Slashdot Mirror
PCMCIA Computer Project Aims Even Higher (and Cheaper) Than Raspberry Pi
lkcl writes "An initiative by a Community Interest Company Rhombus Tech aims to provide Software (Libre) Developers with a PCMCIA-sized modular computer that could end up in mass-volume products. The reference design mass-volume pricing guide from the SoC manufacturer, for a device with similar capability to the Raspberry Pi, is around $15: 40% less than the $25 Raspberry Pi but for a device with an ARM Cortex A8 CPU 3x times faster than the 700mhz ARM11 used in the Raspberry Pi. GPL Kernel source code is available. A page for community ideas for motherboard designs has also been created. The overall goal is to bring more mass-volume products to market which Software (Libre) Developers have actually been involved in, reversing the trend of endemic GPL violations surrounding ARM-based mass-produced hardware. The Preorder pledge registration is now open (account creation required)." Of course, the Raspberry Pi is not only only much further along, but has recently announced an expansion module (the Gertboard).
"Mass-volume" pricing is manufacturer speak for wholesale prices, as in buying thousands of units at a time. You expect those prices to be half or less of retail. So a $15 OEM price will be about $30 at retail, generally speaking. That compares reasonably well to the $25 retail pricing of the Raspberry Pi, given that this new board has somewhat higher specs.
Put my fist through my alarm clock with its ding-dong death inside my ear. - The Blackjacks.
PCMCIA was originally designed as a memory card form factor. It was later thickened up for use as an expansion card form factor.
Also I think being thicker would have doomed it sooner as laptops got thinner.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
These pinouts make no attempt to be electrically or electronically compatible with the legacy PCMCIA standard. 16 GPIO pins, 24-pin RGB/TTL, USB2, I2C, 10/100 Ethernet and SATA-II interfaces are included in the Version 1.0 specification.
good man! feel free to fill in the preorder form http://rhombus-tech.net/allwinner_a10/orders/ i'm a bit reluctant to do it on your behalf [aitch tee tee pee slashdot dot org slash tilde hatta]
please do bear in mind that in the early stage we're *not* going to sell completely untested cards in mass-volume right away, that would be foolish. we're going to follow the process that Dr Schaller has been doing on the development of the GTA04 - http://wiki.openmoko.org/wiki/GTA04_revisions as have various other projects, OpenPandora included.
so, early alpha boards go out to people prepared to take a risk, but who have the money spare (under $100, gosh, wow, break the bank why not) to consider "what the heck, this is cool, let's support this initiative" but at the same time have some expertise in embedded GNU/Linux development, and they might actually get something that works perfectly first time, and they're the ones that got it, before anyone else.
beta boards go out to people who want something that, hardware-wise, is pretty much guaranteed to work 100%, but maybe the software's not all there, and they might have to (gosh) get involved and help write it.
stable boards go out to people who really would "just like something that works, thank you, where's the debian distro image, where's the instructions for putting everything onto an sdcard, heck, where can i buy a pre-loaded MicroSD card so i don't have to do that, even".
so it's a known trade-off: the principles of Software (Libre) Development as applied to hardware: release early, release often. exactly the sort of thing that you never normally see in the development of hardware products, and i think it's pretty damn cool to be able to witness and be part of something that *isn't* GPL-violating. at bloody last.
Great, so this low cost computer can be plugged into the PCMCIA slot of a laptop.
ah NO! :) the mechanical design prevents insertion of EOMA-PCMCIA CPU cards into legacy PCMCIA slots:
http://elinux.org/Embedded_Open_Modular_Architecture/PCMCIA#Deliberate_Mechanical_Non-interoperability
if you tried to force it in, you would mechanically damage the laptop and/or the card, and once you'd done that, the chances are that you'd blow up the card and/or the laptop as well.
Or you couyld just use the laptop. Am i missing something here?
you're missing something :) the design concept is that the EOMA-PCMCIA CPU card *is* the laptop... but only when the modular CPU card is plugged into an EOMA-PCMCIA-compliant laptop Motherboard that's *designed* to take these CPU cards. see example motherboards here: http://elinux.org/Embedded_Open_Modular_Architecture/PCMCIA#Example_Motherboards
to have an x86 CPU in a laptop already (cost of $300+) and to then put in an extra low-cost CPU card that does pretty much the same job as far as 98% of computer users are concerned, well... that would just be silly. why not just have a modular mass-volume laptop plus CPU card that can retail for about $95, eh? :)
hi pedrop,
yeah we thought about that, but ExpressCard means PCI-e. so we investigated how many SoC "embedded" CPUs have PCI-e and it's like about... 5. several from marvell, but they're the high-end "server" style ones, with power budgets wayyyy over the reasonable limit.
then we looked instead at PCI-e "PHY" chips. there's _one_: it's a PCI-e to USB converter, which says it all, really.
the other option is an FPGA: they're just... horrendously expensive. something like the Zynq-7030 would be able to do it in hardware using the on-board FPGA, but that's not out yet.
the only other option would be bit-banging! updating multiple GPIO pins to emulate a 33mhz ISA bus *shudder*. ... yeah overall we just went "ok, forget it - let's just take over PCMCIA at the connector/header/socket level".
re-use of *existing* connectors, housings and assemblies keeps the price right down
lkcl, before you jump into production please make sure that you don't want your SATA and USB work. Because they aren't very likely to; the PCMCIA connector is not a controlled impedance part, and your pinout requires 90 Ohm differential for USB and 100 Ohm differential for SATA. Ethernet is also 100 Ohm, but it has plenty of margin. Even if the board works on the bench, it's not the right thing to do. You need a proper differential connector, something that you can get from Samtec, for example.
I really don't know how much you are an expert in manufacturing, but I built a number of professional designs, and I strongly suggest that you don't pick an old, obsolete connector just because you think it is cool. You need to consider the other side of the connector. How many PCMCIA cages can you find at Digikey? How many of them are easy to solder by hand? Hell, this connector would give *me* trouble, and I can solder 0402 all day long under the microscope. This connector has pin spacing of 0.635 mm, and practically none of your customers can solder it.
I still don't quite understand the business idea of your product. By "business" I don't mean making money; I mean "delivering value," making good things. What value do you expect to deliver if nobody can connect to your board? Your super-small form factor is a problem here. Very few electronic enthusiasts are so much concerned about size and space. They are far more concerned about being able to see the parts without using an electron microscope. If you'd ask me, I'd say you need to think how your customers are going to use your product.
The talk about standard connectors ... if you want it done right, use COM Express. These modules are interchangeable and your product would actually fit into an existing market. You can actually sell the thing without Slashdot. Inventing your own standard, using an obsolete connector and breaking the electrical signaling requirements will not do you any good. You are not large enough to establish a competing standard, and your design is not as good anyway. But if you don't want to deal with COM Express (which is not a pleasure to solder either, I admit) then just forget the unification and use plain vanilla 0.1" headers for everything except high speed interfaces. Or include a CardBus breakout board with your CPU board.