Slashdot Mirror
More Fun With 1 Chip Systems
Anonymous Coward writes "Axis, maker of the ETRAX system-on-chip, has just announced a multi-chip module that has the same pinout as their ETRAX SOC but which contains over 50 components. According to this article at linuxdevices.com, all you need to add is an external 20MHz crystal and power, to end up with a fully functional networked Linux system. It contains 8 megs sdram & 2 megs flash. Now you can really put Linux anywhere! The estimated price (when it hits production) to buy these little goodies is $75 (qty 1) down to $50 (qty 10K). "
> Is there any way to communicate without having an accurate clock?
Yes, you can. I used to work on microcontrollers that had a low power mode wherein they shut off the external crystal, and instead used an internal ring oscillator for a clock. This is a big power savings, but the oscillator made for a poor clock: it was slow and it's period varied wildly. The uncertainty made it unacceptable for anything requiring accurate timing...even, in some cases, memory accesses. Still, it was useful for some things.
I think most commonly, the micro would go into this low power state, and sit in a loop, periodically polling it's serials. If anything interesting happened (say, change in position of a machine assembly), it would kick the clock back on and, once stable, execute the appropriate control routine, which likely required accurate timing.
Microcontrollers are really quite fascinating. It's amazing what you can do with just a very little computational power. The hackers who stayed up nights bumming single instructions out of their ASM routines in the early days of the PC now have a home in the embedded world, where resources are still tight.
--Lenny
Those need low power systems. I don't think this has been designed to be a low power system. They also need cheep, my last GPS was $120, replacing the DragonBallEX ($25) with this would bump the price, and lose the LCD display...and none of them need a ethernet nearly as much as a 802.11 wireless net...
Those would be a better place. However I'm not sure it would be ideal for the PVR, that needs a drive controller, and an interface to the MPEG encoder(s), or at least the decoder and tuner devices...
I don't know exactly where I would use one. I assume they are already in web cameras though. Actually it might be decent for use in a stereo with a dedicated MP3 decoder (however if you do it all in software you could switch to Ogg/Vorbis later, so there is something to be said for faster CPUs there).
It is a safe bet that they pretty much designed this for themselves, and are selling it to anyone that has a use for it. Not a bad way to go, as long as the people buying it don't cut into your web cam market :-)
I'm not sure a 100Mhz 68000 can do great machine vision algos, but it would be pretty good for less intensive tasks. (Yes, it isn't a 68000, but despite them advertising it as a "RISC" CPU, the instruction set is very very similar to the 68000...and I don't imagine the performance will be any better...)
I wonder if the computerized machine tool manufacturers have thought about this?
Yes, for this application the ethernet support would be an asset. But what support does the package have for analog/digitital IO? This is the be-all and end-all in industrial control. Even the lowly 68HC11 has onboard A-D.
Also don't forget, industrial control tends to be the wooden-spokes and horse-powered side of the industry. It takes absolutely ages for anything current to find its way onto the factory floor.
--
Life's a bitch but somebody's gotta do it.
So then, where do you get someone to solder your chip to the bga adapter? it still has the same problem, getting a reliable and solid connection between all pads on the BGA to the board/adapter.
Heck I found a place that has BGA sockets. They were glad to sell me a socket for $160.00USD it had tiny springs and cone points to secure a solid connection to each pin etc... but adding that kind of money to each BGA chip makes the project worthless to try and forces the home hacker to just either buy a dimmpc or a PC-104 board.
On a side note- Thank you Hitachi for not making only bga, every SH series processor is available as QFPA.
Do not look at laser with remaining good eye.
All the new manufacturers are making thse wonder cips as BGA packages. guarenteeing that no-one other than a massive board house can try to prototype with them. I wanted to do a prototype with the machZ desperately. It fit all specifications and would have been perfect. it is only available as a BGA and will never be produced as a QFPA per the sales and engineering people.
I have tried hot-air to get a BGA to solder to the board without luck you never get all pins soldered without either an commercial IR or Hot air rework station.... and for some reason us home hackers dont have an extra $40K lying around to buy one.
