User:Bluefoxicy/FeatureSuggestions
I'm sticking feature suggestions here for the second revision, which I'll roll into the discussions on the mailing list when appropriate.
I am not an engineer. Do not assume I'm engineering awesome stuff or know what I'm talking about at all. Ask a professional.
Expandability
The OLPCs are not very expandable or even repairable. If RAM goes, if NAND goes, if the CPU dies, we're left with an unworking model. Some parts should be replaceable and even expandable. Here are my suggestions.
Note that I don't include the CPU in these. Replacing the CPU serves almost no purpose, aside from actually repairing a damaged board.
NAND Card
An attack on the machine may destroy NAND, leaving the machine crippled to USB2.0 or SD booting. This could be a lot slower; and USB drives will dangle off the side and be exposed to damage. Besides that, 512M bay be too little NAND in the future.
R2 should feature a riser connection for NAND, allowing a PCB containing traces reading to an edge connector to be inserted into the side. The connector for the card would be a simple card slot. Being a simple riser, there would be no controller chip for the card; it would be a 1:1 replacement of the NAND soldered on board to a bunch of traces leading to a card slot at the edge of the board.
When the NAND card is inserted into the laptop, it should slide deeper than flush. A cap can then be placed on which fits snugly due to a rubber O ring around it. The cap could be removed by sliding an object into a handle flush with the edge of the laptop, and tugging it back out; otherwise it should not wobble or fall out, and should be water tight.
It may also be possible to place the controller on riser, so the flash controller itself can be updated. This may be difficult; although it could be slotted in the same spot, see diagram below:
Board |
------------ -------------
|| <-- card =|[XXXXXXXX] |
|| slot =| ________|
|| =| |~-~-~-~ |||||||
|| =| | card |-------|
|| =| | slot |[NAND] |
------------ ----- |_______|
| Controller NAND
| Card Card
| [X] - NAND Ctlr
Shown in the diagram, a riser on the board connects to the NAND controller, which has a NAND controller chip on it. A further riser on the NAND controller connects to a smaller card containing a NAND chip. Note the design layout chosen does not allow NAND to be inserted directly in without removing the controller card; this is intentional. It also prevents having three major "thin" sides and wasted space, the surface area of the NAND controller card is as large as possible.
Problems with removable NAND controllers and cards include generally routing circuitry to the edge of the board. It's convenient to plod the controller and chip on the board somewhere open; when you have to remove the controller from the side, it has to be at the side, and the traces may not want to go that way. The CPU or south bridge would likely have to be closer to the edge of the board to make this work; this is not easy, because of what the CPU has to get to.
One possible design to solve this would be to use a ZIF socket, where you slide the controller/NAND card in at an angle, then tilt it parallel to the board and it locks in place. This would allow the ZIF socket to be placed anywhere on the board, and accessed by removing a back panel; however, it would also increase the depth of the board, unless (not recommended) a hole were made in the board that the card tilts into. Further, a large, removable back panel would have to be added, which would be difficult to get to seal and stay in place as reliably as a small cover; or the laptop would have to be disassembled to access the guts, which leaves the problem of finding a screwdriver and stripping or grinding the screw heads to the point that you can't remove the screws.
In this design I have not considered how the NAND size affects the circuitry. The design should be such that the flash controller chip can interface with a card containing various sizes of NAND. For example, if the controller chip supports up to 4G of NAND, it should be possible to insert a 512M, 1G, 2G, and 4G card without added controller circuitry on the card; further denominations, such as 1.5G, 3G, or "any reasonable iteration" may also be supported, to what is technically feasible and cost effective.
Another not-considered issue is that switching controller cards is probably not feasible; different chips will likely have different pin-outs, creating the upgrade problem present with desktop CPUs.
RAM
RAM is the next biggest thing to consider upgrading. RAM is not likely to be damaged by a DOS; but it is likely that increasing RAM will increase the range of effectiveness of the OLPC, by boosting performance and allowing a larger working set. An expansion option would be nice.
