To be written to document the processing I've done in S+ (now R):

Processing Steps:

Velocity: [The below doesn't quite work.  Velocity data still have biases.  Also, there is a strong correlation between wind and temperature at the frequency of each turn and each complete cycle.  I still have to determine why.  Part of the correlation problem is that, due to the catenary, the trolley samples temperature at different heights between turns which, due to the temperature gradient, introduces a periodic temperature signal.  This can be overcome by using potential temperature, but then I'd need to know the trolley height quite accurately (GPS??).]

• read data for each sensor and QC
• integrate IMU to make vel & angles
• approx() all data into one array [at one sampling rate]
• produce index of beg & end turns and turn metadata:

1. pitch>max | gyroz>max | accX<min for less than 3sec
2. produces time beginning/leaving turns
3. find heading leaving turn
4. compute turn number based on heading (pattern)
5. compute mean GPS position of each turn beg&end

• add turn number, speed, position along track to array
• adjust gyro integrals to mean TCM attitude
• remove g*sin/cos(angles) from linear accelerations
• just subtract trolley cable speed from u in instrument coordinates!
• rotate u,v,w into geographic coords