Imaging Lightning Progression Using VHF Broadband Radio Interferometry

Transcription

1 1892 IEICE TRANS. ELECTRON., VOL.E84 C, NO.12 DECEMBER 2001 PAPER Special Issue on New Technologies in Signal Processing for Electromagnetic-wave Sensing and Imaging Imaging Lightning Progression Using VHF Broadband Radio Interferometry Redy MARDIANA, Takeshi MORIMOTO, Nonmembers, and Zen-Ichiro KAWASAKI a), Regular Member SUMMARY The VHF broadband radio interferometry operated from 25 MHz to 250 MHz has been developed for observing lightning discharge progression. The lightning images are derived by sensing the electromagnetic-waves which are radiated during the discharges. The perpendicular baseline geometry provides the angular locations (azimuth and elevation) of lightning radiation sources. The lightning observations have been conducted in the Hokuriku District in The station consisted of three broadband antennas and an electric field antenna as well as a GPS receiver. The system was able to reconstruct lightning discharge channels in two-spatial dimensions and in time sequence. As one of the observation results, an upward negative cloud-to-ground lightning flash will be presented. key words: interferometry, broadband interferometry, lightning, electromagnetic-wave, remote sensing 1. Introduction Lightning discharges produce electromagnetic radiation over a very large frequency spectrum, from a few hundred khz to the GHz. Hence, during the last two decades a new range of investigation has been opened by the application of VHFelectromagnetic localization techniques to the field of lightning remote sensing. The VHFnarrowband interferometries have been developed for imaging lightning discharge channels [1] [3]. These interferometries are based on the measurement of phase differences of a certain frequency among signals received on different antennas of an antenna array. These phase differences are directly related to the direction of arrival of the signals and are used to calculate the angular locations of the sources in azimuth and elevation. Recently, broadband radio interferometries have been manufactured and its capability to image lightning discharge progression have been demonstrated. Unlike the narrowband interferometry, which is operated at a fixed frequency to locate the sources, the broadband interferometry takes into account all frequency components of the lightning flash emitted during its progression and uses these frequencies to solve the source Manuscript received March 28, Manuscript revised June 8, The Authors are with the Department of Electrical Engineering, Osaka University, Osaka-shi, Japan. The Author is with the Department of Communication Engineering, Osaka University, Osaka-shi, Japan. a) redy@pels.pwr.eng.osaka-u.ac.jp locations. Shao et al. [4] have used a broadband interferometry in a one-spatial dimension. Ushio et al. [5] have demonstrated the results of lightning mapping in a two-spatial dimension. Mardiana and Kawasaki [6] and Kawasaki et al. [7] explained more detail the principle and the system of broadband interferometry for two-dimensional observations. There are two recognized superiority of a broadband interferometry over a narrowband one when imaging the lightning discharges [7]: (1) the broadband interferometry can locate multiple radiation sources, which propagate simultaneously through branching in time and space, and (2) the located sources show less scattered tendency than those resulted from the narrowband interferometry. Since the lightning spectrum has very large bandwidth, the broadband interferometry system needs a high digitization rate to record its broadband signals, but since the memory of the digitizer is limited, the entire radiation from a lightning flash cannot be recorded continuously. Also, it is not practical if the digitizer has to record undetected broadband pulses during a lightning progression. To overcome the above disadvantage, a sequential triggering technique has been applied [5] [7], where the instruments start to acquire data only if the lightning emits broadband pulses exceeding a threshold level. This paper starts with an introduction of the experimental broadband interferometry for imaging lightning discharge progression. This includes a brief review of the principle of broadband interferometry and the method to determine the source location in two-spatial dimensions (azimuth and elevation) as well as phase ambiguity analysis. Then, we describe the experimental system. Last, we present an upward lightning flash as one of the observation results. 