Open Access
Add to BibSonomyAbstract
We implement a BOTDR sensor that combines the complementary coding with the fast Fourier transform (FFT) technique for high-performance distributed sensing. The employment of the complementary coding provides an enhanced signal-to-noise ratio of the sensing system, which leads to high accuracy measurement. Meanwhile, FFT technique in BOTDR is combined to reduce the measurement time sharply compared to the classical frequency sweeping technique. In addition, a pre-depletion two-wavelength probe pulse is proposed to suppress the distortion of the coding probe pulse induced by EDFA. Experiments are carried out beyond 10 km single-mode fiber, and the results show the capabilities of the proposed scheme to achieve 2 m spatial resolution with 0.37 MHz frequency uncertainty which corresponds to ∼0.37 °C temperature resolution or ∼7.4 με strain resolution. The measurement time can be more than tens of times faster than traditional frequency sweeping method in theory.
© 2017 Optical Society of America
1. Introduction
Distributed Brillouin fiber sensors attract many research interests because of the capability of truly distributed sensing of temperature/strain of fiber [1–7]. Brillouin optical time-domain reflectometer (BOTDR), which is one kind of the distributed Brillouin fiber sensors, has good flexibility by accessing only one end of fiber to perform the measurement. Therefore it has good potential in the monitoring of large-scale structures, and a few field applications have been successfully employed [8–11]. However, there is a trade-off between the measurement accuracy, the spatial resolution and the measurement speed for BOTDR. Because the spontaneous Brillouin scattering signal in BOTDR is very weak, one has to improve the signal-to-noise ratio (SNR) to increase the measurement accuracy. The cumulative average method can improve the SNR of the system, but it takes longer measurement time. Increasing the pulse width can also improve the SNR of the system. However it deteriorates the spatial resolution.
In recent years coding method has been proved to be an effective technique to improve the SNR of BOTDR. By using an appropriate coded pulse as probe pulse, one cannot only increase the energy of the incident signal but also maintain the spatial resolution, which is a very practical way to improve the performance [12–15]. One of the popular coding techniques is the Golay complementary sequence. It has good autocorrelation, and has shown good performance in BOTDR system for the improvement of SNR [14].
Meanwhile, in order to improve the measurement speed of BOTDR, FFT technique is proposed to obtain the Brillouin gain spectrum (BGS) [16–19]. Compared to the frequency sweeping method, it can obtain the BGS at one time without sweeping the frequency components of the BGS, which will improve the measurement speed greatly. In [16], it takes only 1 s to measure the temperature by using the FFT technique with 1024 data averages along a 1.5 km fiber.
In this paper, we implement a BOTDR by combining the complementary pulse coding method and the FFT technique simultaneously to circumvent the trade-off between the measurement accuracy, the spatial resolution and the measurement speed of BOTDR. The method of combining the FFT technique and the complementary coding is analyzed. The decoding process for the combination BOTDR scheme is explained. In addition, a method to suppress the distortion of the coded probe pulse (CPP) induced by an erbium-doped optical amplifier (EDFA) is proposed by using a pre-depletion two-wavelength probe pulse for the first time. We experimentally demonstrate that the capabilities of the proposed scheme to achieve high spatial resolution and high measurement accuracy over 10km-long sensing length. Experiment results show that this scheme can achieve ∼0.37 MHz frequency uncertainty with 2 m spatial resolution at the end of 10 km fiber. Based on the calculation ability of field-programmable gate array (FPGA) technique, the measurement time is estimated to be less than 6 s for 10 km sensing range and 20000 measurement points.
2. Fundamentals of the proposed BOTDR scheme
In general, the obtained Brillouin spectrum needs to cover hundreds of MHz for Brillouin based fiber sensor to measure the strain/temperature distribution along a fiber. Thus one has to sweep the BGS step by step to obtain the whole spectrum, which will consume much time. In BOTDR, since the Brillouin scattering signal is detected with heterodyne detection method, it is possible to obtain the Brillouin spectrum directly with FFT technique instead of the frequency sweeping method. The FFT technique utilizes a wide-bandwidth photodetector and a high speed A/D to acquire the Brillouin signal in a single measurement. Then a section of successive signal is extracted out by a window with time length T and is transformed into the frequency domain with the FFT arithmetic, so the BGS for the corresponding fiber section is obtained. A longer time length T of the window can improve the frequency resolution of the BGS, whereas it deteriorates the spatial resolution [17, 19]. Then the window slides with a step of and the FFT is performed continuously for the extracted signals at different time. And finally the whole BGS along the fiber can be obtained, as depicted in Fig. 1. The spatial interval between two adjacent BGSs equals to , where n is the effective refractive index of fiber and c is the velocity of light in vacuum. The spatial resolution is mainly determined by the pulse width and the time length of the FFT window [17]. The spatial interval should be shorter or at most equal to the spatial resolution, otherwise parts of information along fiber will be lost.
