Overview: Silicon-on-chip LiDAR technology

The development of silicon-based optoelectronic technology can integrate discrete active and passive devices in the lidar system's transmitting and receiving modules onto a chip, making the lidar smaller, more stable, and less expensive, thus promoting the application of lidar in areas such as autonomous driving.

According to MEMS Consulting, Chen Xiaolin's team from the Southwest Institute of Technical Physics conducted a review and analysis of the development of this field, including the basic concepts of LiDAR and the ranging principles of common LiDARs, analyzed the scanning schemes of common silicon-based LiDAR systems, and discussed the current challenges and development directions of silicon-based LiDARs. The relevant research content was published in the journal Laser and Optoelectronics Progress under the title "A Review of Silicon-based LiDAR Technology".

Basic concepts of LiDAR

Laser wavelength

Taking into account the atmospheric window, eye safety, and available lasers and photodetectors, the wavelength used by lidar is usually 0.8~1.55 μm, and the corresponding laser and detector types are shown in Table 1. Since the ambient temperature that the vehicle-mounted lidar may face during operation has a large range of variation, when a filter is used in the system to suppress background light interference, the output wavelength of the laser should always remain within the passband of the filter when it changes under the influence of temperature. This is also a problem that needs to be considered when selecting a light source.

Table 1 Laser and detector types corresponding to lidar wavelengths

The current mainstream wavelengths of LiDAR are 905 nm and 1550 nm. Pulse LiDAR uses 905 nm lasers. The main advantage is that cheaper silicon-based detectors can be used, and 905 nm is less hydrophilic than 1550 nm, resulting in less light loss. However, due to the transparent window of silicon material, it cannot be used in silicon photonic systems. Because 905 nm lasers can penetrate the vitreous body of the human eye to reach the sensitive retina, its peak power is limited to avoid retinal damage. 1550 nm is suitable for long-distance continuous wave LiDAR systems. Since the light is absorbed in the front half of the human eye, it will not harm the retina, and the laser output power can be greater. This wavelength is commonly used in communication equipment, has rich technical reserves, and continuous wave laser sources can be obtained at low cost.

Detection distance

The detection distance refers to the maximum distance at which the laser radar can detect the target, which is mainly restricted by the laser signal transmission power. For pulse laser radar, the detection distance is also related to the signal repetition period T, because the correct target distance information can only be calculated when the echo signal returns to the receiving system within time T, and the echo signal returned beyond this time period cannot be distinguished from the subsequent echo, resulting in distance ambiguity. For frequency modulated continuous wave (FMCW) laser radar, the detection distance is also affected by the laser line width Δν. The narrower Δν, the longer the coherence length Lc of the laser signal, and for targets outside the coherence length, the echo signal-to-noise ratio will be greatly reduced and difficult to be recognized by the system. This parameter is generally obtained by the laser radar's maximum detection distance for 10% low reflectivity targets. In actual applications, due to changes in the environment and target surface conditions, the value is not absolute.

Field of view

The field of view (FOV) refers to the area that the LiDAR can detect, and is usually expressed in degrees. For automotive applications, the LiDAR field of view must include the horizontal field of view (HFOV) and the vertical field of view (VFOV). The larger the field of view, the wider the LiDAR's angular coverage of the space and the greater its perception range of the surrounding environment.

Measurement accuracy and measurement resolution

Measurement accuracy is a combination of precision and accuracy. Precision refers to the consistency of distance values ​​measured by the LiDAR under the same conditions, while accuracy refers to the closeness between the mean of the distance distribution measured by the LiDAR and the true distance. The two are affected by random errors and systematic errors in the measurement process, respectively, as shown in Figure 1. Among them, the accuracy of distance measurement is mainly affected by the generation of optical signals and the measurement system, while the accuracy of angle measurement is mainly affected by the accuracy of the laser steering scan at the transmitting end.

