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

On Wed, Jun 20, 2007 at 09:29:30AM +0100, Alan B. Pearce wrote:
> >Would a high current (20ma/node) shared power/data single
> >wire bus be feasible?
> 
> Depends on how you want to modulate it. Are you looking at 2 way comms? If 
> not then you could transmit out by turning the voltage off in short pulses, 
> with a very short pulse for 0 and a slightly less short pulse for 1. If you 
> need to have return data, then the current modulation will need to be a 
> significant (I would look at around 10%) of the max current to be able to 
> detect the change without problems.

Given that I need to have address discovery, then yes, 2 way comms has
to be possible. I might want to put sensors in the network too later on.
Being able to upgrade the firmware of the devices via bootloader would
be really cool too, except that no 12F or 10F part supports it.


Now by current modulation are you assuming that the nodes will work as
some sort of current loop? My understanding is that I'd be doing voltage
modulation for both ways. The slaves would short Vcc/Data and GND to
indicate zero's. Given, say, a 5k pull up and a short through a 200ohm
resistor the slaves would be pulling the line down to 0.2V

> >My speed requierments are very low, 500bytes/sec for the whole
> >bus would be fine for this application. I'll only need to update
> >maybe 100 nodes once per second with say a 1 byte payload,
> >4 byte address.
> 
> Do you really need a 4 byte address, or is that just trying to be compatible 
> to the Dallas 1 wire system? I would have thought that a 1 byte preamble, 1 
> byte address, 1 byte payload, 1 byte checksum would be a better scheme.

Dallas actually uses an 8 byte address, compatibility isn't really an
issue. I just think that being able to assign addresses randomly would
be really usefull in this application and would save a lot of hassle
with keeping track of node addresses. After the fact addresses can be
discovered with a binary search type mechanism and, and then visually
confirmed to figure out where in physical space each node ended up.

I'm going to leave the checksum off as integrity isn't critical. This is
just a light show. :)

In the 1-wire system the "preamble" as I understand it is the master
holding the Vcc line to ground long enough to reset everything. I'll use
a similar system I think, timing wise though, yeah, that preamble would
take up, say, 2 or 3 bytes worth of time.

> >I can't say I know much about emf though, and this will be
> >potentially thousands of feet worth of twisted pair CAT3 strands.
> >Would I be correct in saying that if I keep my slew rates and
> >overall bandwidth down, through low-pass filters, it would help?
> >Is this likely to even be an issue?
> 
> >From this I take it that you are looking to split Cat 3 cable into twisted 
> pairs, and string that around. Personally I would look at stringing around 
> full Cat 3 cable so you have 4 twisted pairs, and use 3 pairs for power, and 
> 1 pair for differential data. I appreciate you are trying to make the nodes 
> as small as possible, but I would seriously contemplate using a CAN 
> transceiver at each node to drive the cable. This is one additional 8 pin 
> SO-8 chip per node, plus a couple of capacitors and maybe a resistor. It 
> will take care of emc, shaping and thresholds for you, while allowing you to 
> use RS485 style comms. Then do the comms at around 1200 baud standard serial 
> should work fine. The 3 pairs for power should handle the 2A you are wanting 
> for 100 nodes at 20mA/node without serious voltage drops, but again you may 
> need to look at using an LDO or switching regulator at each node, and supply 
> the cable with a higher voltage.

Sounds like a very robust system, but I really don't think it's doable
cost wise. For even just two pairs the wiring cost, 50ft per node, will
end up at $2.50 per node assuming I buy spools of 4 pair CAT3 at $200
per 1000ft.

I'm working on the basis that I can implement that each node will have
something like the following part list: (digikey prices)

$1 - 10f200 SOT23-6 package (maybe, depends on solderability by
		students)
$0.25 - led of some sort
$0.10 - diode
$0.10 - capacitor
$1.75 - 1"x1" pcb, qty 100
$?.?? - conformal coat of some sort
$0.00 - student labour to assemble it all...

$3.20 a node.


Another problem with full cat 3 is visuals... it's big and ugly. If
anything I'll may very well have to find a source of jet black twisted
pair somewhere, skinny and unobtrusive will be important during the day.

- -- 
http://petertodd.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)

iD8DBQFGeVlS3bMhDbI9xWQRAvunAJ43ggc1GSFIdahn/NkN1lEaByGQtgCfbDfW
ngTpaiSCi4ygdSEjHfVYNQE=
=ouVe
-----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