Hi All

I have got  2 weeks more data for you - and it looks spectacularly good! The diurnal variation is clear - we have not seen that before.
The ascii data file (2007-03-13_16.57.19_data.dat) is big  (111MB), so I have made a matlab mat file(2007-03-13_16.57.19_data_cols_1-3.mat)  that just contains the first three columns which area the ones you need contained in the variable d.

Load like this:
load('2007-03-13_16.57.19_data_cols_1-3.mat')
t=d(:,1);   %time in matlab datenum format
d1=d(:,2);
d2=d(:,3);

I would recommend binning the data to a slower cadence before doing any analysis (eg 30s), as it is 1.4 million lines. Ensure that you remove the bad data points before (or during) binning, or one bad data point in the entire bin will wreck the binned data - which will result in 30x more bad data.