Coherent Doppler lidar (CDL) is a commonly used remote sensing technology found in a variety of industries for instantaneous ranging and velocimetry, such as in airports for the characterization of wind vortices [1]. Furthermore, it shows good potential for automotive lidar, as it provides ranges far greater than other types of lidar, which is necessary for high-speed automotive applications [2]. Given power restrictions due to safety concerns in automotive lidar, short pulses are required for improved range [3]. Furthermore, wind lidars depend on shorter pulses for improved spatial resolution [4]. Therefore, there is a clear need for shorter pulses in CDL.

Recent years have seen several implementations of CDL with shorter pulses. For example, approaches implementing 40-ns Golay coded pulses [5] or using Gaussian pulse pairs in a differential correlation scheme with a probing pulse width set to 18 ns [6]. Furthermore, random phase-coding has been applied in a 30-ns modulation bit interval [7]. Although the aforementioned solutions have successfully used shorter pulses combined with innovative techniques, they do not address the fundamental problems that occur in short-pulse CDL. The first is that shorter pulses will produce a weaker beat frequency signal, limiting the possible range in CDL [8]. To this end, major organizations such as Mitsubishi and NASA are developing high-power amplification schemes [9,10]. A second problem relates to the fact that short pulses inherently have broader intermediate frequency (IF) and baseband spectra, which could potentially overlap and lead to spectral cross talk and distortion, as depicted in Fig. 1 [11]. As the pulse is shortened, both the baseband and IF spectra are broadened to the extent where they can no longer be differentiated. In addition, shorter pulses have an insufficient number of zero crossings to properly resolve the Doppler frequency shift [12]. However, these issues can be simultaneously mitigated through the use of a higher IF, providing more cycles per pulse and reducing spectral cross talk effects. To shape the pulse and provide the aforementioned IF shift in CDL, the acousto-optic modulator (AOM) has been the modulator of choice [13–17]. However, there have been other modulation schemes proposed, such as a single electro-optic modulator (EOM) [18], multiple EOMs [19], serrodyning [20], and even multiple AOMs [21]. However, because of its ability to simultaneously provide frequency and intensity modulation, the AOM is preferred. That being said, the AOM is both limited by the achievable pulse duration and IF. Commercial AOMs can achieve rise times as fast as approximately $5$ ns and frequency shifts limited to several hundred MHz, with current research only now entering the GHz range [22,23]. Therefore, these limitations make the AOM extremely incompatible for short-pulse regimes in CDL. However, a relatively new modulator currently being used in the optical communications domain, the optical single-sideband modulator (OSSBM), has been shown to easily produce data rates and frequency modulation in the tens of GHz range [24]. Recently, it has been used in FMCW lidar for the first time in [25]. Furthermore, it has been reported in laser Doppler velocimetry, but by purely electrical and semi-optical means for short-range applications [26]. Therefore, in this Letter, we present the first implementation of an OSSBM in CDL. We achieve a pulse width of 4 ns and an IF of 1 GHz, which are respectively several times shorter and higher than previous implementations of a fiber-based CDL.

 

Fig. 1. Spectral cross talk between the baseband and IF spectra for several pulse widths.

Download Full Size | PDF

The OSSBM considered in this CDL implementation is the X-cut lithium niobate single-sideband suppressed carrier modulator (X-cut LN SSB-SC) [27], shown in Fig. 2. It uses a dual Mach–Zehnder modulation scheme, with a Mach–Zehnder modulator (MZM) located in each arm of the original MZM. As per [28], two RF signals equal in frequency and offset in phase are inputted to MZM$_1$ and MZM$_2$, respectively. Thus, if $RF_1(t) = V_1\cos {2\pi f_m t}$, then $RF_2(t) = V_2\cos (2\pi f_m t + 90^{\circ })$, where $V_1$ and $V_2$ are the gain voltages of MZM$_1$ and MZM$_2$, respectively. Next, designating the input light as $E_{in} = e^{i2\pi f_0t}$, $E_{out}$ is approximated as [28]

