[Arm-netbook] Schematic and PCB layout CAD files

Fri May 31 23:49:57 BST 2019

On Saturday, June 1, 2019, Paul Boddie <paul at boddie.org.uk> wrote:

> On Friday 31. May 2019 17.06.14 Luke Kenneth Casson Leighton wrote:
> > On Fri, May 31, 2019 at 3:47 PM Paul Boddie <paul at boddie.org.uk> wrote:
> > > Also, the A10 news page has mentions of these dimensions from several
> > > years
> > > ago:
> > >
> > > http://rhombus-tech.net/allwinner_a10/news/
> > >
> > > However, the PC Card documentation indicates that the actual housing of
> > > such a card is only 54mm wide, so I don't see how existing PC Card
> > > housings would accommodate such a PCB.
> >
> >  that's what cost USD $15k when the engineer we paid didn't read the
> > spec properly.  i told him to source one of the caseworks, measure the
> > INNER dimensions of the area as defined by the plastic, and use those.
>
> The inner dimensions are actually interesting for practical reasons, which
> was
> why I was asking about the range of PCB sizes.

If there is no user facing connector, there is the possibility of a range
of length.

If it is ok to stick the PCB out the end of the Card just like WIFI and Eth
PCMCIA used to do, length is limited only by practicalities.

Otherwise, there is no "range": the casework really does define the PCB
size to around a 0.1 mm tolerance or less.

> [...]
>
> > > I do remember that EOMA68 cards (maybe of an earlier generation) were
> > > produced.
> >
> >  yes.
>
> 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.

>
> > > Then again, looking at the A10 news page, there is a picture of a
> > > PCB layout from 2013 with dimensions of 78.1mm x 47.3mm, although it
> isn't
> > > completely clear what the screenshot is really showing.
>
> Again, it was interesting to see this with regard to what kind of PCB
> sizes
> would actually be produced.

78.1 x 47.3 is what is required for the litkconn casework, and there is no
wiggle room on that.

>
> > > Another thing that I wondered about is the width of the board when
> > > accommodating a board edge connector like the Amphenol 95622-004LF,
> which
> > > seems to be a low-cost and readily-available connector. It seems that
> the
> > > board edge needs to be less than 50.8mm across because such connectors
> > > enclose the contact area on each side.
> >
> > https://cdn.amphenol-icc.com/media/wysiwyg/files/drawing/95622.pdf
> >
> > that's a really weird connector.  it appears to be a socket, however
> > it is one that fits on the *edge* of a PCB, of a very specific height
> > (am having difficulty working that out from the diagram).  it's
> > probably requiring 1.2mm PCB however that is guess-work.
> >
> > no wait, Cross Section C, it's 1.5mm, and that also tells us it's the
> > PCMCIA *header*.
>
> Yes, my interpretation of Section C-C is that 1.5mm is the minimum board
> thickness, with the usual 2.2mm connector.
>
> > btw, 1.5mm is useless because the clearance on TOP components is
> > nowhere near enough.
>
> When you note that 1.5mm is useless, do you mean that within a housing
> (casework), a 1.5mm board takes too much space from, say, 3.3mm (Type I)
> or
> 5.0mm (Type II), and that a thinner board is needed?

Both.

>
> Computer card heights/thicknesses mentioned here, by the way:
>
> https://www.elinux.org/Embedded_Open_Modular_Architecture/EOMA68/FAQ
>
> I could understand 3.3mm being difficult, with 0.9mm left on each side of
> the
> board, but 5.0mm should leave 2.6mm (minus casework) on one side. Then
> again,
> I know the margins can be pretty small with these things.

They are. This was described several times on this list and in at least one
update.

The stainless steel is 0.1mm thick.  That leaves 4.8mm.

 The SDMMC sits 1.9mm above the PCB, as does the mid mount USB OTG.

The 2.2uH Inductors were also very very specifically chosen to fit under a
1.9mm threshold, they are 3.2 x 3.2 mm and are quite expensive compared to
cheaper wire wound inductors which come in around the 2.5mm height mark.

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.

So now we are down to 2.9mm.

On the underside the *MID MOUNT* Micro HDMI and USB OTG connectors sit
1.6mm below the PCB.

