Abstract
The interference between a frequency-modulated continuous-wave (FMCW) light detection and ranging (LiDAR) and other LiDARs or sunlight was theorized, considering the spatial overlap, frequency overlap, and intensity ratio. It has been concluded that the interference probability between LiDARs can be lower than a safety standard value for autonomous vehicles when the number of the resolution points of a single LiDAR is increased sufficiently and that the interference with incoherent sunlight does not occur. Due to the coherent detection of FMCW, such ambient light immunity is much better than time-of-flight LiDAR. The dependence of the interference on the wavelength range, sweep bandwidth, and sweep period was also observed experimentally using a silicon (Si) photonics FMCW LiDAR chip incorporating slow-light grating beam scanners. It was shown that the interference can be suppressed by increasing the number of resolution points and changing their common parameters moderately. Regarding the contamination of sunlight, unwanted beam shift due to heating was observed, although it will be suppressed simply by wavelength filtering.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Light detection and ranging (LiDAR), a three-dimensional (3D) image sensor, has attracted great attention along with the development of autonomous vehicles [1–3]. In particular, on-chip frequency-modulated continuous-wave (FMCW) LiDAR fabricated using silicon (Si) photonics platform has been developed actively due to expectations for low-cost production, compact size, flexibility, and stability [4–8]. The FMCW method generates a linearly frequency-swept laser beam, irradiates it on surrounding objects, and measures their distances from beat frequencies between reflected light from the objects and local reference light. The 3D point cloud image is obtained by scanning the beam and repeating the ranging. In addition, the velocity and vibration of objects can be detected from a Doppler shift, which is often referred to as a four-dimensional LiDAR [4,5,9,10].
LiDAR in autonomous vehicles requires sufficient detectable range, filed of view (FOV), framerate, and so on. In addition, high immunity to ambient light from other LiDARs, sunlight, and streetlight is also crucial for future practical use. Such immunity is often said to be high regarding the FMCW method exploiting coherent detection. However, it has been reported previously that the interference occurs between synchronized FMCW radars [11,12] and LiDARs [13] at the same frequency or wavelength range and with the same frequency sweep rates. It has also been estimated that contamination of sunlight increases the shot noise [14]. These reports suggest that ambient light may degrade signal-to-noise ratio (S/N) and cause ghost images in FMCW LiDAR even using coherent detection, however, they did not assume practical situations, including accidental optical coupling, unsynchronized condition, and different coherence. Detection errors in LiDAR caused by ambient light can lead to a safety hazard in mobiles. Certification Specification 25, published by the European Aviation Safety Agency, has defined the safety standard probability for aircraft as p < 10−9/h = 2.8 × 10−13/s [15]. Although similar standard has not yet been defined, a requirement 30 times looser than that for aircraft has been suggested for autonomous vehicles, i.e., p < 3 × 10−8/h = 8.4 × 10−12/s [16].
In this paper, the interference probability of FMCW LiDAR caused by other FMCW LiDARs and sunlight is theorized. The rest of the paper is organized as follows. Section 2 presents the general expressions of the interference probability. Sections 3 and 4 then present more concrete expressions for that with other LiDAR and sunlight, respectively, and discuss the requirements for the safety standard, showing calculation results. Section 5 also shows some experimental observations of the interference using the Si photonics FMCW LiDAR [17–19] and pseudosunlight from a solar simulator. Finally, Section 6 provides the conclusion.
2. General expressions
Consider the following situation. FMCW LiDAR A transmits a linearly frequency-swept light with the sweep period TA and the sweep bandwidth BA, receives light reflected at an object at distance L with the round-trip time τA = 2 L/c (c is the speed of light in the air), and generates the following beat frequency fA as the ranging signal after mixing with local reference light:
For ambient light B from another LiDAR and sunlight to contaminate into LiDAR A and cause the interference, three probability factors can be considered. These include (a) the spatial overlap pI [/frame], (b) the frequency overlap pII, and (c) the interference intensity ratio pIII, as shown in Fig. 1. Multiplying them and the framerate F [frame/s], the interference probability pMI [/s] is expressed as follows:
This pMI shows the number of error resolution points per second.
