Purpose:
This version of rfplot.c was meant to replace the existing version of rfplot.c in Cees Bassa's STRF, https://github.com/cbassa/strf. I never got around to making some further changes and the present version of rfplot.c has evolved... I may get back to making a few more changes and work with Cees Bassa to incorporate those changes in the offical version. However, Cees is moving to porting much of this code to Python and the project is in limbo.
rfplot.c was modified to include the ability to mark wideband signal Doppler points in a more 'scientific' and 'consistent' manner than guessing. This is particularly important for the analysis of wideband data signals that have no carrier component.
This is not an offical or supported version. It is offered entirely at your own risk and may not work with current versions of STRF.
Instructions:
-
The 'w' and 'W' keys are all you need to control all of the functionality. 'w' selects a point and 'W' resets only the screen size so you can zoom into the data to see individual bins and reset as needed to relocate as you reduce the trace. The two keys like this make for fast work flow... More on that below.
-
First select with 'w' one point immediately above OR below the wideband data you would like to reduce a point from, a '+' will appear at that location on the screen. This selection will anchor the time so it is the one you need to be accurately aligned with over or under the data. Next, go to the other end of the swath of data and hit 'w' again, another '+' will appear. The alg will run and place a 'o' where it figures the centre of the curve is and save the point. Now you are ready to repeat on a new wideband swath.
- If you've zoomed in too much and need to relocate on the curve to reduce more points hit 'W'. This will reset the screen so you can zoom in on different areas of the curve and allow you to accurately select more points. Using 'W' only affects the screen zooming and will not delete your marks and points from the screen. This is so you can proceed with a reference of what you have done and what you need to complete. I find the zoom level in the first image below about right for accurate placement of the first point. The second point can be anywhere along the x-axis as long as it's in the spot on the y-axis you want. I find it helpful to give a little margin at either end of the data selection of no signal data.
Here's the result of reducing an entire curve in the image below.
-
To delete your marks from the screen use 'r' as usual.
-
As you mark your points they are reduced to a file called wide.dat.
A few notes on how the alg works
-
The slice of the spectrum between the points selected is taken and the data has its log taken.
-
The data is then fit to ax^2+bx+c to fit a parabola and the coefficients a,b and c are obtained.
-
From the coefficients obtained, the vertex of a parabola is solved using the equation vertex_x = -b/2a, thus in theory identifying the peak of the data blob.
-
The value is converted to MJD and Frequency and the point logged to wide.dat.
-
There is a certain amount of filtering in the alg. to prevent invalid entry of x,y mouse coordinates to prevent program crashes. Otherwise the code seems robust.
-
One area of improvement could be to implement a Weighted Gaussian solver as they are more immune to noise and could improve the reduction of data overall particularly with noisy data.
-
After writing this I discovered that there was an internal function for the curve fitting already in rfplot.c so my added external routine could be removed.
-
Some error handling should be added to prevent zero's from getting 'logged' etc.