On Friday 06 April 2007 13:57, Philip Levis wrote:
> On Apr 6, 2007, at 12:27 PM, Kurt Peters wrote:
> >     If you seek to estimate the power consumption, you can read
> > through the
> > data sheet for the 128L from Atmel if you're using that group of
> > motes.
[snip]
>
> One of the issues here is that the range of currents make it very
> difficult to measure with a single set-up; there's four orders of
> magnitude difference between the lowest and highest draws.
>
> The approach that being seem to be settling on in the research
> literature so far is to perform microbenchmarks of specific states
> using a precision multimeter, correlate those values with the
> observed but inaccurate oscilloscope values, use an oscilloscope
> trace to determine how long the node spent in each state, and
> calculate the energy that way. This mostly matters for sleep currents
> -- the oscilloscope readings are go good for things like the radio
> being on.
>
> I'm sure there are better and more complex measurement setups to do
> this more accurately, but this approach seems to be pretty effective.

[OT] You mean like this one?  ;^)

We coupled high-bandwidth precision power supply with a precision resistor in 
series with its output *before* the voltage feedback path.  We power a sensor 
node with this guy, then use a high-speed DAQ to capture the voltage drop 
across the precision resistor and the PS output voltage at high speed 
(>10MHz).  We can then graph the voltage, current and power consumption over 
time.  By having the sensor node twiddle a couple of digital outputs sampled 
concurrently, we also get a time reference for notable events.

We didn't build this to create power consumption models but to measure power 
state transistions.  As an example, our board uses a lithium battery, so the 
design includes a voltage regulator.  We initially selected a very efficient, 
low-Q part.  But testing has shown the original VR was woefully inadequate in 
terms of bandwidth and couldn't maintain desired regulation during many types 
of power events.  The moral for me has been: even though we're talking about 
small currents in sensor nodes, the relative magnitude of spikes and slopes 
in power draw can be significant.

Steve
_______________________________________________
Tinyos-help mailing list
[email protected]
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help

Reply via email to