Investigations on the performance of lidar measurements with different pulse shapes using a multi-channel Doppler lidar system

Transcription

1 Th12 Albert Töws Investigations on the performance of lidar measurements with different pulse shapes using a multi-channel Doppler lidar system Albert Töws and Alfred Kurtz Cologne University of Applied Sciences Steinmüllerallee 1, Gummersbach, Germany albert.toews@gmail.com Abstract: Pulse distortion during pulse amplification and the consequential influence on measurement results is a problem for low and high power pulsed lidar systems. We present a multi-channel coherent Doppler lidar system with automatic feedback control of pulse shapes and durations on each wavelength-channel allowing this system to send laser pulses simultaneously with individual pulse parameters into the atmosphere. The influence of different pulse shapes and durations on the lidar performance in terms of the resulting signal-to-noise ratio, Doppler spectrum width, and velocity precision in atmospheric measurements is investigated. A comparison of two main transmitted pulse shapes will be discussed: a Gauß- and square-shaped pulse. Keywords: Coherent Doppler lidar, multi-channel, pulse-shape control, Gauß- and square-shaped pulse 1. Introduction Fiber based Doppler lidar systems are usually designed in master oscillator power amplifier (MOPA) architecture with erbium doped fibers (EDFA) as power amplifiers [i.a. 1, 2]. These fibers are pumped with laser diodes to achieve the population inversion which is needed for stimulated amplification. Usually, pulse lengths of 100 ns to 800 ns are applied in Doppler lidar systems as a compromise between range resolution and velocity precision. Injecting such long seed pulses into fiber amplifiers leads to decreasing population inversion during the pulse duration. Therefore, the gain decreases while the seed pulse passes the fiber amplifier. Thus, fiber amplifiers generate significant pulse distortions at typical pulse lengths and repetition frequencies in lidar systems. In our lidar system, pulses on multiple different wavelength channels are simultaneously generated and one monostatic telescope for all channels is used. Thus, the same line-of-sight can be measured at the same time applying different pulse shapes and durations. 2. Methodology The multi-wavelength lidar system [3] shown in Figure 1 consists of five overall units. The master oscillator unit contains multiple external cavity diode lasers. The wavelength channels are chosen from the ITU-grid near 1.55 µm. One laser light output of the master oscillator is directed to the pulse shaping unit, where the laser light is demultiplexed, so it is possible to shape the pulses for each channel with electro-optical modulators (EOM). Then the multiplexed pulses are frequency shifted by an acoustooptic modulator (AOM). The shifted pulses are amplified with a multi-stage EDFA. The pulse-shape control circuit uses a part of the output power to measure the power and the shape of the output pulses. In order to achieve this, the wavelength-channels are demultiplexed and illuminate separate photodiodes. It is essential to pre-shape the pulses with the AOM for gain saturation compensation. The feedback control unit iteratively calculates a corrected seed pulse shape for each channel generated by the EOMs. Via circulator the amplified and shaped pulses are directed to the collimator, where the light is decoupled. The polarization sensitive transceiver is realized by a quarter-wave plate and the circulator. In the detection unit the wavelength-channels of the local oscillator and the backscattered light are demultiplexed, thus, every channel can be coupled separately to its balanced detector. The signal processing unit extracts the information on the wind velocity and the backscattered power. CLRC 2018, June

