UST030

86 to be particularly fast, and may be easier to implement on-chip. These beam-steering devices are aiming for a 150-300 m range for automotive applications, with a 5 cm resolution at 200 m. A key factor is how narrow the beam is. System designers are looking for angular resolution below 0.1 º , so the beam will have a spread of 50 cm at 200 m. Current devices with 100 antennas provide an angular resolution of 0.3-0.4 º . For a 5 cm resolution at 200 m, the beam needs an angular resolution of 0.01 º , and that requires thousands of antennas on a pitch of 1-2 microns, which will deliver an output (an aperture) of 1 cm. Current developments include chips with 5000-8000 antennas with an angular resolution of 0.05 º to provide a 25 cm beam at 200 m. Receiver architecture An important choice for the Lidar sensor is that of the wavelength. Most Lidars these days are based on the ToF principle, and operate in the near infrared (NIR) range with 850 or 905 nm laser diodes. This is selected because of the receivers. Sensitive APDs and single photon avalanche photodiodes (SPADs) are readily available, based on silicon fabrication processes that provide a high signal-to-noise ratio (SNR). A downside for the NIR wavelengths though is that the maximum permissible energy (MPE) level of the laser is constrained. That means nanosecond pulses from a high-power laser are needed but will ultimately limit range. However, these frequencies are eye-safe – a critical requirement for ground-based driverless car systems. The other choice is to operate in the 1550 nm range. This wavelength has an MPE level that is orders of magnitude higher but requires the use of non-silicon receivers. These tend to use the FMCW signal processing approach. It provides the velocity of the object detected, has a higher SNR, lower power consumption and less susceptibility to interference. However, it is harder to implement because it requires a highly coherent stable and tuneable laser as well as a coherent optical mixer, as a continuous light wave is used. The choice of wavelength also affects the choice of the laser source. For NIR, relatively low-cost gallium arsenide and vertical cavity surface emitting laser (VCSEL) sources are available. Both technologies can achieve hundreds of watts of peak power by stacking or implementing a VCSEL array. Flash architectures A flash sensor uses an array of VCSELs, usually operating at 905 nm. The latest flash-based Lidar sensors use full-waveform (FW) processing, which is fairly common in high-end sensors in low-cost, short-range devices. Processing the entire waveform provides an effective resolution comparable to a 16-mirror system but at much lower processing cost. The optics for the receiver are key. Good optical design will allow lower cross-talk and improve the overall system, and requires a high-speed transimpedance amplifier (TIA) and 12-bit analogue-to-digital converter (ADC). The processing can go beyond detecting a single photon. Sub-pixel processing provides 512 points within each 3D pixel in space, known as a voxel, with each pixel being essentially a returning photon. The output of such sensors is not a point cloud. As the entire scene is illuminated, the equivalent of a point in the point cloud is the collection of 512 samples from each photon. Current flash sensors use a processing chip that supports a pulse from the laser array of less than 10 ns. This gives a 23 m range for a target with 10% reflectivity such as a pedestrian, with a 180 x 16 º FoV, or a 50 m range for a 50% reflectivity object. This can be used for autonomous commercial delivery vehicles to provide a cocoon around the vehicle to cover dead zones and blind spots. The next- generation designs with higher power laser arrays will boost the range to 75 m for a 10% reflectivity object by the end of this year. Faster processing chips that are now sampling will allow for a hybrid Lidar sensor. This will provide a range of up to 200 m by using a 1D micro-mirror to direct the flash array. The micro-mirror is used to orient and direct the light back to the receiver for processing with a faster TIA, and ADCs in sensors and reference designs are expected in 2021. A different solid-state approach February/March 2020 | Unmanned Systems Technology A flash Lidar sensor with specialist processing can be used to create a cocoon of sensing around a vehicle (Courtesy of LeddarTech) Focus | Lidar sense & avoid