the other. Today's version of this technology uses microscopic white spheres in tiny capsules of dyed oil-- in what is known as electrophoretic displays-- or other techniques. They retain their state, but most can display shades of gray, too. A more generic name is "electronic paper," and fittingly such displays are used in electronic book devices such as Sony's PRS-500 Portable Reader System. The Reader uses an electrophoretic display from the E Ink Corporation with 166 dpi (dots per inch).

I have a Sony Reader myself, and I use it a lot, but this generation of display is a long way from looking like paper. The contrast ratio is much lower because many of the capsules appear to be stuck in their white or black states. Also, although the individual capsules are much smaller than a pixel, the pixels on the Reader are much larger than a dot from a laser printer or inkjet. The large pixels makes the text look blocky.

I'm looking forward to the day when the pixels themselves are much smaller than we can see-- when instead of 166 dots per inch, displays have 600 dpi or more.

Some people say there's no need for that kind of display resolution, but just recently we've seen another story about how displays are a long way from good enough. When Apple released its Safari browser for Windows, it came with Apple's own font-rendering engine, which produces very different results from Microsoft's standard renderer. Joel Spolsky blogged about the differences (here) last month. Although I disagree with his conclusion about which approach looks better, my point here is that they're different. With much higher display resolutions-- at least three or four times higher in each direction-- both engines would produce the same results and everyone would be happy.

There are two more related display types that might show up eventually, but they're more of a longshot. FEDs (field emission displays) and SEDs (surface-conduction electron-emitter displays) consist of many tiny cathode-ray tubes, just like old-style CRTs-- but displaying only one spot each. FEDs use many of these per pixel; SEDs use just one.

My guess is that OLEDs will eventually win out in portable applications where rapid updating and high color fidelity are needed (laptops, video players, etc.). Electronic-paper displays may find a niche in e-book readers, but if OLEDs turn out to be as efficient as theory allows, they may sweep the category.

In desktop displays and televisions, where power consumption is not so critical, it's likely that all of these techniques will continue to compete for a long time to come. In these applications, manufacturing cost is the most important thing, which is why CRT-based TVs are still common today.

Display technology has fascinated me ever since I took a job as an engineering manager at SuperMac Technology in 1990. If I recall correctly, my home computer at the time was a 286-based PC clone with a 13" EGA display. At my new job, they sat me down in front of a Mac IIcx with a 19" CRT driven by a true-color accelerated graphics card, and I was instantly hooked. You couldn't even attach displays like this to a PC back then.

Today, displays are a lot better, but there's still plenty of room for improvement-- another 17 years of progress and then some. Keep your eyes open...