The Inquirer-Home

OLPC screen is clever, but very odd

Low power, low price, but low quality too?
Wed Sep 13 2006, 20:20

WHEN I FIRST HEARD about the LCD used in the OLPC project, I was equally intrigued and sceptical. The claimed specifications put it, on paper, well into the league of the best panels available and, if manufactured using conventional technology, would present a major step forward in LCDs, which you would expect to carry a significant price premium.

The specs again for the uninitiated:

  • 7.5" diagonal LCD (6.0" x 4.5")
  • Dual mode operation: Reflective Monochrome or Transflective Colour
  • TFT LCD driving
  • 1200x900, 200dpi resolution
  • < 1W in colour mode, 0.2W in B&W mode
  • LED backlight
  • All with a target $35 cost

Looking at what's available in standard 7" LCDs, what's available is not really all that impressive: most models available are designed for portable media players, so come in 800x480 or lower, so already the 1200x900 is sounding good.

To get this sort of LCD you'd probably need to make the jump to polycrystalline silicon (made with laser beams), rather than amorphous silicon used on most large panels, and 7.5" would be bigger than anything I know that is shipping on this process.

Also, as I've wittered on about before, no TFTs are available that work with any decent kind of performance in sunlight, with the backlight switched off completely, as the colour filter and transistor block out some of the light. Even the transflective models are rare and expensive.

To put $35 into perspective, a 3.8" 320x240 PDA screen was selling to the HPs of this world for around $26, so this is a very tough target for a 7.5" screen of any variety.

I was therefore pretty impressed to spot photos up yesterday of prototype OLPC displays actually up and running, driven from the Geode GX2, and decided to investigate further.

Olpc-screen

Well, if something sounds too good to be true, then it usually is, and the answer is that the OLPC screen can't be directly compared to ordinary panels. Firstly, see that 1200x900 res? Well that's 1 million sub-pixels, not full colour pixels. Most normal screens are RGB stripe, so a 640x480 screen is actually 1920x480 (also 1 million sub-pixels). You can already get that in a 2.8" screen, so it's not a massive step forward from there.

Instead, the lead designer, Mary Lou Jepsen (former head techy at Intel's display division) has come up with an interesting and novel alternative to the RGB stripe and RGB delta pixel formats usually used: instead the colours are arranged in diagonal stripes, and sub-pixels are roughly square, which allows an increase in DPI.

Olpc-pixel-layout

A display pixel, as seen by the operating system, is very easy to form from RGB stripe - for that pixel will be made from one red, one green and one blue sub-pixel, and forms neat square blocks. But the smallest square block capable of displaying full colour on the OLPC screen will contain four sub-pixels, with two of the same colour, i.e. two red, one blue, one green.

Therefore there will need to be some maths behind the screen to convert the 800x600 picture into one capable of being displayed on this screen at all, let alone in good quality. On the OLPC this is done by a clever new chip they call the DCON ("De-swizzling" & timing Controller). As a side note, I love their use of the term "swizzle" in a technical context.

To me, this sounded very, very odd. Why on earth anybody should choose to do this was initially beyond me, as it would lead to pretty strange picture quality. Even the RGB delta used on camera screens looks pretty rubbish when displaying text or straight lines.

Here's a picture from one of the OLPC software guy's work on simulating the picture quality - look at the fuzzy edges around all the lines and text. It should still be fine at displaying photos, though.

Olpc-screen-simulation

However there is a method behind the madness, as Jepsen has managed to do some serious work on the colour filter, eliminating most, and eventually all of it. This enables the display to work in mono or colour mode, and gets out one of the major cost components of a smaller display. Clever stuff.

The part I'm still struggling to understand is how each sub-pixel can emit a colour, rather than just the white of the backlight, as would happen if you removed the colour filter from a normal TFT.

One possibility that came to my mind is that the backlight could actually be supplying diagonal stripes of red light, green light and blue light from individually coloured LEDs down light pipes, which are then transmitted out through the screen. Something like this would explain how the display can only work in colour mode when the backlight is on, and only in black & white when it's off. It would also explain the odd pixel pattern.

Anyway whichever way you look at it, the screen is clearly managing to achieve 1/7 the power, 1/3 the price, higher resolution and sunlight readability that the project claims, which is a very highly impressive achievement, but it's unlikely to replace standard RGB stripe screens, as the usable resolution in colour mode is closer to 400x300 than 1200x900 sadly.

It should, also, look pretty awesome in black and white reflective mode, which may be far more important in the intended environment. And if Mary Lou wants to explain to me how the hell the coloured backlight works, she's more than welcome to drop me an e-mail :o) µ

 

Share this:

blog comments powered by Disqus
Advertisement
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

Advertisement
INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?