TDoA measurements using GPS time-stamped IQ samples from KiwiSDRs (8)

Today I picked up a HF radar signal on 13500+-20 kHz. These are the time-differences between four KiwiSDRs:

TDoA 13490 kHz

This is most likely the Calypso CODAR radar system:

TDoA maps for 14390 kHz

Within the TDoA resolution the three sites in Ta'Sopu (Gozo), Ta'Barkat (Malta) and Pozzallo Harbour (Sicily) cannot be distinguished.

It is still puzzling for me that the TDoA maps seem to provide reasonable results, as for now great-circle distances are used, i.e., delays due to skywave propagation are not taken into account.

De-chirping applied to a HF over the horizon radar signal

Here is another look at the HF OTHR signal from Cyprus already mentioned in this blog post.

This OTHR signal is (at least to 1st order) FMCW modulated (period 0.02 sec, bandwidth 20kHz), as can be seen in the FM demodulated phase which follows a sawtooth pattern with period 0.02 sec.

Although the bandwidth of the KiwiSDR IQ data stream is smaller than the bandwidth of the OTHR signal it turns out that de-chirping can be done: using KiwiSDR GPS time-stamps t, de-chirping can be implemented by multiplying the IQ samples by exp(-2i*pi*f(t)) with f=@(x) 0.5*1e6*(mod(x,0.02)-0.01).**2-50;  (this is octave notation)

The de-chirped signal looks like in the bottom panel below and is flat as expected:

OTHR FM demodulated phase; bottom: de-chirped

The relative range (which includes a contribution from the doppler effect, see, e.g., here) is then obtained by performing a FFT transformation on the de-chirped signal in each 0.02 sec period (~240 samples):

OTHR de-chirped spectrum vs. time

• There is a main reflection in the ionosphere which has an interesting, time-varying structure to the left of it, i.e., towards smaller ranges
• In the light blue regions there is a kind of digital pattern visible; this might be due to the signal processing and limited bandwidth of the KiwiSDR but it could also be caused by some additional modulation of the radar signal

TDoA measurements using GPS time-stamped IQ samples from KiwiSDRs (7)

14370 kHz OTHR 20171211T1040

The autocorrelation function has peaks at multiples of 0.02 seconds: (as is well known)

14370 kHz OTHR autocorrelation

The pulse shapes and the FM demodulated phases are shown below w.r.t. GPS seconds mod 0.02 seconds. Of the full bandwidth of this signal is 20kHz the KiwiSDRs sample 12 kHz. Therefore only part of the chirp signal is seen:

14370 kHz OTHR pulse shape and FM demodulated phase

Correcting for the sample rate the chirpyness becomes 12001*83.326 = 1e6, i.e., one sweep covers 20 kHz.

In the time differences different paths of propagation can be seen; it looks like the ionosphere at this time was quite patchy:

TDoA 14370 kHz

The TDoA location is consistent with the known position of the transmitter:

TDoA maps 14370 kHz

TDoA measurements using GPS time-stamped IQ samples from KiwiSDRs (6)

6030 kHz Radio Romania International in DRM

Time differences:

TDoA 6030 kHz

Note that there are two propagation paths with different delays observed in Khimki. Using the delays indicated in red the following maps are obtained:

TDoA maps 6030 kHz

For the other path the maps look like this:

TDoA maps 6030 kHz

Both are compatible with the known position of the transmitter.

TDoA measurements using GPS time-stamped IQ samples from KiwiSDRs (5)

VOLMET on 5450 kHz 20171208T1734Z

Below are time-differences for a VOLMET signal using USB modulation. The cross-correlation peaks are not very sharp due to the limited bandwidth. For the cross-correlations, intervals of 1 seconds have been used.

time-differences for 5450 kHz 20171208T1734Z (VOLMET)

The resulting TDoA maps look like this:

TDoA for 5450 kHz 20171208T1734Z (VOLMET)

Note that all these results are biased by ionospheric propagation delays and many other sources of systematic errors.

HF over the horizon radar (OTHR)

Here are some notes about an OTHR radar signal (likely the one described  on page 28 here http://www.iarums-r1.org/iarums/radar-2013.pdf) which is audible every evening/night in most of Europe.

Note that there is a bias due to the fact that the KiwiSDR bandwidth (12 kHz) is less than the bandwidth of this radar signal (40-50 kHz)

This is how the autocorrelation looks like:

Autocorrelation

Autocorrelation zoom

• fundamental period: dt = (11573.3 ± 0.2) μs
• the autocorrelation function shows peaks at multiples of dt, 4×dt, and 150×dt

Using GPS time-stamped IQ samples each pulse can be plotted on top of the other pulses:

OTHR pulses

There is a discrete structure in the pulses: pulses with pulse number mod 4 = 0,1,2,3 look more similar:

OTHR pulses - pulse number mod 4

TDoA measurements using GPS time-stamped IQ samples from KiwiSDRs (3)

DCF77 20171127T1041

Shown are the cross-correlations w.r.t. time for DCF77. The red dots indicate the time lags obtains by a pol2 fit.

77.5 kHz DCF77

TDoA measurements using GPS time-stamped IQ samples from KiwiSDRs (2)

20171203T1836 2117 khz STANAG4285

For this signal the cross-correlations look like this:

TDoA cross-correlations 2117 kHz

There are multiple propagation paths with different delays

Using the time interval indicated by the red lines the following plots are obtained:

TDoA 2117 kHz

It would be good to have a method for finding the best interval instead of doing this manually.

Thanks to all KiwiSDR owners who have made an effort to provide GPS signals to their KiwiSDRs.

TDoA measurements using GPS time-stamped IQ samples from KiwiSDRs (1)

Some relevant links are:

• http://kiwisdr.com/
• https://github.com/jks-prv/Beagle_SDR_GPS/issues/130
• http://valentfx.com/vanilla/discussion/940/direction-finding-and-linking-existing-kiwisdrs

Big thanks to John Seamons for creating this excellent receiver and for support with the GPS time stamps, as well as to all KiwiSDR owners who have enabled GPS.

All plots below were generated by a combination of 

• kiwirecorder.py (https://github.com/jks-prv/kiwiclient/tree/jks-v0.1) forked from Dmitry’s https://github.com/dev-zzo/kiwiclient
• http://octave.org/ (+signal +mapping)
• http://www.gnuplot.info/

using the code here https://github.com/hcab14/TDoA which for now is the only documentation.

Note that this is preliminary work at a very early stage.

In particular many systematic errors are for now being neglected, such as

• different delays (antenna + frond-end) per KiwiSDR
• delays due to propagation
• not using consistently WGS84 coordinates
• having only approximate locations for the KiwiSDRs

More information will be added later on.

77.5 kHz 20171127T104156Z DCF77

http://www.mwlist.org/vlf.php?kHz=77.5

DCF77 TDoA

225 kHz 20171128T104940Z Polskie Radio 1

http://www.mwlist.org/mwlist_quick_and_easy.php?area=1&kHz=225

TDoA 225 kHz PR2

10611 kHz 20171128T171256Z OTHR

https://en.wikipedia.org/wiki/RAF_Akrotiri

TDoA 10611 kHz

65.8 kHz 20171127T160437Z FUE

http://www.mwlist.org/vlf.php?kHz=65.80

TDoA 68.5 kHz