If they want wide acceptance and use... produce it in a hand solderable version (Not DIP, good grief no... DIP is evil) it will make life easy for us home prototype builders and the corperate guys that dont have to take over a rework station just to try a design.
Do not look at laser with remaining good eye.
Isn't buckshot like a Beowulf cluster of rocks?
/me ducks and runs
Enjoy your job, make lots of money, work within the law. Choose any two.
Mainly because it's very hard to build an accurate oscillator in an IC. Some microcontrollers have a "crystal-less" oscillator but they use an external RC (resistor-capacitor) network for setting the clock period. You'd probably get around +/- 20% if you're lucky. They're probably using one oscillator to drive everything (VGA circuitry, USB host, etc.) with a number of PLLs to get the various individual clocks they need, and things like USB and VGA need accurate clock frequencies.
Enjoy your job, make lots of money, work within the law. Choose any two.
Cheap competitor to Palm.
Homebuilt robotics.
Arena-grade laser tag SBC.
Keychain network diagnostic tool.
Keychain security client.
Wearable computer system.
"Avast! Prepare for the rodgering!" THWACK! "Arrr.. me nards.."
I've just finished a design using the LX100.
7 -07-005-21-OS-BZ-EM )
( I'm one of the people quoted in the recent EEtimes article about Axis and Linux :
see http://linuxtoday.com/news_story.php3?ltsn=2001-0
You can have SCSI / IDE etc - you just need to add some resistors and cheap buffers.
The rules about what can be combined with what are complicated and you have to work through them.
I've managed to lay my board out so it can be I built with :
4 serial + 1 IDE
or
4 serial + 1 parallel + 1 USB
or
3 Serial + 1 IDE + 1 USB
or
3 Serial + 2 USB
I could have got some SCSI options in there too - but didn't bother because I don't need them.
It's a very flexible chip and very capable at what it was designed for - doing I/O
AJB
You can build an on chip oscillator (this is done for virtually any modern RF circuit), however in order to make it slow enough to drive digital circuits you would need large valued passive components on board, these are expensive to implement as they will consume a large die area.
Concievably you could build a prescaler to handle the fast oscillator and scale it down to a managable speed but since a free running oscillator will have a rather imprecise speed, this prescaler must be designed very carefully.
The lack of a predictable clock will also be a detrimental for the end user as performance will vary rather wildly.
The use of an external timing source will enable you to stabilize and normalize the clock speed, which is exactly why you want a crystal.
Use of higher level timing sources like a heartbeat over ethernet is not apropriate as proper reception of such data really requires an accurate timing in the first place.
Generally reception of data without sunchronization is possible as long as the data is recieved much slower than the local clock. This way we can sample the data signal fast enough to make sense of it. I doubt such a scheme is appropriate for ethernet. (since the ethernet driver is on board you need to sample the ethernet signal directly which means you need a timing insensisitive way to transform the ethernet symbols to simple logic level signalling. I doubt this is a trivial task)
There are a bunch of oscillator technologies that don't involve crystals that can be mounted on an MCM. They are either less precise or much more expensive than a crystal.
The ethernet interface needs a precision clock, or it won't work right. So a cheap truly solid state oscillator won't do.
Wow, "27mm x 27mm PBGA IC package". Now I can put Linux into my Nike's and boot my win98se POS out the window....
Is there any way to communicate without having an accurate clock?
Morse Code has no accurate clock, and works just fine. (I'm not joking or patronizing in pointing this out. Your question is a good one.)
A "latch" signal is one which tells a related circuit that it is okay to perform some task: "the data is ready to read", "the data is ready to write", "the inputs are ready to combine", etc.
A "clock" is just a way to organize many circuits on the same latch signal. All data must be ready at the same time, and then the clock strikes, latching all the circuits that use that data. Most simple data circuits will work if you clock them once a year, instead of once every few nanoseconds.
A clock has to be set to a period that the worst-case circuit can cope with. If it takes a circuit a long time to generate its results, the next circuit had better not be latched too early, so the clock is slowed to make sure the results are always ready in time.