2 Figure 1. Overview of the multi-channel coherent Doppler lidar system. 3. Atmospheric observations with multiple pulse durations To compare different pulse durations under atmospheric conditions, two channels of the multiwavelength system can be utilized. Sending identically shaped pulses on two wavelength-channels, the evaluated return-signals show minor differences concerning estimated velocity and width of the Doppler-spectra. In contrast, by applying different pulse-shapes there are significant changes in the evaluated return-signals. In the first step we investigate the influence of different pulse durations of square-shaped pulses on the atmospheric return signal. The chosen FWHM durations are 300 ns and 600 ns. The measured pulse energies at SBS threshold are 2.2 µj and 4.3 µj, respectively. Figure 2c shows the two regulated square-shaped pulses with different pulse durations. The resulting range gate weighting function (RWF), Figure 2b, is the convolution of the probing pulse and the FFT-window, and the effective longitudinal size of the sensing volume is defined at z = RWF(0) 1 [4]. The calculated RWFs for both measured probing pulses are z = 46 m for the shorter pulse and z = 91 m for the longer pulse. The line-of-sight velocity, Figure 2a, is estimated by applying the FFT with zero-padding and a peak finding algorithm. 30,000 spectra (3 s) were averaged for this measurement. Because of the longer sensing volume, the longer probing pulse channel (red line) shows the wind structure more smoothed along the line-of-sight than the shorter pulse. The channel with the shorter pulse determines the wind structure in more detail. This can especially be seen around 250 m and 450 m, where the blue channel estimates the velocity up to 0.4 m/s lower than the red channel due to wind gradients. Figure 2. Atmospheric comparison measurement of line-of-sight velocity for different pulse durations. The Doppler spectrum width over the range for both pulse durations is shown in Figure 3a. This spectral width contains information on the instrumental broadening and the broadening caused by changes of the wind velocity within the sensing volume. This might be wind shear or turbulent fluctuations. By comparing Figure 2a to Figure 3a it can be obtained that the Doppler spectrum width increases in the CLRC 2018, June

3 vicinity of 250 m and 480 m and the range dependent wind velocity shows a significant decrease of velocity. The change in bandwidth is more pronounce for longer pulses. Therefore, the increased Doppler width in Figure 3a is an indication for a stronger wind gradient. The Doppler spectrum of longer pulses experiences stronger broadening due to the larger contribution of different velocities. Therefore, the difference between those spectral widths is not constant over range and depends on the level of wind gradient. Figure 3. Measured a) FWHM bandwidth and b) narrowband SNR for different probing pulse lengths. The measured SNR is proportional to the pulse energy of the probing pulse. Compared to the 300 ns pulse the pulse energy of a 600 ns pulse is higher by a factor of two due to SBS limitation. This results at least in a 3 db higher SNR for the longer pulse. Additionally, the spectral width of the longer pulse is smaller by a factor of two. According to the lidar equation [5], the higher pulse energy and the lower spectral width should result in a 6 db higher SNR. Figure 3b shows that the SNR of the longer pulse is 4-5 db higher. In regions where the Doppler spectrum experiences stronger broadening due to wind gradients within the sensing volume the difference in the SNR is reduced. According to the variance summation model [6], both, the reduced spectral width and the higher SNR, lead to a higher velocity precision for longer pulses. To measure the velocity precision, the difference between two successive registered peak-frequency values is evaluated to prevent long-term influences of wind velocity changes. The time between those measurements was around 0.05 s to ensure that the change in velocity of the particles is negligible in such a short time window. The difference of measured peak-frequencies of successive registered spectra is displayed in Figure 4. The distributions of each channel are fitted with the Gaussian function. For comparison the range gate located at a range around 350 m at high SNR-level was used. The standard deviation of the Doppler peak-frequency of the longer pulse is 0.14 MHz and 0.5 MHz for the shorter pulse for that range gate. Figure 4. Comparison of velocity precision of two different square-shaped pulse durations. The calculated velocity precisions are also listed in Figure 4. The results using the variance model are well comparable to the measured velocity precision and show that the velocity precision of the longer pulse is better by more than a factor of two. CLRC 2018, June

4 According to the velocity precision, the statistical error of the determined velocity in Figure 2a is in the order of the line thickness in the case of applied 600 ns pulses and in the order of some minor oscillations for the 300 ns curve. There are significant differences of the measured wind velocity in case of wind gradients in the region around 250 m and 480 m, which is also confirmed by the increasing linewidth of the 600 ns curve in Figure 3a. 4. Atmospheric observations with different pulse shapes This multi-wavelength system can also regulate different pulse shapes on every channel. The chosen pulses are a Gauß- and square-shaped pulse with a FWHM duration of 300 ns for both pulse shapes. Both regulated output pulse shapes are shown in Figure 5c. Each pulse energy is limited to a pulse shape dependent stimulated Brillouin scattering (SBS) threshold. The pulse energy reached with a standard EDFA for the Gauß-shaped pulse is 3.9 μj, for the square-shaped pulse it is 3.4 μj. The calculated RWFs for both measured probing pulses are depicted in 5b, which result in z = 46 m for the square-shaped pulse and z = 58 m for the Gauß-shaped pulse. In Figure 5a, the measured velocity over range of both channels is plotted. Both estimated wind velocities are consistent. The sensing volume of the Gauß-shaped pulse is about 25 % wider. The Gaußchannel shows a stronger averaging effect over the spatial velocity development than the square-shaped pulse, especially seen in the range near 200 m and 450 m. Figure 5. Atmospheric comparison measurement of line-of-sight velocity for different pulse durations. The Doppler spectrum width (FWHM) of both pulse shapes is shown in Figure 6a. At 500 m, the spectral width of the Gauß-channel is enhanced due to stronger wind gradient. Also around 260 m, a weaker gradient merely increases the spectral width of the Gauß-channel. The squared-pulse channel was not affected by the smaller fluctuations in radial wind velocity. Consequently, the Gauß-shaped pulses are more effected by wind changes due to the wider sensing volume. Figure 6. Measured a) FWHM bandwidth and b) narrowband SNR for different probing pulse shapes. CLRC 2018, June

