(HTML mode is broken for me: preview okay, end result nasty. Text mode preview doesn't highlight links, end result: links are clickable. So now I know how to post links and the next post is the real post. I had some issues editing: links started behaving strangely. Sorry for the noise.)
We finally fitted the Pixel Qi screen to my MSI Wind U100.
As I already mentioned, the Wind has to be disassembled further to get at the display. You have to screw the hinges of the display lid off the mainboard before you can open the display casing.
Because the connectors mismatch, I ordered a display cable for a Lenovo S10-2 to be a donor for the connector. My friend Sprite_tm is a very capable modder: http://spritesmods.com/ has many interesting smart hacks and mods. So he could quickly solder the original Wind display cable and the donor cable together. The resulting cable connects the Wind motherboard to the Pixel Qi display. All the wires were simple wires, albeit very small. No mini-coax or hassles like that. The ground wires had no insulation inside the cable, only at the end at the connectors.
The interesting part was the backlight. The Wind can't turn it off normally; it doesn't even have the LED_EN signal the Pixel Qi panel wants. On experimentation, Sprite_tm noticed the LED_EN input exhibited some Schmitt-Trigger like behaviour. The Pixel Qi datasheet only mentions TTL-level signals, no more details. I'd appreciate a comment from @pqsupport regarding this. Anyway, Sprite_tm built a very small piece of electronics[1] to exploit the behaviour of the LED_EN input. If the lowest backlight level is chosen on the Wind, the output voltage of the small circuit is low enough to de-assert the LED_EN signal, and the backlight goes off. Increasing the backlight by one step asserts LED_EN and we have a very low brightness backlight. It's not completely to spec, but it works very nicely.
Note that if you want to use the Pixel Qi panel without the LED_EN signal, you'll have to pull it up to some supply voltage; leaving LED_EN floating means the backlight is off. It might not be a good idea in any case; just pull it to a defined level. Be careful with the supply voltage for the backlight; it might be very high, so it is not a good choice in general.
The small circuit needed for LED_EN can be seen on photos [2] and [3].
You can see everything working, but disassembled, on photo [4].
I mentioned before that the physical dimensions of the panel are slightly different. I decided to do nothing about the amount of padding at the back of the panel; it is only half a millimeter difference. However, part of the padding at the back was on the old LCD panel rather than the casing. I experimented a bit with the location for this padding in the new setup. The back casing is rounded at the bottom, so when I put the padding too far to the lower edge, the display would not fit in the case. In the end I rather randomly stuck it to the Pixel Qi panel as seen on photo [5]. Because the cable runs behind the panel completely differently from the original setup, I had to cut up the padding and keep holes for the cable to go through. Starting from the connector at the back of the panel, the cable first goes downwards, and then the smaller, round cable goes back up again through the hole in the padding that can be seen. Behind the panel, it connects to the remnant of the original cable. The original cable is very flat and is not bothered by the padding on the panel; it ran like that originally too.
The front of the panel also has slightly different dimensions. Part of the metal bezel of the Pixel Qi panel is visible at the top and the bottom of the screen. A black felt-tipped pen obscured this very satisfactorily. The bezel of the casing obscures part of the leftmost and rightmost pixels, but this is to such a small extent that I am completely not bothered by it. A Linux textmode console still reads fine; since the text on such a console runs right up to the edges, it is one of the cases where obscured pixels would stand out the earliest.
I really like the Pixel Qi panel. I did not like the sticker that says "Sample For Evaluation Purpose Only / No Guarantee For Sales". I'd have really appreciated it if this was mentioned and explained /before/ you order the panel, instead of it coming as a surprise when you open the box. That would have been a lot more sincere.
- It's wise to cover the screen when using a pen to make the bezel black. One slip of the hand and you have a black stripe on your screen!
- Even though the keyboard of the Wind does not touch the screen with the laptop shut, I did see scratches in the shape of keys on my old screen. So apparently it is compressed a bit further during transport; at least the way I transport it. So I made a screen protector out of a piece of felt to put between it. It is slightly too thick, so the magnets keeping the lid shut do so with less force.
@DigitalBrains There is a weak pulldown on the LED_EN signal. That may partly explain the behavior you see. If the Wind doesn't have a pull-up on LED_EN, the backlight should never turn on, but if you've made a custom cable, you can connect LED_EN to any driver you like. It looks like your PWM to LED_EN circuit is a simple PWM to voltage converter. It's preferable to add gain (just a buffer) to the output to make sure you're driving valid voltage levels. Intermediate voltage values generally don't make good input signals. Both LED_EN and PWM should be rail-to-rail digital signals, but I understand your fix uses only passive components and wraps up nicely. As you say, it's not in spec, although anything from 0 to 1 V on LED_EN should shut off the backlight and anything from 1.6 to 5 V should turn it on. Also, you lose a little bit of range on the backlight adjustment (you lose the 10% brightness setting), but that may be ok for some users.
As for connecting LED_EN to a dc value, the 3.3 V power supply is a good choice if you want to leave it permanently on. Definitely do NOT connect it to the LED power supply, which may be as high as 21 V.