[Home]
[QEX] [GNU radio NC]
Finding Signals in the Noise using Two Antennas
This page is an addition to the article in the March/April 2019 issue of QEX.
This is courtesy ARRL and posting here is allowed because I am the author.
Signals and noise are discriminated by their polarization behavior.
The polarization statistics of noise is different from that of sky-wave signals (and that of identifiable local noise sources). Noise is the sum of multiple uncorrelated sources received from different directions with varying power and polarization. A sky-wave signal drifts slowly in time, in direction of arrival and in polarization. Signals and noise are separated by filtering over time, frequency and polarization.
It contains additional information and noise reduction audio examples of SSB and of CW signals. The examples start with screen recordings of one mouse click noise cancelling, of displaying propagation polarization behavior and of a very weak CW signal.
Results of the VERON man-made noise measurement campaign (K.Fockens, PA0KDF, et al) clearly indicate a significant increase of the noise floor over the years:
Screen recordings:
1---- Screen_recording_of one_mouse_click_noise_cancelling shows how a local man-made noise source is cancelled by clicking on the location in bulkspace. Bulkspace is the sum of all subspaces. It is resized and interpolated for a higher resolution in location needed for the noise cancelling. Phase difference P horizontally (X axis), amplitude ratio R vertically (Y axis).
The recording starts in bypass mode passing the raw antenna signals. The noise cancelling is enabled by switching the bypass off. The narrow subspace LPF is (for demonstration purposes) selected to enhance the location of the local man-made noise source. The man-made noise source is cancelled by a single mouse click on the indicated location. The S-meters indicate the decreasing noise level.
Note: if not only the man-made noise is present, but also a signal, the location will still be visible in the breaks of the conversation. And the location will equally be visible in the detected Noise Floor. So the best setting for the noise cancelling can be found even in the presence of a signal.
2---- Screen recording_FSF_in_ABspace shows an example of a Frequency Selective Fading NVIS signal on 80m in the A/B linear polarization space. Both the ordinary (O) and extraordinary (X) propagation mode are present and multiple reflections. The resulting elliptical polarization is frequency selective drifting/rotating in orientation and direction.
It starts with showing in gray the frequency selective polarization drift in the subspaces. The drift is visible as a frequency selective rotating location of the signal. Next bulkspace shows in color the frequency selective fading over all the subspaces (frequencies).
Each one of the 64 squares is a subspace of 4 bins (4×15.625=62.5Hz). Audio band is 0-4kHz from top left to bottom right.
3---- Screen recording_FSF_in_LRspace shows the same example of a Frequency Selective Fading NVIS signal on 80m in the L/R circular polarization space.
It starts with showing in gray the frequency selective polarization drift in the subspaces. The drift is now visible as a frequency selective horizontally drifting location of the signal. The Left and Right circular components are not exactly equal in amplitude. The Ordinary mode is slightly stronger. Next bulkspace shows, also in gray, the frequency selective fading over all the subspaces (frequencies).
4---- Screen_recording_20m_CW_very_weak_signal_NC_NR shows the presence of a very weak CW signal by the emerging red dots in the subspaces. See also audio example 11 below.
Each one of the 64 squares is a subspace of 1 bin (15.625Hz). Audio band is 250Hz-1250Hz from top left to bottom right.
Noise reduction audio examples:
Different versions of each audio sample are indicated in the name by:
Note 1: The slow AGC is active in all samples, but at a low knee level. Bypass and Stereo diversity signals are controlled by a single AGC gain.
Note 2: Using (two) speakers can result in notches in the audio spectrum even with mono signals. Readability, especially for CW, can be listening location dependent. A headphone gives the best results not only for stereo diversity.
5---- An easy 40m band day time SSB signal with Frequency Selective Fading in stereo diversity (using minimum frequency span setting in the noise reduction), the carrier is locally generated man made noise.
40m_SSB_G_day_DiversityStereo_minSpan_NR
6---- An easy 80m band day time SSB signal with a weak local man-made noise source (NVIS Dutch stations).
80m_SSB_demo_weakest_parts_Bypass
80m_SSB_demo_weakest_parts_NC_NR
80m_SSB_demo_weakest_parts_NC_NR_AGC
7---- A very weak 40m band day time SSB signal with some Frequency Selective Fading using noise cancelling, the carrier is locally generated man made noise.
40m_SSB_G_day_weak_NC_minSpan_NRwide
8---- A weak 20m band SSB signal using mono and stereo diversity.
20m_SSB_weak_DiversityStereo_NR
9---- Combined real band SSB signal and real 80m day time band noise with some local man-made noise. The signal is set as a circular polarized signal. In the Netherlands during day time NVIS signals on 80m are most of the time circular polarized. The peak signal level (RMS time constant 125msec) in the next examples is 3dB below the noise cancelled noise level.
Combined_SSB_signal_source: the original signal
Combined_SSB_NC_NRwide: using a wide LPF
Combined_SSB_NC_NRwide_AGC: using a wide LPF and higher AGC knee
In the next two samples the peak signal level (RMS time constant 125msec) is reduced to 6dB below the noise cancelled noise level.
Combined_SSB_NC_NRwide_AGC_min3dB: using a wide LPF and higher AGC knee
10---- Combined real band SSB signal and real 80m band noise with strong local QRM.
QRM @180degrees, signal @270degrees, signal shows some QSB.
Combined_SSB_strong_QRM_Bypass
Combined_SSB_strong_QRM_NC_NRwide: using a wide LPF
11---- A real 20m band very weak CW with QSB at the threshold of the noise reduction.
Bandwidth is 500Hz.
20m_CW_very_weak_signal_Bypass
12---- A circular polarized artificial CW signal @270degrees (12wpm) with Gaussion/Normal distributed noise. Bandwidth is 300Hz. RMS time constant 125msec.
The indicated key down SNR (S/N) is in each antenna signal. For average SNR subtract 3dB.
Carrier at -8dB/300Hz = -3dB/100Hz = -17dB/2500Hz = +3dB in 1 bin:
CW_carrier_750Hz_random_noise_Bypass
CW_carrier_750Hz_random_noise_DiversityStereo_NR
CW V at 12wpm at -8dB/300Hz = -3dB/100Hz = -17dB/2500Hz = +3dB in 1 bin:
CW_V_750Hz_random_noise_Bypass_1
CW_V_750Hz_random_noise_DiversityStereo_NR_1
CW V at 12wpm at -5dB/300Hz = 0dB/100Hz = -14dB/2500Hz = +6dB in 1 bin:
CW_V_750Hz_random_noise_Bypass_2
CW_V_750Hz_random_noise_DiversityStereo_NR_2
Read what a good weak-signal EME operator can copy: The Weak-Signal Capability of the Human Ear by Ray Soifer, W2RS
Note about noise reduction for digital modes
If the digital mode software is properly implemented the signals are filtered already in the best way matching the mode. Noise reduction basically only filters and so cannot improve (much) the filtering for digital modes like FT8.
Only if the digital mode software is not optimal implemented noise reduction could help.
Noise cancelling and diversity however can and will improve weak signals reception for digital modes.
[up]
Last update: October 8, 2024
© PA0SIM