Here are the data files as well as some graphs and a tool for extracting the relevant information from the raw data files. These graphs show that the median us much more consistent and reliable than the mean or mode. See for yourself: median.ps mean.ps mode.ps
The data was collected on March 20, 2001 in the hallway on the 5th floor of Soda. The experiment was setup as follows: the base station was set on the ground at the end of the hall and a test mote (programmed to transmit at different signal strengths) would broadcast a test packet. The base station would measure the signal strength of the packet, and report the received packet along with the signal strength. The packet format is as follows:
Each packet is followed by a signal strength readout (printed as a float).
The signal strength is determined by sampling the BBout pin from the TR1000 only when two consecutive bits in the packet are identical. This assures that we are not sampling the transitions between bits.
Note that the test pattern is designed to encode into something which looks like: 110011001100... so that we have a balanced number of ones and zeros. Consult the SEC_DED code for details.
Based on this data it should be possible to analyze bit error rates, signal strengths, packet loss rates and so forth.
The data files are contain the raw captured packets. The name of the file indicates the measurement point, eg. 20ft_20_mod.dat indicates a set of measurements taken at 20 feet with the potentiometer modification and the potentiometer set to 20.
I also wrote a simple analysis program to extract more meaningful aggregates. This program is newprocess.c The *.stat files contain these aggregates, each contains a bunch of different aggregates for a particular pot setting. The format is:
distance packet reception rate mean variance max_strengthmin_strength median, mode
There is also a bunch of postscript graphs with these results.
I hope that this will be of some use to you.