Some thoughts under experiment:
It could be based on 4mm*4mm MU series,
It could be stackable by FPC,
It could reuse current shield whenever necessary,
It could be as small as a stamp,
It could be as thin as a bookmark,
It could be scrolled around your wrist,
It could be attached in your belt,
And it could still be a full functional arduino...
Sounds interesting. My favorite points:
- And it could still be a full functional arduino...
- It could be as small as a stamp
Lefty
It could be as small as a stamp,
It could be as thin as a bookmark,
It could still be a full functional arduino,
It could be under $10
Small means it can be put in more things - a tiny lilypad the size of a button perhaps. Could all the pins be in a 0.1" pitch grid, packed in 2 rows round the edge? LED in the centre on top, chip on underside.
We might have only 10 0.1" pitch holes for easy prototyping, and keep the rest for FPC connection, like stacking shields but in a chain way.
What I would like to see is a pin-for-pin replacement of the ATMega8/168/328 - basically a small PCB (could be larger than the socket), with 0.1" spaced male-header "legs" that could be fit into the socket. It wouldn't have to be more powerful than the standard ATMega.
I guess what I worry about is what happens to prototyping with our Arduinos and breadboards when we can no longer get the DIP versions? What happens when we can no longer get them, and we accidentally blow our chips? I would hope that somebody somewhere would have created a replacement that could easily plug into the Arduino PCB. If it were more powerful in some manner, that would be a bonus.
With that said, even a chip-like, "Basic Stamp"-like "chip" PCB that could be used for prototyping on a breadboard would be welcome; something that could be easily embedded into other products and such - maybe it could become the core for a successor to the Arduino?
I can't wait to see what you guys come up with!
If it is so small it is likely to go into battery powered devices. I would want:
- switching regulator
- option for watch crystal instead of 16 MHz crystal
Udo
Yeah, I think you're on to something there... If it's small, it should have some buck up or whatever to be used with a single cell lipo or 2 AAs or even a 3V coin cell. A modular concept, small and thin modules that can be mounted everywhere would be great to have.
- Wrist-watch format, with 2x12 LEDs all around
- on board radio module (e.g. RFM12)
- powered by 3V (running at 8MHz).
Definitely radio ! 
I think that radio seems like a bit much to ask at that size... that might have to be a shield. And I would love something like this in the form of a breakout board, sort of... as long as I have .1" through-hole sockets I'll be happy. And Udo, are you thinking of low-power applications or timekeeping? I don't think that anything from atmel will actually save much power running at lower speeds (TI has some stuff for that) and if you want a clock I'm sure they could make a small RTC shield. I like the dc/dc converter idea. Perhaps a boost converter to run off a very small LiPo, Li+, or one of those tiny little things you get in dollar store laser pointers?
If it features a 644P it can have a separate crystall for time keeping AND achieve low power. My experiments show that it is quite feasible to average achieve power consumption <10uA. I actually went down to 4uA and I think this is not the end of the pole. But I agree that an RTC module would be simpler to use for timekeeping.
So I am aiming for low power here.
Udo
Interesting... I was under the impression that the AVRs would draw more or less the same power in between operations as during operations. However, one of the kinks to work out would be getting arduino to work with that, as well as serial communication. The best baud rate you could hope for is 1200 baud, or 2400 if you can have a prescaler of 0 (I'm not sure on this one) but that would have significant error. Perhaps a normal mode and then a power saving one as well? And on the note of clocks, to save space on such a small board I think using the internal oscillator might be good. I think there is a way to get it to about 16mhz, but I don't know how. Lots of calibration and fuse changing.
Wyager: you should have a look into the datasheets. The AVR family has quite different power consumption characteristics depending on the model, the voltage, the frequency and the fuse settings / modes.
Udo
I make a point of avoiding any datasheet from atmel... lol. Any datasheet from atmel leaves me more confused than when I started, so I just look on the internet.
Well, then you will face a hard time. I figured that the datasheets are actually the best sources of information and that it is worth to learn how to read them. The information on the internet is most of the time derived from the datasheets and often very much less accurate. Especially if you want to go for the low power modes you will need to consult the datasheet and the application notes. Otherwise you will get a solution that is most probably much worse than it could be.
Udo