2. Broadband Interferometry 2.1 Principle The basic idea of this interferometry technique is to estimate the phase differences at various frequency components of Fourier spectra between a pair of broadband antenna sensors. The simplest radio interferometry consists of two separate antenna sensors. Consider two antenna sensors separated horizontally above the

2 MARDIANA et al.: IMAGING LIGHTNING USING BROADBAND RADIO INTERFEROMETRY 1893 Fig. 1 Two antenna sensors separated horizontally. ground by a distance d, as shown in Fig. 1. The broadband signal which is originated from a common source impinging on antenna 1 is r k (t) and on antenna 2 is r l (t). Assuming a total record length of T, then deriving the signals in the frequency domain by using Fourier transformation, we have R k (f) = T 0 r k (t)e j2πft dt R l (f) = T 0 r l (t)e j2πft dt. (1) The cross correlation between both signals is related to the cross Fourier spectrum function by the well-known relationship G kl (f) =R k(f) R l (f) (2) where * denotes the complex conjugate. The G kl (f) will generally be a complex number such that where G kl (f) =G Re (f) jg Im (f) = G kl (f) e jφ(f) (3) G kl (f) =[G Re (f) 2 + G Im (f) 2 ] 1/2 ( ) GIm (f) φ(f) = arctan. (4) G Re (f) The term φ(f) of Eq. (4) corresponds to the phase difference of signal r k (t) and r l (t) for each frequency component. For the case where the source is sufficiently distant to be approximated by a plane wave at the sensor locations, the angle of incidence θ(f) for the received signal can be interpreted as φ(f) given by or φ(f) = 2πfd cos θ(f) c (5) ( ) cφ(f) θ(f) = arccos (6) 2πfd where c is the electromagnetic propagation speed in free space ( m/s). It has been proven in literature [8] that Fourier spectral amplitudes of broadband lightning pulses decrease with increasing frequency, and frequency components having low spectral amplitudes give less contribution to the lightning image. To improve the lightning Fig. 2 Sensor arrangement: radiation source localization (top) and position of sensors at the observation site (bottom). image, a weighting function has been introduced to determine the estimated incident angle. The estimated incident angle θ est can be solved through the following equation, w(f) = G kl (f) θ est = w(f).θ(f) w(f) (7) where w(f) is a weighting function and it depends on the cross Fourier spectrum function G kl (f). In practice, the continuous signal will be digitized into digital data. For digital data processing with a recording time T and a sampling interval t there will be T/ t = N data points. Thus, the discrete frequencies will be separated by f =1/T =1/( t.n), and the upper frequency limit of the analysis is f c = 1/(2 t). Hence, for the Eqs. (1) (7) the frequency increment will be f with the highest frequency f c. 2.2 Baseline Geometry Employing two-antenna sensors, only a one-dimensional location can be resolved. To provide a two-dimensional system, the third antenna is added to determine azimuth and elevation of the source. The first and second antennas form the first baseline, and the second and third antennas form the second baseline. These two baselines are perpendicular to each other. Figure 2 shows the geometry of a perpendicular baseline interferometry. Consider a source in the first quadrant (the

3 1894 IEICE TRANS. ELECTRON., VOL.E84 C, NO.12 DECEMBER 2001 direction from antenna 2 to antenna 1 is 0 and the direction is counterclockwise). Each baseline provides a direction cosine between the source and the respective baseline. Let θ 1 and θ 2 be the estimated incident angles of the source for the first and the second baselines, respectively. The azimuth (α) and elevation (β) can be determined using trigonometric equations and cos θ 1 = cos α cos β (8) cos θ 2 = sin α cos β. (9) By dividing Eq. (9) with Eq. (8), with the constraint θ 1 90 and β 90, we obtain ( ) cos θ2 α = arctan. (10) cos θ 1 Finally, we determine β as ( ) cos θ1 β = arccos (11) cos α