From Eq. (2), the output light has a single-frequency component shifted by the initial RF frequency $f_m$. In our implementation, the OSSBM, capable of 10-GHz frequency modulation and measured to have an insertion loss of approximately 4.1 dB, is tuned using the setup in Fig. 2. The setup is an interferometer with one arm being the OSSBM at $f_m = 1$ GHz and the second arm being an 80-MHz AOM. Therefore, theoretically speaking, a single beat frequency should be measured from the AC current exiting the BPD, representing the difference between the two arms as

As can be seen, there are three main beat frequencies present. They are the original AOM frequency $f_{AOM} = 0.08$ GHz, the LSB frequency $f_m - f_{AOM} = 0.92$ GHz, and USB frequency $f_m + f_{AOM} = 1.08$ GHz. Ideally, as in Eq. (3), the OSSBM is tuned such that a single sideband remains, which in our case is the USB at $1.08$ GHz. Thus, this method presents a straightforward manner in differentiating between upper and lower sideband components, since without $f_{AOM}$, they would appear along the same spectral line. With this setup, the optimal gains $V_1$, $V_2$ as well as biases $\phi _1$, $\phi _2$, and $\phi _3$ were determined and the results of the tuning are shown in Fig. 3. In Fig. 3(a), the OSSBM has not been tuned and the three major spectral components are clearly visible. After tuning, it can be seen that an extinction ratio of over 35 dB has been achieved between the upper and lower sidebands. Furthermore, the upper sideband has an on/off extinction ratio of 25 dB, which is seen by the difference in amplitude between the USB at $1.08$ GHz and the AOM beat frequency at $0.08$ GHz. That being said, when the AOM is eventually removed for the CDL implementation, the unmodulated light from the OSSBM will have the same frequency as the local oscillator and this beat frequency will no longer exist.

 

Fig. 2. Tuning setup for the X-cut LN SSB-SC OSSBM. MZM, Mach–Zehnder modulator; CW, continuous wave; AOM, acousto-optic modulator; BPD, balanced photodetector; OC, optical coupler.

Download Full Size | PDF

 

Fig. 3. Tuning results of the OSSBM: (a) OSSBM spectrum before tuning; (b) OSSBM spectrum after tuning.

Download Full Size | PDF

To test the effectiveness of CDL with OSSBM under a short pulse regime, a setup with the characteristics of Table 1 was fabricated and is shown in Fig. 4. Object velocities were measured using a turntable and the return signals were gathered and post-processed. In a bi-static system, the tangential velocity $v_t$ of the turntable is related to the Doppler shift frequency $f_d$ by [8]

 

Fig. 4. CDL configuration using the OSSBM. EDFA, erbium-doped fiber amplifier.

Download Full Size | PDF

 

Fig. 5. Average velocity estimates of a CDL with OSSBM compared to CDL with AOM, at a 300-ns pulse width.

Download Full Size | PDF

Table 1. CDL Parameters

View Table

Next, to analyze the OSSBM CDL in the short-pulse regime, a tap was added to the transmitter arm and the shortest pulse observed by the OSSBM is shown in Fig. 6, having an FMHW of 4 ns. The main limitation in achieving this pulse width in this implementation comes from the driving electronics of the OSSBM and not the OSSBM itself, as the electronics have rise times of several nanoseconds. In fact, with the proper electronics, the X-cut LN SSB-SC has been shown to achieve modulation bandwidths of over 20 GHz, representing the potential for 50-picosecond pulses [27]. Also in the figure, for comparison, is the minimum pulse duration achieved by the G&H T-M080-0.4C2J-3-F2P, a high performance commercially available AOM. The 4-ns pulse width achieved by the OSSBM is not only much smaller than the G&H AOM, but also any other commercially available AOM, as was mentioned previously having minimum pulse widths in the 10-ns range. Furthermore, a 4-ns pulse width is shorter than any previous implementation of a fiber based CDL, with the shortest implementation, to the best of the authors’ knowledge, being 18 ns [6]. Conversely, current commercially available units are limited to pulse widths of approximately 100 nanoseconds [11].

 

Fig. 6. Instantaneous power of the transmitted pulse, both for the AOM and OSSBM.

Download Full Size | PDF

The major advantage of the OSSBM CDL over the AOM version comes with the possibility of a higher IF, because short pulses inherently have broader IF and baseband spectra, which could potentially overlap and lead to spectral cross talk if not enough bandwidth separation is provided. However, there should be some consideration made on how high an IF is necessary, as a higher IF requires a faster digitizer, which often are limited in the dynamic range. Therefore, the IF chosen should provide the minimum bandwidth necessary for the mainlobe of the IF spectrum to remain unaffected by the baseband spectrum.

To confirm the advantage of a higher IF, a CDL with a 1-GHz OSSBM was compared against the traditional CDL with an 80-MHz AOM in the short-pulse regime, as shown by Fig. 7, for a stationary target. Note that in the case of the AOM, the 4-ns pulse was simulated by using a 4-ns window on a 21-ns pulse because, as was previously mentioned, AOMs are currently not capable of producing pulse widths in this range. As is expected, the shorter pulses have broader spectra, which can be seen by the accumulated spectra shown in Figs. 7(a) and 7(b). The 100-ns pulses, which represent current commercial standards, have narrow baseband and IF spectra so there is no cross talk between them. Therefore, their respective likelihood functions in Figs. 7(c) and 7(d) estimate the proper Doppler shift peaks. That is, in the case of a traditional CDL with an 80-MHz AOM, the maximum likelihood is observed at 80 MHz as well, shown by the “|” located at the likelihood function peak in Fig. 7(c). Similarly, in the case of the OSSBM CDL operated at 1 GHz, the maximum likelihood occurs at 1 GHz. Reducing the pulse width to 21 ns, the observed limit of the traditional CDL, it is clear to see from Fig. 7(a) that the spectrum is now wider than the available bandwidth provided by the low IF of 80 MHz. Consequently, the baseband spectrum overlaps with the IF spectrum and distorts the spectral shape as well as the likelihood function. In the last scenario with a 4-ns pulse width, this overlap has completely distorted the shape of the IF spectrum for the traditional CDL. However, for the CDL with OSSBM [Fig. 7(b)], despite the broader spectra, there is little to no spectral distortion observed and a proper maximum likelihood estimate is made in Fig. 7(d). This is because using a larger IF provides the bandwidth necessary to avoid cross talk between baseband and IF spectra, which are broader in shorter pulses. That is, in the 4-ns pulse width case, the influence from baseband cross talk produced an error of 17 m/s for the traditional CDL, whereas for the same pulse width using the same CDL scheme but with an OSSBM, the velocity estimate error remained as 0 m/s. These results clearly show the necessity of using a higher IF if shorter pulses are to be used in CDL and that, to this end, the OSSBM provides an appropriate solution to the problem.

 

Fig. 7. Doppler shift peak estimation for various pulse widths in both the AOM CDL and OSSBM CDL: (a) Return spectra from the AOM CDL; (b) return spectra from the OSSBM CDL; (c) matched filter likelihood function for the AOM CDL; (d) matched filter likelihood function for the OSSBM CDL. The likelihood estimates are depicted using the “|” symbol.

Download Full Size | PDF

Thus, in this Letter, we have introduced for the first time the use of optical single-sideband modulation in CDL. Furthermore, we have shown through live experiments that compared with a traditional CDL, using an OSSBM lends much more favorably when used in congruence with shorter pulses. In fact, with the OSSBM CDL, we were able to achieve a pulse width of 4 ns, which is at least four times shorter than any previous fiber-based CDL implementation. Additionally, we achieved an intermediate frequency of 1 GHz, which has also never been implemented. This, all while confirming the fact that the OSSBM appropriately addresses baseband and IF spectra cross talk, which was shown to not be possible with the AOM. Moreover, the potential for OSSBM CDL is yet still higher, as pulse widths and intermediate frequencies several times higher than those presented in this Letter are possible. To this end, further analysis of OSSBM for CDL is needed, such as outdoor experiments.