The complementary coding would remarkably enhance the SNR of the measurement, since the coded pulses can significantly improve the total energy of the probe pulse than a single-pulse. The Golay complementary sequence has characteristic of pseudorandom and autocorrelation with low side lobes. In addition, it is easy to be generated and replicated. Generally, a group of Golay complementary sequence consists of two bipolar coding sequences. However, since only unipolar optical pulse can be transferred in fiber, bipolar complementary sequence cannot be used directly. Therefore, we need to transform each bipolar coding sequence into a pair of unipolar coding sequences. So two pairs of unipolar codes A1(t), A2(t), B1(t), and B2(t) are generated as follows,
where codes A(t) and B(t) are the original bipolar Golay complementary pair.In the decoding process for the signals obtained with Golay complementary sequence, it requires the impulse response of the fiber to the probe pulse to keep constant. However, since the heterodyne signal is derived directly with a high speed A/D in the FFT based BOTDR, the phase of the signal is determined by the phase difference between the Brillouin scattering signal and the reference lightwave. The finite coherent length of the laser source and the influences of environment to the fiber make it impossible for the fiber to keep a constant phase response to each probe sequence. So it will lead to failure when decoding the time-domain signals directly. Instead of decoding the time-domain signals, the frequency spectra along the fiber are firstly obtained by using FFT. Then the decoding procedure is carried out to the power distribution curve of each of the frequency components, as depicted in Fig. 2. Because the power distribution curve is robust to the phase noise, the decoding procedure can be completed successfully. And finally the impulse response of the Brillouin frequency spectra along the fiber is obtained. The method of decoding is depicted as follow:
where SA1(f,l), SA2(f,l), SB1(f,l) and SB1(f,l) represent the power distribution curves of the frequency component f obtained with coding pulses A1(t), A2(t), B1(t), and B2(t) respectively. The SNR improvement compared to a conventional BOTDR sensor with the same measurement times is , where N denotes the length of the codewords [12].Before inputted into the sensing fiber, the CPP needs to be amplified by an EDFA to achieve high SNR for the signal. However, the power of a long CPP is not flat because of the transient effect of EDFA. Therefore, the decoded BGS distribution may be distorted, because the latter part of the CPP has smaller impulse response than that of the front one. A simulation for a 64-bit Golay return-to-zero (RZ) CPP (pulse width is 20 ns, and the duty cycle is 10%) is shown in Fig. 3. The simulation is conducted with the software of Optisystem 7.0. The EDF Dynamic model in the software is used to simulate the amplification of the coded optical pulse. This model can simulate dynamic effects presented by EDFA by calculating the powers and population densities as a function of the time variation at each point of the EDF. From Fig. 3, it can be seen that the CPP after amplified by the EDFA has significant decay shape.
In order to avoid the distortion of CPP induced by EDFA, some works generate the CPP after the EDFA [20]. However, this method decreases the peak power of the CPP because the pulse modulator has a limitation of the maximum input power and has a non-negligible insertion loss, resulting in the SNR of the signal decreases consequently. Some works decrease the duty cycle to decrease the transient effect [21, 22], but the distortion of the coding pulse after the EDFA cannot be eliminated thoroughly and the measurement time will be increased because the whole coding pulse lasts for longer time. Some other works use a weighted code sequence to decode the BGS, which leads to complex decoding process [23].
The distortion of the long CPP is mainly because of that the front part and the latter part of the CPP encounter different upper-level population cumulated within different periods in EDFA. In coding BOTDR, The accumulation time of the upper-level population for the front pulse may be thousand times longer than that for the latter part. Thus, we propose using a preceding long pulse, which is called pre-depletion pulse, to deplete the cumulated population within long time period to make the whole CPP experience the same amplification. The scheme of the proposed pulse is shown in Fig. 4(a). In order to avoid the influence of the Brillouin scattering generated by the pre-depletion pulse, a pre-depletion two-wavelength probe pulse is proposed. The wavelength of the CPP (1550 nm) is different with the pre-depletion pulse (1553 nm). Thus, the Brillouin scattering generated by the pre-depletion pulse can be easily filtered out by the heterodyne detection scheme. So the pre-depletion pulse and the CPP can exist in fiber simultaneously, which means the measurement time can be the same as that without the pre-depletion pulse. Figure 4(b) shows the simulation result of the pre-depletion two-wavelength probe pulse amplified by an EDFA. The parameters of the CPP are the same as that in Fig. 3. By comparing Fig. 3(b) and 4(b), it can be seen clearly that the distortion of the CPP after the EDFA is eliminated thoroughly.
Fig. 4 (a) Simulation results of the pre-depletion two-wavelength probe pulse (a) before amplified by EDFA, and (b) after amplified by EDFA.
In order to get a flat amplified CPP, the width of the pre-depletion pulse, the peak power of the pre-depletion pulse, and the time interval between the pre-depletion pulse and the CPP should be adjusted for different EDFA. We make a group of simulations to find out the relationships between the three parameters of the pre-depletion CPP. In the simulation, the period of the CPP is 100 μs which corresponds to a maximum sensing length of 10 km. The combinations of the parameters which can obtain the flat CPP are shown in Table 1.
Table 1. Parameter Combinations of the Pre-depletion Pulse Obtaining the Flat CPP
From Table 1, we can see that in order to keep a flat CPP, the peak power and the width of the pre-depletion pulse change in the opposite direction; whereas each of them changes in the same direction as the time interval. Increasing the peak power or increasing the width can deplete more upper-level population, so they will decrease the amplification of the front part of the CPP; whereas increasing the time interval can cumulate more upper-level population to improve the amplification of the front part of the CPP. Clearly, once the CPP reaches the flat state with the help of pre-depletion pulse, each codeword of the CPP is amplified by the upper-level population cumulated within only the off time of the codeword. So the method is suitable for the CPP with any length.
3. Experiment and discussion
The experiment setup of the BOTDR system using the FFT technique and complementary coding is shown in Fig. 5(a). The CW light from a distributed feedback (DFB) laser diode, operating at 1550 nm with ∼11 dBm output power, is split into two branches through a 90/10 optical fiber coupler. The linewidth of the laser diode is about 10 kHz, which is much narrower than that of the Brillouin gain linewidth (typically 30 MHz). In the upper branch, a high extinction-ratio (>40 dB) EOIM2 driven by an AWG is used to generate complementary CPP. The CW light from another DFB laser diode, operating at 1553 nm, is modulated by EOIM1 to generate the pre-depletion pulse. The two arms are then combined together through a 50/50 optical fiber coupler to generate the final pre-depletion two-wavelength probe pulse. A PS is used to eliminate the polarization-induced noise. The peak power of the CPP is amplified to about 20 dBm by EDFA1 before being launched into the sensing fiber.
Fig. 5 (a) Experiment setup of the BOTDR system using FFT and complementary coding. PC: polarization controller; EOIM: electro-optic intensity modulator; AWG: arbitrary waveform generator; PS: polarization scrambler; EDFA: erbium-doped fiber amplifier; CIR: circulator. (b) The layout of the sensing fiber consists of two different SMF’s.
In the lower branch, the EOIM3 is driven by a microwave generator to generate the frequency-shifted reference lightwave. The frequency of the microwave is 10.4 GHz. The reference lightwave is amplified to about 0 dBm by EDFA2. Then it is combined with spontaneous Brillouin scattering lightwave through a 50/50 optical fiber coupler. Finally, the light is directed to a 1 GHz balanced photodetector, which is connected to a high-speed digital oscilloscope with sampling rate of 2 GSa/s. The beat frequency between the valid Brillouin signal generated by the CPP and the reference lightwave is about 440 MHz, and the beat frequency between the Brillouin scattering generated by the pre-depletion pulse and the reference lightwave is more than 370 GHz. Thus only the valid Brillouin signal can be finally obtained. The sensing fiber is composed of two SMF’s with slightly different BFS: fiber-II with 1.25 km length is spliced at the end of fiber-I with 8.64 km length, as shown in Fig. 5(b). The section from 1055 m to 1145 m of fiber-II is heated to 35 °C in an oven. Fiber-I and the rest part of fiber-II are at room temperature (~12 °C).
