Time Variability in Ambient Noise - Assignment Example

1.4 Time Variability in Ambient Noise Notable time variability in the noise spectrum happens when sound reaches below 100 kHz. Variability occurs mainly due to: Shipping Intensity – There is a normal diurnal or daily periodicity when there is increased shipping intensity within the day. Wave induced turbulence - Huge energy level needed to come up with a fully risen sea and the gradual decrease of wave height after a storm due to inertia are two reasons why wave induced turbulence varies over much longer time periods. A breaking wave in coastal regions can lead to an increase of up to 10dB in deep water predictions. Currently, there is no proof for seasonal variability in ambient noise levels. However, one could expect noise level to increase because of surface agitation during winter. Generally, ambient noise in deep oceans decreases with depth because of the waves coming from ocean surface, which is the principal source of noise. An exception to this principle is sound channels. Herein ambient noise levels are nearly constant with depth. The usual assumption is that ambient noise is isotropic (equal on all directions). This is commonly considered in sonar equations. We therefore attempt to reduce the effective noise level by utilising of the directional capabilities of the hydrophone (NL-DI). However, in deep waters the sea surface is considerably the principal source of ambient noise. It these areas the ambient level is not truly isotropic. Conversely, in shallow waters ambient noise is essentially isotropic due to reflections from the sea floor (Davidson, 2006). The features of the parametric approach are best illustrated on an example taken from a hypothetical sonar application. The approach was to apply time-frequency circumstances for detection. The detection conditions are simply: A signal level high above the ambient noise as measured in time. Amplitudes around the expected sonar frequency that is sufficiently high above the ambient noise levels in frequency at the same time. Only when both conditions apply can we detect a signal. The second level, in particular, enables signal detection of a very low false alarm rate on narrow-band sonar signals. It acts as a narrow band pass filter at the input of detection algorithm, which efficiently eliminates out-of-band noise. Figure 1 shows a 50 kHz sonar return from shallow water. Simulated water-depth is 3 m. the transmit pulse has a duration of 0.4 Ms. The noise is of uniform distribution restricted to the signal band. Signal to noise ratio (S/N) is 20 dB while the signal’s full waveform was digitised at the rate of 1 MHz. Only the envelope of the signal is shown in Figure 1. Synthetic Echo Figure 1: Envelope of a Synthetic Echo from a 50 kHz Sonar (Acoustic Research Laboratory, 2006) Without any further processing, the echo time series, which consists of 1000 sample, is now submitted to the detection algorithm. While the parametric model is calculated for every sample, the spectrum from 48 kHz to 52 kHz is computed for every 10th interval. The 4 kHz band is evaluated at 200 frequencies with a resolution of 20 Hz. A three dimensional view of the sonogram (Figure 2) may illustrate how well the spectral peaks in the sonar return are defined in a time-frequency representation. 3D Sonogram Figure 2: 3D-Sonogram from a 50 kHz Sonar Return [Acoustic Research Laboratory, 2006] The algorithm is very fast and real time processing capable. It does not require that the entire signal should have been acquired before the process begins. The parametric spectral estimate is computed on a sample-by-sample basis. True real-time performance depends only on signal digitalisation rate in relation to processor speed. To evaluate detection reliability, signals with diverse S/N ratios were submitted to the algorithm. For each of the thousand experiments in each noise level, the signal was constant yet contaminated with different realisations of pseudo-random noise. The noise was evenly distributed and added to the signal. Signal plus noise were then band-pass filtered to simulate a band pass in an analogue signal pre-conditioning circuit. Signal and noise characteristics are summarised in Table 1. The results of two trial runs with 1000 signals for each noise level are elaborated in Table 2. A single trial was considered a hit when a signal was detected within the interval of 0.0036 and 0,0046 seconds. The true first break occurs at 0.00004 s. A miss was scored when the signal was detected outside the interval. When the signal was not detected at all, the outcome was summed into the no detection column. The hits, misses and no detection counters sum up to 1000 independent runs for the initial and repeat experiments for each S/N ratio. Remarkably, performance is regular with over 99% hits down to an S/N ratio of 12 dB where the