From: Eric Fort <[email protected]> Reply-To: <[email protected]> Date: Thursday, June 5, 2014 at 12:56 PM To: beagleboard <[email protected]> Subject: Re: [beagleboard] Do as Raspberry - Make a Beaglebone Black - Compute Module !? - Why not
> These pretty much already exist as a module.... check out gumstix which ave > been around for quite some time. The best one I seen so far is this one from Maxim Podbereznyy in that it maintains compatibility to BBB. http://www.mentorel.com/product/usomiq-am335x/ Regards, John > > > Eric > > > On Thu, Jun 5, 2014 at 11:11 AM, John Syn <[email protected]> wrote: >> >> From: CEinTX <[email protected]> >> Reply-To: <[email protected]> >> Date: Thursday, June 5, 2014 at 7:22 AM >> To: <[email protected]> >> >> Subject: Re: [beagleboard] Do as Raspberry - Make a Beaglebone Black - >> Compute Module !? - Why not >> >>> John, >>> >>> You definitely make some good points, however... >>> I don't really feel I missed the mark completely. >>> So when you say 'developer', do you mean hobbyist or product developer. >>> If you mean hobbyist, I understand your comments completely. >>> If not, that's a totally different situation entirely. If you are developing >>> a product for sale, >>> then I have little sympathy / empathy / whatever for not wanting to deal >>> with the >>> technology required to build a product. It's not cheap, easy, simple, >>> _________ fill in >>> any of a long list of adjectives to describe the difficultly of designing a >>> product much less >>> building a business around it to market and sell it. >>> >>> One is not entirely constrained by the BBB design >>> in doing your own - see what comes below.... >>> >>> (When I say 'you' in my comments - it's not pointed at you John but a >>> general you from >>> the community standpoint) >>> >>> I personally don't see these as problems as I've been doing this my whole >>> professional career. >>> So, yes I'm probably trivializing some of this as I don't see any of these >>> items as potential >>> stumbling blocks. >>> >>> By the BBB statements of use - this is not for commercial use. Although I >>> know from speaking >>> with Gerald, that many are using this for commercial apps. That's also one >>> of the reasons that >>> the supply is being gobbled up. So for those of you out there that are >>> complaining about not being >>> able to get BBBs, blame those that are not following the licensing and terms >>> of use and eating >>> up the supply for their commercial needs versus developing their own board. >>> >>> Here's a thought for you Gerald, require your distributors/resellers to have >>> a reverse discount >>> model. So as the volume goes up, so does the price - this should discourage >>> volume buys >>> without the organization's consent, which you are suppose to have if using >>> this product >>> commercially, and would make the distributors happy by increasing their >>> margins. >>> Or would require a custom 'factory' price to get a volume purchase at a >>> discount to get >>> around this. This is done the other way around - via product/design >>> registrations all the >>> time in the electronics distribution model. Just a thought. >> Gerald & Co need to keep their costs as low as possible and this doesn¹t only >> refer to the cost of materials and labor, but also the cost of money. >> Therefor they need a consistent supply model with minimal supply >> interruptions because the last thing they need is to sit with large inventory >> when the demand dries up. Don¹t mess with pricing because there could be all >> kinds of unintended consequences. While the inventory shortfall is really >> irritating to most, it is because of the demand that is helping to keep the >> pricing down. Gerald has to manage a find balance between delayed delivery >> and maintaining demand volumes. Gerald & Co are continuing to add resources >> to increase monthly volume and I think that is the best approach. >>> >>> So really it's suppose to be either for small non-commercial projects or >>> using as a shortcut >>> to figuring out what you need and don't to roll your own. It's not a one >>> size fits all or trying >>> to be everything to everybody - like a lot seem to think it is or should be. >>> So really the capes >>> and position of connectors, to me, are really a moot point as one is