Band Survey

What is this?

Using an RTL dongle you can create a survey that shows signal strength across the band over time. This is useful to determine if any patterns for interference are visible, like a specific time, or a recurring time, or some other aspect.

Last edited: 27 June 2019

Band Survey as a waterfall graph.

Band Survey as a time-frequency graph.

Hardware

The hardware is a Raspberry Pi connected to an RTL SDR v3 dongle with a Diamond HF10CL antenna. The data is gathered on the pi and transferred to my workstation for visualisation.

Software

These charts were created using rtl_power and gnuplot.

The rtl_power command comes from Kyle Keen, who has documented how to create a band survey. Kyle's page provides access to a python script that creates a chart. It works, but it's very slow for large data-sets and it's non-trivial to adapt the display.

In addition to the python script, Kyle refers to work done by superkuh, who documents how to use the rtl_power band survey data to make a plot using gnuplot.

Source Code

The source for this project is available on github: https://github.com/vk6flab/band-survey and can be cloned from: https://github.com/vk6flab/band-survey.git

Details

Various tools are combined to create the visualisation. rtl_power gathers data, gnuplot creates a chart.

rtl-power

The command to gather data suitable for use as a "really wide band survey":

rtl_power -f 0M:1766M:1M -i 1m -g 50 -e 240h "data.csv"

Warning: The above command runs for 240 hours, that's 10 days. The charts shown cover 85 hours and the data file for 85 hours is approximately 318 Mb. (This is a work in progress.)

gnuplot

The gnuplot file to create the time-frequency chart:

# Source: http://superkuh.com/rtlsdr.html#keenerdappnote

# rtl_power output format

# 2013-08-19, 20:53:16, 55000000, 57000000, 1000000, 4096, -29.60, -29.60,

set timefmt '%Y-%m-%d, %H:%M:%S'

set view map

# Load a color palette

load 'default.plt'

set output 'data-survey.png'

set term png size 1920, 1080

set key off

set title "Band survey: 0 - 1766 MHz"

set xlabel 'Time'

# Label every 3 Hours

set xtics 3 * 3600

set xdata time

set format x "%H:%M\n%d/%m"

set ylabel 'Frequency (Hz)'

# Label every 100 MHz

set ytics 100 * 1000000

set format y "%.1s %cHz"

# Set the range from -25dB to 25dB

set cbrange [-25:25]

unset colorbox

plot 'data.csv' using 1:3:7 with points ps 0.1 pt 5 palette

The gnuplot file to create the waterfall chart (essentially swapping the axis from the time-frequency graph):

# Source: http://superkuh.com/rtlsdr.html#keenerdappnote

# rtl_power output format

# 2013-08-19, 20:53:16, 55000000, 57000000, 1000000, 4096, -29.60, -29.60,

set timefmt '%Y-%m-%d, %H:%M:%S'

set view map

# Load a color palette

load 'default.plt'

set output 'data-waterfall.png'

set term png size 1920, 1080

set key off

set title "Band survey: 0 - 1766 MHz"

set xlabel 'Frequency (Hz)'

# Label every 100 MHz

set xtics 100 * 1000000

set format x "%.1s %cHz"

set ylabel 'Time'

# Label every 3 Hours

set ytics 3 * 3600

set ydata time

set format y "%d/%m %H:%M"

# Show time backwards

set yrange [*:*] reverse

set cbrange [-25:25]

unset colorbox

plot 'data.csv' using 3:1:7 with points ps 0.1 pt 5 palette

Both the plot files rely on a palette file called default.plt:

# New default color palette after Moreland (2009)

# see: http://www.sandia.gov/~kmorel/documents/ColorMaps/

# For the gnuplot implementation have a look at

# http://bastian.rieck.ru/blog/posts/2012/gnuplot_better_colour_palettes/

set palette defined(\

0       0.2314  0.2980  0.7529,\

0.03125 0.2667  0.3529  0.8000,\

0.0625  0.3020  0.4078  0.8431,\

0.09375 0.3412  0.4588  0.8824,\

0.125   0.3843  0.5098  0.9176,\

0.15625 0.4235  0.5569  0.9451,\

0.1875  0.4667  0.6039  0.9686,\

0.21875 0.5098  0.6471  0.9843,\

0.25    0.5529  0.6902  0.9961,\

0.28125 0.5961  0.7255  1.0000,\

0.3125  0.6392  0.7608  1.0000,\

0.34375 0.6824  0.7882  0.9922,\

0.375   0.7216  0.8157  0.9765,\

0.40625 0.7608  0.8353  0.9569,\

0.4375  0.8000  0.8510  0.9333,\

0.46875 0.8353  0.8588  0.9020,\

0.5     0.8667  0.8667  0.8667,\

0.53125 0.8980  0.8471  0.8196,\

0.5625  0.9255  0.8275  0.7725,\

0.59375 0.9451  0.8000  0.7255,\

0.625   0.9608  0.7686  0.6784,\

0.65625 0.9686  0.7333  0.6275,\

0.6875  0.9686  0.6941  0.5804,\

0.71875 0.9686  0.6510  0.5294,\

0.75    0.9569  0.6039  0.4824,\

0.78125 0.9451  0.5529  0.4353,\

0.8125  0.9255  0.4980  0.3882,\

0.84375 0.8980  0.4392  0.3451,\

0.875   0.8706  0.3765  0.3020,\

0.90625 0.8353  0.3137  0.2588,\

0.9375  0.7961  0.2431  0.2196,\

0.96875 0.7529  0.1569  0.1843,\

1       0.7059  0.0157  0.1490\

Notes

If you need to change a time offset, because your local time isn't correct, or different from the time on the data capture, you can replace the time column with a formula: (timecolumn(1)+8*3600). This adds 8 hours worth of seconds to each time stamp.

For example, to add 8 hours to the time:

plot 'data.csv' using 3:1:7 with points ps 0.1 pt 5 palette

Becomes:

plot 'data.csv' using 3:(timecolumn(1)+8*3600):7 with points ps 0.1 pt 5 palette

This project is not yet completed and the information on this page will be updated as it progresses.

Enjoy.

73, de Onno VK6FLAB