signal is not detected at all in about 2% of the trials. Detection breaks down when S/N drops to 10 dB and in 35% of the cases no signal is detected. However, even the false alarm rate is still less than 5%. If the signal is detected, the dependability of the “time-of-flight” amount is still high. The noise lenience of the detection algorithm can be tuned to a particular function. For a demonstration, the algorithm is tested with a synthetic 50 kHz echo from a depth of 180 m. and break-off S/N of 6dB. The echo was sampled at 125 kHz, which is only 2.5 times the Nyquist frequency. Therefore a high over sampling ratio is not a requisite for the effective performance of the algorithm. The higher noise tolerance obviously trades off against a loss in accuracy even at higher S/N ratios where the number of misses increases. The masking background noise level above which the signal echo limit must be detected is the summation of: Self Noise, Ambient Noise and Reverberation. Both NL and RL are expressed on the dB scale relative to the standard intensity (Io). For successful detection of a target the echo level must exceed the general noise level. Whilst it is true that both noise and reverberation can co-exist, it is a common occurrence that one or the other will predominate. Self-noise consists of electrical noise generated by the hydrophone, noise from boat’s machinery and hydrodynamic noise. Self-noise may reach a receiving hydrophone via various paths. For example, noise from the ship’s engine, transmission and propeller may reach the hydrophone via: The superstructure of the boat Water column (directly) Reflection from the sea floor Backscatter from particulates in the water column. Self-noise generally increases systematically, with the speed of the vessel and it is usually accounted for via empirical measurements of self-noise versus vessel speed. We have already stated that within the operational frequency band of most sonar systems the ambient noise level (in dB ref. 1mPa) is dominated by surface agitation (waves). Also we have already realised that the magnitudes of the ambient noise due to waves “rolls off” or decreases linearly with frequency. 1.4.2 Noise Level Calculation There are essentially two stages of calculating Noise Level (NL): a.) Estimate the ambient noise level (Nf) at the operational frequency of the sonar (f in Hz). This is performed using the following equation: Nf = N1 - 17 log (fx10-3) (dB) [Davidson, 2006] Here N1 is the spectrum level at 1kHz. The value of N1 is dependent on the wind strength and can be obtained using simple graph or equations. b.) Also, correction must be made for the bandwidth of the receiving hydrophone. The term bandwidth here refers to the frequency band over which the hydrophone is sensitive. The broader the bandwidth of the hydrophone, the larger the range of frequencies in which the hydrophone can hear. Thus, a hydrophone with a broad bandwidth, hears more noise than one with a narrow bandwidth. The bandwidth (Df in Hz) correction is given by the following equation: NL = Nf + 10 log (Df x 10-3) – DI (dB ref. 1 mPa) [Davidson, 2006] 1.4.3 Reverberation Level The term reverberation refers to the sum total of the scattered sound energy. If a sound pulse is emitted into a room (e.g. a church), the sound can be heard reverberating around exponentially decaying in intensity with time. It is noted in the previous discussions of target strength that scattering (or texture) of reflector is small compared to the wavelength of the sound pulse. Possible causes of reverberations are: Volume Reverberation: Scattering by marine life, air bubbles and turbulence Bottom Reverberation: Scattering from the seabed. Surface Reverberation: Scattering from the sea surface. Thus, reverberation in limited conditions (requiring the reverberation limited sonar equation) occur when trying to sense a target close to the sea bed, surface or in a medium containing a large amount of particulate matter or bubbles. Volume reverberation level: Follows the inverse square law (decreases in proportion to 1/Range2) Increases with pulse length Increases with the source level (SL) Increases with the beam width Thus, in reverberation limited conditions are important to design a sonar system that is directive, has a short pulse length and uses low source intensity. Increasing the source level above a critical level merely increases reverberation and therefore does not improve echo detection. Fortunately, volume reverberation is generally low (-100dB to –80 dB) compared to typical target strengths (-25 to +25dB). However, larger (-70 to -60 dB) reverberation levels can occur in the deep scattering layer (DSL). 1.4.4 The Deep Scattering Layer Deep scattering layer is a diffuse band of densely packed biological material that includes: Phytoplankton and zooplankton (>10 kHz) Cephalopods Pelagic fish ( Read More