or equivalent with ( ) cos θ2 β = arccos. (12) sin α 2.3 Phase Ambiguity There is a dependence of the calculated φ(f) onθ(f) because of the large bandwidth and antenna separation geometry. From Eq. (5), to illustrate the relation between φ(f) onθ(f), a computer simulation has been made. The two antennas are assumed to have d =10m with the corresponding two antennas being identical. Figure 3 shows φ(f) as a function of f for three different θ(f) values and f is limited to a range of 0 to 250 MHz. Because Eq. (5) and also Eq. (4) are trigonometric functions, the output of φ(f) varies from π to π as the θ(f) varies from 0 to π. It is shown in the first and third panels that φ(f) goes through more than one cycle with increasing f, except for θ(f) = to radian (see the second panel). This means there are at least two frequencies that have the same φ(f) and, therefore, they provide ambiguous indication of θ(f) (see Eq. (6)). These cycles in φ(f) are often referred to as fringes. As the radiation source moves overhead (θ(f) π/2), the number of fringes decreases. However, φ(f) has a unique value or provides no ambiguity for each θ(f) atanyf 15 MHz, as seen from Eq. (5) for any d c/(2f c ) [6]. Note again that if the output of φ(f) is not limited from π to π, then φ(f) for f 250 MHz should cover a range of about or 17π <φ(f) < 17π Fig. 3 The first three panels from the top show the phase difference as a function of frequency for three different incident angles. The bottom panel shows the phase difference after displacing the fringes. π 2mπ < φ(f) <π+2mπ. (13) Equation (13) shows that the φ(f) will have a maximum of eight fringes (m = 8) and it varies from π to π or π to π as θ(f) varies from 0 to π. Therefore, it is obvious that to eliminate the ambiguity, φ(f) should have a unique value for each θ(f) for the entire frequency range. Hence, the φ(f) function should be linear as f increases (see the fourth panel in Fig. 3). The procedure to remove this ambiguity is proposed as follows: (1) displacing φ(f) for the high frequencies over ±2nπ, where n =1, 2,..., m; (2) selecting only the φ(f) function that crosses the origin, because it is known that for f =0,φ(f) =0. The procedure of removing phase ambiguity is as follows [6]. Consider a typical broadband pulse which was received by the three antenna sensors, as depicted in Fig. 4(a). The Fourier spectrum of the signal of each sensor is shown in Fig. 4(b). The φ(f) between antenna 2 and antenna 1, and also between antenna 2 and antenna 3 are shown in Fig. 5(a). The φ(f) is derived using Eq. (4). Antenna 2 is the reference. Here, only the frequencies higher than 25 MHz are analyzed due to the bandwidth of this system. From the Fig. 5(a), it can be seen clearly the fringes for the f below 100 MHz of each baseline. For the f above 100 MHz the φ(f) seems to scatters, this might be due the low Fourier spectrum in that frequency range (see Fig. 4(b)). Figure 5(b) shows the displacing process of φ(f) of the first baseline (antennas 2 and 1) over ±6π only with the in-

4 MARDIANA et al.: IMAGING LIGHTNING USING BROADBAND RADIO INTERFEROMETRY 1895 Fig. 4 (a) The extended scale of a typical broadband pulse, and (b) the associated Fourier spectrum. Fig. 6 (a) Phase differences after removing the fringes, and (b) the incident angles for both baselines. Fig. 7 Block diagram of the experimental system. Fig. 5 Removing phase ambiguity: (a) phase differences before the removing process, and (b) displacing the phase difference of antennas 2-1 up to n = 3. The arrow indicates the phase difference function that crosses the origin. crement of ±2π (or n = 3). The displacing of phase difference is continued until n = m = 8. Then, only the φ(f) function that crosses the origin is selected, as indicated by an arrow in the Fig. 5(b). The same procedure is also applied to the second baseline, and the results for the entire frequency range are shown in Fig. 6(a) for both baselines. From here, we can see that the φ(f) for f above 100 MHz still has a tendency to increase linearly as f increases, although the deviation is larger than the φ(f) for f below 100 MHz. To find θ(f), Eq. (6) is applied to those values of φ(f), and the results of each baseline can be seen in Fig. 6(b). Equation (7) is then applied to the θ(f) of all frequencies so the estimated incident angle θ est of both baselines can be determined. The location of this pulse in azimuth and elevation can be estimated using Eqs. (10) (12). 3. Experimental System Figure 7 shows the experimental broadband interferometry system. The system comprised three antennas which were separated horizontally with the baseline s length of 10 m and aligned at three apexes of a triangle (see Fig. 2). Each antenna is a flat circular capac-

