Embedded Linux Hardware Resources?
jessecurry asks: "I've recently come up with a conceptual Linux based piece of hardware and have been able to find a huge amount of information regarding getting Linux on to a device, but almost nothing about creating the device itself. I'd like to know if there are any books, online guides, or software that would help in designing a device that would accept some flavor of Linux. I really don't want to go 'off the shelf', but I need something that can at least display graphics, respond to positional input, and play sound. Also, is there a good place to have all of these components put together once I have a finalized design?"
On how many you want! The answer is very different even for 1 or 20, let alone two million.
The RTAI mailing lists are a good place for this kind of question however, and handhelds.org has a lot of links to embedded Linux projects.
Beep beep.
Some folks swear by 'em, some folks hate 'em. Check out http://www.oreilly.com/catalog/dbhardware2/ Covers a range of embedded devices from PICs to DSPs and talks about the various buses you'll see.
If you want to make something useful you should be concerned about what the device is supposed to do, how compact it is, how much it should cost etc. Whatever OS is selected (if any) it should logically follow from the goals of the project. Unless this is an academic or work experience building effort, the OS selected is just a detail. Otherwise, it's a bit like asking "how can I build an embedded systems that uses 2k resistors".
You'll never be able to make something on your own as cheaply as an off-the-shelf system that's mass produced due to economies of scale. Focus on the end-product and don't get caught up on trying to do everything yourself. I'd buy an off-the-shelf hardware system.
Look at the gumstix: http://www.gumstix.org
Their boards will do everything you've mentioned, they have good documentation, the schematics are under creative commons, and it runs Linux.
Why do you want to build it yourself?
While Linux runs on ARMs, be assured that building such a thing is not much different from designing a mainboard with bridges... yourself.
That's why I'm asking why.
- either you want to make it for fun, for learning. that's fine, go ahead
- or you have your idea about some software running on embedded Linux which you want to make into money. Then don't waste your time creating new hardware. Two links for you: <a href="http://www.pcengines.ch/wrap.htm">PCEngines WRAP</a>, <a href="http://www.soekris.com/">Soekris</a>.
Linux is an Operating System - it has drivers for *existing* hardware. It is really hard to write drivers (or modules in Linux-speak) for dealing with non-existing hardware - I see real problems with testing it. :)
So obviously if you plan to build hardware that will run Linux you need to use existing chips that have support in Linux. On the other way you may be willing (but there is no economical point for you really) to design some brand new hardware - just make sure it is atractive so few linux-heads-devs will use it and also make sure you release full specification of it.
...to my finish my robot. My GIRL robot. *wink*
PS: Anybody know a good inexpensive source for lubricated latex sleeves?
See http://www.linuxdevices.com/ for probably the single best central resource for what you can buy off the shelf already.
Get a reference board from one of the chip vendors. You'll find that they've generally already built a board that does anything you want to do and more. Then you can strip the features you don't want, and the design work is essentially done.
In my experience, Cirrus Logic is very Linux friendly, with good driver support for their ARM processors and dev boards.
www.mini-itx.com has a bunch of stuff to build small computers. Starting with something like a standard mobo means you are more likely to get a working system sooner. A mini itx board plus flash memory (instead of a hard drive) plus a tiny dc power supply plus www.damnsmalllinux.com gives you a working system with minimal expenditure of time, effort and money.
When you have something working, you can think about a smaller board, less power consumption, etc.
What you're describing sounds an awful lot like the new Hot-E from http://www.thinlinx.com/ This product has been in development for a while and it's now in production. I'm using it to do Linux/X touchscreen point of sale for restaurants, bars and retail.
Sound like you want a Chumby.
(Saying from the get-go that you want your device to be linux-based is probably a bad
idea, however.)
There's like 4 levels of "embedded" systems. Depending on who you talk to, some of these don't qualify. But "embedded" has been unclear for the past few years.
Lowest level - True microcontroller. AVR or PIC. They can easily display LCD graphics, MP3 chips exist for sound. If you can get away with a true microcontroller, DO IT! It isn't Linux, but the AVR line is heavily supported by the open source community. Some of the best developer tools for it are GCC based. It can also easily do DAC and ADC. Super cheap per unit costs (some systems under a dollar in high enough bulk)..
Next level - Gumstix style system. Has a heavier OS, but much more hardware support. Also, you have the advantage of many premade libs for you to use. They usually use 200-400 mhz processors. Also have the advantage of much more storage. Can usually access CF and SD cards.