We use the setup depicted in Fig. 5(a) to verify the feasibility of combination of FFT technique and complementary coding. Meanwhile, the effect of the proposed pre-depletion two-wavelength probe pulse is also verified. The 64-bit Golay RZ CPP after the EDFA with and without pre-depletion pulse are measured and shown in Figs. 6 and 7. The pulse width is 20 ns, and the duty cycle is 10%. The width of the pre-depletion pulse is 3.2 μs, and the time interval between the pre-depletion pulse and the CPP is 30 μs. Clearly, the CPP with pre-depletion pulse has a much flatter profile. The maximum power difference for the CPP without the pre-depletion pulse is 1.59 dB, whereas the one with the pre-depletion pulse is only 0.16 dB. So by using the pre-depletion two- wavelength probe pulse, the distortion of the CPP after EDFA is improved substantially.
Fig. 6 Measured results of Golay sequence (a) before amplified by EDFA, and (b) after amplified by EDFA.
Fig. 7 Measured results of the pre-depletion two-wavelength probe pulse (a) before amplified by EDFA, and (b) after amplified by EDFA.
For each 64-bit Golay CPP, the corresponding BGS distribution is obtained through FFT to the time domain signal. The time length of the sliding window is 16 ns which will extract 32 points (zero padded to 128 points) for each FFT operation and the slide step is 0.5 m. Each BGS distribution is averaged 1024 times under the same coding sequence to improve the SNR. Then the four BGS distribution curves for a set of Golay complementary sequences are decoded according to Eq. (2) and the actual BGS distribution is obtained. Figure 8 shows the obtained 3D view of the BGS distribution curve as a function of distance and frequency shift. Then the BFS of fiber is obtained by fitting the BGS with Lorentzian function for each position of fiber and finally the BFS distribution along the fiber is obtained.
In order to evaluate the effects of the proposed BOTDR scheme, the BFS distribution of the fiber is also measured with a single pulse of 20 ns pulse width. And the number of averages is 4096 to make it has the same total measurement times as the proposed BOTDR scheme. Both the results obtained by the proposed scheme and the single pulse are shown in Fig. 9(a). It can be seen that both results can reflect the temperature difference between the room temperature and the oven. However, the frequency measurement uncertainty is only 0.37 MHz for the 64-bit Golay CPP, which corresponds to a temperature and strain resolution of 0.37 °C and 7.4 με respectively, whereas the uncertainty for the single pulse is 1.45 MHz. Apparently, the proposed scheme improves the SNR of BOTDR significantly. Figure 9(b) shows the BFS distribution of the fiber section just entering the oven which is obtained with the proposed scheme. The time length of the sliding window is 16 ns and the pulse width of a single codeword is 20 ns. If the spatial resolution of the coded probe pulse is determined by a single codeword, then according to [17] the spatial resolution is about 2.7m. From Fig. 9(b), it can be seen that the spatial length of the rise slope from 10% to 90% is about 2 m, which demonstrates that the proposed scheme also maintains the spatial resolution of single codeword. The difference of the spatial resolution between this paper and that in [17] is mainly because of the different definition of the length of the transition section.
Fig. 9 (a) The BFS distribution curves obtained by the CPP and single pulse. (b) BFS near the beginning of the heated fiber sections obtained with the proposed scheme.
In our experiment, the BGS distribution of fiber is obtained with software, which takes 10 minutes if doesn’t count the time consumed by the data transmission from the oscilloscope to the computer. However, as given in [24], when FPGA is used the time cost for FFT to 128 points is about 71 ns. So for our experiment condition - 128 FFT points, 0.5 m slide step, 10 km sensing length, 2 GSa/s sampling rate, and 4096 averages - the time cost can be estimated to be 6 s in theory, whereas the frequency sweeping method would take tens of times than that. If the slide step is increased, the time cost can be reduced further. However, the slide step cannot be larger than the spatial resolution as said in section 2.
So the experimental results indicate that the proposed BOTDR scheme can be successfully applied to distributed strain and temperature measurement and the trade-off between the measurement accuracy, the spatial resolution and the measurement speed is circumvented.