No components above that height are permitted, it has meant a couple of
redesigns, moving some large 0805 capacitors topside.

These underside capacitors, particularly under the CPU and DRAM, are
absolutely essential for stabilising power during peak loads, and they
clearly cannot go on TOP because they have to be extremely short tracks.
Being NEXT to the SoC would NOT be okay. They HAVE to go underside, as
close to the BGA pin where power is drained as physically possible.

Aside from the Micro HDMI and OTG mid mount connectors there is nothing
over a 1mm height. Still, that is enough.

2.9 minus 1.6 is 1.3mm

That leaves a mere TENTH of a millimetre spare clearance if using a 1.2mm
PCB.

Can you see that for this type of design a 1.5mm PCB would be completely
impractical?

If on the other hand there were no mid mount connectors it MIGHT be
possible.

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.

> > > I am only really asking these questions because I have been looking at
> > > making some footprints and other resources, and at least the
> fundamental
> > > board dimensions should be an obvious thing to discover, but I just
> > > didn't see them mentioned as prominently as I had thought they would
> be.
> >
> >  the PCB has to fit inside the casework, and the casework's *external*
> > dimensions are required to conform to PCMCIA.
>
> Yes.
>
> >  PCMCIA, is, obviously
> > (https://www.google.com/search?q=PCMCIA+card+dimensions) 5.0 x 85.6 x
> > 54.0 mm
>
> This is what I found, but I wondered why it wasn't mentioned on the
> specification pages, or why 55mm appears in the diagrams. I did find two
> pages
> with the 85mm x 54mm x 5mm dimensions on the Rhombus Tech site, however:

If you see any that are wrong please do correct them.

>
> http://rhombus-tech.net/allwinner/a31/orders/
> http://rhombus-tech.net/freescale/iMX6/orders/
>
> >  however the reason why there's no prominent mention of the *inner*
> > dimensions is because, obviously, they're critically dependent on what
> > *casework* is chosen, *NOT* repeat *NOT* on a hard spec.
> >
> >  it is *PCMCIA case size conformance* that is the hard requirement,
> > *NOT* the actual PCB size.
>
> Yes, it was really for an indication of what kind of PCB sizes result from
> these hard requirements that prompted me to ask. That and a need to
> question
> why there were boards made that were wider than the apparent external
> width of
> the card housing.

Engineer whom we paid $15k to didn't listen to instructions.

We got access to a Senior Engineer after that, at a massive discount.

That was Mr Xu and he was brilliant. The PCB he did was visually stunningly
beautiful.

>
> >  example: the PCB size (its length) will also critically depend on the
> > PCMCIA header dimensions.  if the header is N mm deep, then obviously,
> > if you make a PCB that is exactly 85.6 mm long, it will stick out the
> > end of the PCMCIA casework by N mm, won't it?
>
> Yes, so my impression is that a PCB using the Amphenol part referenced
> above
> would give up 6mm of the total board length budget, yielding a 79.6mm PCB,
> perhaps minus whatever any rear part of the casework might require.

>
Sonething like that, yes.

>  Again, not
> having any experience with the casework, it would be interesting to know
> what
> the margins are, especially if ports would be exposed.

0.1mm would be an acceptable tolerance.

>
> >  so this is why you don't see "PCB dimensions" mentioned as a fixed
> > quantity anywhere in the EOMA68 specification, because the PCB
> > dimensions must be *CALCULATED*, based on the PCMCIA *PARTS* that are
> > selected for use in production.
>
> Yes, so my query was mostly motivated by a lack of familiarity with the
> variables unknown to me, which are mainly related to how bulky the
> casework
> is.

The case is very very specifically tied to the connector.

It's no good ordering random headers and hoping that the plastic and metal
case will fit it.

The ends of the connector for example may have a different shape from
another manufacturer part.

Bottom line, it is basically absolutely critical to get a matched set of
casework and connector.

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.

I deliberately overordered litkcon cases and headers, as a just in case.
And the AMP socket and rails, too. Meaning, some can be ordered from Mike
direct, if you really need them.

And overordered somewhat on the JAE mid mount micro HDMI Type D, after
going through FOUR redesigns due to HDMI Connectors going EOL, sigh

L.

> Paul
>
> _______________________________________________
> 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

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