[Arm-netbook] rk3399 what full schematics does lkcl want?

[Luke Kenneth Casson Leighton]

---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68


On Thu, May 18, 2017 at 8:51 PM,  <ronwirring at safe-mail.net> wrote:

> lkcl says, he cannot make use of the rk3399 cpu

 no, i did not say that.  i specifically said that it would be a total
waste of my time to *CONSIDER* using the rk3399 until *full reference
design CAD source files are available*

>  because he does not have the full schematics.

 no, i said full reference design PCB and schematic *CAD* files:
that's completely different, as full schematics are usually provided
in "PDF" format which are completely f*****g useless as you have to
re-enter absolutely everything into the CAD design system and it takes
weeks if not months... *and then you have to do the PCB as well*.

> What are these schematics?

 you mean full reference design PCB and schematic CAD files.

 examples include:

 * http://beagleboard.org/hardware/design  search that page for the
words "allegro" and "orcad"
 * https://www.kosagi.com/w/index.php?title=Novena_PVT_Design_Source
search that page for "altium source"
 * https://github.com/NextThingCo/CHIP-Hardware and again look for the
".DSN" (orcad) and ".brd" (allegro) files
 * https://github.com/radxa/oshw/tree/master/rock_pro likewise

 and many, many more.  Jetson TK1, Sabre Lite, Sabre iMX7 - there are
absolutely loads of examples.


>  Can you tell in advance what schematics are required?

 you mean full reference design PCB and schematic CAD files.  yes you
can tell: the best ones contain both an example PMIC layout as well as
the DDR RAM layout.  the *absolute* best ones are a fully-working
board (typically named EVB - evaluation board) but a "beaglebone" or
somesuch will do just as well.

 a full reference design basically allows a transformation into a
fully working Card to take place in as little as three weeks.

 the ones that are a pain are the ones that *only* contain an example
DDR RAM layout to the processor.  but this alone saves LITERALLY
months of effort.

> Can you name them?

 name them... sort-of.

 words to use when doing google searches include "BSP" or "EVB" but if
that doesn't work try "{processor name} space {PCB CAD Design Package
name}"

> Or do you require all of the full schematics?

 the more that is available the less time and money is wasted.

> You do not know what schematics you require until you have read the full schematics?

 if you just want to READ the schematics you can do that with the
auto-generated (read-only) PDF that is otherwise f******g useless for
doing actual modifications because it's a read-only output from the
CAD package.

 obviously this will however allow you to assess whether the SOURCE of
the CAD files will be any good, such that you can decide whether
you're wasting your time or not.

 for example, it's no good if you have a tiny board and the PDF shows
that the design has (and requires) 8 DDR RAM ICs.   or, if it uses
LPDDR RAM or uses POP (package-on-packagee) RAM, you need to make a
decision about that (see the neo900 nightmare they just had because
they picked a POP RAM)

 but once you have reviewed the *READ-ONLY* PDF, even if you obtain
the schematic SOURCE file, you STILL cannot make a final decision
because you NEED TO SEE THE PCB CAD LAYOUT FILE AS WELL.

 the PCB CAD file contains the footprints of the ICs, it contains
tracks laid out, ground planes all done and completed, and a huge
range of engineering expertise including R.F. and E.M. expertise of
the person who did the layout, all done, all of which you DO NOT HAVE
TO DUPLICATE.

 but sometimes the distance between the SoC and the RAM ICs will be
too great: in a small PCB the ICs would not fit, so you would need to
consider *modifying* one of the most difficult and challenging areas
to get right (the DDR layout).  that's just nuts: you'd be wasting
your time, so it would be better to find a completely new layout that
someone else has already done.

 sometimes the PMIC (power management ICs) are in the wrong place for
the target PCB size you want, so you have to work out how to move (or
re-lay-out) those, or replace them entirely.  that is a whole can of
worms on its own, so again it would be better to see if you can find a
new layout that's been tested and known to work.

 basically it's a hell of a lot of work even to *evaluate* a Reference Design.

l.