22/07/2004 11:56:49, Tim Froggatt <
tgf20@cam.ac.uk> wrote:
>Okay that sounds good. So would it be feasible to do the
Balloon3 in, say,
>a 4 layer board?
Err, no. If you look at the BGA package, if I only had 2 routing
layers, I'd need to get 2 or 3 wires between each pair of pads.
Not going to happen, given the fine pitch of the package. I'll be
trying to get the layer count down to 6 (2 planes, 4 routing), but
that's not guaranteed either - the outer layers tend to be used for
components, and can't be used for anything other than extremely
local routing. That leaves me 2 layers for real routing, and that's
not much at all, especially as they're corrupted with vias.
>Obviously that would make it a good deal cheaper to
>manufacture.
Well, yes. It's not as if on Balloon2.0X I sat there and thought "how
can I make this more expensive?". Routing was absolute hell on
the layer stack we ended up with, and I had to push via pad sizes
rather further than I liked, just to get it to route at all. (The
problem is that, after a while, there's no point adding more layers,
since there's no room to put vias to get signals onto those layers).
-layer and 8-layer PCBs aren't particularly expensive or hard to
source, and will be an ongoing feature of Balloon.
>Yes that would be great. Although if we start using the existing
2.05
>boards initially, we may have to come up with our PWM add-on
board anyway.
If you don't need audio, and aren't planning to use the top 8 bits
of the NAND flash interface, then there may be sufficient room in
the Xilinx device to fit some simple PWM engines. However, I think
you want rather more than that. Hanging an extra Xilinx off the
GPIO resources (LCD connectors) would be my favoured route for
such simple addons)
>But the FPGA stuff would offer great potential. Do you have a
rough idea
>of how long it will be until the version 3 boards are
manufactured?
Not yet, no. Before the end of the year, I'd hope. I do have other
(paying...) projects on at the moment. Balloon isn't yet funded,
and is a heck of a lot of hours. I expect about 2-300, by the time
it's done - there's just a _lot_ of work to do.
>Another issue is that we were planning to network our robots
using
>wireless cards in the CF slot. So our plans would change if there
was a
>move away from the CF slot,
CF stays, since it's the way that many random things attach. The
PXA27X's USB interface may mean that CF gets less use this time,
but CF stays.
There are plenty of other ways to attach wireless gubbins, not
least via the synchronous serial ports, but that's not standard, and
there's no way I'm putting wireless silicon onto the Balloon board,
with all the regulatory hassles that accompany such a thing.
Steve
(text/plain)