I’m searching the right hardware for an embedded project. One of the Beagle variants is likely to be the right base to start with. Some point I couldn’t figure out:
Why are some boards using the TI AM335x like BBB and others use the same core, but wrapped in a Octavio SIP. I know that it already contains a bunch of components board designers place next to the AM335x. But what are the main reasons to choose one or the other? Price? Design complexity? Is it fully software compatible on image level?
The PocketBeagle uses the Octavio SIP and the Debian images for the BBB work
just fine on the PocketBeagle, with no compatiblity issues. I think the
Octavio SIP allows for a more compact board because all of the support
components are incorporated into a single package. This makes for a smaller
(and cheaper) PCB.
I forgot to mention that I’m looking at the BBW, but I think your answer refers to this one as well.
What about the availability between both solutions? My intended development should be for industry, where they expect very very long product availability. Would a SIP be better over single components, or does Octavio face the same problems when any part gets end of life status?
Is the image really completely the same? No diff config in uboot or somewhere else? Not easy to compare, because the pinout for both ICs is different.
I forgot to mention that I'm looking at the BBW, but I think your answer
refers to this one as well.
What about the availability between both solutions? My intended development
should be for industry, where they expect very very long product
availability. Would a SIP be better over single components, or does Octavio
face the same problems when any part gets end of life status?
I don't really know. You'll have to talk to the people at TI about that.
Is the image really completely the same? No diff config in uboot or
somewhere else? Not easy to compare, because the pinout for both ICs is
different.
The stock Beagle image works without mods on either. I don't know if there is
a difference in the boot loader or if the uboot srcript in the chip EEPPROM is
different. If there is, there is suitable firmware included on the image.
We have designed a board based upon the Octavo Systems cSIP, which is a step beyond the SIP. They added eMMC, EEPROM and a MEMS oscillator. Many layout issues go away using their solution. I have worked extensively with the Octavo folks. Very nice people. What I understand is that the company was founded by ex-TI people. They provide free design review services that uncovered problems with our design. They say they will manage obsolescence of the components inside of the cSIP.
I’ve not build my own board. The project is still in an early development stage. But you can find the design including layout for the boards. I guess they’ve managed thermal management.
Sorry. Can’t share our design. Proprietary. No thermal issues as of yet, but we are not running intensive software. I do plan on inspecting using a thermal camera in real time sometime soon.
The board is loosely based upon the Red board. Added a second Ethernet. The Red is a good starting point. I have a Red that we used to evaluate. I also have a Blue. Not a bad board, but uses funky connectors. I do not know of a dev board based upon the cSIP.
I’m prototyping on a pocketbeagle, but how to scale up for production or ~1000 units is unclear. I can buy pocketbeagle retail for $25, but the Octavia SIP lists on Digikey at $36 and I have to but 500 parts to get that price . So it seems like integrating a pocketbeagle into my product would be easier and more cost effective. The only bit I don’t know is how to solder down a pocketbeagle onto a larger board with automation, any ideas?
I'm prototyping on a pocketbeagle, but how to scale up for production or
~1000 units is unclear. I can buy pocketbeagle retail for $25, but the
Octavia SIP lists on Digikey at $36 and I have to but 500 parts to get that
price . So it seems like integrating a pocketbeagle into my product would
be easier and more cost effective. The only bit I don't know is how to
solder down a pocketbeagle onto a larger board with automation, any ideas?
I have soldered male headers (pins) on the bottom of a PocketBeagle and
installed female headers (sockets) on the "base board". The PocketBeagle is
not soldered to the base board, but is mounted in a pair of sockets.
I ended up with the same conclusion of just using the PocketBeagle. My “capes” have the male pinheaders on them and we solder female headers on the underside of the PB.
Thanks Daniel, (cool site btw!) I have headers exactly the same on my prototype board set up. Do you have any automation for soldering the female headers or do you hand solder them? Someone has made an automatic header soldering machine which is interesting, but quite a project https://hackaday.com/2015/05/05/open-source-diy-soldering-robot/ . I have also heard of ‘pin in paste’ headers, which would require some kind of reflow process only on the edge of the board. I’m trying to avoid as many manual assembly processes as possible.
Thanks Daniel, (cool site btw!) I have headers exactly the same on my
prototype board set up. Do you have any automation for soldering the female
headers or do you hand solder them? Someone has made an automatic header
soldering machine which is interesting, but quite a project Open Source, DIY Soldering Robot | Hackaday . I have
also heard of 'pin in paste' headers, which would require some kind of
reflow process only on the edge of the board. I'm trying to avoid as many
manual assembly processes as possible.
It is possible to get SMD versions of the headers (they might have locating
pins). The PB itself has to have through hole headers and would have be hand
soldered in any case, but there are options for your cape/base board to be all
SMD parts, machine placed and reflow soldered, with no manual soldering.
What about taking one step further back and integrating the TI AM3358 into your product?
I have doubts this solves the problem either however. I think you can’t buy the processor and PMIC and LDO and DDR3 for less than this $25.00 price tag.
Its always a scaling issue. The small developer cannot afford the gamble that his product will be successful enough to justify the outlay of the volume needed to get the initial price low.
When volume was lower, I actually paid my 13yo son to solder them. He wanted to save up for a new composite Baseball Bat so I paid him $3 for each one that he soldered. At one point, he got good enough to do about 8 an hour (I can do about 9/10) so he made out pretty well. However, you could use this idea and maybe find a couple enterprising high school students or something that would like a socially distant work from home job that could be just a couple hours a day. Give them $3 or $4 each and they’d make out really well once they got good at it. (Assuming they have the soldering station or you could loan them one or something)
With higher volume, I’m just having the cape manufacturer do it. They are more than happy to solder headers on and mount them onto the capes. The trick here is getting them to the board manufacturer since the PB’s aren’t available everywhere or is very expensive in some locations. Thus, it kind of depends on where you have your boards made. I’ve had bunches of PB’s shipped to me, I’ve ripped them out of the retail packaging (to save weight/space) and then shipped them off to manufacture. Anyway, talk to your board manufacturer.