pI is the probability that ambient light is coupled into the receiver of LiDAR A and is expressed as follows:
where Γf is the overlap efficiency of LiDAR’s image frame and pd is the probability of matching the directionality of LiDAR’s beam and ambient light.pII is the probability that the beat frequency caused by the interference occurs in a frequency bandwidth used for measuring the ranging signal and is expressed as follows:
pIII is the probability that the signal intensity caused by the interference, $\left\langle {{i_{\textrm{MI, meas}}}^2} \right\rangle$, becomes larger than LiDAR’s original ranging signal, $\left\langle {{i_{\textrm{A, meas}}}^2} \right\rangle$, at each resolution point. Generally, the FMCW method measures the distance from a peak frequency in a spectrum obtained after the Fourier transform of a beat signal. Therefore, defining the intensity ratio r as
3. Interference with another LiDAR
Consider a situation where that frequency-swept light with a sweep period TB, sweep bandwidth BB, and sweep rate γB ≡ BB/TB of FMCW LiDAR B is contaminated into FMCW LiDAR A and the interference occurs.
3.1 Spatial overlap
Figure 2 shows two contamination situations of FMCW light from LiDAR B. A direct interference occurs when light from LiDAR B is coupled into LiDAR A directly, while an indirect interference occurs when two LiDARs are observing the same point of an object. Since light of the FMCW LiDAR interferes with the reference light after receiving signal light, all light is fundamentally processed under the single-mode condition. Therefore, direct interference occurs only when the beam direction of LiDAR B coincides with the directionality of LiDAR A. Thus, pI is expressed as follows:
In addition, the indirect interference can occur at any resolution points when LiDAR A observes a point irradiated by light of LiDAR B simultaneously. Therefore,
Compared with the direct interference, pI in the indirect interference is larger by the number of the resolution points.
3.2 Frequency overlap
Equation (12) can be used to calculate the temporal response of the beat frequency caused by the interference (Supplement 1):
Figure 3 shows the frequency temporal response of FMCW light in LiDAR A and B and an example calculation result of fMI. It should be noted that fMI becomes a periodic function whose period is the least common multiple TAB between TA and TB. Whether fMI appears within the measurement bandwidth fAmax or not depends on the frequency sweep and the contamination timing. Moreover, whether the wavelength ranges used in two LiDARs are close or not is the most crucial factor since fi includes the optical frequency. If the wavelength ranges of future LiDAR products are set to be as different as possible within the available range, ρII(f) becomes an almost uniform distribution. Thus, pII is expressed as follows:
3.3 Intensity ratio
ηB/ηA is an important factor in Eqs. (7)–(9). A Lambertian scatterer is often modeled as a standard object for LiDAR. Hence, it is considered at a distance L with the reflectivity RA affected by the surface absorption and angular dependence. When the reception aperture of LiDAR A is denoted as S, ηA is approximated as
where ηc is the coupling efficiency into LiDAR A, including the round-trip propagation efficiency in the air. In the direct interference, the beam from LiDAR B slightly spreads and reaches LiDAR A. Consequently, ηB is approximated asWhen Tmeas ≈ TA, as aforementioned, the frequency broadening of the ranging signal, δfA, is written as
As shown in Eq. (12), the frequency broadening of the interference signal, δfMI, is written as
However, δfMI converges to 1/TA when Δγ decreases, as understood from the Fourier series expansion of a periodic function. From Eqs. (7)–(9),
If two LiDARs have similar characteristics (PA ≈ PB, τc,A ≈ τc,B, Δγ ≈ 0, TMI ≈ TA), and the coherence time is sufficiently longer than the propagation time of light (τA << τc,A), then r can be represented as follows:
Now, r is calculated for typical values of parameters, such as laser linewidth of 200 kHz (τc,A ≈ 1.6 µs), TA = 10–100 µs, RA = 0.1, and δθB, δφB = 0.1° = 1.7 × 10−3 rad. Equation (20) then gives r ∼ 105–106.
