-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sat, Jun 23, 2007 at 10:01:03AM -0300, Gerhard Fiedler wrote:
> > There will two things driving the data/vcc line. A high-side fet,
> > connecting directly to the 5A supply, the strong pull up, and an IO pin
> > from the master PIC, connected through a 5k resistor, the weak pull up.
> 
> Yes, this works if you can shut down supply when you need to communicate
> back. OTOH, it has some potential of disaster: if a slave, for what reason
> ever, tries to communicate while the supply is on, at least that slave's tx
> driver will pop.

One 200ohm resistor on the TX will fix that I think.

> >> Depending on your requirements, there may also be the possibility to use
> >> some 5 or 10 sub-centrals, controlling 10 or 20 LEDs each. These can then
> >> be connected to the central with more wires (separate supply and comms). 
> > 
> > Well assuming that I simply wire each RGB LED up directly to the master
> > I'll need 4 pairs. At rock bottom prices that's going to cost me about
> > $1.36/50ft.
> 
> Did you mean 4 pairs or 4 wires?
 
Doh, yeah, 4 wires, 2 pairs, but I did get the price right.
 
> > Only a $0.37 difference, 
> 
> Well, if cost minimization is your goal... You didn't count the PCBs,
> though. At this end of the price scale, the number of PCBs does matter
> (unless the plan is to etch them at home). 

Well, I assumed that the "dumb node" system wouldn't need a PCB, and
that the smart node system would, and that the pcb would cost $0.50
each. I didn't include the cost of pcb's for the master, but that's more
of a one off expense, and could be done with cheap vector board.

> > although the latter is a lot more "interesting" in terms of getting it to
> > work. 
> 
> Definitely :)

Admittedly the interest does have a bias on me... :)

> > The former though would be more labour in making it, I bet plugging parts
> > into a pcb is easier than shrink wrapping four leads per led.
> 
> I don't think so. Populating 100 PCBs with small micros is probably more
> work than populating a bigger one with a few big micros or a few bigger
> ones with one micro each, even considering the mounting of the LEDs. Don't
> forget the chance for mounting errors. Using the latter approach (a
> two-level star) would get the wire cost still down (even though it seems
> that your node cost is determined mostly be the LED).

In my previous experience having students plug stuff into boards they're
surprisingly good at making no mounting errors, but fidgity stuff like
stripping wires and doing up shrink wrap they tend to be slow at. Heck,
if there such such a thing as a cheap, pcb mounted punch down block, I'd
be very tempted to use it. In my experience it's really tough to get
students to strip wires well. It takes ages to teach them not to nick
the wire, even with proper tools.

Another key thing is that with pcb's you can set up a nice efficient
production line system, with one person handling one task, like putting
the parts in the board. That's not so easy to do with attaching leads to
an led.

> > On another note though... Cheapest RGB leds at digikey are $3.50 each
> > (qty 100) and they are going to be non-stocking. Eek! Someone said
> > they'd gotten really cheap RGB leds, like $0.50 each or soemthing,
> > where?
> 
> Can you use individual LEDs?

Probably not, there wil be no way to mix the lighting, the leds will
bare other than some sort of conformal coating.

- -- 
http://petertodd.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFGfdzJ3bMhDbI9xWQRAlGxAJ9kD8s9sfpP8yPbtrPn2Z2bBBMqJACeNtbW
FDRGsWHu7rofL/Ew+t8CYt4=
=vhnY
-----END PGP SIGNATURE-----
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist