Open Access
Add to BibSonomy
Abstract
A microwave photonic system which can simultaneously realize the functions of rapidly tunable Doppler frequency shift (DFS) and high fidelity storage of broadband RF signals is proposed and verified. Single-sideband carrier-suppression modulation combined with dual-AOM frequency shifting ensures large-range and fast-responding DFS. And time-gated semiconductor amplifier (SOA) based fiber delay loop can realize high-fidelity RF pulse storage with high extinction ratio switching and amplification characteristics of time-gated SOA. A spurious rejection ratio greater than 40 dB, tuning range of DFS greater than ± 3 MHz, response speed of DFS less than 30 ns, and high fidelity storage of 4 GHz-12 GHz RF signals with greater than 381 circulations (corresponding 80 us delay time) are realized by the proposed structure. The maximum signal-to-noise ratio (SNR) is 13.6 dB within 381 circulations. Based on the experimental data, the simulation results show that the delay time also could be extended to 10 times more.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Radar deceptive jamming plays an increasingly important role in the field of electronic countermeasures with its flexibility and concealment [1,2]. High fidelity jamming to the distance, speed and quantity of moving targets is one of the important research contents in deceptive jamming [3,4]. The use of range gate pull off jamming or velocity gate pull off jamming alone can only work on radars with capabilities of range or velocity detection and tracking, but cannot deceive radars with simultaneous detection and tracking capabilities of range and velocity information, such as pulse Doppler radar. Therefore, it is necessary to use simultaneous range and velocity jamming technology to effectively counter the new radar system [5,6]. Meanwhile, the jamming scenarios are different, and the parameters such as the maximum storage time and storage precision are needed differently, especially for short pulse signals which require high storage precision and signal-to-noise ratio (SNR).
However, high-fidelity storage and Doppler frequency shift (DFS) of broadband RF pulse signals is a problem that has been difficult to solve for a long time. Digital RF memory (DPRM) is the most widely used solution currently. The technology has good flexibility and coherence, but is limited by A/D conversion sampling speed and resolution. It is difficult to achieve a low spurious DFS and a large storage time of RF signals with wide instantaneous bandwidth, and is also difficult to achieve high fidelity processing of broadband and high frequency RF signals [7–9]. The instantaneous bandwidth of commercial DPRM is about 500 MHz-2 GHz, and the number of quantization bits of DPRM operating in the X-band and Ku-band is 3∼4 with 13 dB-20 dB of the spurious suppression ratio [10].
The photonic method can effectively solve the bottlenecks such as the problems of speed and bandwidth existing in electronics. Optical fiber has the advantages of large bandwidth, low loss, and anti-electromagnetic interference. Therefore, photonic broadband RF signal transmission and processing have attracted extensive attention of researchers [11–13]. In the field of microwave photonics combined with electronic countermeasures, researchers have proposed and validated photonic RF signal storage structures [14,15]. In [14], a method of RF storage based on active fiber loop is proposed. And the relationship between the performance parameters of the key components and the number of circulations is simulated and the feasibility of using active fiber loop as a storage unit is verified. In [15], a new RF storage method based on fiber-optic frequency shifting fiber loop is proposed. The optical frequency shifter in the frequency shifting fiber loop is used to suppress the coherent lasing of ASE noise in the fiber loop. When the number of circulations is 186, the SNR of 5 GHz RF signal is degraded by 10 dB. Meantime, there are some other optical buffer or optical delay line structure can provide the similar time delay function for the improvement of delay controlling and high fidelity of the RF signals [16–18]. Regarding photonic DFS controlling, a DFS compensation scheme using the dispersion effect of frequency modulated laser is proposed in [19], and it verifies the feasibility of DFS compensation for wireless communication on high-speed railway by simulation.
Therefore, in order to realize synchronous jamming of the range and velocity of wideband RF signals, a photonic rapidly tunable DFS and high-fidelity broadband RF storage structure is proposed and implemented. As far as we know, this is the first simultaneous realization of the two functions by photonic method. Firstly, a DFS structure based on single sideband carrier suppression modulation and dual-AOM frequency shifting is proposed. The spurious suppression ratio in frequency shift process exceeds 40 dB, the tuning range of DFS is greater than ± 3 MHz and the frequency shift response speed is less than 30 ns. In addition, a RF storage structure using a high extinction ratio gated SOA based fiber loop structure is also proposed. With a 42 m delay fiber used in the fiber loop, the number of circulations exceeds 381 times, the storage time is greater than 80 us, the SNR is less than 13.6 dB and the storage of 4 GHz-12 GHz RF pulse signals is also experimentally verified. Finally, the extensibility of storage time is also analyzed by simulation based on the experiments.