For the indirect interference, the light beams of both LiDARs are assumed to be irradiated on a point of the same Lambertian scatterer; otherwise, LiDAR A cannot receive light of LiDAR B. When the beams of both LiDARs are received, the range signal can give a wrong distance, depending on the intensity ratio, frequency setting and received timing. Let us denote the reflectivity for the beam of LiDAR B as RB. Therefore, ηB/ηA = RB/RA. Considering LiDARs with a similar performance and assuming the above conditions,
When RA ≈ RB, Eq. (21) gives r = 0.016–0.16 for the above values. From these results,
Next, consider the condition that δfMI = |Δγ|TMI >> 1/TA. The spectral peak intensity is reduced by the spectral broadening and inversely proportional to δfMI. In addition, as shown in Fig. 3, the interference occurs in a certain duration TMI within the measurement duration TA. Applying Eqs. (7)–(9) and (20) to this condition, r can be written for the direct interference as follows:
The right term assumes the above typical values. When BA = 10 GHz and TA = 100 µs, BATA2 = 100 Hz·s2, and α > 0.2. This means that if sweep bandwidths of LiDAR A and B change by more than 20%, r < 1, and the interference probability can be suppressed. However, in practice, the sweep bandwidth cannot be changed so widely for each LiDAR; therefore, the reduction of pIII is not quite significant. If BA > 100 GHz or TA > 1,000 µs can be set, the right term of Eq. (24) becomes on the order of 0.01 or 0.001, respectively. Then, the choice of the sweep bandwidth is greatly enhanced, and pIII can be suppressed by 2–3 orders of magnitude. However, for BA > 100 GHz, fAmax increases from Eq. (1), pII increases from Eq. (14), and the suppression of pIII is canceled. Therefore, increasing the bandwidth is not effective. On the other hand, for TA > 1,000 µs, fAmax decreases from Eq. (14), and pII also decreases, indicating no cancelations. However, such large TA degrades the LiDAR framerate.
For the indirect interference, r is represented as
Only even for the later term, r is likely to be < 1, and pIII = 0, but the front term further ensures r << 1; thus, pIII = 0 definitely.
3.4 Example of calculations
Figure 4(a) shows the dependence of pI on the number of resolution points Nx. When NA = NB, pI is inversely proportional to NA2 and NA for the direct and indirect interferences, respectively. Therefore, the indirect interference shows larger values, but it is not an issue eventually because pIII = 0 as mentioned above.
Figure 4(b) shows the dependence of pII on the sweep bandwidth BA and the sweep period TA, which determine the sweep rate γA, assuming τAmax = 1.3 µs. Here, a future situation is considered, where LiDAR products are fabricated using Si photonics technology and their operating wavelength ranges are chosen randomly in the C-band (λ = 1,530–1,565 nm, Δλ = 35 nm, fall ≈ 4.4 THz). Obviously, a narrower BA and a longer TA suppress pII. Figure 4(c) shows the dependence of r on Δγ and TA on the above conditions. A longer TA and a larger Δγ reduce r. For the indirect interference, r < 1 for any Δγ and pIII = 0. Even for the direct interference, Δγ > 2 × 1013 Hz/s when TA = 100 µs. Moreover, r < 1 when α > 0.2. As a result, the interference can be suppressed.
A main lobe of transmitted and received beams of LiDARs has only been considered so far, but the interference between the main lobe and sidelobes and between sidelobes may occur. However, for example, if the sidelobe intensity is 10 dB lower than that of the main lobe, the interference intensity is reduced by 10–20 dB, and r becomes lower by that amount. Besides, pMI has been calculated for the interference of one LiDAR, but pMI will be enhanced when many LiDARs interfere, which increases pI.