Finally, the effect of the pre-depletion pulse in the coding method is examined with experiment. The experiment setup is the same as that depicted in Fig. 5(a). Only Fiber-II is used as the FUT. Two measurement results by using 64-bit Golay CPP with and without pre-depletion pulse are shown in Fig. 10. It can be seen that the obtained BFS distribution curve without the pre-depletion pulse has larger fluctuation, and the measured frequency difference between the room temperature and the oven is about 2 MHz smaller than the actual value of 23 MHz. On the other hand, the result obtained with the pre-depletion pulse can obtain the heated temperature accurately and the measurement frequency uncertainty is about 0.2 MHz. So the experimental results indicate that the proposed pre-depletion two-wavelength probe pulse is beneficial to the improvement of the frequency measurement accuracy.
Fig. 10 The BFS distribution curve of fiber-II obtained by using 64-bit Golay CPP with and without the pre-depletion pulse
4. Conclusions
In summary, we have demonstrated a BOTDR scheme of combination of FFT technique and complementary coding which can achieve high measurement accuracy, high spatial resolution and fast measurement speed simultaneously. A proof-of-concept experiment has been implemented in a ∼10 km long SMF, and we finally obtain the temperature measurement result with ∼0.37 °C measurement uncertainty and 2 m spatial resolution. The measurement time is estimated to be 6 s in theory which is tens of times faster than frequency sweeping method. In addition, we have demonstrated the feasibility of the proposed pre-depletion two-wavelength probe pulse to significantly suppress the distortion of the coding probe pulse induced by the EDFA. Experiment result indicates the proposed scheme of the pre-depletion two-wavelength probe pulse is beneficial to the improvement of the frequency measurement accuracy.
Funding
National Natural Science Foundation of China (NSFC) (61627816, 61540017, 61405090, 61107074); The Key Lab of Advanced Transducers and Intelligent Control System, Ministry of Education and Shanxi Province, Taiyuan University of Technology, Taiyuan, China (201404).
References and links
1. T. Horiguchi, K. Shimizu, T. Kurashima, M. Tateda, and Y. Koyamada, “Development of a distributed sensing technique using Brillouin scattering,” J. Lightwave Technol. 13(7), 1296–1302 (1995). [CrossRef]
2. M. A. Soto, J. A. Ramírez, and L. Thévenaz, “Intensifying the response of distributed optical fibre sensors using 2D and 3D image restoration,” Nat. Commun. 7, 10870 (2016). [CrossRef] [PubMed]
3. D. Ba, B. Wang, D. Zhou, M. Yin, Y. Dong, H. Li, Z. Lu, and Z. Fan, “Distributed measurement of dynamic strain based on multi-slope assisted fast BOTDA,” Opt. Express 24(9), 9781–9793 (2016). [CrossRef] [PubMed]
4. F. Wang, W. Zhan, Y. Lu, Z. Yan, and X. Zhang, “Determining the change of Brillouin frequency shift by using similarity matching method,” J. Lightwave Technol. 33(19), 4101–4108 (2015). [CrossRef]
5. Z. Li, L. Yan, L. Shao, W. Pan, B. Luo, J. Liang, H. He, and Y. Zhang, “Precise Brillouin gain and phase spectra measurements in coherent BOTDA sensor with phase fluctuation cancellation,” Opt. Express 24(5), 4824–4833 (2016). [CrossRef]
6. M. Li, W. Jiao, X. Liuwu, Y. Song, and L. Chang, “A Method for Peak Seeking of BOTDR Based on the Incomplete Brillouin Spectrum,” IEEE Photonics J. 7(5), 7102110 (2015). [CrossRef]
7. D. Yongkang, B. Dexin, J. Taofei, Z. Dengwang, Z. Hongying, Z. Chengyu, L. Zhiwei, L. Hui, C. Liang, and B. Xiaoyi, “High-spatial-resolution fast BOTDA for dynamic strain measurement based on differential double-pulse and second-order sideband of modulation,” IEEE Photonics J. 5(3), 2600407 (2013). [CrossRef]
8. H. Ohno, H. Naruse, M. Kihara, and A. Shimada, “Industrial applications of the BOTDR optical fiber strain sensor,” Opt. Fiber Technol. 7(1), 45–64 (2001). [CrossRef]
9. H. Ohno, H. Naruse, T. Kurashima, A. Nobiki, Y. Uchiyama, and Y. Kusakabe, “Application of Brillouin scattering-based distributed optical fiber strain sensor to actual concrete piles,” IEICE Trans. Electron. 85(4), 945–951 (2002).
