[Arm-netbook] [EOMA26] example product layout

[joem]

> >> >> > does anyone know of any other SoCs that would suit EOMA-26?  i've
> >> >> > listed the TI AM3389 already
> >> >> > http://elinux.org/Embedded_Open_Modular_Architecture/EOMA-26
> >> >
> >> >
> >> > Sqeeeeeeek!!!
> >> >
> >> > This type of IO expander scheme chips are a waste of money for the
> >> > price they are being sold in relation to the much higher functions
> >> > that are expected of them.
> >> >
> >> > Its cheaper to get a $2 cortex-3 lets say with 48 pins or 64 pins
> >> > and then use its i2c to talk to the chip, and then have inside
> >> > the chip a little bit of software for keyboarding, led driving,
> >> > interrupt generation, adc, usb, ethernet, lcd, etc.
> >>
> >>  yes - i mention that further down.
> >>
> >
> > Sorry, yes, I were trawling through list of datasheets and didn't get to
> > bottom of page.
> 
>  i've put a summary at the top.
> 
> > So would it not be easier to start with an arm cortex m3 lets say
> > and stick with it. The shipping cost would not pay for the difference
> > between a $2 chip and a $0.5 chip. Such an expansion board would cost
> > under $10 but do far more than anything competing with it.
> 
>  what happens when people look at that page and start making I/O
> boards in quantity 100k and 1m, with local supply chains?

They would go for dice - $0.2 to $0.5 max.
Make 10k boards per day easy.

>   this isn't
> a small-scale project joe where we're writing this solely and
> exclusively for tinkerers and arduinoists.
> 
> > There is also the LPC1768 costing $5
> > http://www.aliexpress.com/item/LPC1768FBD100-LPC1768-NXP-QFP100-100-brand-new-original-authentic/990248235.html
> 
>  $5 is way too much, given that the SoCs on the EOMA68 CPU Cards are
> likely to be around the $2 to $3 mark.  how can we find the 10k / 100k
> pricing?


Talk to NXP direct I guess.
Again dice if possible if you got the qty.
Alternatively Nuvoton who has similar ARM cortex chips.
There are numerous other ARM players who are too eager to sell dice.
But qty is what matters to them. Haven't been in CN for a while
but if I remember back to last visit, the dev kits for these
things are in CN and generally not available in English.
I don't know if that has changed.
But dice is your best friend if you got qty and a dev system in English.
If you got dice most small factories could turn out 10k boards in a day.

The other advantage of dice is that the same chip
can be bonded to different PCBs that bring out different pins.
So a 100 pin or 144 pin ARM dice would cost not much more,
but you could make use of 10 pins or 20 pins or 50 pins
and no one would actually need to know what is under that blob
of goo that covers the dice.
So you can settle for best dice and go for qty
instead of several different CPUs or different IO chips
that eat away at resources.

> 
> > I got KiCAD boards that I make already - the SoM1 CPU board takes 2
> > of these puppies and brings them out to 200 pin SODIMM.
> > http://www.gplsquared.com/SoM1/SoM1.html (the chip it takes is
> > marked up as LPC1764 - same pinout more memory in LPC1768 - 512K!)
> > It will take next to no time to modify it and make it specific for
> > EOMA26 needs.
> 
> that'd be fantastic.


Good - post what you need - lets see how long it takes to get it
together. The programmers are under $20 and software is free for
code under 128k. Anyone can knock up the code - if I have time, I might
be able to start it.