From these results, pMI was calculated for the direct interference as a function of the number of the resolution points, as shown in Fig. 5, where NA = NB, χ = 1, BA = 10 GHz, τAmax = 1.3 µs (L = 200 m), fall = 4.4 THz (C-band), pIII = 1, and F = 10 frame/s. Figure 5(a) shows the result for a single LiDAR A. Here, we assumed TA = 1/NAF, which means that the sweep period is changed with NA to maintain the framerate. Then, pII in Eq. (14) becomes proportional to NA and this makes pMI to be inversely proportional to NA after multiplying pI in Eq. (10). For example, the nonmechanical beam scanner based on slow-light grating (SLG), which will be described later, can achieve NA = 400 × 32 = 12,800, resulting in pMI ≈ 2.3 × 10−11/s. With this NA value, the scanner does not satisfy the standard for autonomous vehicles, which is shown by the red solid line. However, increasing NA to more than 35,000 leads to the standard satisfaction. Figure 5(b) shows the result for the parallel operation of nA units of LiDARs. In this calculation, we fixed TA and assumed nA = TA/(1/NAF), so that NA can be increased even with a fixed TA. Then, pII becomes independent of NA and pMI becomes inversely proportional to NA2 under the assumption NA = NB. For TA = 100 µs, the parallel action of nA = 11 is required for NA = 12,800, and pMI is suppressed to 1.8 × 10−12/s, satisfying the autonomous vehicle standard. Furthermore, NA ≥ 33,000 also satisfies the aircraft standard, which is shown by the red dashed line.
These estimations are based on the most severe criterion that only an error of one resolution point in one frame causes failure of the sensing. Here, we define this condition as class-S, and also define class-A, class-B, and class-C, which permit errors of up to 3, 10, and 30 times, respectively. The requirements are then much more relaxed, as shown in Fig. 5. Given the same parameters as above, the single LiDAR can satisfy class-A for autonomous vehicles, as shown in Fig. 5(a), and the parallel LiDARs can satisfy class-S and class-B for autonomous vehicles and aircrafts, respectively, as shown in Fig. 5(b).
Let us compare the above results with those for time of flight (ToF) LiDARs. Similar to the estimation for FMCW, we assume the irradiation and detection of each resolution point one by one for ToF. Due to direct detection of ToF, the interference occurs even in any contamination timing within the bandwidth of a wavelength filter, which is usually inserted in front of a photodetector to pass laser light and eliminate ambient light. Thus, pII is expressed as δλ/Δλ, where δλ and Δλ are the bandwidth of the wavelength filter and available wavelength range, respectively. ToF LiDARs use a high-power pulsed laser diode whose wavelength is shifted with temperature, and δλ is usually set wider than 10 nm to cover a wide temperature range (e.g., δλ = 20 nm). The laser wavelength is set in the range of λ = 930–970 nm due to the weak sunlight intensity, as shown in the next section. Therefore, we set Δλ ≈ 40 nm, resulting in pII = 0.5 here. Regarding pI and pIII, there are no significant differences between FMCW and ToF. Then, we obtain green lines in Fig. 5 for ToF. Even for a single LiDAR action condition TA = 1/NAF, pII is independent of NA, as mentioned above. Then, pMI becomes proportional to NA−2 and is calculated to be 3.1 × 10−8/s at NA = 12,800. This pMI is three orders of magnitude higher than that for FMCW. Due to the different dependences on NA, this difference in pMI decreases with increasing NA and pMI takes the same value when NA = 17,000,000. For parallel LiDAR action, pMI for ToF is four orders higher than FMCW’s for any NA. These results vary with the configuration and parameter settings of LiDAR, but it is definite that the immunity of FMCW is many orders better than ToF’s, mainly owing to the frequency selectivity.
4. Interference with sunlight
This section discusses the probability that sunlight is directly coupled into a LiDAR chip. The spatial overlap with direct sunlight occurs when its solid angle ΩSUN ≈ 6.8 × 10−5 sr [22] overlaps with LiDAR’s FOV. Then,
Since the frequency overlap with the measurement band occurs definitely due to the wide spectrum of sunlight,
The interference frequency broadens as wide as BA because the frequency of sunlight is not swept like FMCW, and r is expressed as
Some literatures have defined coherence standards as |j12| = 0.88 [27] and 0.5 [24]. However, this study sets |j12| = 0, which means that the coherence sufficiently disappears, and then v = 3.83 and Sc = 1.17(λ/ϕs)2. It was calculated that Sc = 0.032 mm2 from ΩSUN ≈ 6.8 × 10−5 sr, ϕs ≈ 9.3 × 10−3 rad, and SSI = 0.26 µW/mm2/nm around λ = 1,550 nm. As shown in Fig. 6(b), the sunlight intensity coupled into the LiDAR is calculated, where ηSUN = 0.5 is assumed for random polarizations of sunlight. For the SLG scanner, the wavelength range, corresponding to one resolution point, δλ, is ≈ 0.1 nm, then PSUN < −67 dBm from Eq. (30). It was also found that PSUN << Pref when Pref is set to be ∼ 0 dBm. The shot noise of sunlight does not affect the ranging signal because the shot noise is mostly determined by Pref. Besides, it should be noted that although not only direct interference but also indirect one may also occur, indirect sunlight is much weaker than direct sunlight and can be ignored completely.
