[net2-wg] 600 pps, high throughput

[Joe Polastre]

If you are sampling at a really high rate constantly, you are no
longer a low power sensor network which makes this conversation moot.

You have a choice:
(1) use DMA on the radio
-OR-
(2) use DMA on your sensors, where sensors is plural.

If you don't like that choice, employ a second microcontroller, or an
external sampling agent.  This really isn't that hard of a problem.

I can't believe how unnecessarily complicated things have become when
a simple piece of additional hardware simply solve the problem, or
when you can tradeoff one resource for another.

the way sukun's hardware was designed was flawed; that's irreleavant
of the software design.

-Joe

On 3/13/06, Jonathan Hui <jwhui at cs.berkeley.edu> wrote:
> DMA makes it easy to minimize jitter between samples. It would've
> taken some of Sukun's headaches away.
>
> If you want to share the DMA, then you have to consider whether its
> better off to block until the resource is available, or just send SPI
> commands in software without having to block for the DMA. It's
> dependent on the app and how it uses its resources. In many apps,
> including Moteiv's demo apps, it is generally feasible to share the
> SPI as samples are done in bursts with lots of dead time in between.
>
> --
> Jonathan W. Hui
> jwhui at cs.berkeley.edu
> http://www.cs.berkeley.edu/~jwhui/
>
> On 3/13/06, Joe Polastre <joe at polastre.com> wrote:
> > you don't need DMA for an accelerometer.  microphone/sound, yes.
> > accelerometer, no.
> >
> > if (call Resource.inUse()) {
> >   call SPIInterrupt.enable();
> > }
> > else {
> >   call DMA.setupTransfer();
> > }
> >
> > Is that really so complicated?
> >
> > -Joe
>