New Life in Openmoko Phones

Nils Faerber nils.faerber at kernelconcepts.de
Wed May 20 12:37:17 CEST 2009


Werner Almesberger schrieb:
> Nils Faerber wrote:
>> This would be one of the details I am interested in, i.e. would OpenMoko
>> Inc. help in making (read as "producing") this new design? With its part
>> stock, manufacturing capabilities, etc.?
> Access to components is currently under discussion, yes. There are
> at least some logistical issues, i.e., the GTA02 components seem to
> be at a place where it's difficult to move them. But we're working
> on it ...
> The idea is indeed that we can get most of the components from
> Openmoko. It's not only about the cost of the material but also the
> difficulty of sourcing certain parts and the errors that could be
> introduced in the process.

That would be cool, yes!
I also know from experience that some parts are really nasty to get -
either you do not get them at all or you have to buy large quantaties of
them.

>> Many of the parts in the GTA02 cannot be reasonably placed by hand.
>> There are almost a dozen (or more?) BGA chips which are extremely hard
>> to handle (you do not see if the balls match the pads).
> 
> Hehe, this reminds me of the usual "SMT sucks, where can I get this
> chip in DIP ?" discussion. This question is usually followed by
> someone suggesting some more or less crazy scheme that actually does
> yield a DIP component, and a number of people explaining their
> techniques for soldering SOIC and even SSOP. Then usually someone
> chimes in describing how to solder QFN and the like with often
> grossly inadequate equipment. And often enough, this ends with hints
> for how BGAs can be done with kitchen utensils :)

For sure this *can* be done - you can solder even large BGAs using a
hot-air gun. This may work, sometimes. But even if the probability of a
failure is just 10%, which is small in this procedure, and you have 10
chips like this on one board, well calculate yourself - you are already
scratching the 100% for a board failure. And repair of BGA is even more
nasty than first time soldering them ;)

[...]
> Anyway, back to reality. I agree that this needs a real SMT production
> line. There are some parts that can be difficult to SMT (buttons,
> connectors) that are better hand-soldered, but for most of the items,
> you want a properly quality-controlled and automated process.
> 
> Please bear in mind that the objective of gta02-core is not to make a
> design that's immediately ready for mass-production but to set up the
> process and make a small number of prototypes.
> 
> If some company should find the result appealing enough to turn this
> into a real product and make the corresponding inventments, that would
> of course be very welcome. But we can't count on this happening so
> far.

And here my realism kicks in and sees exactly this as the most critical
point since it involves quite high initial costs even for a fist small
run (or especially for this).

> If you have contacts with companies that make prototype SMT runs, it
> would be interesting if you could get rough cost estimates from them.
> Let's assume the following parameters:
> 
> - 150-200 different components, all of them in reasonably common
>   packages, on tape.
> - most difficult component is a 332-FBGA with 0.5 mm pitch (the
>   S3C2442B MCP)
> - 500-600 components in total.
> - 10-20 units produced.

That's tough.
I know at least three such companies, one beeing in my home town. For
such a large number of different components 10-20 units will be
*extremely* expensive. As a very rough number I once got ~10EUR
setup-cost per part type (and this was a friendship price for a much
simpler board), i.e. taking the above numbers this would roughly be
1.500 to 2.000EUR just for setting up the placer. After that it is some
small Euro-Cent number per placed part. Printing the paste and soldering
is more or less for free. To probably put this into perspective imagine
that such a placing machine easily costs several hundred thousand Euros,
up to millions of Euros (depending on speed and precision). Then
calculate an average life-time of ~10 years and you get an estimate of
the machine hour cost. And then imagine that setting up the placer for a
board like the GTA02 will take around a day - still not counting service
personnel needed to do the actual setup.

Those are of course just very rough numbers. It also depends on type of
parts, how many of which type, etc. But as a first rough figure it could do.

>> Then there are
>> almost microscopic parts like resistors and capacitors - which pitch?
>> 0402 at least if not even 0201 or smaller.
> 0402 is the smallest. For manual soldering (e.g., rework), size is
> less of a problem than density.

0402 is OK - we can do this in our work-shop, but that's smallest we can
do ;) Density is indeed another critical issue. And pitily we cannot do
any BGA at all. What we have is the "Expert" from Essemtec:
	http://www.essemtec.com/products.asp?ArtNr=EXPERT
For BGA we would need the extension "Microplacer"
	http://www.essemtec.com/products.asp?ArtNr=MPL3100%2F3200
which is ~10kEUR alone :(
And a new reflow oven at ~5kEUR.

Most "professional" manufacturers do not even have the above prototyping
equipment anymore - too expensive and too seldomly used. Because of the
high device price pressure they rather set on high volume.

> - Werner
Cheers
  nils faerber

-- 
kernel concepts GbR        Tel: +49-271-771091-12
Sieghuetter Hauptweg 48    Fax: +49-271-771091-19
D-57072 Siegen             Mob: +49-176-21024535
http://www.kernelconcepts.de



More information about the hardware mailing list