As the frame overlap efficiency Γf depends on situations, we consider it as the average value in one day. Assuming an FOV of 40° × 8.8° expected for the SLG LiDAR [17] and overlap time of 0.75 h/day and adding an angular divergence corresponding to ΩSUN, Γf is calculated to be 3.6 × 10−4. From Eq. (26), pd ≈ 80 /frame and then pI = 2.9 × 10−2/frame. From Eqs. (8), (15), (28), and (29), r is given as
Assuming ηc = 0.5, RA = 0.1, S = 0.06 mm2, L = 200 m, and PA = 10 dBm, r roughly becomes 10−10 for the above PSUN, thus, pIII = 0 and pMI = 0. It can be concluded that the interference with sunlight does not occur in FMCW LiDARs due to the weak intensity.
Let us compare the result with ToF LiDAR’s again. Due to direct detection of ToF, S and δλ in Eq. (30) are simply given by the LiDAR reception aperture (e.g., S ≈ 300 mm2 for 20 mmϕ aperture) and filter bandwidth (δλ = 20 nm), respectively. For ηSUN = 0.5 and SSI ≈ 0.3 µW/mm2/nm averaged from λ = 930 nm to 950 nm, PSUN is estimated to be ∼0 dBm. Although only a part of this PSUN is detected at one resolution point, its intensity easily exceeds the intensity of the signal light returned after the Lambertian scattering at a long distance. Therefore, pIII = 1, resulting in pMI = 0.29 /s for F = 10 frame/s. This large value of pMI is due to multicounting of resolution points that overlap with the sun, but even without multicounting, pMI is as large as 3.6 × 10−3/s. This means that ToF LiDAR is easily affected by sunlight.
5. Experiment
5.1 Device
Figure 7(a) and (b) shows the LiDAR chip with the SLG scanner and its schematic structure, respectively. It was fabricated on a 200-mm-diamter Si-on-insulator using Si photonics complementary metal-oxide-semiconductor process, including a stepper exposure. Reference [19] describes the details, and the chip size is 9.1 × 5.5 mm2. Transverse-electric polarized light from a laser is coupled to a spot size converter (SSC) on the chip from a polarization-maintaining core-shrunk fiber attached using an ultraviolet curing resin. A 2 × 2 coupler splits coupled light into reference and signal paths after passing through a ON/OFF switch (SW) that consists of a Mach–Zehnder interferometer equipped with a thermo-optic heater. Signal light is emitted into free space (Tx) from one SLG after passing through six serial SWs, one of which selects the incident direction on the SLG (left/right SW), and five of which selects one SLG (SW trees). The same SLG receives the reflected and returned light from objects (Rx). It is mixed with reference light, whose intensity is adjusted by an attenuator (ATT), and detected by germanium (Ge) balanced photodiodes (BPDs). These SWs and Ge BPD were electrically connected to a printed circuit board via wire bonding for external control and signal acquisition.
Figure 7(c) shows the measurement setup for the FMCW ranging. CW light from a tunable laser (Santec TSL-550, linewidth = 200 kHz, τc,A ≈ 1.6 µs) was input into lithium niobate (LiNbO3) in-phase/quadrature-phase (I-Q) modulator (Thorlabs LN-86-14-P-A-A). It was driven by sawtooth frequency-swept sinusoidal electrical signals of BA = 10 GHz and TA = 100 µs, which were generated by an arbitrary waveform generator (AWG, Keysight M8195A). The frequency-swept light was produced via the single sideband (SSB) modulation. An erbium-doped fiber amplifier (EDFA, Thorlabs 100P) amplified light to 10 dBm, which was then coupled into the chip. The emitted fan beam from the SLG was collimated by a prism lens [23] and irradiated on a target at a distance of ∼3 m. Trans-impedance amplifiers (TIAs) and a differential amplifier were used to amplify the ranging signal, which was observed using an electrical spectrum analyzer (ESA, Rohde & Schwarz FSW43).