2. Experiment setup
Figure 1 shows the experimental diagram of photonic broadband RF pulse signal storage and DFS system. A continuous wave light emitting from the laser was split into two, one passed through a dual parallel Mach-Zender modulator (DPMZM) which was modulated by input RF pulse signal with carrier suppression and single sideband (CS-SSB), then the modulated light was frequency-shifted by the acousto-optic frequency shifter 1 (AOM1), the center frequency of AOM1 was fixed. The other one light passed through AOM2, the center frequency of AOM2 was adjustable, then the output RF frequency detected by the photodetector can be expressed as:
According to Eq. (1) and the principle of DFS, the DFS range and response speed of the RF pulse signal are only related to the frequency shifting range and speed of AOM2. The spurious suppression capability is related to performance of the DPMZM. It can be seen from [8] that when a fiber loop is used as the RF storage unit, the input light needs to be a pulse light with high extinction ratio. And only the modulated part by RF pulse can be input into the fiber loop in order to suppress other light leaking into the fiber loop. AOM1 and AOM2 were not only used as optical frequency shifters, but also as input optical switches with ultra-high extinction ratio and fast response speed.
Fig. 1. Experimental diagram of photonic RF pulse DFS and storage system. EDFA: Erbium-doped fiber amplifier; DPMZM: dual parallel Mach-Zender modulator; AOM: acousto-optic frequency shifter; SMF: single model fiber; SOA: semiconductor amplifier; PC: polarization controller; PD: photodetector; OSA: optical spectrum analyzer; ESA: electronic spectrum analyzer; OSC: oscilloscope.
A narrow linewidth laser was used as the optical source with a linewidth of 1 kHz, and its wavelength was 1550.12 nm and output power was 13 mW. The laser was amplified by EDFA1 to about 70mw, and injected into the 1:99 optical coupler1, 99% of the laser was input to the DPMZM (photline, 25 GHz) and AOM1 which had 80 MHz fixed shift frequency. 1% of the laser was injected into AOM2, the frequency shift range of it was about 70 MHz∼90 MHz. The trigger signals for controlling the ON and OFF of AOM1 and AOM2 were generated by a delay control board, and the trigger signal was synchronously triggered by the control signal of the input RF pulse signal. The spurious suppression ratio and the extinction ratio of the two AOMs were both greater than 50 dB, and the rising time was less than 30 ns. When the two AOMs were turned off, the leakage of laser into the fiber loop had little effect on the storage of signal light pulses, which could be ignored because of the ultra-high extinction ratio of AOMs [8]. The lengths of the two arms between coupler1 and coupler2 should be roughly the same to minimize the difference of delay time between them. The optical pulses were then injected into the fiber loop formed by a 1:1 coupler3, which comprised of an optical filter (100 GHz bandwidth) with a center wavelength of 1550.12 nm, a polarization controller (PC), a time-gated SOA and a 42 m (corresponding to 210 ns delay time) single model fiber (SMF). The insertion loss of the optical filter was about 0.2 dB. The polarization controller minimized the polarization loss of the time-gated SOA. The time-gated SOA had a response speed of less than 5 ns and an extinction ratio greater than 50 dB. When an optical pulse circulated in the fiber loop and passed through the SOA, a control signal from the delay time control board would accurately control the on and off of the SOA to only amplify the optical pulse signal within the pulse width, and greatly suppress the noise out of the pulse width with 50 dB extinction ratio. The characteristic of time gate also enabled flexible control of the storage time in the loop. Of course, the main purpose of setting the SOA gain close to the total loss in the loop and amplifying only the modulated optical pulse signal within its pulse width was to make the RF pulse have extremely little loss and additional noise for storing a long time, so the quality of storage performance is mainly characterized by the gain and SNR of the stored signal. The optical pulse signals were a series of pulses when they were output from the fiber loop. In order to flexibly control the delay and number of the output optical pulses, they needed to be selected by the output optical switch (AOM, 80 MHz, 50 dB extinction ratio) after amplified by an EDFA, and detected by a wideband photodetector. The output RF pulse signals were measured by an electronic oscilloscope (OSC, Keysight 6000X, 4 GHz bandwidth) and an electronic spectrum analyzer (ESA, Keysight N9030B, 2 Hz to 13.6 GHz). The photodetector had a bandwidth of 20 GHz and a response rate of 0.65 A/W.
3. Experimental results
3.1 Doppler frequency shifting function
First, the performance of the DPMZM and the frequency shifting of optical carrier was tested. The input RF frequency was set to 4 GHz, the pulse width was set to 10 us, and the repetition period was 15 us. The two AOMs were synchronously triggered with the input RF signal. Optical signals outputting from the upper and lower arms of coupler2 were respectively read by optical spectrum analyzer (OSA) and ESA as shown in Fig. 2. The results with CS-SSB modulation and 80 MHz fixed frequency shifting are shown by the black curve in Fig. 2. And the red curve is the optical spectrum of the optical carrier shifted by AOM2 with 83 MHz. What is actually needed is the beat signal of the 83 MHz frequency-shifted optical carrier and the negative first-order sideband signal, but the signals including the beat signal of positive first-order sideband and frequency-shifted 83 MHz optical carrier and the beat signal of negative first-order sideband and frequency-shifted 80 MHz suppressed optical carrier are the spur signals near the spectrum of the wanted RF signal, and they are difficult to be filtered out in the RF spectrum. Therefore, it is necessary to suppress it in the optical domain as much as possible. The optical signal from the coupler2 is detected by a photodetector as shown in the inset of Fig. 2. Comparing the optical spectrum and the RF pulse spectrum, it is shown that the positive first-order sideband is 42 dB lower than the negative first-order sideband, and the suppressed optical carrier is 41 dB smaller than the shift-frequency 83 MHz carrier, which achieves ideal spurious suppression in the RF spectrum.
Fig. 2. Spectrum before entering the fiber loop and corresponding output RF signal spectrum (inset).
The performance parameters of DFS mainly include the range and response speed of frequency shift, and spurious suppression capability during frequency shifting. The range and response speed of frequency shift are mainly limited by the performance of AOM2 and its driver. Figure 3(a) shows the relationship between the output optical power of AOM2 and its driving frequency. The center frequency of AOM2 is 80 MHz, but the actual measured driving frequency with minimum insertion loss has a slight deviation, which is about 79 MHz. When the tuning range of frequency shift is greater than ± 8 MHz, the power attenuation will be greater than 3 dB. The power variation within ± 3 MHz is less than 1 dB with high power consistency. And when the input signal frequency is also 4 GHz, the SOA in the loop is turned off, the RF pulse spectrum is read from the photodetector as shown in Fig. 3(b). It can be seen from Fig. 3(b) that the power of the radio frequency and the spurious signal remain substantially the same during the tuning process in the range of ± 3 MHz, and the spurious suppression ratios are greater than 40 dB.
Fig. 3. (a) Relationship between output optical power of AOM2 and its drive frequency and (b) the spectrum response of frequency shift when the input RF signal was 4 GHz.
The response speed of the frequency shift is directly related to the real-time processing capability of RF pulse signals, and is an important parameter in the photoelectric countermeasure. In this paper, frequency shift control is realized by AOM2, and the driving source of AOM2 (Gooch&Housego) has two functions, which are respectively controlled by two voltage signals. One function is switching the amplitude of drive source. When the voltage is greater than 1 V, the driving source has essential power to drive AOM2, and when the voltage is 0 V, AOM2 is turned off. The response speed of the amplitude switch related to the response speed of the AOM itself is less than 30 ns. Another function is the control of frequency shift, the magnitude of driving voltage controls output frequency of drive source. Therefore, when implementing the frequency shift function, the AOM2 has to turn on the amplitude control switch and the frequency shift control function. The delay control board needs to synchronously trigger the amplitude control switch and the frequency shift control function to measure the frequency shift response speed. The experimental results shows that the response speed of the frequency shift is mainly related to the starting frequency (corresponding to the low level of the frequency control switch). When the termination frequency (corresponding to the high level of the frequency control switch) is fixed, the starting frequency is lower (the low level is lower), the response time of frequency shift will be slower. When the shifted frequency is 80 MHz (the high level is 5.6 V), the relationship between the response speed of frequency shift and the low level is shown in Fig. 4(a) and the corresponding pulse waveform is shown in Fig. 4(b). The frequency of the input RF signal in Fig. 4(b) is also 4 GHz. It can be seen that when the low level is 0 V, the response time exceeds 300 ns. When the low level is greater than 3 V, the response speed of the drive source frequency shift will be less than 30 ns, and the frequency shift speed at this time will only be limited by the response speed of the AOM itself.
Fig. 4. (a) AOM2 frequency shift response speed and drive source low level and (b) its impulse response waveform
Therefore, the total response speed of frequency shift can be defined as the fastest speed from the trigger of the two functions to the full implementation of the frequency shift. In order to make the total response speed fast enough, the low level of the frequency control switch should always be greater than 3 V, and the response speed of frequency within 80 MHz ± 3 MHz (corresponding to a high level of 5.4 V ∼ 5.8 V) will reach the AOM's own response speed limit, which is about 30 ns. The response speed meets the practical application requirements of many scenarios.
3.2 RF pulse storage function
First, the storage performance of RF pulse signal with frequency of 4 GHz, pulse width of 200 ns and repetition period of 150 us was measured. The center frequency of AOM2 was set to 80 MHz, and pulse signals and their waveforms after 0, 190, and 381 circulations were measured by the oscilloscope and are shown in Figs. 5(a)–5(c). From the capability of waveform fidelity, the amplitude inconsistency of the waveform after 381 circulation is worse than it after 191 circulations. The ratio of the amplitude peak-to-peak value and average value after 0, 191 and 381 circulations is 0.278, 0.583, and 0.8 respectively.
Fig. 5. Stored 4 GHz pulse signals and their waveforms after (a) 0, (b) 191, and (c) 381 circulations.
In addition, the spectral performance of the RF pulses and their accumulated noises were measured by the ESA. The ESA worked in pulse mode, the center frequency of it was set to 4 GHz and the frequency span was set to 0 Hz which is due to that ESA needs to work in pulse mode with maximum responding speed. The relationship between the power of the pulse signal and the delay time was measured and are shown in Fig. 6. It can be seen that the pulse signal power attenuation within 80 us delay time (381 circulations) is less than 6.1 dB, and the power after 500 circulations is 15 dB less than the power after 0 circulation. Moreover, the spectrum analyzer was operated in the pulse spectrum mode, the frequency span was changed to 30 MHz, and the pulse spectrum after 0, 191 and 381 circulations were respectively measured, and are shown in Fig. 7(a). By comparison, the maximum attenuation of pulse power within 191 circulations (40us delay) is only 2.6 dB. In order to measure the noise spectrum and calculate the SNR of pulse signal, the spectrum of noise of the system with different circulations was measured without loading the input RF signal and is shown in Fig. 7(b) and the SNR was calculated according to [20] which is shown in Table 1. The noise in the time-gated SOA based structure mainly comes from the beat of optical signal and accumulated spontaneous emission noise in SOA (s-sp beat noise). The noise accumulates rapidly within 191 circulations, but its accumulation slows down in the logarithmic domain after more than 191 circulations. The SNR after 191th and 381th circulations degrade 8.9 dB and 13.6 dB compared with the pulse signal after 0th circulation. The average single circulation degradation in the 381 circulations is 0.036 dB.
Fig. 6. Time domain display of ESA measuring the output pulse signal from the time-gated SOA based fiber optic loop
Fig. 7. Pulse spectrum of (a) RF pulse signal and (b) accumulated noise under different circulations
Table 1. Relationship between number of circulations and SNR degradation
The storage capacity of broadband RF pulse signals was also measured. 4 GHz, 8 GHz and 12 GHz RF signals were used as input and kept the power and parameters of pulse unchanged. The pulse storage performance is shown in Fig. 8. As can be seen from Fig. 8, the trends of the 8 GHz and 12 GHz RF power with the storage time are basically the same as that of 4 GHz RF, and the storage performance does not deteriorate as the frequency increases. In addition, the pulse and noise spectrum at 8 GHz and 12 GHz were also measured, and the SNRs were calculated, and are shown in Table 2. The noise at different RF frequencies remains basically the same in the optical domain (the front-end devices such as modulator, phase shifter and cables have little effect on them). But the pulse power attenuation of 8 GHz and 12 GHz are less than that of 4 GHz, so the SNR degradations are less, but the SNRs are lower with the same delay time.
Table 2. Relationship between SNR degradation and number of circulations at different RF frequencies
In addition, storage and DFS functions simultaneously realized were also measured. Input RF frequency was also set to 4 GHz. After 191 circulations, the DFS frequency were respectively set to −9.2 MHz, −4.7 MHz, 4.2 MHz, 8.7 MHz, and −300 kHz. The RF pulse spectrum is shown in Fig. 9(a). It can be seen that the pulse power has a certain attenuation when the DFS frequency is greater than 3 MHz. When the DFS frequency exceeds 8 MHz, the attenuation will be greater than 4 dB, which is basically consistent with the measured result of Fig. 3(a). However, this method can only reflect the pulse peak power near the spectral interval of 5 MHz (the reciprocal of the pulse width ∼200 ns), and the spectrum cannot reflect its capability of spurious suppression. Therefore, in order to measure the DFS within ± 3 MHz, continuous 10 pulses after 191 circulations were measured by the oscilloscope, and the high-level parts of 10 pulses were spliced into a pulse signal with 2 us pulse width, and then the result of FFT processing is shown in Fig. 9(b). The spurious power has been flooded by the pulse signal power, which verifies the high spurious suppression ratio of the frequency shifting process within ± 3 MHz. And the RF pulse power remains relative consistent during the frequency shift process.
Fig. 9. Spectrum of output pulse signal with (a) 200 ns pulse width and (b) 2 us pulse width at different Doppler frequency shifts after 40 us delay time.
Compared with [15], the time-gated SOA based fiber loop is used as the storage unit, and the optical frequency remains unchanged with the increase of number of circulations. Therefore, the RF power decreases less within the self-filtering bandwidth, especially when storing wideband signals. Meanwhile, the wideband RF signals are minimally influenced by phase distortion with a long storage time because of the SSB modulation. Of course, when transmitting a high-frequency signal, the microwave photonic link will have a relatively large loss in the RF front-end, modulator, and photodetector etc., so the output pulse signal power will decrease at high frequencies. In order to acquire a better power consistency of wideband RF signals, the modulator, phase shifter, cables and detector with wider bandwidth should be used.
4. Discussion
It can be seen from the experimental results that the SNR of the pulse signal decreases with the increase of the number of circulations. In order to study the limit level of SNR degradation of the time-gated SOA based RF storage structure, and further guide the optimization of storage performance, such as extended storage time etc., this paper makes the following theoretical simulations on the relationship between SNR degradation and storage time. Assuming that the average power within the pulse width before entering the fiber loop is expressed at P0, the RF power output from the photodetector after n circulations is:
Fig. 10. The relationship between SNR degradation and the number of cycles for different losses, gains and lengths of fiber in the loop. The parameters settings are as follows: γ=0.5, P0=13 mW, η=0.65, ψ=0.65 A/W, nsp=8, γ*(1—γ1) = 0.601 when the length of fiber is 42 m and γ*(1—γ1) = 0.603 when it is 420 m.
5. Conclusion
In this paper, a fast tunable DFS and high fidelity storage simultaneously realized structure for broadband RF pulse signals is proposed and verified experimentally. First, we construct a DFS structure based on CS-SSB modulation and dual-AOM frequency shifting, which achieves a high flat DFS with a spurious rejection ratio greater than 40 dB, a frequency shift range greater than ± 3 MHz, and a frequency shift response speed less than 30 ns. In addition, a time-gated SOA based fiber loop is proposed to achieve broadband RF high-fidelity storage with 42 m fiber in the loop, which suppresses the coherent lasing of the noise in the loop greatly using a controlled SOA with high extinction ratio, and minimizes the ASE accumulation in the loop with low loss. And the consistency of the optical frequencies effectively increases the number of circulations at 4 GHz-12 GHz RF signals, the number of circulations is more than 381 circulations (80 us) with maximum SNR degradation less than 13.6 dB, and the RF power loss is less than 6.1 dB. Moreover, the relationship between the SNR degradation and the number of circulations with different losses in the loop is theoretically analyzed. The results are consistent with the experimental results, and it is confirmed that the storage time can increase to 800 us with little additional SNR degradation.
Funding
National Natural Science Foundation of China (61535014, 61775225, 61875214).
Disclosures
The authors declare that there are no conflicts of interest related to this article.
References
1. D. C. Schleher, Electronic Warfare in the Information Age (Artech House, Inc., 1999).
2. M. Soumekh, “SAR-ECCM using phase-perturbed LFM chirp signals and DRFM repeat jammer penalization,” IEEE International Radar Conference191–205 (2006).