not >>> suppose to be relying >>> on this as a product platform. If it's not where you want or need for your >>> product - roll your own. >>> >>> Embedded system design is not for the faint of heart. You need tools and a >>> lot of patience to get >>> it done - education, experience, and skill doesn't hurt either. If you don't >>> have the tools or skills >>> needed to leverage what Gerald and Co have done here, maybe as part of your >>> business model, >>> you should have some budget for tools and an engineer or hire a design >>> house, um maybe like >>> CircuitCo or another, to help fill the void. Just a thought. If you're just >>> a hobbyist, then most of >>> this discussion is moot, because you should just try to make what's >>> available work or develop >>> a much simpler cape to do what else you need. And if that doesn't work - >>> you're back to roll your >>> own - or find a different development platform that will support your needs. >>> I even did a cape to >>> vet out what I wanted to do before starting my design. So, use the resources >>> available to you. >>> >>> Either using a SOM or not with this is probably not a project for a >>> beginner. But this is also very far >>> from the most complicated or demanding designs I've ever done. Even doing an >>> I/O board is not trivial >>> and that's the point I'm making. You still have a controlled impedance and >>> probably controlled dielectric >>> PCB to design and have fab'd. The I/O board will need to be a min of 4 >>> layers to be able to control the >>> impedances adequately and reliably. The diff pairs for the USB and Ethernet >>> as well as the MII interface do >>> require some work to get right - then there's the memory if you're not using >>> the SOM. As far as the 3 mil >>> min trace/space - the only part on their that needs this is the stupid eMMC. >>> What a poor package >>> design for this - following the JEDEC standard for the complete module was a >>> poor choice - but I digress. >>> I found a different package for my design so I didn't have to push the >>> limits of reasonable board fabrication. >>> My design is 5 mill trace / 4 mill space - could be 5 space if I wanted to >>> spend a bunch more time working >>> around the processor's BGA. But 5/4+ is good enough for me - some of that is >>> legacy from leveraging the >>> layout design / info from the BBB. My previous design was based on the >>> BeagleBoard. That was a lot more >>> complicated design the the BBB. But I rolled my own and went away from the >>> POP and did a design that >>> was 5/5 in 4 layers - so it is possible - just requires some effort. >>> >>> Thanks Gerald for at least making my life easier. >>> >>> So, to reiterate, the SOM has it's place - they exist and some do buy them. >>> I just don't think it's the really >>> the market segment the BeagleBoard/Bone/Black is trying to play in and >>> support - so I'm not at all surprised >>> that Gerald and the developers don't want to pursue that path. Again >>> leverage what Gerald has provided - he's >>> done a good portion of the heavy lifting - you have even been given gerbers >>> to be able to get it right or even just >>> use what he's done in many cases. If you think the SOM is really a good >>> idea, build one yourself and sell it - >>> again, Gerald has provided you with what you need to even do that. Hell, if >>> you want to throw some money at >>> me, I'd probably do the design for you - but I don't have much time for such >>> endeavors - so the money would >>> have to be larger for me to even consider it - I'm rather busy designing and >>> building and marketing my own >>> products to industry. >>> >>> As a community, do your homework/research - do your own due diligence - then >>> go down the path that will work >>> for your needs. There are a lot of people around here who are willing to >>> help solve problems - show what they've >>> done and help you get down the your path - whatever that might be. Just >>> don't be surprised when Gerald gets >>> tired / fed-up / or worse when he's kept on being asked to fit square pegs >>> into round holes so this organization can >>> become everything to everyone. >>> >>> I wish everyone in the community the best of luck in their endeavors. >>> >>> >>> On Wednesday, June 4, 2014 3:47:50 PM UTC-5, john3909 wrote: >>>> I