5.2 Interference from another FMCW LiDAR
The behavior of LiDAR A affected by unsynchronized LiDAR B was experimentally investigated as a simple experimental simulation of the direct interference. Figure 8 shows the setup for LiDAR B, where FMCW light was generated using another laser (Santec TSL-570), I-Q modulator (EO SPACE IQ-0DKS-35-PFA-PFA-LB), AWG (Keysight M9502A), and EDFA (Alnair Labs CPA-100-CL) and emitted to the free space through a fiber collimator (Thorlabs F810FC-1550). In the adjustment process, the FMCW light of LiDAR B was first split into two. One was input into the fiber of LiDAR A, and the other was input into the fiber collimator, whose angle was adjusted so that the emitted beam was well coupled to the SLG of LiDAR A and the beat intensity in LiDAR A was maximized. The maximum intensity was almost the same as the ranging signal intensity for a target object when the FMCW light of LiDAR A is input into LiDAR A. This was much lower than that expected in Section 3.3 for the direct interference. This low interference intensity was due to the slight offset of the fiber collimator from the optical axis of the beam of LiDAR A. However, it was still sufficient for the investigation of the interference qualitative behaviors. After the adjustment, the setup was returned to that described in Fig. 8. Two LiDARs were driven independently, with default parameter settings, such as λ = 1,539 nm, Bx = 1 GHz, and Tx = 4 µs (these Bx and Tx were chosen particularly in this experiment due to limited spec of the AWG for LiDAR B). These parameters were then slightly changed for LiDAR B to simulate situations where pI = 1 and pII and r are changed.
Figure 9(a) shows examples of the measured beat spectra. Here, the spectra were observed under max-hold condition in ESA because the interference between unsynchronized LiDARs occurs at various frequencies and timings and always changes (Fig. 3). The ranging signal peak for the target appeared around 5.5 MHz. On the other hand, similar peaks due to the interference did not appear, while the noise floor increased, as shown in the right panel of Fig. 9(a). This is thought to exhibit spectral broadening due to asynchronization, causing random contamination timing, slight difference of Tx, and jitter between devices. Figure 9(b) plots the increase of the noise floor after taking the difference in the intensity with and without the interference and the average around the target beat frequency as a function of the wavelength difference Δλ = λB – λA. The gray region shows the range in which the noise floor increased by more than 1 dB. The wavelength range δλ is 9 pm, approximately corresponding to BA = 1 GHz. This is a reasonable result because a small wavelength shift out of this range suddenly increases the beat frequency beyond the measurement bandwidth. Random contamination timing may have caused the shift of the maximum point from Δλ = 0. Figure 9(c) and (d) shows the interference intensity for the sweep bandwidth ratio BB/BA and the sweep period ratio TB/TA, respectively. The noise floor also increased when BB/BA and TB/TA became close to 1; however, the increase was limited for BB/BA < ±1% and TB/TA < ±0.05%. It is known from Eq. (23) that r decreases by approximately 5 dB when BB/BA increases from 1.002 to 1.006 in this experiment condition, which was well agrees with the result shown in Fig. 9(c).
5.3 Pseudosunlight contamination
A situation was simulated, where pI = 1, pII = 1, and r is changed, and the pseudosunlight contamination was investigated experimentally. In addition to the measurement setup displayed in Fig. 7(c), pseudosunlight from a solar simulator (Asahi Spectra HAL-320W) was irradiated on the LiDAR chip from a distance of approximately 20 cm, as shown in Fig. 10. This solar simulator emits a spectrum approximating sunlight of air mass 1.5 G up to λ = 1,800 nm.
