New kid on the block the pcDuino ...

I've the Wandboard Quad on the bench, worked out of the box !

Added the WiFi antenna, few minutes configuration via the ubuntu system settings, and joined my WiFi network without a glitch.

Very, very impressed with the quality of video, even playing videos through vnc. Nicely done. Blog article coming soon ...

Cheers

Jorge

Thanks for very sharing - very cool!  You seem to have all the cool single board computer toys

Drew Fustini wrote:

 

That sounds like a fun career

 

Off-topic... when I advocate OSHW, one of the points I makes is that it enables a professional engineer to modify a devkit or SBC to optimize for their use case.  Do you ever do this?  It would be interesting to have some examples to back up this theory.

Yes it is a lot of fun, never took a job that I didn't have fun doing it

About your question, it depends on the application and the volume. Most of the time we look for off-the-shelf modular solutions that can help  doing quick prototyping, proof of concept development and small production runs, in those cases it is not worth the extra time adding  uncertainity, risk and cost driven by a new set of design files, parts procurement, pcb fabrication, assembly, etc. When you get a ready made SBC you have a high probability that it will work out of the box at no additional cost, and if it doesn't work you return it and get a new one.

Now if the application requires a lot of customization, the volume justifies the extra design and production costs or you are trying also to increase reliability, having an OSHW design gives you a better starting point than starting from scratch, but again it has a lot to do with the application and the costs.

One of the great benefits of OSHW besides that you can modify an existing design, is that you can learn a lot by looking at the deteails on the design files, for example on dense pcb layouts a bga fan out and scape routing strategy that you know it worked, or placement of decoupling caps, parts used on the design like a PMIC that you can consider using in your designs, etc.

Also having full access to design files helps to perform a more educated analysis of a particular SBC, most of the time in serious engineering you can't keep folks happy with the taste of the sausage, we really want to know what is on it and how it is made.

My .02

Jorge

jamodio wrote:

 

Most of the time we look for off-the-shelf modular solutions that can help  doing quick prototyping, proof of concept development and small production runs, in those cases it is not worth the extra time adding  uncertainity, risk and cost driven by a new set of design files, parts procurement, pcb fabrication, assembly, etc. When you get a ready made SBC you have a high probability that it will work out of the box at no additional cost, and if it doesn't work you return it and get a new one.

Do you get any sense of the embedded industry moving (or even wanting to move) towards modular standardisation of the EDM, SOM or even EOMA-68 kind?  Unfortunately, I don't.  Almost everyone seems so intent to go their own sweet way through proprietary form factors for entirely isolated competition that the concept of partial cooperation on standards to create a massive global ecosystem appears to be entirely foreign to the sector.

I find it very blinkered.  The few companies that are bucking the trend have almost no mindshare and so the few modular systems remain niche and expensive.  As a result, there isn't even a cost carrot to break the vicious cycle.

Morgaine Dinova wrote:

 

Do you get any sense of the embedded industry moving (or even wanting to move) towards modular standardisation of the EDM, SOM or even EOMA-68 kind?  Unfortunately, I don't.

EDM seems to have some mindshare, but the wandboard seems to be the cheapest and most powerful example.  I know some folks here like EOMA, but I'm skeptical. EDM clearly isn't gaining acceptance, so what killer feature would EOMA have that EDM doesn't in already existing products ?

Perhaps we should be aiming for R-Pi form factor instead

Nope, I don't see any clear signs of any coordinated effort to generate an industry standard for modular SBCs, also I don't know how feasible that could be with such a diverse universe of SoC or application processors.

What I see as modular is something that can be taken as a "module" and integrated into another board for a specific application, good solutions are those where most of the interesting signals/interface/power are brought to some sort of expansion connector.

The Raspberry Pi for example is not good at all, I'd say that the Beagle Bone [Black] is one of the best given that strikes a good balance on type and quantity of signals in a pair of standard and cheap header connectors.

The Wandboard is on the other side of the spectrum, almost all signals are on the connector to the motherboard but it is not as cheap or standard to be integrated.

Olimex is very good on this front for simple applications.

The Cubie board is an interesting option but documentation is very scarce.

I still have the pcDuino on my list of things to test, but this is another Allwinner board so documentation is not good either, and not that many IO and as easy to integrate like the BBB.

You also have to give some points for quality of design and production, and very very important what development tools and support community you have around.

Adding everything up if I have today to pick a winner it will be the Beagle Bone Black.

My. 02

Jorge

You're both confirming the impression I'd gained of the state of play in modular boards.  That's a pity.

Although we have no power to influence the direction of industry in any significant way, perhaps there is room for enthusiast-level design (or promotion) of OSHW modular systems independent of industry, if only one major barrier could be overcome.

BGA is the big elephant in the room, a nearly impossible hurdle placed between SoC manufacturers and enthusiasts (yes I know that a few enthusiasts can handle BGA, but that doesn't change the statistics much).  There is however a way of cutting the elephant down to size, and that is to choose a cheap intermediate adapter form that is easier to handle, and then getting the BGA-on-adapters mass produced for not a lot more than the SoCs themselves.  Based on these pre-mounted BGAs, enthusiasts could then take module design in directions where the embedded industry has not found a reason to go.  Coupled with 3D printing, modular enthusiast OSHW computers would then be not too far-fetched an idea.

The current direction in which every board is different and requires its own case design and mandates almost wholesale reinvestment when you change manufacturers is really bad.  The OSHW community could chose to do better than this, even if industry doesn't want to help.