think you missed the most important part. Most developers here are not >>>> able or do not want to deal with 6 layer boards with 3 mil trace and >>>> spacing (high tech boards). Working with 2 or 4 layer boards with 5 or 6 >>>> mil trace and space (standard tech boards) is low cost (< $40 in small >>>> prototype qty). As you pointed out, the cost to prototype and manufacture 6 >>>> layer high tech boards is expensive and requires a high level of expertise >>>> to make any modifications. As you know the cape concept doesn¹t always work >>>> because of the I/O conflict between capes but it would be easier to develop >>>> a standard tech board with all the I/O designed to work together. Also, the >>>> position of the connectors on the BBB may not be suitable for a specified >>>> enclosure so a module would provide that flexibility as well. >>>> >>>> Just my two cents worth to add a little balance to your comments. >>>> >>>> Regards, >>>> John >>>> >>>> From: CEinTX <[email protected]> >>>> Reply-To: <[email protected]> >>>> Date: Wednesday, June 4, 2014 at 6:41 AM >>>> To: <[email protected]> >>>> Subject: Re: [beagleboard] Do as Raspberry - Make a Beaglebone Black - >>>> Compute Module !? - Why not >>>> >>>>> Being a design engineer for close to 30 years now - doing mostly embedded >>>>> systems - I don't really see the appeal to this approach. >>>>> So - they (R-Pi) are saying the module is $30 in qty 100. A Pi is $35 - a >>>>> little more if you want an SD card. The BBB is $45. >>>>> So a compute module based on the BBB might be $35-40 based on the price >>>>> difference, I don't know. >>>>> >>>>> Let me get this straight, your paying just about as much just for the >>>>> processor and memory as you can get a complete system. >>>>> But you say, I want to develop my own. OK - you've just paid someone else >>>>> to do the processor side - you still have to have a >>>>> connector to make that connection to your processor. Then you get to >>>>> design and build your specific I/O card. >>>>> That, I'm sure, will be easier but at what cost. What's going to be more >>>>> reliable in the long run, a system with or without that connector? >>>>> If you've got to do the design anyway, why not save the money and keep it >>>>> in your pocket. >>>>> >>>>> From my experience, the people who benefit the most from the compute >>>>> module/ SOM approach are for those who know they need a long >>>>> time system life and also know that they will need to upgrade the >>>>> processor and memory capabilities down the road. Of course you also >>>>> need to be willing to accept what processor and memory choices they've >>>>> made - who knows maybe they will have different options for >>>>> different memory sizes and speeds. >>>>> >>>>> The most common place I've seen this approach in the past was with VME and >>>>> Multibus systems. These are expensive systems to begin with. >>>>> So it makes sense to be able to upgrade a portion of the system at a lower >>>>> cost. The only other option for this being a benefit is if someone already >>>>> has an I/O card that meets your requirements. Then it's off to the races. >>>>> How much is that I/O module? I didn't see a price, hum. Bet the two >>>>> combined are more than $35-$45. Also, is the compute module / SOM done to >>>>> a standard so that you can replace it with another down the road - even >>>>> a different architecture? >>>>> >>>>> I have done the cost analysis many times and most embedded systems do not >>>>> need the ability to upgrade the processor and memory down the road. >>>>> They usually have a specific purpose and once designed to that will >>>>> function that way for the life of the product. >>>>> >>>>> I understand that doing the processor and memory design on an embedded >>>>> system can be tough, challenging even, but Gerald and Co have already done >>>>> the lion's share of the work - leverage that effort. >>>>> >>>>> I do small runs on my projects all the time. In fact my current project is >>>>> an industrial temp spin on the BBB. Not 100% compatible, but that's the >>>>> point. >>>>> I'm priced out, for components and pcb, at less than $80 - I couldn't >>>>> justify spending $30-$40 for the processor and memory and still have to do >>>>> the rest. >>>>> Additional costs - NRE for stencils and production programming