Figure 11(a) shows the measured ranging spectra for a retroreflective sheet as a target. Figure 11(b) plots the intensity and noise floor with the irradiation intensity of the pseudosunlight ISUN. 1SUN = 1,000 W/m2 is a standard irradiation intensity in summer, which is obtained by integrating the spectrum in Fig. 6(a) over the entire wavelength range. The signal intensity decreased by maximally 12 dB, and the noise floor slightly fluctuated within < 5 dB with increasing ISUN. However, as expected in Section 4, the obvious increase of the noise floor, similar to that for the interference of another LiDAR, was not observed. This fluctuation may be neglected practically. On the other hand, the pseudosunlight heated the SLG; the temperature measured by a thermistor (10 kΩ at 25°C) rose to 46℃ and 73℃ for 1SUN and 2SUN, respectively. This heating caused the emitted beam to be shifted by 0.04° and 0.16°, respectively, as shown in Fig. 11(d) and (e). The sensitivity of the SLG to the heating power was evaluated independently to be ∼10°/W. These shifts correspond to the heating of 4 mW and 16 mW for 1SUN and 2SUN, respectively. The decrease of the signal intensity might be due to the shift of the irradiation point, resulting in the slight change of the reflection condition on the target surface.
Since we usually target angular resolution of < 0.1°, the beam shift of around 0.1° caused by heating leads to a shift of one resolution point. However, such heating can be suppressed by filtering useless wavelength components.
6. Conclusion
This study discussed the ambient light immunity of FMCW LiDAR by constructing a theory of the interference from another LiDAR and from sunlight. For usual parameters of the developed Si photonics SLG FMCW LiDAR, the probability of the interference from another FMCW LiDAR is estimated to be 2.3 × 10−11/s, which does not barely satisfy the safety standard for autonomous vehicles. However, increasing the number of resolution points to three times larger and/or employing parallel LiDARs can satisfy the standard. On the other hand, sunlight does not affect FMCW LiDARs with respect to the signal interference due to the incoherency of sunlight.
Compared with FMCW, the probability of interference between ToF LiDARs is 3–4 orders higher mainly due to the much lower spectral selectivity. Moreover, since ToF uses the direct detection, sunlight can be coupled into the LiDAR more efficiently and its interference probability can be as high as 3.6 × 10−3/s even without multicounting the resolution points overlapping with sunlight. These estimations may vary depending on the configuration and settings of LiDAR, but in any case they show that FMCW LiDAR is much more immune to ambient light than ToF LiDAR.
In the experiment, the interference from another LiDAR appears as an increase in noise floor but disappears with a slight difference in the used wavelength and frequency sweep rate. However, the interference will not be suppressed when light from another LiDAR is contaminated directly with close wavelengths. For the pseudosunlight, signal interference is negligible, but the beam shift occurs due to heating, which should be suppressed by filtering useless wavelength components.
Funding
Japan Society for the Promotion of Science (22H00299); New Energy and Industrial Technology Development Organization (JPNP14004).
Acknowledgments
The authors would like to thank Prof. Ozeki from University of Tokyo for his technical advice.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time, but may be obtained from the authors upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. J. Hecht, “Lidar for Self-Driving Cars,” Opt. Photonics News 29(1), 26–33 (2018). [CrossRef]