10. K. Komatsu, K. Fujihashi, and M. Okutsu, “Application of optical sensing technology to the civil engineering field with optical fiber strain measurement device (BOTDR),” in International Society for Optics and Photonics (Photonics Asia, 2002), pp. 352–361.
11. F. Wang, Z. Sun, F. Zhu, C. Zhu, Y. Pan, J. Dong, X. Zhang, and L. Gao, “Research on the leakage monitoring of oil pipeline using BOTDR,” in Proceedings of the 2016 Progress in Electromagnetic Research Symposium (PIERS, 2016), pp. 4907–4910. [CrossRef]
12. M. Nazarathy, S. A. Newton, R. P. Giffard, D. S. Moberly, F. Sischka, W. R. Trutna, and S. Foster, “Real-time long range complementary correlation optical time domain reflectometer,” J. Lightwave Technol. 7(1), 24–38 (1989). [CrossRef]
13. J. Park, G. Bolognini, D. Lee, P. Kim, P. Cho, F. D. Pasquale, and N. Park, “Raman-based distributed temperature sensor with simplex coding and link optimization,” IEEE Photonics Technol. Lett. 18(17), 1879–1881 (2006). [CrossRef]
14. Y. Li, J. Wang, and Z. Yang, “A method for improving BOTDR system performance,” in Symposium on Photonics and Optoelectronics (2012), pp. 1–4. [CrossRef]
15. Z. Yang, M. A. Soto, and L. Thévenaz, “Increasing robustness of bipolar pulse coding in Brillouin distributed fiber sensors,” Opt. Express 24(1), 586–597 (2016). [CrossRef] [PubMed]
16. J. Geng, S. Staines, M. Blake, and S. Jiang, “Distributed fiber temperature and strain sensor using coherent radio-frequency detection of spontaneous Brillouin scattering,” Appl. Opt. 46(23), 5928–5932 (2007). [CrossRef] [PubMed]
17. F. Wang, X. Zhang, Y. Lu, R. Dou, and X. Bao, “Spatial resolution analysis for discrete Fourier transform-based Brillouin optical time domain reflectometry,” Meas. Sci. Technol. 20(2), 025202 (2009). [CrossRef]
18. G. Tu, X. Zhang, Y. Zhang, Z. Ying, and L. Lv, “Strain variation measurement with short-time Fourier transform-based Brillouin optical time-domain reflectometry sensing system,” Electron. Lett. 50(22), 1624–1626 (2014). [CrossRef]
19. Y. Yu, L. Luo, B. Li, L. Guo, J. Yan, and K. Soga, “Double peak-induced distance error in short-time-Fourier-transform-Brillouin optical time domain reflectometers event detection and the recovery method,” Appl. Opt. 54(28), E196–E202 (2015). [CrossRef] [PubMed]
20. M. A. Soto, G. Bolognini, F. D. Pasquale, and L. Thévenaz, “Long-range Brillouin optical time-domain analysis sensor employing pulse coding techniques,” Meas. Sci. Technol. 21(9), 094024 (2010). [CrossRef]
21. H. Chang, X. Jia, X. Ji, C. Xu, L. Ao, H. Wu, Z. Wang, and W. Zhang, “DBA-based BOTDA using optical comb pump and pulse coding with a single laser,” IEEE Photonics Technol. Lett. 28(10), 1142–1145 (2016). [CrossRef]
22. Q. Sun, X. Tu, Y. Lu, S. Sun, and Z. Meng, “High-accuracy and long-range Brillouin optical time-domain analysis sensor based on the combination of pulse pre-pump technique and complementary coding,” Opt. Eng. 55(6), 066125 (2016). [CrossRef]
23. S. L. Floch, F. Sauser, M. Llera, M. A. Soto, and L. Thévenaz, “Colour simplex coding for Brillouin distributed sensors,” Proc. SPIE 8794, 879437 (2013). [CrossRef]
24. M. Shin and H. Lee, “A high-speed four-parallel radix-2 4 FFT/IFFT processor for UWB applications,” in IEEE International Symposium on Circuits and Systems (IEEE, 2008), pp. 960–963.