is >>>>> estimated at $500. Not sure what assembly/test costs will be yet, but I >>>>> expect ~$20-30 >>>>> hopefully less. Yes, I'm just about to do my prototype on the board - so >>>>> I'll soon get to see what the actual costs are. >>>>> >>>>> So cost each for the 1st batch of 100 will be ~$110. Not too shabby for an >>>>> I-temp board in that quantity. Future runs will be less without the NRE >>>>> costs >>>>> and hopefully larger build quantities. Of course there are engineering >>>>> costs to be absorbed too, but that's an exercise for the accounting people >>>>> to >>>>> figure out what budget that belongs to. >>>>> >>>>> So, yes the compute modules / SOMs are cool ideas and have their place - >>>>> but they are not that cost effective for most. So do your homework and see >>>>> if that approach will work for you and what you need. I suspect that the >>>>> PI community will not see the compute module as widely bought / accepted >>>>> as the >>>>> base R-PI. I do suspect the the R-PI and the BBB will see strong sales as >>>>> a base platform at those price points. >>>>> >>>>> Good luck in all your endeavors. >>>>> >>>>> >>>>> On Wednesday, June 4, 2014 7:17:20 AM UTC-5, Gerald wrote: >>>>>> We are not interested in getting into the module business as a >>>>>> BeagleBoard branded device. Feel free to do it yourself however. All the >>>>>> information is there. Some people have already made these modules and are >>>>>> out there in the market in various forms.. >>>>>> >>>>>> Gerald >>>>>> >>>>>> >>>>>> >>>>>> On Tue, Jun 3, 2014 at 8:00 PM, <[email protected]> wrote: >>>>>>> I think the Raspberry idea of a compute module is a brilliant one. Now >>>>>>> they will be able to sell, not just to individuals but also to industry. >>>>>>> They will probably reach 5 mill. boards produced before the end of the >>>>>>> year. >>>>>>> >>>>>>> Why not do the same with Beaglebone. The profit margins could probably >>>>>>> be higher then on the Beaglebone Black and each extra $ could help get >>>>>>> rid of the terrible shortage of Beaglebone Black boards - that never >>>>>>> seams to go away. >>>>>>> >>>>>>> Accept that the Beaglebone Black is a huge success and that you probably >>>>>>> have to produce at last 50.000 boards a month to cope with the huge >>>>>>> demand. In the long run we'll all probably get tired of waiting for >>>>>>> boards, and eventually be forced to turn our attention to something >>>>>>> else. >>>>>>> >>>>>>> /Bo >>>>>>> -- >>>>>>> For more options, visit http://beagleboard.org/discuss >>>>>>> --- >>>>>>> You received this message because you are subscribed to the Google >>>>>>> Groups "BeagleBoard" group. >>>>>>> To unsubscribe from this group and stop receiving emails from it, send >>>>>>> an email to [email protected]. >>>>>>> For more options, visit https://groups.google.com/d/optout >>>>>>> <https://groups.google.com/d/optout> . >>>>>> >>>>> >>>>> -- >>>>> For more options, visit http://beagleboard.org/discuss >>>>> --- >>>>> You received this message because you are subscribed to the Google Groups >>>>> "BeagleBoard" group. >>>>> To unsubscribe from this group and stop receiving emails from it, send an >>>>> email to [email protected]. >>>>> For more options, visit https://groups.google.com/d/optout >>>>> <https://groups.google.com/d/optout> . >>> >>> -- >>> For more options, visit http://beagleboard.org/discuss >>> --- >>> You received this message because you are subscribed to the Google Groups >>> "BeagleBoard" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to [email protected]. >>> For more options, visit https://groups.google.com/d/optout. >> -- >> For more options, visit http://beagleboard.org/discuss >> --- >> You received this message because you are subscribed to the Google Groups >> "BeagleBoard" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> For more options, visit https://groups.google.com/d/optout. > > > -- > For more options, visit http://beagleboard.org/discuss > --- > You received this message because you are subscribed to the Google Groups > "BeagleBoard" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. -- For more options, visit http://beagleboard.org/discuss --- You received this message because you are subscribed to the Google Groups "BeagleBoard" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