2. C. Pulikkaseril and S. Lam, “Laser eyes for driverless cars: the road to automotive LIDAR,” OFC, Tu3D (2019).
3. D. J. Yeong, G. V. Hernandez, J. Barry, et al., “Sensor and Sensor Fusion Technology in Autonomous Vehicles: A Review,” Sensors 21(6), 2140 (2021). [CrossRef]
4. C. V. Poulton, M. J. Byrd, P. Russo, et al., “Coherent LiDAR with an 8,192-Element Optical Phased Array and Driving Laser,” IEEE J. Select. Topics Quantum Electron. 28(5), 1–8 (2022). [CrossRef]
5. C. Rogers, A. Y. Piggott, D. J. Thomson, et al., “A universal 3D imaging sensor on a silicon photonics platform,” Nature 590(7845), 256–261 (2021). [CrossRef]
6. A. Martin, D. Dodane, L. Leviandier, et al., “Photonic Integrated Circuit-Based FMCW Coherent LiDAR,” J. Lightwave Technol. 36(19), 4640–4645 (2018). [CrossRef]
7. X. Zhang, K. Kwon, J. Henriksson, et al., “A large-scale microelectromechanical-systems-based silicon photonics LiDAR,” Nature 603(7900), 253–258 (2022). [CrossRef]
8. C. Li, K. Wu, X. Cao, et al., “Monolithic coherent LABS lidar based on an integrated transceiver array,” Opt. Lett. 47(11), 2907–2910 (2022). [CrossRef]
9. J. Riemensberger, A. Lukashchuk, M. Karpov, et al., “Massively parallel coherent laser ranging using a solution microcomb,” Nature 581(7807), 164–170 (2020). [CrossRef]
10. S. Suyama, H. Ito, R. Kurahashi, et al., “Doppler velocimeter and vibrometer FMCW LiDAR with Si photonic crystal beam scanner,” Opt. Express 29(19), 30727–30734 (2021). [CrossRef]
11. G. M. Brooker, “Mutual Interference of Millimeter-Wave Radar Systems,” IEEE Trans. Electromagn. Compat. 49(1), 170–181 (2007). [CrossRef]
12. T.-N. Luo, C.-H. E. Wu, and Y.-J. E. Chen, “A 77-GHz CMOS Automotive Radar Transceiver With Anti-Interference Function,” IEEE Trans. Circuits Syst. I 60(12), 3247–3255 (2013). [CrossRef]
13. I.-P. Hwang, S.-J. Yun, and C.-H. Lee, “Mutual interferences in frequency-modulated continuous-wave (FMCW) LiDARs,” Optik 220, 165109 (2020). [CrossRef]
14. P. A. M. Sandborn, “FMCW Lidar: Scaling to the Chip-Level and Improving Phase-Noise-Limited Performance,” Doctoral Dissertation, Univ. California Berkeley (2017).
15. EASA CS-25 Amendment 24, “ https://www.easa.europa.eu/en/document-library/certification-specifications/cs-25-amendment-24,” accessed on Sep. 7, 2023.
16. WADA ENGINEERING, “https://www.wadae.jp/recruit/member/index.html,” accessed on Sep. 7, 2023.
17. H. Ito, Y. Kusunoki, J. Maeda, et al., “Wide beam steering by slow-light waveguide gratings and a prism lens,” Optica 7(1), 47–52 (2020). [CrossRef]
18. T. Tamanuki, H. Ito, and T. Baba, “Thermo-Optic Beam Scanner Employing Silicon Photonic Crystal Slow-Light Waveguides,” J. Lightwave Technol. 39(4), 904–911 (2021). [CrossRef]
19. T. Baba, T. Tamanuki, H. Ito, et al., “Silicon Photonics FMCW LiDAR Chip With a Slow-Light Grating Beam Scanner,” IEEE J. Select. Topics Quantum Electron. 28(5), 1–8 (2022). [CrossRef]
20. T. Kim, “Realization of Integrated Coherent LiDAR,” Doctoral Dissertation, Univ. California Berkeley (2020).
21. K. Sayyah, R. Sarkissian, P. Patterson, et al., “Fully Integrated FMCW LiDAR Optical Engine on a Single Silicon Chip,” J. Lightwave Technol. 40(9), 2763–2772 (2022). [CrossRef]
22. L. Wald, Basics in Solar Radiation at Earth Surface,https://hal-mines-paristech.archives-ouvertes.fr/hal-01676634, accessed on Oct. 23, 2023.
23. R. Kubota, M. Kamata, R. Tetsuya, et al., “High NA and Size Reduction in Prism Lens for Silicon Photonics SLG Beam Scanner,” CLEO/PR, CWP17A (2022).
24. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics Part 1: Optics, 3rd edition (Wiley,2019), Chap. 12.
25. NREL, “https://www.nrel.gov/grid/solar-resource/spectra-am1.5.html,” accessed on Oct. 23, 2023.
26. M. Born and E. Wolf, Principles of Optics: Electromagnetic Theory of Propagation, Interference and Diffraction of Light, 7th (expanded) edition (Cambridge University, 1999), Chap. 10.
27. E. Hecht, Optics, 2nd edition (Addison Wesley, 1987), Chap. 12.