Mini itx / Nano itx / PC 104- The manufacturing industry uses these a lot. Still solid state, but basically a small PC. REALLY small PC. Many times 1/4 the size of a laptop motherboard.
PC - Some people consider limited resource PC's embedded. I generally don't, but that's me. The manufacturing industry uses these. Think shuttle pc or microatx.
As I said before, depending your programming experience, it's almost always best to start lowest and go up. Your unit costs are cheaper. Devices use less power and not as hot. Much smaller. Etc, etc. Basically, all these things mean more profit to you.
If an officer ever threatens to taze you, say you have a pacemaker.
Uh, most likely is that his vision is that his device *will* in a second generation be mass produced to take advantage of economies of scale.
I expect this is so he can have a working prototype/first-gen device when he goes to see VCs for funding version 2.0.
And this makes it an excellent plan; since VCs often screw you if you go to them when you only have a concept.
I'd take a wild guess that greater than 90% of Linux embedded systems use some kind of ARM processor. Since that design has been licensed far and wide, you can probably find a customized one with just the peripherals you are looking for, from some company or other (e.g. Phillips, Atmel, Freescale, etc. etc. - just search digikey or mouser for some ideas). It uses a 32-bit instruction set (good for Linux, and good for porting code in general), typically has an LCD driver on the chip, and you might even be able to find one with a DAC. For high-quality audio output, you can use a Wolfson DAC (again there are many choices depending on whether you need a headphone amp, ADC, line-in, mic pre-amp, mixer, how many channels etc.) The ipods tend to use these chips (the 2-channel kind with a headphone amp) so some of them are really cheap because of the economies of scale. For even better quality, the best is probably Burr-Brown (as used in the SlimDevices Squeezebox).
As another poster suggested, you can do prototyping with a Gumstix. Just the display may be a pain with those because they use that teeny-weeny surface-mount Hosiden connector for most of the I/O lines, and you will probably need to build your own display adapter board. But they do have an audio daughterboard already. For something more expensive but a little more ready-to-go (as a prototype), check out the offerings from Arcom; I've used those as well. Or look for some other SBC (linuxdevices.com is a good place to search). You can develop software that way, make sure you know how to use all the hardware you're going to need, and in parallel be working on your final board design (suitably miniaturized if necessary, all on one board, and leaving off the features that you don't need).
All info that you need can be found in eDonkey or other p2p networks.
You've asked about how to create a device that can run Linux. That is an embedded question, nothing specific to Linux itself.
Look for embedded forums, start with the sparkfun, philips semiconductors, lpc2000 and similar forums.
Any hardware you work on or will build will highly depend on the type of CPU and system (including memory size and design constraints like power). I suggest you start with just making root and kernel images for something mundane like cisco 2500 routers and Linksys WRT54GS routers, then getting on with your own designs.
Go buy a cirrus logic board from Olimex.
"Give orange me give eat orange me eat orange give me eat orange give me you." -Nim Chimpsky
I've used the Soekris Engineering boards for m0n0wall based firewalls before. They come in different models for different purposes: wireless, vpn (encryption acceleration), general network/communication, etc.
A friend of me is working on an -El Cheapo - design using cirrus logic EP9302 SoC. You basically need some RAM and flash...and thats it. This device has all you need, including USB,LAN,SPI,...
Other variants (series 931x) even have VGA-out!
Our main problem is the flash memory needed, since its out of production. You can use some substitute, fiddling around with the address lines, but we decided to have a second look at the bootloader, and try to boot this baby directly from USB or SPI.
As for the assembly...well, we found a guy in Serbia who has the necessary gear to manufacture 2 and 4 layer boards. His price for a 2 layer board (90x90mm) with ~30 components soldered was 60 Euros for the prototype, 60 for the solder mask, and around 5 Euros for the board and assembly. (Components are on your tab, of course!)
Should you need info about him, send me a message.
Delta-Mike November Bravo Tango
First of all, most embedded systems don't use the same processors as non-embedded systems. They often use microcontrollers that provide additional peripherals in a compact package. Secondly, choosing the appropriate processor is a key element of most embedded designs. Thirdly, even though embedded systems use CPUs that can be used for a variety of purposes, that doesn't make those systems general purpose.
Sure, at a very low level all computer applications do the same sort of things, but embedded product design isn't done at that level. There are real, significant differences between the goals of specific embedded systems and those of general purpose computer systems. Ignoring those differences will make it much harder to successfuly design, build, and sell your product.
Quite a few devices (Wifi etc) often have WinCE drivers available and not Linux drivers. Running Linux is something you want to identify up front.
Engineering is the art of compromise.