Arduino to XMOS chip ?

[williamk]

Not sure if this is even possible, but lets say I already have an Arduino kit. Would it be possible to use it to program a flash memory that is then read by a xmos chip? Or even interface with a xmos chip? Could I do all that WITHOUT any xmos dev kit? (yes, I'm cheap, I know)

Just wondering if I could just order one xmos chip and go nuts with an arduino + xmos chip... ;-)

Wk

[leon_heller]

Both are feasible. Buying an XTAG2 for $50 would be a lot easier, and will give you debugging as well as programming.

[bsmithyman]

It would depend on your level of experience and PCB assembly skills. Even the 64-pin XS1-L1 is difficult (though certainly not impossible) to hand solder, and requires careful board layout and a reasonable number of external components. It would be very difficult to roll your own board for less than the XMOS boards (or the Sparkfun board) unless you have your own PCB manufacturing equipment; something like BatchPCB or PCBPool might be another option.

There's no reason you couldn't implement software to program the SPI flash on an Arduino; note that if it's a 5V model there's a good chance it requires level conversion circuitry to talk to the SPI flash safely. However, this wouldn't play well with the XMOS tools, so you're back to writing your own USB drivers for the programmer (or serial drivers, I suppose, if using the Arduino's FTDI chip). Again no reason you can't make it work, but it's an additional expenditure.

I'm not trying to discourage you, creating your own board could be a great way to customize it to your exact needs; however, at $100 the XC-1A or the XK-1 kit are both very good deals. Especially the XC-1A: the XMOS chip alone is $35. I suspect that in single-unit quantities to build the same thing yourself would run to several hundred dollars, unless you have $20k of equipment available to you. Possibly others on the forums could give you some more insight into the hidden costs of doing a board design: examples would be the Amino group and the XStamp work that's already been done.

Hope that helps :)

[Folknology]

I would agree with bsmithyman that the XC1A represents good value, particularly if you want a kit that is multicore. I do not agree that the XK1 kit is good value, far from it in fact but I won't go into that here.

Even low cost alternative proposed like Xstamp and the minimum L1 board still suffer from a lack of programming/debugging ability, as does Sparkfun's board. These require you to pay the Xmos $50 tax in the form of XTAG 2.

As for building your own board its rather tricky, there are many gotchas but it can be done. In the end however you will save very little money building your own (due to experimenting required). You could do everything by hand but you have to be careful about component choices and it will require some soldering skillz! You could even use hotplates and toaster ovens with a pcb from batch-pcb or similar to give yourself more flexibility, but again there are many places where this can go wrong. However if you really getting into this stuff then its a worth while exercise.

Over all the easy route is to buy a kit and if you can stretch to the XC1A its a seriously good board with a lot of room for development/expansion etc...

Just my $0.02 on the question

regards
Al

[bsmithyman]

Thanks Al :)

Williamk: Folknology's done a lot more board design than I have, so I'd tend to trust his feelings over mine on the XK-1 issue. I also know that there are a large number of people interested in the custom board options, so if you're serious about trying it, I suspect they'd be happy to have more discussion and interest.

Also don't forget that you can try out programs in the simulator and the timing analyser. That can give you a better feeling about exactly what the performance will be like for your application, and e.g. how many cores you need.