5 The Gauß-shaped pulse has a higher pulse energy and a lower spectral width. Simulations show that the spectral width at FWHM of the square-shaped pulse is about 1.3 times broader than the spectral width of the Gauß-shaped pulse. With the higher pulse energy of the Gauß-channel, this should result in a 1.8 db higher SNR for the Gauß-channel. However, the measurement results show just a slightly higher SNR, Figure 6b. Around 500 m, the SNR of the Gauß-shaped pulse equals the SNR of the square-shaped pulse due to the spectrum broadening caused by the stronger velocity gradients. The distribution of the line-of-sight velocity and the Gaussian function fit for both pulse shapes are shown in Figure 7. For comparison the range gate located at a range around 150 m at high SNR-level was used. The standard deviation of the Gauß-shaped pulse is 0.19 MHz and 0.31 MHz for the squareshaped pulse. The figure also shows the measured and calculated parameters for the comparison of two different pulse shapes. As expected, the velocity precision of the Gauß-shaped pulse is higher. However, measurements in the region of 500 m show a worse velocity precision of the Gauß-channel as the Doppler spectrum is strongly broadened due to wind gradients. Figure 7. Comparison of velocity precision between Gauß- and square-shaped pulse. 5. Conclusion Different pulse shapes and durations have an impact on pulse energy, Doppler spectrum width, SNR, and velocity precision. The Gauß-shaped pulse has a slightly higher pulse energy and a better velocity precision. Due to the wider sensing volume, the Gauß-shaped pulse is more sensitive to wind gradients. The sensing volume of the square-shaped pulse is smaller. Hence, the wind gradients do not affect the performance of the square-shaped pulse as strong as they affect the Gauß-shaped pulse. In this paper we described and discussed representative atmospheric measurement examples which show the usefulness measuring with multiple pulse shapes simultaneously. 6. References [1] S. Kameyama, T. Ando, K. Asaka, Y. Hirano, and S. Wadaka, Compact all-fiber pulsed coherent Doppler lidar system for wind sensing, Appl. Opt. 46, (2007). [2] N. S. Prasad, R. Sibell, S. Vetorino, R. Higgins, and A. Tracy. An all-fiber, modular, compact wind lidar for wind sensing and wake vortex applications, in Laser Radar Technology and Applications XX; and Atmospheric Propagation XII, (Society of Photo-Optical Instrumentation Engineers, Baltimore, 2015), 94650C. [3] A. Töws and A. Kurtz, A multi-wavelength LIDAR system based on an erbium-doped fiber MOPAsystem, in Lidar Technologies, Techniques, and Measurements for Atmospheric Remote Sensing X, (Society of Photo-Optical Instrumentation Engineers, Amsterdam, 2014), 92460T. [4] V. Banakh and I. Smalikho, Coherent Doppler wind lidars in a turbulent atmosphere (Artech House, 2013). [5] S. Kameyama, T. Ando, K. Asaka, and Y. Hirano, Semianalytic pulsed coherent laser radar equation for coaxial and apertured systems using nearest Gaussian approximation, Applied Optics 49, (2010). [6] S. W. Henderson, P. Gatt, D. Rees, and R. M. Huffaker, Laser Remote Sensing, (Taylor & Francis Group, 2005), Chap. Wind Lidar CLRC 2018, June