Parallel printers work on a single latch, and no clock is needed. Other communications would be possible quite easily without the assumption that latches always occur at a fixed frequency. Ethernet may be trickier, but it's the same principle.
[
Ok, I understand that the crystal is external because it can't easily be integrated, and the oscillators which are easy to build in silicon aren't very accurate.
But I wonder, could an inaccurate on-silicon clock be used to drive the system slowly until it could pick up an accurate time signal from some other source? I'm assuming that the easily built oscillators are relatively stable.
The "obvious" idea of using NTP to query external clocks in order to compute one's own frequency obviously wouldn't work, since without an accurate clock you can't communicate... or can you? Is there any way to communicate without having an accurate clock?
Yes, this is all windy speculation, but I'm hoping that someone out there will have more of a clue about this stuff than I have.
Tarsnap: Online backups for the truly paranoid
I've designed boards with microprocessors on them, and I appreciate the external crystal. The more I look at Axis's products, the more it seems like they actually know what they are doing.
-- ;-)
Kuro5hin.org: where the good times never end.
You can build 'clocks' without crystals, but it involves doing Bad Things with feedback loops etc, and will vary wildly in speed based on tiny temperature shifts, the particular chip used, etc, etc. Some experimentally-computer-generated FPGA configurations have used these, but they are not recommended for sane human designers.
10 of those chips is only 1000 MIPS; that's less than an 800 MHz x86 CPU. Why not just get a microATX motherboard with a Duron?
I wonder if the computerized machine tool manufacturers have thought about this? This is the general trend in automated manufaturing systems anyway, the ever increasing rate of production and the need to reduce overhead costs have always conflicted. To still be able to telemeterize whole systems to minimize downtime could use a small self contained unit such as this. I have always been interested in a CNC machine tool control system that would run on linux. as the current systems all run on purely propriatary operating sysyems or even (gag) M$ win98 or some such, it would be nice to run an open source control as the current controls are bulky and often designed by engineers or computer software companies with no concept of real world necessities or possibilities. A completly open-source, compact control would be a blessing to every real machinist out there.
-- Defenestrate Microsoft!
I've got a great application for this device. Similar systems would do what I want, but would cost over $1000.
I've dreamed of being able to "clip" a tiny device onto an ethernet port (or even put it inside the wall box behind the faceplate) and have it run dsniff and or mailsnarf. Then at 2am it would e-mail the day's results to a throwaway hotmail account, or irc into a secret channel, or somesuch.
This device could be called a "bug". But now with an SOC that costs $75, someone good with a soldering iron could probably build a bug for $150. This puts such a computer network bug within the budget of a teenage hacker. Wonderful. Just imagine the potential.
--
"Linux is a cancer" -- Steve Ballmer, CEO Microsoft.
I'll see your senator, and I'll raise you two judges.
here that already does all that for you.
"A door is what a dog is perpetually on the wrong side of" - Ogden Nash
Hmmm, seems like one or two early posters seem to think that we will see these chips in desktop PCs (asking about USB, ATA, SCSI support, etc). But if you expect to find these ETRAX system-on-chip units in a desktop PC, then you'll be looking in the wrong place - because they'll be in embedded systems.
With 8MB SDRAM and 2MB Flash memory, these chips could be running your next souped-up GPS device, your next smart (landline) phone, your internet-savvy fridge, your second generation PVR, etc, etc.
As we move to a "chips in everything" society, lower power, highly integrated processors like this one will run countless devices. Today they may cost $50, tomorrow they will cost a fraction of that and will be just about everywhere.
"Accept that some days you are the pigeon, and some days you are the statue." - David Brent, Wernham Hogg
1. It doesn't matter how blank-boxed they are, as long as they are small and cheap?
2. There isn't a huge interest in hacking computers per se; rather, what we care about is modularity-for-the-buck? (I suspect this is the case for me: what I want are things I can buy for under $100US that can be hooked to other things in the same price range in some meaningful way to build (random-string (cons '(a gravity meter) '(an automatic bubble maker) '(a marsupial trap) '(a woodpecker conditioner) future-projects-list)).
3. Most people here only read every fifth thread these days?
-- MarkusQ