I’m basically looking for some pointers on moving away from prototype systems in the field, to a standardized mass produced board. After reading thru the forums, I’ve seen references a few times to people not liking the 10 pin connectors / ribbon cable for industrial systems. In addition, if mass producing a board, they are a very expensive way to connect things together. I’m basically made some assumptions, and would appreciate any feedback on if they are valid, or if I’m out in left field somewhere…
This is all based on production runs of 1000 boards that are basically dev kits with the non needed parts removed, and a few fail safes added in. Assuming an estimated 2-5 production runs per year for the next 3 years.
Correct Processor? Even that it is overkill for most of the applications, I’ve standardized to ONLY using the G400 (currently as a Raptor), simply because its speed makes worrying about timing between different functions a non issue, and only have 1 code base that gets maintained. That said, with Octavo systems being developed, will it run .NETMF if not interested in running Linux?
Correct form factor? There are system on System on Modules, and system on Chips? I do know that with the Raptor, I’ve had issues with the solder joints between the module and the main board, (but to be fair, they were thrown in a backpack, and not handled with the care they should have), but given that the environment they work in can have a significant amount of vibration & temperature swings, it does highlight a point of failure. A benefit to using a module, is that as devices go obsolete (as several have in the short time I’ve worked with Netmf & GHI), new ones could be made to have the same pin outs? Other than extra pins, is there any other benefit to the mechanical Sodimm?
Obsolescence, I understand technology is changing at an incredible pace, and was wondering what is the expected lifetime of these system on modules? Is there any plan to maintain pin compatibility on future modules, which would help justify a socket?
Assuming a production run by GHI, If a pick & place machine is all ready running & set up, would it not make sense to simply include all the components that make up the module, and simply populate them onto the same board, rather than have the module populated separately, and then soldered on?
I do realize some of this falls under the customized service from GHI, and have put in a request for information, but also some of it falls in the generic category of simply picking the right solution for anyone developing a system.
@ michaelb - I will add a few comments based on my limited run production experience.
Keep your connectors to a minimum, they are expensive. I have never used the 10pin connectors in production but I think the problem is repeated use that wears them out. If you only plug it in once and leave it then it probably will give you many years of service. If you constantly are plugging/unplugging then you will have problems.
1 - If you think there will be improvements or new software features added down the road then defiantly plan on the larger processor. However, I would caution getting too carried away. The price difference between a G120E and G400-S is only $10.02, but at 1000pcs that is $10,020 you are giving away if you don’t need it.
2 - I think properly soldered the G120/G400 should be pretty robust to vibration. My guess is that in your backpack it was getting stressed which could break solder joints.
For short runs I prefer plug in modular approach. When having your circuit boards produced in quantity you don’t have to put up the money for the expensive module until closer to the time you ship. It is an easy way to spread your production costs out. Another advantage we found was that we can ship replacement modules to customers if there are problems with the module. We also have had circuit boards fail but the module is still good so we can move it to a new circuit board.
3 - GHI did keep the pin out compatible when it went from the ChipworkX to the G400. Others on this forum are more qualified to answer how the upgrade went from ChipworkX to G400. So yes, it is possible that in the future there will be an upgraded version, but there is no guarantee. However, you could create your own module if necessary.
4 - If you have the circuit design experience to move the module design to your circuit board then yes it could be cost effective. The circuit design of the G400 and G120 are much more intense then a standard 2 or 4 layer board. Routing high speed memory address and data lines needs to be done very carefully. If you want to keep it compact you will require more circuit board layers which increases cost. Using GHI Hardware Engineering services would be helpful in that regard. It looks like the new Octavo SoM might be an exception to that.
no, Octavo is unlikely to be offered with netmf, it’s a higher-order product. G400 vs other more cost effective products, depending on the life you expect from the product that $10,020 extra per 1000 may give you greater life expectancy, but you’re going to be best situated to understand if that’s valuable or not. There’s certainly architectural differences that also change the profile of each of G120 and G400 that you’d need to look at to better understand if that leaves you with the capability you need. Get a Spider II and move your app over to it, and see if it feels a good target; certainly a cheap way to compare IMO.
As Gus says, GHI have strived to keep their modules easily replaceable with the same layout, so if one product did go obsolete you’d likely have a way forward. You’d best work with GHI directly though on making sure you were noted as a commercial customer and to be informed of “last buy” opportunities etc.
see above
you’re going to need to custom design a board - not an insignificant job, but also not overly hard given the starting points you have available to you (module designs, mainboard / reference designs). Like most design processes you’re going to work on those designs to optimise for manufacture, which will include BOM optimisation and looking at second-source parts options. You’ll also need to have industrial design considerations since you have said that there’s temperature and vibration issues to consider. I’d suggest the Gadgeteer sockets/cables as more prone to issues in those environments (they’re mini-antennas too for EMC related issues) so I’d be steering away from thinking about any of those for your final design.
Thanks for the answers so far, I definitely appreciate it. I just want to clarify that price isn’t the most important consideration, especially if its in the $10-$50 dollars / board range. If I pick a worse case scenario for a site that I’ve worked with, the spanking from the local power company would be in the $50,000.00 range for a 2 second oops turning all the outputs on. The biggest concern is reliability, and even if the boards fail, its not an issue, as long as they fail to a dead state vs turning things on.
Why not hire GHI for consulting so we can study your needs and give you a design advice, or even create the design for you. We are an engineering and manufacturing firm in addition to our pre engineered products.
@ Gus - I’ve phoned in to GHI on at least 3 occasions, and get told to email GHI. I write an email, and never get a response. Last one I sent was Thursday Morning.
@ michaelb - I got your email Thursday morning around 11 AM our engineering manager is off until Monday so it’s on the list discuss with him on Monday morning to give you some advice .
