[sim-wg] Status of Power TOSSIM-2

[Philip Levis]

On May 30, 2007, at 4:23 AM, Venkatesh S. wrote:

>
> Hello everyone,
>
> I am giving some final touches to the power profiling in Tossim for  
> T2.
>
> Modules Completed:
>   1) CPU power profiling - Monitoring the MCU Control register of  
> the microcontroller ATMega128 and computing the power consumption  
> of each CPU state.
>   2) LED Power profiling - Monitoring the PortA pins (A0, A1 and  
> A2) to which Leds are connected, and computing the power  
> consumption for each Led.
>   3) Radio power profiling - Monitoring the SPI read and write to  
> the radio, and computing the power consumption of Radio transmit 
> (Tx) and receive (Rx) states.
>
> Other Modules:
>
>   a) Noise characterization using MCU's ADC
>
>         We use the existing Closest Pattern Matching (CPM)  
> developed by HyunagJune Lee et al to generate the noise samples for  
> a  simulation (say X).  We then obtain the Tx power level (say Y)  
> of the remaining nodes by peeking into the CC1000's internal  
> registers. The "gain" parameter specifies the power level at the  
> receiver (say Z). Using X, Y, and Z we finally compute the RSSI 
> (dBm) for a reciever to sense the medium based on which the MAC  
> layer decides "busy" and "free" channel.
>
>   b) Support for B-MAC (CC1000 MAC) functionality for simulation.
>         Three files were modified in order to run the B-MAC  
> simulation.  At the HPL level, each byte of transmission is  
> signalled to all the other nodes.  It is a byte level simulation to  
> suit the CC1000 radio. B-MAC with acknowledgements also works  
> satisfactorily.
>
>   c) Support for obtaining node expended energy at the end of the  
> simulation. (modified tossim.c, tossim.i files and used SWIG tool).
>
> ToDo :
>
>     i) Thorough testing and comparision of power profiling with  
> respect to AVRORA.  I am currently working on this.
>
>    ii) Should add the functionality of getting only connectivity  
> based node Tx Power level.
>
>   iii) Should add the functionality of signalling the SPI read  
> events only to the nodes that are connected to the transmitting node.
>
>    iv) Monitoring the External Memory access and computation of  
> energy for read and write to memory.
> Specifically, it is adding the AT45DB functionality for simulation.  
> Since this is restricted to power profiling, we do not intend to  
> create or emulate the memory. Any pointers/feedback will help a lot.
>
>     v) A packet level simulation module for CC1000. Any pointers on  
> this will help immensely.
>

It might be useful to separate out all of these contributions;  
introducing them separately might make it easier to manage.

For example, one first step might be to incorporate mica2 simulation  
without power support. That is, make it so a user can type 'make  
mica2 sim' and have it work.

The next step can be to incorporate the power calculations. By doing  
this in a slightly more general way, it might be possible to easily  
incorporate support for the micaz.

As for a packet-level simulation module for the CC1000, the current  
one might work if you tweak some of the constants. One major issue is  
the SNR/PRR curve, but it shouldn't be a big deal to figure this out.

Phil