5 1896 IEICE TRANS. ELECTRON., VOL.E84 C, NO.12 DECEMBER 2001 itive antenna with a diameter of 30 cm and acts as an isotropic receiver over the frequency bandwidth. This isotropic radiation pattern is needed since the system should enable to locate lightning radiation sources in all directions. The electric field derivative de/dt was measured by sensing the displacement current intercepted by the antennas. These three antennas were connected to a digitizer, after passing through analog band-pass filters, via 50 Ω coaxial cables with 30 m length each. The broadband de/dt pulses were digitized at a rate of 500 MS/s and 8 bit resolution by a 4-channel LeCroy 9374L. The digitizer memory was divided into an adjustable number of segments. One segment stored a broadband pulse for 1 µs (502 points), each time a trigger event occurred at Channel 2 of digitizer. One segment corresponds to one de/dt angular (azimuth and elevation) location. Consecutive trigger events were digitized sequentially with a 70 µs dead time between segments. The trigger time of each segment was recorded automatically by the digitizer with an accuracy of 2 ns. In the observations, a total of 800 segments was allocated per lightning flash within 1 second. Every channel was delayed to have a pretrigger of 50%, and Channel 2 was set as the reference (connected to antenna 2). Channel 4 was not used. The system was operated to record radiation sources from f = 25 MHz to f c = 250 MHz. Of 502 points sampled for one segment, only the center 256 points (N = 256) are analyzed due to the fast Fourier transform (FFT) restriction. This means T =0.5 µs and f = 2 MHz. A sensor to measure the electric field (E) was also equipped. This E measurement was used primarily to reveal the polarity of lightning discharges and the charge motions. The E was measured by integrating the displacement current through the sensing plate. The integrator circuit had a 10 s decay time constant. The E waveforms were then sampled by an analog-todigital converter (ADC) with a rate of 1 MS/s, 12 bit resolution and a pretrigger of 30%. This ADC was externally triggered by the digitizer. The synchronization between E and de/dt measurements was within 10 µs. A 1 pulse-per-second output of an external GPS receiver with an accuracy of 1 µs was also setup at the site. The GPS receiver sent a time tag each time a lightning flash occurred, and this time tag was set to be the same as the trigger time of the first segment. The digitizer, ADC card, and GPS receiver were connected to, respectively, IEEE port, bus slot and serial port of an NEC PC MHz pentium-based personal computer. A dead time of at most 10 s, depending on the number of occupied segments, was resulted when a lightning flash data were transferred from the digitizer, ADC, and GPS to the computer hard-disk. Later the data were archived on magneto-optical disks. The system described above was similar to that described in the previous literature [5] [7]. The C/C++ language and Matlab software package were used during post processing data and for displaying the results. The accuracy of this system has been tested on site during the 1997 observations by comparing the results with the narrowband system, which was already established, and with an all-sky video camera. The results showed that the average discrepancy for azimuth direction is less than 1 with standard deviation of 4.5 and for elevation direction is less than 2 with standard deviation of 5.0 [7]. 