[Contiki-developers] Tmote sky low-power modes

[Jonathan L. <jo...@cs...>]

Hi guys,

To set the scene, I am part of a team implementing a high-level
component-based language called Insense for use on resource constrained
devices such as the Tmote Sky nodes. 

	http://www-systems.cs.st-andrews.ac.uk/wiki/Dias/InsenseLanguage

Thanks to Contiki our work has been fairly straight forward in not having to
deal with very low-level issues, well done ... ;-)


Recently I've been playing about with a Scheduler component that we want to
provide as part of our language and we would ideally like to incorporate
some power-management features. So I started playing around with low-power
modes under Contiki.

As far as I can tell, sending the processor to sleep when nothing is to be
done using the 

		_BIS_SR(GIE | SCG0 | SCG1 | CPUOFF); 

	(not sure why Contiki leaves out the SCG1 flag,
	seems to work fine when it is added, but that is
	not my main concern here, although I am intrigued)

statement in the contiki-sky-main.c scheduler loop saves roughly 1mA when
there is nothing to do. I get a pretty constant usage of around 2.5-2.6mA
without the sleep statement and roughly 1.5-1.6mA with the sleep statement
in place (excluding current spikes due to radio listen as set in xmac.c).
Even sending the processor into LPM4 doesn't make a huge difference, maybe
takes off another 0.05mA.

To clarify how I measure current: I am using an oscilloscope to measure the
voltage drop over a small resistor that is in series with the battery supply
to the device. I have experimented with 10 and 100 Ohm resistors, both show
the same 1.5-1.6mA results.


1) My main query is why we don't seem to get anything near the low μA usage
as outlined in the tmote datasheet (values shown below). Maybe I'm reading
the figures in the datasheet wrong or the figures in the datasheet are
somewhat optimistic, maybe some other parts of the tmote can be switched off
but aren't, maybe ... ?  I was wondering if anyone had had similar
experiences and/or has an explanation.

2) It is also odd that I can hardly detect the wake-sleep cycles. I can just
about see small blips (of less that 100μA) roughly every 15ms (64 times a
second) within the fairly constant 1.5mA (1.6mA for tmote with sensors)
trace. I am no electronics expert, but my initial thought for this one is
that I'm running into some sort of resolution issue in that the scope isn't
picking up the change in voltage drop over the resistor for very small times
(when the processor wakes up for some micro-seconds and goes back to sleep
as nothing needs doing).

3) Has anyone experimented with waking the tmote up from LPM4? If anyone has
any info here that would be great.

4) Apart from setting XMAC_CONF_OFF_TIME at compile time, is there any way
(without hacking Contiki or duplicating code) of altering the CC2420 power
cycle at runtime?


If anyone can shed any light on the above, that would be great. The values
taken from the tmote datasheet are below.

Cheers

	Jon

-----------------------------------------------------------------

According to the tmote datasheet we should be able to expect around the
following for the device as a whole (I think)

						NOM	MAX	UNIT
Current Consumption: MCU on, Radio RX		21.8	23	mA
Current Consumption: MCU on, Radio TX 	19.5	21	mA
Current Consumption: MCU on, Radio off		1800	2400	μA
Current Consumption: MCU idle, Radio off	54.5	1200	μA
Current Consumption: MCU standby		5.1	21.0	μA

-----------------------------------------------------------------

And for the MSP430 with Vcc = 3V
						NOM	MAX	UNIT
Active current at 1MHz 				500	600	μA
Sleep current in LPM3, 32.768kHz active		2.6	3.0	μA

-----------------------------------------------------------------

No virus found in this outgoing message.
Checked by AVG. 
Version: 7.5.524 / Virus Database: 270.7.0/1680 - Release Date: 19/09/2008
08:25