Figure 1 LiDAR Precision and Accuracy

The measurement resolution specifically includes distance resolution and angular resolution. Distance resolution refers to the minimum distance at which a target can be distinguished in a single laser radar measurement. For pulse laser radar, the narrower the pulse width, the higher the distance resolution; for FMCW laser radar, the distance resolution is inversely proportional to the modulation bandwidth and is affected by the frequency modulation linearity. Angular resolution refers to the minimum angle at which a target can be distinguished in a single laser radar measurement. Increasing the transmit aperture is an effective means to improve angular resolution.

The ranging principle of silicon-based laser radar

Pulse time-of-flight method

The pulse time-of-flight method (TOF) uses direct detection to measure the target distance by measuring the time it takes for the laser signal to be reflected by the target and collected by the detection system. The ranging principle and system design of pulse laser radar are relatively simple, as shown in Figure 2.

Figure 2 Working principle of pulsed laser radar

Pulse lidar has an ambiguous distance because there is uncertainty as to whether the echo signal is offset by one or more cycles relative to the transmitted signal. In addition, due to the scattering loss of optical pulse energy on the link from transmitter to receiver, the signal-to-noise ratio is also the main factor limiting the detection range of pulse lidar. In order to increase the effective range, the pulse transmission power needs to be increased, and for autonomous driving applications, the eye-safe power limit needs to be considered. One method is to use pulse trains to reduce the high power required for a single pulse, and to improve the signal-to-noise ratio and accuracy by integrating and averaging the received power. Despite these limitations, the simple detection principle and implementation of pulse lidar make it highly competitive.

AM Continuous Wave Ranging

The amplitude modulated continuous wave (AMCW) ranging method is also called the indirect time-of-flight method. In the AM CW lidar, the laser is amplitude modulated before it is emitted, and the modulation period is greater than the round-trip flight time, and the echo signal is compared with the emission signal. For the AM CW measurement method, the distance resolution is determined by the ranging signal frequency and the phase meter resolution, and the distance resolution increases with the increase of the AM signal frequency. Like the pulsed lidar, the phase of the echo signal of the AM CW lidar begins to repeat after a 2π phase shift. To avoid the problem of multiple solutions, the ranging range will be reduced accordingly. One solution is to select a modulated light wave with a high modulation frequency as the basic scale, and then introduce one or more modulated light waves with a lower modulation frequency as auxiliary scales, and combine the measurement results of each scale to obtain an accurate measurement value.

Random Modulation Continuous Wave Ranging Method

The random modulated continuous wave (RMCW) ranging method modulates the pseudo-random bit sequence (PRBS) onto the amplitude or phase of the outgoing laser, and obtains the flight time of the received light by calculating the correlation between the received return laser signal and the original template of the PRBS using methods such as matched filters, as shown in Figure 3. Since the PRBS is only correlated with itself, the RMCW lidar is insensitive to sunlight, lights, and light from other lidars. However, this technology is sensitive to relative velocity, laser phase noise, and speckle, which is a major technical challenge. In the field of lidar applications, Baraja, an Australian startup lidar company, is a typical representative of RMCW technology. The company uses RMCW technology combined with a unique prism dispersion spectrum scanning technology to develop a vehicle-grade radar model called Spectrum HD.

Figure 3 Schematic diagram of RMCW lidar system

FMCW laser ranging

FMCW laser radar uses a modulated signal whose frequency changes periodically with time for detection. The returned light signal is coherent with the local oscillator light signal. The distance of the target object can be measured according to the frequency of the intermediate frequency signal generated by mixing. Its signal modulation forms include triangular wave, sawtooth wave, sine wave, etc. Among them, the sine wave modulated signal needs to adjust the signal frequency deviation when detecting an object, so it is mostly used in situations where there is only one detection target. For the detection needs of multiple targets in autonomous driving applications, triangular waves or sawtooth waves are generally used. Triangular waves can obtain the distance and speed information of the object at the same time, while sawtooth waves are mainly used to measure the distance of the object. For triangular wave detection, when the target moves, the signal undergoes a Doppler frequency shift, and the frequency difference between the reflected signal and the local oscillator signal has different beat frequencies in the rising and falling sections of the linear frequency modulation, as shown in Figure 4.