FMCW vs ToF LiDAR: Leading the Autonomous Vehicle Revolution

Developments in LiDAR for autonomy accelerated a decade ago as the excitement of driverless cars built up, with rosy projections of revenues, lower number of accidents and giving passengers back the luxury of time as they commuted to work. Significant funding was committed to the endeavor with about 70 independent companies getting billions of dollars in funding.

Table 1 is one way of understanding the LiDAR landscape. It groups companies across operating wavelengths, how the FOV is addressed and the LIDAR physics (ToF and FMCW).

Notably, there are no companies doing Flash LiDAR anymore (the few that were have wound up operations). Also, there are no companies doing FMCW LiDAR at the 8XX-9XX nm wavelengths. The one company doing 15XX nm scanning LiDAR using Geiger Mode Photon Counting detectors was Argo (which purchased Princeton Lightwave in 2017), which has wound up operations. The dominant concentration of companies are in Cell ❶, with about ~50 companies originally funded (many have gone bankrupt at this point). Today, many of the actual design-in successes for vehicle autonomy are in this bucket (Hesai, Valeo, Robosense, Innoviz). Ouster has been successful as well (part of ❹), but primarily in non-automotive applications. On the FMCW front (❷), ~10 companies were funded with players like Mobileye and Baraja winding down their operations. Today, the active players in the automotive space include Aurora (which purchased Blackmore, a pioneer in FMCW LiDAR), Aeva, Analog Photonics and Voyant (primarily for nonautomotive applications). Cell ❸ includes Luminar, an important player with successful automotive design-ins.

ToF vs FMCW

Table 2 presents the relative advantages and disadvantages of the 2 approaches.

1.  ToF LiDAR is more mature and lower cost than FMCW LIDAR today, with all of the vehicle design-ins to date occurring with ToF LIDARs (No. 1, 2)


2.  Performance of FMCW LiDAR is superior in terms of range, axial velocity measurement and lower latency in image segmentation computing time (No. 3, 4, 5). It also performs better in fog and dust, primarily because of the higher operating wavelength (No. 6).


3. Laser simplicity is an advantage for ToF LiDAR with reasonable availability of VCSELS and EEL diode lasers (No. 7,  8). FMCW LiDAR requires high power, single mode coherent lasers, a capability that is only possible with expensive and bulky fiber lasers today.

4. 
   
   
   
   4. Simpler PIN based detectors can be used with FMCW LIDAR (No. 9) because part of the outgoing laser power is tapped out as a local oscillator and provides signal gain. ToF LiDARs require more complicated Avalanche Photo Diodes (APDs)
   
   
   5. FMCW LiDAR holds the promise of chip scale LiDAR (No. 10, 11, 12) with diode lasers, detectors and passive components like waveguides and splitters integrated into a single silicon photonics chip, with non-mechanical optical scanning techniques like Optical Phased Array (OPA) or switching (optical or MEMs) techniques used to transmit-receive through waveguides without any mechanical scanning. There are challenges to achieving this although Analog Photonics has advertised significant progress in this area. Voyant Photonics also announced chip scale FMCW LiDAR recently and advertise it for sale now on their website.
   
   
   
   

Sabbir Rangwala provides consulting services in perception, sensing, LiDAR and AI for enabling the Autonomy of Things (AoT™) revolution (https://autonomyofthings.co/)

He is also a Senior Contributor on Transportation and Innovation at Forbes.com (https://www.forbes.com/sites/sabbirrangwala/)