4. Observation Results The observations have been conducted in the Hokuriku District, Japan in the winter The observation site was about 1 km from the Mikuni Power Plant, and it was almost at zero meter above sea level of Japan Sea. Twenty-four good lightning data could be identified, and thirteen of them had branches. As a case study, a cloud-to-ground lightning flash will be analyzed. This flash occurred on December 15, 1999 at 06:34:38 local time as recorded by the GPS receiver. The flash propagated upward and a total of 790 broadband pulses were successfully recorded. Figure 8, on the right panels from top to bottom, shows a broadband de/dt signal of the first segment captured by antennas 1, 2, and 3, respectively. This de/dt waveform is typical of VHF broadband lightning signal. It is clearly seen the difference of arrival-time delay among the received signals. The left panels from top to bottom in Fig. 8 show, respectively, the electric field change, the azimuth, and the elevation of all de/dt radiation sources as a function of time. The locations of such de/dt sources define the spatial and temporal development of the upward lightning discharges. Figure 9, on the right panels, shows the de/dt signal for the segment number of 200. As mentioned earlier, one segment will correspond to one de/dt angular location. The left panels in Fig. 9 are the extended scales of those in Fig. 8 to show the angular location of this segment number and its associated E waveform. It can be seen on the left panels that the occurrence of each de/dt source coincides with that of small changes in the E waveform. This means that both de/dt and E measurements are synchronized. In the display monitor, all panels can be extended via zooming facility. Every angular location is numbered according to its segment number to simplify the analysis. Figure 10 shows the total activity of the flash in angular-time and in azimuth-elevation as well as in hemisphere projection formats. The flash began from the azimuth direction of about 280 and split into different directions at approximately a constant azimuth of 310 with the corresponding elevation of about 25,37, and 45 as we can see from azimuth-time and elevationtime panels. This event indicates that the lightning had branching channels during its progression. Part of the branching channels moved upward and the other chan-

6 MARDIANA et al.: IMAGING LIGHTNING USING BROADBAND RADIO INTERFEROMETRY 1897 Fig. 8 A broadband lightning de/dt signal captured by each antenna is shown on the right panels. Electric field change E and all angular locations of de/dt sources as a function of time are shown on the left panels. nels moved downward. The channels progressed downward, probably because the discharges moved horizontally away the observation site. The E was negativegoing, as seen at the top panel on the left, indicating that there was removal of negative charge from ground to neutralize positive pocket charge in the cloud. This also indicates that the channels had negative polarity. The lightning propagated from the azimuth direction of about 280 to 30 (see the hemisphere projection), corresponding to the movement from south to east. In this instance the lightning required about 140 ms to progress from ground to the cloud, and the activity culminated after that. The occurrence of the lightning flashes which are initiated near ground and propagate upward is very common on the Japan Sea Coast during the winter. 5. Conclusions The VHFbroadband radio interferometry technique for imaging lightning discharge progression has been introduced and demonstrated. The electromagneticwaves, which are emitted during the discharges, are remotely sensed by three antenna sensors to realize twodimensional images. Each of the two unknowns for a radiation source location (azimuth and elevation) was resolved by arranging the antenna sensors to form a perpendicular baseline interferometry. The polarity of lightning can be recognized by employing an electric field sensor. The system applied a high digitization rate to overcome the large lightning frequency bandwidth with a range from 25 MHz to 250 MHz, and a sequential triggering technique was used to overcome the limitation of digitizer memory. The memory of the digitizer was divided into segments. One segment stores broadband electromagnetic pulses for 1 µs, and it corresponds to an angular location of the radiation source. An upward cloud-to-ground lightning flash has been presented. The system recorded a total number of 790 radiation sources within 140 ms during the flash progression. The lightning channels had negative polarity. From the spatial and temporal development, the lightning began from ground level and created three branching channels during its progression. The occur-