3. P. Gonzalez-Blanco, E. D. Diego, E. Millan, B. Errasti, and I. Montiel, “Stepped-Frequency Waveform radar demonstrator and its jamming,” International Waveform Diversity & Design Conference (2009).
4. Q. Yang, Y. Zhang, and X. Gu, “Wide-band chaotic noise signal for velocity estimation and imaging of high-speed moving targets,” Prog. Electromagn. Res. 63, 1–15 (2015). [CrossRef]
5. S. D. Berger, “Digital radio frequency memory linear range gate stealer spectrum,” IEEE Trans. Aerosp. Electron. Syst. 39(2), 725–735 (2003). [CrossRef]
6. M. Greco, F. Gini, A. Farina, and V. Ravenni, “Effect of phase and range gate pull-off delay quantisation on jammer signal,” IEE Proc., Radar Sonar Navig. 153(5), 454–459 (2006). [CrossRef]
7. A. Partizian, “Airborne Pulse-Doppler Radar,” IRE Trans. Mil. Electron. MIL-5(2), 116–126 (1961). [CrossRef]
8. J. Baldwinson and I. Antipov, “Prediction of electronic attack effectiveness against Maritime Patrol Radars,” International Conference on Radar (2008).
9. M. Greco, F. Gini, A. Farina, and V. Ravenni, “Effect of phase and range gate pull-off delay quantisation on jammer signal,” IEE Proc., Radar Sonar Navig. 153(5), 454–459 (2006). [CrossRef]
10. B. Manz, “DRFM grow to meet new threats,” J. Electron. Defense 33(8), 43 (2010).
11. J. Capmany and D. Novak, “Microwave photonics combines two worlds,” Nat. Photonics 1(6), 319–330 (2007). [CrossRef]
12. J. Wang, P. Hou, H. Cai, J. Sun, S. Wang, L. Wang, and F. Yang, “Continuous angle steering of an optically- controlled phased array antenna based on differential true time delay constituted by micro-optical components,” Opt. Express 23(7), 9432–9439 (2015). [CrossRef]
13. Z. Ding, J. Zhao, F. Yang, Z. Feng, B. Lu, and H. Cai, “Wide Bandwidth Reconfigurable and Frequency Tunable Microwave Photonic Filter Based on Tunable Ultra-sharp Roll-off Optical Filter,” 2018 Asia Communications and Photonics Conference (ACP)1–3 (2018).
14. L. V. Nguyen, “Photonic radio frequency memory-design issues and possible solutions,” DEFENCE SCIENCE AND TECHNOLOGY ORGANISATION SALISBURY (AUSTRALIA) SYSTEM (2003).
15. T. A. Nguyen, E. H. Chan, and R. A. Minasian, “Photonic radio frequency memory using frequency shifting recirculating delay line structure,” J. Lightwave Technol. 32(1), 99–106 (2014). [CrossRef]
16. J. D. LeGrange, J. E. Simsarian, P. Bernasconi, D. T. Neilson, L. Buhl, and J. Gripp, “Demonstration of an integrated buffer for an all-optical packet router,” IEEE Photonics Technol. Lett. 21(12), 781–783 (2009). [CrossRef]
17. D. Liu, S. Sun, X. Yin, B. Sun, J. Sun, Y. Liu, W. Li, N. Zhu, and M. Li, “Large-capacity and low-loss integrated optical buffer,” Opt. Express 27(8), 11585–11593 (2019). [CrossRef]
18. Q. Song, Z. Hu, and K. Chen, “Scalable and reconfigurable true time delay line based on an ultra-low loss silica waveguide,” Appl. Opt. 57(16), 4434–4439 (2018). [CrossRef]
19. R. Zheng, Y. Kong, E. Chan, Y. Cao, X. Wang, X. Feng, and B. Guan, “Photonics based Microwave Frequency Shifter for Doppler Shift Compensation in High-Speed Railways,” Conference on Lasers and Electro-Optics/Pacific Rim F2C. 4 (2018).
20. Agilent Technologies Application Note 150–2, “Spectrum and signal analysis...pulsed RF,” (2012).
21. E. Chan and R. Minasian, “Single sideband suppressed carrier modulator based frequency shifting recirculating delay line microwave photonic filter,” Opt. Laser Technol. 45, 160–167 (2013). [CrossRef]
