[Arm-netbook] iMX6 EOMA-68 CPU Card

[jm]

On Fri, 2013-02-22 at 02:39 +0000, luke.leighton wrote:
> http://rhombus-tech.net/freescale/iMX6/news/
> 
> just a brief update - i'm going counter-clockwise round the iMX6 from
> top left, and have managed to convince the autorouter to make all the
> critical connections: GPIO, LCD, SDCard, NAND, and so on.  i've found
> someone with experience with PADS whom i will be paying for their time
> and advice, to ensure that the layout is correct before they get it
> printed and populated.

I didn't realize you were doing this by yourself.
(I assumed a third party like Wits-Tech was doing this.)

Casting a critical ee eye over the project.

The usual way these things are done for this level of complexity
is to have break out boards and sample/prototyping boards done
up as quickly as possible to do the testing, and then when they are
working move for the final board.

To jump to the final board is for the gods that know what they
are doing and merely redoing what they do already every day
for different customers based from their experience with prototyping
boards and sample boards they did earlier.


> i thought people might appreciate the pretty colours.  those routes
> aren't done by hand, but i've spent six to seven hours a day
> re-running the auto-router, adding in *every single* BGA pin manually
> and in specific order, changing the order one or two pins at a time
> until it routes.  the density is so high that the optimise phase just
> completely fails to make anything but the smallest of cleanups.

Hmm...

My advise
1. Make a minimal break out board - it should only take about 2 days.
   The board should not be constrained for space.

2. Make second board with RAM, flash and power to be manually wired to
first board - about 7 to 14 days. The boards are not constrained for
space or number of layers.

3. third board with everything on it - but with prototyping area, some
breakout pads, and possibly extras on the board. About 2 weeks. The
boards are more constrained, and take on final form, with added bits 
that can be cleanly broken away if required.

4. When 1,2,3 are working, make the final board.


At each stage you will have boards and hardware that could be being
worked on and built by others and extended as needed
in parallel with your effort. Its less time to get the sample boards
built.

The MOST IMPORTANT THING in a hardware project is to have
working hardware. Even a semblance of life such as a flashing LED
controlled by software is a huge step forward for a project.

As hardware grows more complex, and as software gets into the
system, 'failure' become more common.
You must rely on your previous prototype hardware and software to
isolate the problems.
By building lifeline like this all the way back to simple hardware
and software, it is easier to get projects working quickly.


> i should ask, however, if anyone knows about schematics, particularly
> NAND flash, if they could check the schematics because i don't
> entirely know how that's supposed to go: the iMX6 appears not to have
> some of the pins needed (WP).  a PDF version is here:
> http://hands.com/~lkcl/eoma/iMX6/SABER_LITE_4.pdf

Page 158 - ball E15, IMX6DQCEC.pdf
( http://www.gplsquared.com/SoM3/IMX6DQCEC.pdf )