[Arm-netbook] Plug computer and FreedomBox-related devices

[Luke Kenneth Casson Leighton]

On Wed, Sep 16, 2015 at 11:49 PM, Paul Boddie <paul at boddie.org.uk> wrote:
> On Tuesday 15. September 2015 19.18.26 Luke Kenneth Casson Leighton wrote:
>> On Mon, Sep 14, 2015 at 4:47 PM, Paul Boddie <paul at boddie.org.uk> wrote:
>> >
>> > It occurred to me that if only we had EOMA-68 boards out there, maybe
>> > people wouldn't be so enthusiastic to go to the trouble of making new
>> > boards and running the gauntlet of crowdfunding.
>>
>>  i knooooow :)
>>
>>  here's the thing though:
>>
>>  (1) i have to get the standard right for a 10-year-old and greater period
>>  (2) i have to get CPU Cards (plural) designed, sponsored, and tested
>>  (3) i have to prove that it is, indeed, simpler and lower-cost to
>> make carrier boards.
>>
>> so that's what i'm doing.
>>
>> to explain, first: this project has an absolute top priority of being
>> *right* (defined as "viable long-term") over being "rushed to
>> profitability".
>
> Sure, I understand that. But what worries me a little is that experience isn't
> being gained to possibly refine the standard

 paul, you misunderstand the concept of a simple long-term standard.
it goes from "being developed and refined" to "absolutely locked
irrevocably without fail absolute without fail absolute 100% in stone"
with no phase in between.

 once there are any end-users out there, there *is* no "second
chance". that's *IT*. i cannot make that complex enough to be
understood and accepted.

 the reasons why there are no second chances are clearly laid out in
the white paper i wrote.

 so, the opportunity, if anyone wants to get involved with
"refinement" of the standard, that's what they've had over the past
four years, and now it's almost over.  they can talk to me, they can
read the white paper (in which i describe the justification of the
interface selection), and they can try to argue for additions or
changes, but right now the window of opportinity to do that is
closing, as i am about to send off two CPU Cards with the latest
(last) revisions.

 anyone beyond that point in about 10-14 days time, if they want
modifications to the standard, it had better come with a cheque for
$5,000 attached to it in order to repay both my clients for the
samples that will have been made by that time.


>  or develop for it.

 that's straightforward "critical mass", which comes once the
standard's finalised, first products out the door, etc. etc.

> Of course, I'm
> writing this with only a superficial knowledge of what has gone on in the
> past, and it is possible that people have done things with previously-produced
> hardware that has informed the effort.

 http://rhombus-tech.net/allwinner_a10/news/

 that's a brief timeline for the a10 cpu card (which became the a20
cpu card).  the standard has been revised about... i think it's three
times since then, maybe four.

 first revision was to remove SATA and replace it with a 2nd USB2.
second revision was to add VREFTTL, add SD/MMC and UART,  third (or
possibly still part of the 2nd) was to reduce 24-pin RGB/TTL to 18-pin
RGB/TTL and use the 4 spare lines for an SPI interface, also USB3 was
added at some point.  PWM and an extra EINT were also added.  the
fourth - and almost certainly final revision - has been very recent:
removal of Ethernet, upgrading to being able to do USB 3.1, as well as
add 2 more EINT lines and 3 more GPIOs.

 those interfaces have all been very carefully considered, especially
when developing the ICubeCorp IC3128 CPU Card, where, due to its low
pincount and being a QFP, there's *literally* only 2 spare unused pins
left on the *entire* processor that don't go to the EOMA68 interface
or the SD/MMC boot card.

>> second: it's no good having just the one CPU Card out there.  people
>> won't comprehend the modularity concept if there is only the
>> additional cost of having a single processor available.
>
> I understand this, too. It's also useful to have different devices they can be
> used in as well, and I worry that these won't come about without any cards
> being available.

 that's why i'm doing more than one product - micro-desktop as well as
the laptop.  i've actually designed around... 5 or 6 different
products, all at different stages.  the tablet is "on hold" because of
lack of interest [and, because it's a complex and dense 4-layer board
it's going to be tough converting it to 2-layer].  the 15in laptop is
sponsored so that's ok.  and the micro-desktop, what with the PCB now
being 2-layer, is actually low-cost enough for me to put together a
new revision on a very low budget.

 irony is all these will probably hit all at the same time.

> One thing in the back of my mind (and part of a long list of things that I
> could consider doing)

 :)

> is helping to design such a device, and for that I
> suppose I need to collect links to documentation that might help me get
> started.

 honestly, finding the parts (ones that are affordable, not
end-of-life, or fell off the back of a lorry and i do mean that
literally, even though they had to have a bit of a push to fall
of....) and/or finding the contacts _willing_ to source the parts -
that's the hardest bit of the entire job of designing any board.

 oh.  that and, of course, being able to get hold of the damn datasheets.

 one cool product that would be awesome to have would be a 3d printer
controller that took an EOMA68 CPU Card.  in some ways this would be a
bodge-job of taking a libre-hardware-licensed PCB design, expanding
its size a bit and connecting 5V power and USB up to the on-board USB
interface of the Micro-Controller.

 that way you would have the actual computer *on* the same board that
had the (usually unreliable) power line... the one that gets spiked
due to 50hz mains fluctuations and lack of proper earth loops
associated with USB cables,500 watt PSUs, and amateur-designed PCBs.

 a second cool product would be a robotics platform / educational platform.

 another would be to track down a suitable libre-toolchain FPGA and
make a CPU Card out of it.  i heard that someone has actually managed
to reverse-engineer one of the commonly-available FPGAs, to the point
where the toolchain is stable.  saw an article about it recently.
that _would_ be awesome.


>> third: even i was unable to move the micro-desktop board (which is
>> only 4in x 4.5in) forward because i had designed it as a 4-layer PCB -
>> costs are around $400 for qty 5 4-layer prototype boards on a 3 week
>> turnaround  by complete contrast, a 7-day turnaround for qty 5 2-layer
>> prototype (bare copper) boards with larger vias is around $40 for qty
>> 2, and around $100 for qty 5.
>>
>> so.... it's getting there, paul.
>
> Yes, prototyping gets expensive for advanced stuff, I suppose, and since we're
> not part of the normal industrial operations that can do this efficiently (in
> terms of costs and other things), we're at a disadvantage.
>
> Still, I wonder what those of us reading this list might be able to do to move
> the effort forward in our own way.

that would be great.