7 1898 IEICE TRANS. ELECTRON., VOL.E84 C, NO.12 DECEMBER 2001 Fig. 9 Broadband de/dt signal for the segment number of 200 (right panels) and its angular location indicated by its segment number (inside the circles) along with its associated electric field change (left panels). rence of each radiation source coincides with that of small changes in the electric field waveform. The results confirm that this broadband radio interferometry is a useful tool to understand the electrical discharge processes in the atmosphere. Acknowledgments We are grateful to T. Ushio and Y. Ota for help in interferometry design and construction. The author RM would like to thank the Hitachi Scholarship Foundation for the generous supports of his stay and study in Japan. This work was supported by the grant of TRMM Second Research Announcement/NASDA. References [1] C.O. Hayenga and J.W. Warwick, Two-dimensional interferometricpositions of VHF lightning sources, J. Geophys. Res., vol.86, no.c8, pp , Aug [2] P. Richard, A. Delannoy, G. Labaune, and P. Laroche, Result of spatial and temporal characterization of the VHF-UHF radiation of lightning, J. Geophys. Res., vol.91, no.d1, pp , Jan [3] X.M. Shao, P.R. Krehbiel, R.J. Thomas, and W. Rison, Radio interferometricobservations of cloud-to-ground lightning phenomena in Florida, J. Geophys. Res., vol.100, no.d2, pp , Feb [4] X.M. Shao, D.N. Holden, and C.T. Rhodes, Broad band radio interferometry for lightning observations, Geophys. Res. Lett., vol.23, no.15, pp , July [5] T. Ushio, Z. Kawasaki, Y. Ohta, and K. Matsuura, Broadband interferometricmeasurement of rocket triggered lightning in Japan, Geophys. Res. Lett, vol.24, no.22, pp , Nov [6] R. Mardiana and Z. Kawasaki, Broad band radio interferometer utilizing a sequential triggering technique for locating fast-moving electromagnetic sources emitted from lightning, IEEE Trans. Instrum. & Meas., vol.49, no.2, pp , April [7] Z. Kawasaki, R. Mardiana, and T. Ushio, Broadband and narrowband RF interferometers for lightning observations, Geophys. Res. Lett., vol.27, no.19, pp , Oct [8] R. Mardiana and Z. Kawasaki, Dependency of VHF broad band lightning source mapping on Fourier spectra, Geophys. Res. Lett., vol.27, no.18, pp , Sept

8 MARDIANA et al.: IMAGING LIGHTNING USING BROADBAND RADIO INTERFEROMETRY 1899 Fig. 10 Spatial and temporal development of an upward lightning discharge. Azimuth vs. elevation and hemisphere projection are on the right panels. The electric field change, azimuth, and elevation as a function of time are shown on the left panels. Redy Mardiana received the bachelor and master degrees in electrical engineering from Institut Teknologi Bandung, Indonesia in 1992 and 1997, respectively. From 1993 to 1994, he was with the Forschungzentrum Jülich, Germany. Currently, he is finishing his Ph.D. thesis in the Department of Electrical Engineering, Osaka University. His interest is in lightning phenomena and its detection. He is a member of IEEE I&M and SAEJ. Takeshi Morimoto received the bachelor degree in electrical engineering from Osaka University in Now, he is pursuing his master degree in electrical engineering at the same university. His main interest is in lightning detection. He is a member of IEE of Japan and The Society of Atmospheric Electricity of Japan (SAEJ). Zen-Ichiro Kawasaki received the B.S., M.S. and Dr.Eng. degrees in communication engineering from Osaka University, Japan in 1973, 1975 and 1978, respectively. From 1979 to 1988, he was with the Research Institute of Atmospherics, Nagoya University. From 1989 to 2000 he joined the Department of Electrical Engineering, Osaka University. Currently he is a full professor in the Department of Communication Engineering at the same university. He is chairman of the International Union of Radio Science (URSI) working group of Terrestrial and Planetary Lightning Generation of Electromagnetic Noise. His research interests are in EMC and the electromagnetic of lightning. He is a member of IEEE, IEE of Japan, American Geophysical Union, and SAEJ.