Integrating XMOS processors with custom PCB board designs

Technical discussions around xCORE processors (e.g. xcore-200 & xcore.ai).
User avatar
MatCat
Active Member
Posts: 40
Joined: Wed Dec 18, 2013 11:51 am

Post by MatCat »

segher wrote: U and A series chips should be easy to integrate on two-layer
boards (the datasheets show examples, looks quite straightforward).
Are you saying 0.8mm ball pitch is too small? That's what, 16
mil space per trace?
The 0.8mm pitch isn't the issue really, at least not for the 2 rowed 96 pin ones.
User avatar
MatCat
Active Member
Posts: 40
Joined: Wed Dec 18, 2013 11:51 am

Post by MatCat »

mon2 wrote: Err..think you mean pitched :) Sorry, spelling police now off.
typo hehe.
User avatar
segher
XCore Expert
Posts: 844
Joined: Sun Jul 11, 2010 1:31 am

Post by segher »

MatCat wrote:0.8mm pitch isn't the issue really, at least not for the 2 rowed 96 pin ones.
So what is? The ball pitch (and the escape pattern) determines
what trace width you can use, which sounds like it will fit with
your board manufacturer just fine?

I'm not a hardware guy as you might have noticed; but I a) like
to learn stuff, and b) I do need to know basics like this ("will it
work on a cheap board"), esp. after some less than overwhelming
experiences with the QF124 thing ;-) I was under the impression
these FB packages are excellently workable, and you're saying they
are horrible.

FB217 is just two rows as well, fwiw.
mhelin
Active Member
Posts: 52
Joined: Fri Dec 24, 2010 10:54 am

Post by mhelin »

It would interesting to try making (meaning ordering by using ready made design) a PCB for the FB217 parts and soldering the parts. There are two problems though, one is that there are not any design files for FB217 parts (startKIT would be one but I haven't found any desgin files for it) and the other is the availability of the parts (see https://www.xmos.com/en/buy). There are none at Digikey. It would be fine if XMOS could give us the PCB design presented in U16A datasheet (supplemented with headers) so we could try it. It would though be nice if those 0402 parts would be replaced with larger size packages (0603 or 0805) first. Regarding the available PCB manufacturers the size of the board should fit to 100x50mm or at least 100x100mm size to keep the costs down. BOM parts should be available from Digikey.
User avatar
MatCat
Active Member
Posts: 40
Joined: Wed Dec 18, 2013 11:51 am

Post by MatCat »

mhelin wrote:It would interesting to try making (meaning ordering by using ready made design) a PCB for the FB217 parts and soldering the parts. There are two problems though, one is that there are not any design files for FB217 parts (startKIT would be one but I haven't found any desgin files for it) and the other is the availability of the parts (see https://www.xmos.com/en/buy). There are none at Digikey. It would be fine if XMOS could give us the PCB design presented in U16A datasheet (supplemented with headers) so we could try it. It would though be nice if those 0402 parts would be replaced with larger size packages (0603 or 0805) first. Regarding the available PCB manufacturers the size of the board should fit to 100x50mm or at least 100x100mm size to keep the costs down. BOM parts should be available from Digikey.
The datasheets actually are some of the better ones I have found for showing how to layout the board design.

As far as what I was talking about is the 144 and 512 BGA's, those are not getting done at any of the 'cheap' PCB companies, you would HAVE to have multiple layers and the ability to get sub 10 mil holes drilled to get to all the balls.

I am unaware of a 217 ball BGA, neither Digikey nor Farnell list one of those. There is the 124QFN which probably could be done on a 2 layer with careful VIA work, but would be iffy.
User avatar
Bianco
XCore Expert
Posts: 754
Joined: Thu Dec 10, 2009 6:56 pm

Post by Bianco »

XMOS plays in the high end microcontroller ball park, certainly their higher end SKUs are not optimized for cheap hobbyist/semi-professional use. And they shouldn't!. The G series are for either high volume or low volume high end products. In the first case the additional costs for PCB are negligible, in the second case, the production costs do not matter as much anyway. If those customers are happy, then it's good.
Increasing the pitch to 1.0mm would increase the package size to at least 25x25mm and it was already quite big. That said, G-series are not really recommended for new designs anyway.

To put it rude, hobbyists and semi-professional hobbyists are only good for one thing: they potentially may generate revenues in the future by convincing their work environment to use XMOS chips. They don't generate enough revenue by their own usage. To give an example: there are maybe 2 million arduinos + clones sold. Say that each AVR is sold for 2 euros, then the total revenue is about 4 million euros. This is nothing compared to the yearly revenue of 1400 million that Atmel has. And in the case of Atmel I even believe that it didn't even helped THAT much to get new design wins that they otherwise wouldn't have had. (Parallax is the only company that might prove me wrong :p).

This view may sound very economic but in the end everything is about money.
User avatar
segher
XCore Expert
Posts: 844
Joined: Sun Jul 11, 2010 1:31 am

Post by segher »

Bianco wrote:To put it rude, hobbyists and semi-professional hobbyists are only good for one thing: they potentially may generate revenues in the future by convincing their work environment to use XMOS chips. They don't generate enough revenue by their own usage.
I agree, except with the very narrow point of view that they are
only good for convincing employers. Hobbyists broaden the
ecosystem, which helps find bugs, exposes opportunities, grows
mindshare, what have you. Even if they do not generate revenue,
they still generate value.
User avatar
Bianco
XCore Expert
Posts: 754
Joined: Thu Dec 10, 2009 6:56 pm

Post by Bianco »

You certainly made a valid point.
User avatar
LyleHaze
Experienced Member
Posts: 71
Joined: Wed Apr 11, 2012 6:21 am

Post by LyleHaze »

Bianco wrote:...This is nothing compared to the yearly revenue of 1400 million that Atmel has....
edit: Those numbers are confirmed, my original comment was in error.

Regarding the cost of supporting "hobby" developers, I can see both sides of the argument. Comparing the cost against the purchases doesn't account for the value of public exposure for your product and your company. On the other extreme I know of a few engineers that flat out refuse to use certain brands because of their poor support policies. If you alienate the product designers, you will certainly lose opportunities for profit as a result.
User avatar
MatCat
Active Member
Posts: 40
Joined: Wed Dec 18, 2013 11:51 am

Post by MatCat »

Ultimately I like XMOS, I don't think it's that bad for design, my real gripe is with I/O on these U6A and U8A :), I think 32A could be better served to have 8 bit and 4 bit ports as secondaries since most of the chips break those pins out, at least on the UxA line. Personally I am going to develop a couple of xcore dev boards, mainly because I see good potential for future product and hobby use. Sure I am not a big company, I am not going to move millions of dollars in chips on a production run, but the thousands of products I may one day get out there could influence others to consider xcore, plus my company develops all of our products open source, so it contributes to the eco-system, which of course the more hobbyists there are the more there will be in that eco-system. I think in the long term it leads to a lot of people doing small runs of products that total up to a lot of total chips.

I do not know the stats of Atmels sales, I would love to see some stats on how many of their sales are for large production runs vs small batches and vendors.