Future revisions of OLPC may simply be the same board as their immediate predecessor with more NAND and RAM. If the CPU is the same or is still an x86, or in any way capable of running the new OS natively, then it would make sense that an old machine could simply get a RAM and NAND boost and run the new OS. For target countries where kids are poor, a $20-$30 RAM and NAND upgrade to get a newer, faster machine may become interesting; especially if the new OS and software has some substantial benefits for education.
Another interesting thought is that some users in higher grades may simply wish to boost their storage space and memory, in the same way some of us get an Athlon 64 X2 with 2 gigs of RAM just so we have a beefy system. The laptop is given to the children to take home, to do schoolwork on, and to generally carry around; there is no reason they should not be able to buy upgrades through the school if they actually have the money (aside, of course, from the fact that Jack Thompson would condemn us the first time a 14 year old realized he could get his system to run Doom).
Adding more RAM would be similar to adding more NAND, with the RAM moved to a riser card on a ZIF or side socket; commercial laptops have a lovely design already. The second revision should allow for expansion through adding to the existing RAM, and the existing RAM should be removable in case A) the laptop breaks and it can be reused in another machine; B) a 256M expansion is added (384M total), then another is purchased and the stock memory is in the way. Stacking RAM by having cards that accept cards may be a cute idea, but not likely useful.
Keyboard and Touchpad
An easy target, the keyboard could potentially break, or the laptop could be rendered useless and the keyboard could be still useful. The same goes for the touchpad, which is likely to stay functional due to lack of mechanical stress but could still be useful once i.e. the screen is destroyed.
The keyboard and touchpad could be secured by rubber seal around the edges, and be replaceable by removing it and inserting a new one. A proprietary form factor with standard USB connector on the bottom could be used; this is similar to HP media center drives, which are standard form factor hard drives enclosed in a proprietary USB enclosure that slides into the HP media center and connects with a USB port.
The keyboard and touchpad would have to latch tight into the machine so that it wouldn't fall out when dropped. Control for the latch could be a physical toggle on top the keyboard, appropriately sealed from moisture; when one side is depressed the latch is in place, depressing the other side disengages the latch.
----------- ,----, -----------
|/______\|
Loop to insert lanyard through,
flush with surface
The keyboard and touchpad would stay in place with the latch disengaged due to the O ring seal; but could be removed by pulling, with an appropriate handle in place to slide a lanyard through, allowing for any thin, strong, fibrous strap to be used to remove the keyboard. Inserting the keyboard should disengage the latch; if the latch breaks down, it should still be possible to remove the keyboard.
The touchpad and keyboard could be integrated; but the keyboard is much more likely to break due to mechanical stress on the keys. Isolating the two components will allow repairs to be made more cost effectively once in production.
Enhancements
These suggest enhancements, such as durability enhancements.
Non-Mechanical Keyboard
Keyboards eventually break. Mechanical parts are inherently flawed because they wear quickly; cite the iPod with a spinning wheel versus the fifth generation iPod with the touch sensitive wheel, scrolling the 5G causes almost no wear while scrolling the original iPod causes mechanical stress on gears inside the iPod itself. Keyboards are the same way, and keys eventually weaken from being mashed and eventually fall off.
A non-mechanical keyboard would last much longer than an equivalent mechanical model due simply to the keys not wearing out. It would be possible to produce one based on a number of technologies, ranging from sonics to pressure sensitive diodes that change resistance when the keys are tapped to simply making each key a touchpad. The biggest advantage would be that the model would be completely sealed, except for the connector on the bottom in the above replaceable keyboard suggestion; this would further keep moisture out of the laptop.
A solution where the sensors are inside the keyboard would be most favorable. If small microphones are under the keys, or if the plastic has a diode in it that changes resistance when the key is tapped due to sudden compression of the diode from the vibration of the plastic, then scratches and dents in the surface of the plastic are largely harmless. Surface sensors as used on touch screens can be scratched and otherwise damaged.