A python script called in the tests/NEC Borealis directory contains functionality to produce the correct geometry and other inputs for a NEC engine program like 4nec2 to simulate/model SuperDARN antenna arrays.

This script can be used to generate some common orientations of the SuperDARN antenna arrays for use with a NEC engine. Some programs that can read NEC inputs are 4nec2 or eznec. This script has been tested with the free, latest version of 4nec2, 5.8.17, updated January 2020 and available here:

In order to use this with 4nec2, simply open 4nec2 and go to File->Open 4nec2 in/out file and select the file (it must end in ‘.nec’). Then hit F7 or go to the Calculate->NEC Output-data option. Interpretations of the results are beyond the scope of this help message. Note that 4nec2 requires DOS line-endings, which is why this script outputs DOS line endings.

By default, if you run this script with no options, it will create TTFD main and interferometer arrays with 21-wire reflectors, oriented like the Rankin Inlet radar with the main array in front of the interferometer array. This can be changed with the --int-x-spacing, --int-y-spacing and --int-z-spacing options which take in a floating point number of meters to offset the interferometer from the main array in 3d space. By default, it is 100m behind (y == -100) and centered in both x and z dimensions. Turn off the reflector via the --without-fence flag. In summary, the boresite is in the +y direction, the arrays are typically along the x axis, and the +z axis is distance from the ground.

The number of antennas in both the main and interferometer arrays are controlled via the --antennas and --int_antennas options. Defaults are 16 and 4, like the Rankin Inlet array. Set the --int_antennas value to 0 to remove the interferometer array.

The antenna spacing can be modified from the 15.24m default via the --antenna_spacing option.

The beam and frequency used can be changed from the defaults of boresite and 10.5MHz via the --beam and --frequency options, which take floating point values.

Finally, you can provide a custom name for the file generated by this script using the --output_file option. Just be aware that 4nec2 input file naming requirements are strict, no periods, and it must end in ‘.nec’.

NOTE There are some options that are not supported yet, like log periodic arrays, yagi arrays, as well as different power and phase inputs for the arrays. As well, baluns and feedlines are not implemented, so the signal sources are currently modeled directly where the balun would be on the antennas.

Here are two images that show what the default geometry looks like when importing the default output file into 4nec2. In the first image you can see a wide bird’s eye view of the main and interferometer arrays. In the second image you can see a close-up of the main array center two antennas, with the 21 reflector wires in the background. The blue rectangles are the loads and the pink circle on the antenna’s is the current source modeled in NEC.

ttfd array, wide view, from 4nec2 ttfd array, closeup, from 4nec2

If you calculate this array geometry with the defaults, you’ll see this window from 4nec2: This shows the horizontal gain in red, and the vertical gain in blue. You’ll notice that the most power is in the main lobe at 0 degrees off azimuth (boresite). This is effectively a beam 7.5 in the standard SuperDARN configuration and shows that the radar array has a F/B ratio of 22dB, a beamwidth of 8 degrees and a gain of 17.47dB. Note this is at a frequency of 10.5MHz.

ttfd array modeled in 4nec2

So, why is this considered a tool? What happens when a transmitter goes down? What happens when two transmitters go down? What about if the power output from one transmitter is half of what it should be? How about phase errors? All of these questions are possible to answer with tools like this one. Here’s a real example from Rankin Inlet, where transmitters #6 and #12 (indexed from 0) are both down:

../_images/pattern.png ../_images/pattern_no_tx6_no_tx12_boresite.png

The above two images are generated for the radar at Rankin Inlet, the first image shows the standard pattern if everything is working properly at boresite. The second image shows the pattern resulting from transmitters #6 and #12 not contributing to the system. The effects are immediately visible in the higher power sidelobes. The main lobe gain is reduced from 17.47dB to 16.92dB. The main lobe remains the same shape and width in both azimuth and elevation angles.

../_images/pattern_bm1.png ../_images/pattern_no_tx6_no_tx12_bm1.png

The above two images are generated for the radar at Rankin Inlet, the first image shows the standard pattern if everything is working properly at beam 1. The second image shows the pattern resulting from transmitters #6 and #12 not contributing to the system. The effects are immediately visible in the higher power sidelobes. The main lobe gain is reduced from 16.66dB to 16.13dB. The main lobe remains the same shape but is slightly smaller (~1 degree) in elevation angle.

Filter Testing

A Jupyter notebook called filters.ipynb is located in the tests/dsp_testing directory of Borealis. This notebook describes in detail the default DecimationScheme used by Borealis, the helper functions in for creating a digital filter, and creates an alternative filter with comparison to the default. This notebook also benchmarks the performance of the filter schemes on the GPU, both in runtime and in memory usage. Finally, the ramp-up and ramp-down of transmitted pulses is looked at, for characterization of the expected transmission spectrum.

This notebook is intended to make it easy to design and prototype new filtering schemes, which is useful for experiments which, for example:

  • Have a non-standard pulse length (the default is 300 microseconds)

  • Use range gates of differing size (the default is 45 km)

  • Use pulse compression (this increases the transmission bandwidth)

  • Are listening experiments (e.g. to measure the frequency spectrum)

  • Use a different receiver bandwidth (i.e. not the default 5 MHz)

In any of these circumstances, it is important to design a filter which works for the situation at hand.


A python script called located in the tests/NTP borealis directory contains functionality that can be used to plot some common statistics that the ntpd program can produce.

It requires that you’ve set up ntpd to log statistics. Currently supported plots are basic, but still useful. This script also requires the ntp configuration file to be able to accurately calculate the Allan deviation for PPS drivers.

The Allan deviation can be plotted if you have a clockstats file. The subject of Allan deviation is beyond the scope of this documentation, but it can give you an indication of your short, mid and long-term stability of your oscillator. In short, if you see a negative relationship between the y axis and the x axis that means that over the long term your oscillator is more stable than it is over the short term. Phase noise and Allan deviation are closely related.

Here is an example of an Allan deviation plot:

NTP stats Allan Deviation plot

Looking at the above image, it’s clear that the clock stats indicate the clock is more stable the longer you view it. This is generally true for GPS disciplined clocks. If you have a piezo crystal oscillator and generated an Allan deviation plot for it, you might see the opposite relationship. Combining the two types of clocks into a GPS disciplined oscillator will get you the best of both short and long term stability.

If you have a loopstats input file then you can plot two quantities:

  • The ntpd estimated time offset from true time in seconds vs time smaller values are better.

  • The ntpd estimated frequency offset in PPM from a ‘true oscillator’ (ideal UTC clock) vs time, smaller values are better.

Here are example plots of the loopstats offset and frequency offset:

NTP stats loopstats offset NTP stats loopstats freq offset

If you have a peerstats input file then you can plot three quantities for each peer:

  • The ntpd estimated time offset from true time in seconds vs time, smaller values mean ntpd thinks it’s closer to true time.

  • The estimated round-trip time for ntpd packets vs time. Very small values would indicate the peer is on the local network.

  • The dispersion value (seconds) indicates how spread out the offsets are for this particular peer.

Here are examples of the above three plots:

NTP stats peerstats offset NTP stats peerstats delay NTP stats peerstats dispersion

That dispersion plot looks like there are a few outliers, so lets zoom in on a smaller section:

NTP stats peerstats dispersion zoom