This experiment verifies the wireless-card power consumption measurements made by Feeney. This experiment gives us a framework for performing other wireless experiments. The results are intended for internal use but would be publishable if we find an unexpected anomaly.

Background reading

Table of contents
Method
Results
Discussion
Conclusion

When make measurements on an ad-hoc network connecting two wireless machines. Both machines booted in Linux can switch to ad-hos mode even when access points are available. 

Laptops:
Wireless cards:
PCI adaptor to measure V,I:
Oscilliscope: 

The measurements are made from UDP traffic between the two laptops. One device is connected to the adaptor from which the voltage and current consumption of the wireless-card is measured.

Experimental code. We should work on automating the power calculation as much as possible in the program that communicates with the oscilloscope. For example, the program can select the part of waveform where packets are sent or received. Those points will be distinguishable from the ideal times when the current is above a threshold (around 180 mA for receive and 280 mA for send).

Code

• udp -- Send and receive packets
• DSATalk -- Program communicates through network with oscilloscope.
• power_format -- Format DSATalks into format read into Excel.
• e_power -- Excel spreadsheets that loads and plots the formatted files in a batch process. Some parts are hardcoded.

Summary of results. Link to complete results and plots (includes a macro to process data).

 

Mode	Current (std) mA	Feeney mA 11 Mbps	Idle mA (Feeney Idle  156 mA)
Sleep Mode	--	10 mA
Unicast send	264	284 mA	152 (3.3)
Multicast send	264 (2.2)	"	154 (3.1)
Broadcast send	264 (4.2)	"	153 (2.3)
Unicast recv	191 *	190 mA	153 (1.7)
Multicast recv	168 (2.9)*	"	154 (4.3)
Broadcast recv	198 *	"	153 (1.1)

* Spike before recv (maybe from send)

The multicast recv measurements should be measured again because it differs from the other recv measurements.

Discussion [top]

One observation is that the voltage dips when the current spikes while sending. The current is at the highest when sending.

Our results compare to Feeney's results. Our wireless card seems to consume less power than Feeney's wireless card. No other anomalies were discovered.