[Arm-netbook] EOMA-68 development cycle

[luke.leighton]

On Sat, Aug 17, 2013 at 11:05 PM, Craig Ballew
<arm-netbook2013 at mymemopad.com> wrote:
> I admit I haven’t reviewed every page at http://rhombus-tech.net and I have
> not been able to read every post on this mailing list

 there's a lot there, and a lot of different things going on

> but I love the concept
> and design and believe the potential is great for a successful product

 ... products plural.  welcome to the list :)  it's a common
misconception that there will be only one product.

> used
> by OEM and hobbyist customers if it is released in a timely manner

 indeed.  the first product in the first of the standards has been...
a long time coming, whilst we worked out... *deep breath*.... SoC
vendors, _legal_ SoC vendors (ones who respect copyright), standards,
interfaces, components, suppliers, designers, procedures, processes,
PCB sizes, manufacturers, deals with manufacturers, contracts with
manufacturers - i haven't gone into full details but you get the
general picture.

 luckily through this learning process there is a significant amount
of re-use.  for example after asking and obtaining the boundary
devices ORCAD and PADS files it took me a mere 5 days to turn that
into a 95%-ready iMX6 EOMA68 PCB [the remaining 5% requiring a
competent electronics engineer and somewhere around $20k to finish it
off].  the ingenic jz4750 one was about the same amount of time, using
virtually the exact same components.

 so if we stick to the formula, having got the parts up, supply chains
sorted etc. then the introduction of new CPU Cards will be a matter of
a few weeks... *if* they're funded.


>
>
>
> I apologize if it has been covered here or if it is on the website and I
> missed it or it’s considered confidential, but what is the funding model for
> the development for the EOMA-68 A20 prototype boards, I/O Interface Boards
> and related accessory products?

 there isn't any funding.  there's no bank loans, no investment,
meaning that we're beholden to absolutely noone - no debts, no
shareholders to demand dividend payments.

 in some ways that's good: it means we can go straight into profitable
stable production with no overheads to be paid off.  in others it's a
pain because it's clearly...  ... well, one way to put it is that it
requires "creative solutions" to substitute for what finance would
normally help deal with immediately.

 we _have_ asked our manufacturing partners for funding: they're
sufficiently large that it's just simply... not the way they do
things.  not even a $50k R&D budget.  normally, on presentation of a
working prototype they put in a cash order for say 100k units and
that's the end of it.   they haven't changed their procedures, so
we're doing this on a boot-strapping process, leveraging one deal to
get to the next phase, then the next, then the next.

 once we're actually in production, we'll have available profits, and
things change from then on.  .... but we have to _get_ to that point,
first.

 so.  the phases are: a) work out a way to get the prototype done b)
do it c) make it work d) present it e) receive cash order for tens to
hundreds of thousands of units f) profit g) use profit to fulfil
long-term goals of project.

 in the meantime there are other people - most of them here on this
list - who are cooking up schemes to get I/O boards done, helping out
by buying prototype boards in order to test them and so on.


> I think I understand the concept for a CIC but does Rhombus Tech provide the
> capital to fund any portion of the development cycle

 if we had capital we would have had the products finished 18 months ago.

> or are the developers
> shouldering the costs of producing the prototypes

 we got the first prototype paid-for, i can't say by whom.  the second
prototype of the first EOMA68 CPU Card in the series we got lucky
windfall amount of cash which allowed us to pay for a rework.

 etc. etc.

> and relying on the alpha
> testers to reimburse them for actual costs when purchasing the prototypes as
> they are developed?

 the alpha testers paid for the boards and the NREs for the production
of those boards.  we made no profit on their sale.  it would be nice
to have made some profit because we could use that to fund further
rounds of fixes, or product development.


> If so, how much is this funding model slowing the development cycle

 it's slowing things down _a lot_.  in some ways that's a good thing,
but in others it's bad.  for example because of the embarrassment of
WITS messing up the 1st revision CPU Card PCB (using a junior
engineer), we were able to get a discount on the 2nd run.  however,
_because_ of that discount, the more senior engineer i think was told
"get it done fast", he didn't communicate properly with us, so made
unauthorised changes that had to be discovered and corrected... we now
find that the ethernet's not fully confirmed working...

 .... ordinarily we'd go back to WITS and say "oi!  fix it!  we paid
you!!" but we know what they'll say: they'll say "please pay us some
more money".

> and what would speed it up?

 about $30k per CPU Card would _drastically_ speed things up.  we'd be
able to pay professional people who would work hard, instead of having
to either do deals or use people who may have their own agendas.


> Who is going to fund the initial and subsequent
> production run? Everybody pre-pay?

 of which product?  there are several.  there's the flying squirrel,
the MEB, the carrier board, 15in tablet, EOMA68-A20, EOMA68-iMX6, and
i'd like to start an EOMA26-A13 and other sets of products based
around EOMA26.  then also there's EOMA-200 which is more
desktop-orientated...

 let's assume you're referring to the EOMA68-A20.  once the ethernet
and sata issues are resolved (we've tracked sata down to wrong
capacitors, joe's going to try replacing the ones on his board) then
we'll be ready to do a pre-production run.

 at that point (once i'm confident that the cards will actually work)
as there are over 900 people on the preorders page i'm sure we'll be
able to do a run of at least 1,000 which would test our factory
partner's brand new fully-automated PCB and assembly facilities.  that
will need to be pre-pay.

 basically, with the U.S. govt insisting on complete insane oversight
across the entire world's banking system the U.S. could use that to
seize funds during the middle of a Letter of Credit transaction [*1],
leaving the bank completely out of pocket.  the world's banks - all of
them - have solved this problem by terminating Letters of Credit.
world-wide.

 so - yes, pre-pay by some means.  this isn't as bad as it sounds, in
fact it's pretty normal in asia, and through e.g. crowd-funding it's
becoming the norm across the rest of the world as well - just in a
different guise.

 my responsibility however is to ensure that the CPU Cards are
actually working *before* we take peoples' money :)

 the MEB as another example - the MEB was done on a quid-pro-quo
basis.  i helped someone out, they helped me out: no money changed
hands.  unfortunately the advice i gave them allowed them to make the
very sensible business decision to end the R&D project they were
doing!  we then transferred the MEB's design files over to
christopher, who at cost-only put them into a test / sample production
run.

 this then allowed us to test the CPU Card, we were able to find that
there were errors (in both the MEB and the EOMA68-A20), we're sorting
those, learning from this, etc. etc.

 now that we've learned from the experience of the MEB - it's served
its purpose - we can do MEBv2 as a much more "user-ready" product
rather than as a test/engineering sample.  but, again, these will need
to be paid-for on pre-pay.

 anyway.  this is probably more than you were expecting but gives you
a quick round-up (yes, really: quick!) of what's going on and where
we're going.

l.


[*1] Letters of Credit btw used to be the way that you could get say
60% of the funds to pay the factory and components, based on the money
being held in your client's bank account as security.  once the
products were finished and at the docks, the client's money would be
transferred in full and the bank that put the 60% up would be paid off
from that.  the problem comes if the U.S. Govt decides to freeze the
client's funds in between the transfer to the factory and before the
client's money gets transferred rather than held...