[Arm-netbook] iMX6 EOMA-68 CPU Card

[joem]

Luke,

I'm only looking at it from an EE issue.

If you have EEs to hand, and plenty of confidence and certainty,
then debugging is a minimal problem.

However, if your infrastructure doesn't have EEs, then even a single
capacitor value mistake such as using a 1uF for a 10uF can
kill the board. Shorted wires, open wires and accidental wiring
mistakes are totally unforgiving on multilayer boards. In worst case,
the thing will smoke, and it won't offer up clues for a long time.

I've been there are done it - and my experience says build it
in small modules that are known to work (like Unix modular software),
and then join them up.

The equivalent of that is also to build final board, and then
to separate out various modules, put some distance between them
and put in break out pads. That only takes about one or two days.
(But disadvantage is that EEs don't get their boards until a very late stage.)

Either way, the EEs have something to work on when debugging.
Either way, there is no need to throw out anything.


________________________________________
From: arm-netbook-bounces at lists.phcomp.co.uk [arm-netbook-bounces at lists.phcomp.co.uk] on behalf of luke.leighton [luke.leighton at gmail.com]
Sent: 22 February 2013 20:36
To: Linux on small ARM machines
Subject: Re: [Arm-netbook] iMX6 EOMA-68 CPU Card

... to put that another way, jm: the boundarydevices layout is a known
quantity.  both power management (one of the parts i've left mostly
alone) and the DDR3 (the other part i've left COMPLETELY alone) are
already known and proven to work.

if i were to split the board up (processor only on break-out) what
would it achieve?  it would actually be a step *backwards*.  i would
learn a great deal, but i do not have to time to "learn a great deal"
- i want to get results.

so - all i have had to do is:

* take the boundary devices PCB layout
* remove lots and lots of parts and connectors.
* add a NAND flash chip (by cut/paste from other schematics i.e. the
A10 EOMA68 PCB)
* change the connectors (by cut/paste from other schematics i.e. the
A10 EOMA68 PCB)

the technique you're recommending would be:

* throw everything away
* start from scratch
* make a breakout PCB for the processor
* learn how to do DDR3
* make a PCB with DDR3
* learn how to do the Power Management IC layout
* do the Power Management IC layout
* everything else

... and then once that's done, then having been on such an exercise, i
would then have to go *back* to the boundary devices PCB layout and
continue from there.

so... i wouldn't *actually* be gaining anything by doing break-out
boards, because in this case, the hardest part - the DDR3 RAM - has
been done, and the bit that needs some expertise and knowledge of
power circuit layout has *also* been done.

basically what i am doing is applying the hacking techniques that i've
learned for free software engineering.  "that's beyond me, ergo don't
touch it, don't alter it, don't look at it, but most importantly TRUST
IT".  there are so many people - so many programmers - who cannot
comprehend that, let alone do it.

and i've found this technique to save me vast amounts of time.  it's
why i'm going after processors that have full schematics available as
well as a full known working PCB layout.

l.

_______________________________________________
arm-netbook mailing list arm-netbook at lists.phcomp.co.uk
http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook
Send large attachments to arm-netbook at files.phcomp.co.uk