[Arm-netbook] Schematic and PCB layout CAD files

Luke Kenneth Casson Leighton lkcl at lkcl.net
Sat Jun 1 21:18:34 BST 2019


On Sat, Jun 1, 2019 at 7:05 PM Paul Boddie <paul at boddie.org.uk> wrote:

> > Otherwise, there is no "range": the casework really does define the PCB
> > size to around a 0.1 mm tolerance or less.
>
> It was mostly needing to know how wide a PCB might need to be. Obviously, one
> can play it safe by keeping the width well within 54mm.

 the connector end clearly has to be wide enough to actually fit the
connector pins.  that's somewhere around the 41 mm mark.

> [...]
>
> > > There was a run of cards done at one point (or more). Did anyone actually
> > > do anything with those cards at the time? I seem to remember confusion
> > > about engineering boards, people having one kind of board and not the
> > > other, and so on. Did they all end up in people's desk drawers or
> > > something?
> >
> > Pretty much.
>
> Do you have any reflections on that? Was there a realistic expectation that
> people might develop software or hardware to take advantage of those boards?

 the vivaldi project - where its leader aaron seigo lied to a *lot* of
people in order to keep his standing with his team, after agreeing to
the hardware spec *and* the approach taken (which then didn't work
out, and he ran out of money to do another iteration) - that really
f*****d things up.

 it didn't help that the people that he'd picked to work on the
hardware and the website failed to listen to what i was saying: they
made a board that didn't conform to the spec, and started publicly
making unauthorized statements about what would go into the EOMA68
specification.

 that was the big opportunity to reach a lot more people - hardware
and software engineers.

 it also didn't help that tsvetan has consistently and repeatedly lied
to people in the sunxi community (and other businesses) about the
goals and business collaboration that we proposed to him.  his lying
has caused several people in both the hardware and software
communities to walk away, because they believed his lies.

they didn't come and ask me, "is this true?", meaning that they
*wanted* to believe the lying.

> Could more have been done to accelerate the initiative with this access to
> actual hardware?

 that was the idea: unfortunately, as those people were volunteers,
they didn't take it further.  someone will, at some point.

> Knowing your policy on information existing only on the list,

 que?  no, the wiki's there for people to edit and make the
information more accessible.

> > Also the SoC is around the same height, plus various large capacitors (1206
> > 10uF) and one large diode have all had to be special item "Low Profile"
> > orders, all 1.9mm or lower.
>
> I must admit that I don't have a wide experience of SoC heights, with the only
> one to hand here being an Ingenic JZ4780 that doesn't really seem very tall.

 it doesn't... and yet if you check, you'll find that the ethernet
port is 10mm high (5x higher than the limit for components inside an
EOMA68 Card on the TOP layer), a double-height USB2 stack is just over
12mm, and even the TOP mount Micro USB-OTG port will be somewhere
around the 4mm mark.

 this is what has made the EOMA68 project so ridiculously difficult
compared to "Just Yet Another SBC".  availability of TOP mount
standard-sized components like USB2, ETH, USB-OTG, SD/MMC, they're
dead easy to get hold of, plus if the locations are slightly "off", so
what?

 it was pure fluke for example that the MicroSD, Micro-HDMI and
Micro-USB-OTG happen to fit in between the "teeth" of the Litkconn
PCMCIA stainless steel shells (the ones that notch into the plastic).

> > Anyway this is why Litkconn header and casework were selected because it
> > has all 68 PCMCIA pins on TOP instead of a twin pair staggered in height.
>
> Are such headers (I wasn't aware of the terminology) easy to find?

 no they're not - PCMCIA is only now used in some obscure locations:
France still has Satellite "Conditional Access Module" cards, and the
supplier that i found in shenzhen has a big customer in Korea that
still buys PCMCIA parts.

> Searching
> for PCMCIA headers in the mainstream channels, even using convenient sites
> like Octopart, is excruciating.

 try finding a mid-mount Micro HDMI Type D, or a mid-mount Micro-USB-OTG.

 then check to see if they're not "last available supply in the world"
(having been manufactured over 5-8 years ago, and Digikey is the last
company in the world with the last remaining parts)...

> > If you see any that are wrong please do correct them.
>
> The diagrams are the challenge since they are not readily editable. However,
> they possibly need reworking, anyway, partly due to their vintage.

 i hand-edited them anyway.

> > Bottom line, it is basically absolutely critical to get a matched set of
> > casework and connector.
>
> Yes, I understand that although the connector should be standardised, how the
> "non-mating" side of the connector fits into the housing might be wildly
> different from one kind of connector/header to the next.

 they're all designed to slot together according to *external*
dimensions, because that's what PCMCIA specified.  the *internal*
dimensions are what you cannot just go and mix-and-match random parts
on, because the *internal* dimensions were never part of the PCMCIA
specification.

> > The casework forms the basis of fitting in the rails.
> >
> > No casework, the bare PCB can be misaligned on insertion, not just
> > horizontally by a couple of pins, it can even be misinserted by an entire
> > row.
>
> This seems pretty unfortunate, really.

 PCMCIA was never designed as a "bare PCB mechanism" for engineers, it
was designed as a *mass-volume* standard, easy for *users*.

> It does seem that the Amphenol
> connector tries to guard against this - as well as I interpret the drawings -
> and I wonder if others also do so.

 in combination with the correct matching casework, of course.  being
incompatible with the PCMCIA *external* specification would have been
disastrous.
> Do you think that, ultimately, some other connector standard would be more
> accessible? That is, easier to develop for without substantial tooling
> investment.

 if one existed i would have found it and written a spec based around it.

 as the number of pins would be completely different, it would *be* an
entirely different standard, and would take a good year or so to
properly review and get right.

 and would require basically leaving all other standards unfinished
and require additional funding to complete.

 best to get this standard out the door, first.

l.



More information about the arm-netbook mailing list