Addendum: I've just realized that this is the Pi group, singularly the wrong place to be having a discussion about OSHW modules. The chances of getting a BCM2835 on a bare BGA-adapter boardlet are not very good.

jamodio wrote:

The Wandboard is on the other side of the spectrum, almost all signals are on the connector to the motherboard but it is not as cheap or standard to be integrated.

As I understand it, the wandboard follows the EDM 'standard'.  I'd been kind of hoping it might help drive some more adoption.

Morgaine Dinova wrote:

 

The current direction in which every board is different and requires its own case design

Consumers want shiny cases, not me

You're right about BGA, it makes the barrier quite high, both from the assembly side and as jamodio points out from the pcb design angle. Once you designing 6 or 8 layer boards in order to get your signals out from under the BGA you're probably as well thinking about volumes and getting them assembled by someone with the proper equipment.

For a while there Olimex seemed to be trying to support people wanting to hand assemble stuff by using TQFP's where possible, but I see most of their Allwinner based stuff is BGA now.

So while a standard module with all the bga stuff contained would be a wonderful idea, in practise the commercial entities building these things are in it for their own (or their shareholders) gain, so 'value add' and 'lock in' is far more important than keeping us OSHW proponents happy.

Ultimately, it's not just BGA though. Any sufficiently high density connector will cause problems for enthusiasts, the higher the density the harder it is to handle and the number of people capable of doing it reduces as the cost and complexity of the equipment needed rises.  So a balance is needed, otherwise people just end up buying a commercial carrier for the module anyway, at which point you're back to the argument of 'why bother'

selsinork wrote:

 

As I understand it, the wandboard follows the EDM 'standard'.  I'd been kind of hoping it might help drive some more adoption.

An OSHW kickstarter campaign to put one (or all) of TI, Allwinner, Samsung and ST applications processors on the same EDM form factor as was used by Wandboard for Freescale's i.MX6 aren't beyond the bounds of possibility.  Kickstarters have been successful with far stupider goals.

selsinork wrote:

 

Ultimately, it's not just BGA though. Any sufficiently high density connector will cause problems for enthusiasts, the higher the density the harder it is to handle and the number of people capable of doing it reduces as the cost and complexity of the equipment needed rises.

To some extent, high density is a commercial affectation, driven by the need to look modern and high tech and also because real estate costs money.  Most enthusiasts don't need high density though, and as you say, it's just a liability to them.  As long as signal paths are properly balanced and well within all the electrical specs, spread them out to low-density edges of adapters.  Modules and connectors being a bit bigger is an advantage for OSHW.

And that in turn would obviate the need for the commercial carrier that you mentioned, because the carrier can then also be low-density OSHW.

selsinork wrote:

 

jamodio wrote:

The Wandboard is on the other side of the spectrum, almost all signals are on the connector to the motherboard but it is not as cheap or standard to be integrated.

As I understand it, the wandboard follows the EDM 'standard'.  I'd been kind of hoping it might help drive some more adoption.

There is no such thing as an "EDM standard".

The wandboard and the "EDM standard" are creations of the same company.

It may look confusing and misleading but behind all of them is TechNexion LTD of Taiwan. The only company to use EDM is them so there is no "standard" as you can see from this detailed list of products that support it http://www.edm-standard.org/index.php/product-list or the exensive list of companies that adhere to the standard http://www.edm-standard.org/index.php/members

Those guys may succeed tyring to fool some newbie but not an old dog

-J

selsinork wrote:

 

jamodio wrote:

The Wandboard is on the other side of the spectrum, almost all signals are on the connector to the motherboard but it is not as cheap or standard to be integrated.

As I understand it, the wandboard follows the EDM 'standard'.  I'd been kind of hoping it might help drive some more adoption.

There is no such thing as an "EDM standard".

The wandboard and the "EDM standard" are creations of the same company.

It may look confusing and misleading but behind all of them is TechNexion LTD of Taiwan. The only company to use EDM is them so there is no "standard" as you can see from this detailed list of products that support it http://www.edm-standard.org/index.php/product-list or the exensive list of companies that adhere to the standard http://www.edm-standard.org/index.php/members

Those guys may succeed tyring to fool some newbie but not an old dog

-J

jamodio wrote:

 

There is no such thing as an "EDM standard".

there was a reason I put 'standard' in quotes    anything called a standard is only of use if you can convince others to follow it, and short of being a government and being able to pass a law forcing people to use your 'standard' none are any better than others, they're mostly cooked up by some company anyway.  but does that matter ?  as long as it's free to use and you can build a consensus that it's a good idea.  I'd venture to think that any 'standards' that come with patent claims, licensing fees etc. are a bad thing, but the world has no shortage of those.

The only company to use EDM is them

Doing some research a few months ago I did come across both TechNexion and a couple of places in Germany using the same non-standard while the wandboard appears to be a barely disguised version of the EDM2-CF-iMX6, the german company was using older SoC's like the imx233 clearly aimed at industrial use with price tags to match.  Simple searches for 'EDM' aren't bringing me back to their page, so I fully understand you're not likely to believe me

I did however turn up a couple of competing 'standards', SMARC from Kontron http://uk.kontron.com/about-kontron/news-events/ulpcom+is+now+smarc.6720.html and Qseven, both sponsored by "Standardization Group for Embedded Technologies" http://www.sget.org/ who do seem to have some members and Qseven has a bunch of products form several manufacturers.

Interestingly SMARC uses the same connector as the EDM one.

To add to the amusement, there's the AMD version of the Minnow http://www.gizmosphere.org/

And it seems likely we should get used to angstrom like setups or be prepared to roll our own:

http://linuxgizmos.com/gumstix-adopts-yocto-for-overo-modules/