9kHz Tests, 2010-10-09 to 2010-10-15

Long transmissions of constant carrier at 8970Hz from 49.4159N 11.1347E by amateur radio station DF6NM were found to be detectable at Todmorden (range 1027.2km, 306 degrees). The frequency of the vertically polarised 5uW ERP transmission was maintained sufficiently constant for spectrum analysis at bin widths of 20 to 60 uHz to reveal the signal at a few dB above the sferic-removed background in a receiving loop antenna oriented towards Nürnberg.

Three transmission were made and the received signals are described below. Sferics were removed using a blanker having a threshold of 1.2 times the mean amplitude in the band 8-10kHz, where the mean amplitude is maintained by an exponential moving mean of time constant 1 second. Flux densities are accurate to about +/-50% and are compensated for the effects of the sferic blanker. Statistical significance of the spectrum peaks are given assuming an ideal Rayleigh distribution of the background bin magnitudes. The significance is the inverse of the probability that the signal bin would reach this level by chance, given the prevailing mean noise level.

2010-10-09 09:50 to 18:00 UT, 8969.998 Hz

Mean noise: 0.067 fT Signal: 0.38 fT SNR: 5.7 Significance: 1 * 10^11

2010-10-10 07:00 to 15:04 UT, 8969.997 Hz

Mean noise: 0.074 fT Signal: 0.29 fT SNR: 3.9 Significance: 1.7 * 10^5

2010-10-14 08:52 to 2010-10-15 09:00 UT, 8970.002 Hz

For this 24 hour test, the signal is analysed over 6 hours segments in 46uHz resolution, overlapping by 3 hours.

	Mean noise: 0.081 fT Signal: 0.42 fT SNR: 5.2 Significance: 1.5 * 10^9
	Mean noise: 0.10 fT Signal: 0.41 fT SNR: 4.0 Significance: 5.4 * 10^5
	Mean noise: 0.19 fT Signal: 0.49 fT SNR: 2.6 Significance: 190
	Mean noise: 0.22 fT Signal: 0.50 fT SNR: 2.3 Significance: 58
	Mean noise: 0.20 fT Signal: 0.24 fT SNR: 1.2 Significance: 3.1
	Mean noise: 0.20 fT Signal: 0.27 fT SNR: 1.4 Significance: 4.2
	Mean noise: 0.16 fT Signal: 0.32 fT SNR: 2.0 fT Significance: 23

Noise

Notes

• The flux density predicted using a simple propagation model described in Tests at 9kHz is 0.13fT for daytime and 0.11fT for nighttime.
• It is probable that the loss of signal between 21:00 and 09:00 in the third transmission is mostly due to spreading of the signal energy out of the 46uHz bin due to Doppler resulting from changing path length.
• Data collection was interrupted during the third test when the PC which handles the East/West loop developed a hardware fault.
• Receiving antenna located at 53.703358N 2.07491W. System noise is 2.8fT in 1Hz at 9kHz. Received data timed to 30nS RMS error.
• Current injection calibration performed at 9kHz on 2010-10-23 indicates that all the signal and noise levels reported above need to be increased by 6%.
• When examining many high resolution spectra it appeared that bin amplitudes of high significance were occuring randomly more often than might be expected for a Rayleigh distribution of the same mean amplitude. The presence of a 'high tail' in the bin amplitude distribution was confirmed by running some histograms, and then reproduced using artificial Gaussian noise fed through the same signal processing chain. The effect only appears at resolutions below about 10mHz. For example, at 1mHz, peaks with amplitude between 2.8 and 3.0 times the mean occur roughly twice as often as expected. This effect disappears when the sferic blanker is removed from the chain. It is suspected that the random peaks are due to chance periodicities occuring in the blanking waveform.