OLPC hardware: what if there was an SDR modem / chipset?

[Luke Kenneth Casson Leighton]

hi scott, thanks for joining in, here.

On Tue, Jan 26, 2010 at 3:01 PM, C. Scott Ananian <cscott at laptop.org> wrote:
>>  so, there's therefore a wide-open opportunity to deploy SDR modems
>> which can be reprogrammed as a poor-man's Base Station in a pinch,

> I'm not sure why you think this is a radical idea.

 i don't.  well, i do.  ok.  clarify.  i believe it's a "possible"
idea - one with quite radical benefits.

> OLPC is sitting 1
> floor above Vanu (http://vanu.com/) which has spent 12 years working
> on software-defined radio for base station applications.

 that's far too long!!

> truth, OLPC is very very very small, and can barely afford to manage
> 802.11 networks in its deployments, let alone anything more
> complicated.  OLPC is still around because (and only because) they've
> managed to shrink their vision from the "7 new things" to the 1 or 2
> new things they can actually afford to develop and field.

 good plan!  get something out the door, for goodness sake.  excellent strategy.

> Good luck with your SDR dreams.  I think the most you can expect from
> OLPC is that, *if* the 10 improbable things you described actually
> come to pass, OLPC will someday include a miniPCI slot so that they
> can plug in your magic daughtercard.

 miniPCI assumes there's a PCI bus.  ARM SoCs don't have PCI (that i
know of).  the only SoC x86 chip i've heard about is the 1ghz IEC
100HV from RDC.  but with ARM it'd have to be a mini PCI-e slot with
the USB2 option - pins 36 and 34 - but that's been ruled out because
john's removed all internal USB paths.  the next gen x86 clone from
RDC that they're planning will have full PCI-e.

>  (And even this step would be
> horrifically expensive, in terms of retooling costs & etc.)

 but on the scheme of things, not _that_ expensive.  even the
certification costs, with the mass-volumes involved, can be sort-of
crossed out / recovered (i hope!) leaving the highest priorities being
the BOM and the power budget.

> that's not going to happen until *after* that magic daughtercard is in
> mass production.  So get busy!  It's only a couple months work, right?

 :)  one of the chip's out Q2 2010 est.

>  --scott
>
> ps: It might also be worth your time to study manufacturing processes
> in detail.

 sorry, i didn't mention: i have a friend and mentor who worked for
LSI Logic and then samsung.  over the past ... 7 years in which i've
known him, he's told me all sorts of stories and information about
chip manufacturing, RF design issues, the works.

>  The "dozen  RF connectors" you describe *by themselves*
> add a significant amount to the cost of the product.

 that was an evaluation "breakout" kit for limemicro's chip.  you
wouldn't put those kinds of RF connectors and certainly not that many
onto a low-cost finished product.

 ok.  lots to do.  thank you to everyone who's been responding.  can
we draw this to a close, let people get on with work: i have enough
information now to make the necessary enquiries.

 thank you for everyone's time, i'll be back again when i have more
concrete news.

 l.