[Arm-netbook] microdesktop casework as DXF files for laser-cutting

[Luke Kenneth Casson Leighton]

On Sun, Apr 2, 2017 at 10:51 PM, zap <zapper at openmailbox.org> wrote:
> I am really though glad you take this sort of thing seriously. We should
> all thank you for such zeal towards keeping things "Libre"

 mmm... zeal's a dirty word, now :)

> I hope you are successful Luke. And really, if normand sees this post, I
> hope he has a better understanding of the consequences of such actions.

 let's hope so

> Anyways though, I look forward to when you get the rockchip processor
> working to your specifications.

 i'll be happy it works _at all_!

> One last thing though, you said the a20 doesn't support 4.9 kernel...

correction: it does... but not with 100% hardware functionality
supported (which is no good for what i promised and demonstrated in
the updates)

> but what about the rk3288 processor?

 mainline for everything.  u-boot, kernel - everything.  it's been
that way for some considerable time.

 i'm still struggling with reverse-engineering the proprietary DDR3
200mhz initialisation binary, working out what the discrepancies are
between that and u-boot.  it doesn't help that i don't know if i
soldered the processor on incorrectly or any one of the four 96-pin
DDR3 RAM ICs...

 i've bought a stack of cheaper RAM ICs with which to experiment:
they're the same ones as on the RK3288-Firefly so in theory i should
be able to use exactly the same dtb file for the dram settings...
except the damn impedance is only (appx) 50 ohms not 100 like it is
for a 6-layer 1.6mm PCB...

 basically there's too many unknowns in order to triangulate exactly
where the problem might be when any one of them could be the source of
the failure.

 at some point i'm just going to say "fuck it", work out how to load
the proprietary DRAM initialisation, then u-boot with DRAM
initialisation *disabled*, work out how to get the RK3288 *out* of
DRAM initialisation mode and to continue running u-boot.

 apparently the internal ROM USB-loader can be asked to "shift"
whatever's loaded down into a different address in memory, but how
it's then executed i don't know.

l.