Ionosphere frequencies

I computed effective ionosphere frequencies by two ways: 1) using weights proportional to the number of APs in each channel; 2) using weights proportional to the number of processed samples per channel, per sideband. The contribution to the ionosphere free linear combination of X-band and S-band group delays due to the differences in effective ionosphere frequency at X-band was computed as 
  Delta taugr = 2 Delta feff / feff * taugix
where feff -- effective ionosphere frequency, taugix contribution to X-band group delay due to ionosphere.

I used experiment c1010 ( $01MAR05XA ). The table and the plot of the contribution shows quasi-random function with the average -0.5 psec and wrms 4.1 psec. (The argument of the plot is the observation counter and units of values are picoseconds). Using new ionosphere frequencies slightly improves fit: 26.5 psec with respect to 26.8 -- about the same level which we may expect when we remove source of noise which rss at 4 psec level. At the same time I cannot claim today (11-JUL-2001) that the problem is solved. As a test I tried to reproduce band synthesis by computing group and phase delays using fringe phases, amplitudes, frequencies and the number of recorded samples per channel. The differences in group delay at X-band for c1010 experiment between the values whcih I computed and the values supplied by Fourfit are presented in plots and tables. Units are picoseconds.

Without using #SAMPLES With using #SAMPLES
Table of values Table of values
Plot of differences Plot of differences

We see remarkable improvements: rms of the differences without information about the number of samples (and using weights which are proportional to the number of AP instead of) is 12.2 psec while with using #SAMPLE it reduces to 1.0 psec. There are about 2% of outliers (about the same as the number of postfit residual outliers rejected in solution, accidental coincidence?). I gazed at fringe plots. The common feature of fringe plots of the observation for which computation group delays differ from Fourfit results is excessive scatter of residual phases: 50-120 degrees instead of usual 5-15 degrees. Outliers at S-band are about 10 times larger.

Back to Leonid Petrov's discussion page.

Last update: 11-JUL-2001 20:03:15