05/17/2019 | News release | Distributed by Public on 05/17/2019 15:24
By Ergun Canoglu, Ph.D. on May 17, 2019 | Leave a Comment
Part 3: Introduction to Coherent Lidar
Lidar used in today's autonomous driving has its roots in laser/radar guns used by traffic law enforcement for speed limit compliance. Most laser speed guns emit a single beam and measure the reflection time of the laser pulse (Time-of-Flight, TOF, measurement) from a vehicle at a distance. An advanced laser gun operation is similar to a Doppler radar in which the frequency shift of the reflected laser beam is measured. In this method, a continuous laser beam is sent to the target and frequency shift of the reflected beam is measured to identify a target vehicle's distance and speed. This is also the basis for coherent lidars.
Pulsed laser measurements operate similar to AM radios in which the pulse amplitude and timing are used to determine distance, whereas Doppler laser measurements are similar to FM radio operation in which the transmitting laser frequency is modulated and the reflected laser beam's frequency is compared through coherent mixing of the transmitted frequency (also called a local oscillator) and reflected laser signals. By mixing the reflected signal with an optical local oscillator, the full field can be detected in a Doppler laser measurement, including both phase and amplitude, as compared to pulse laser measurements, where only the intensity (amplitude) of the reflected laser signal is measured. In a coherent lidar, target range is obtained through the frequency difference of the local oscillator and return signal, enabling high fidelity signal detection. Measuring the phase of the return signal through Doppler shift enables the simultaneous measurement of velocity and range.
The primary difference between laser speed guns and lidars for autonomous driving is that lidars provide distance/speed measurement over a 2D scene as opposed to a single point distance/speed measurement from a laser speed gun. By combining distance measurement at each point in the 2D scene, lidars provide a 3D image of the scene. Lidars achieve this either by illuminating a 2D scene with a powerful laser and taking a picture of the scene with high speed cameras (flash lidar) or scan a single or multi laser beams (beam steering lidars) over the 2D scene.
Laser Eye Safety
Due to eye safety concerns, the maximum output power of lasers is limited and regulated.Thus both flash and beam steering lidar solutions need to keep laser energy/output power below eye safety limits defined by the regulations. This impacts selection of laser wavelength, its operating mode (pulsed or continuous) and drives the selection of detection method.
Figure 1. (a). Maximum Permissible Energy for pulsed laser operation (b) Maximum allowed optical power for continuous wave (CW) laser operation.
Lasers operating in shorter wavelengths in near-infrared (NIR) regions have lower output power/energy limits due to the fact that human eye focuses shorter NIR wavelengths onto retina thus concentrating the laser radiation onto a small point. Longer laser wavelengths in NIR however are absorbed in the cornea and have higher output power/energy limits. For example in a 1 nsec laser pulse, laser safety limit for 1550 nm is 1,000,000 times higher than a laser operating at 905 nm. Similarly, for the continuous laser operating mode, Class-1 laser safety limit is 10x higher for 1550 nm than 905 nm operating. Thus operating in 1550 nm wavelength range is preferred for long reach applications due to eye safety considerations.
Optical received power from a lidar target not only depends on the laser output power but also the 2D scene illumination method. In the flash lidar schemes where a 2D scene illuminated all at once, the received power is decreased much faster (1/R4, R: distance) as a function of distance, whereas in beam steering, lidar optical power is decreased by the square of the distance (1/R2). Thus beam steering lidar solutions are better suited for long range applications.
Received Signal & Detection Method
Electrical signal level derived from photodetectors at the lidar receiver depends on the photodetector type and detection method.
In direct detection methods utilized in pulsed lidar configuration, the electrical signal level is proportional to the detector sensitivity and gain (PIN detector gain=1, APD gain=50-100). On the other hand, in coherent detection, the electrical signal level is proportional to the local oscillator power and can provide signal gains > 109 when compared to direct detection methods utilizing PIN detectors. Because of this large signal gain in coherent detection, transmitter power levels can be small, allowing the use of low power semiconductor lasers that are readily available from high volume telecom industry.
It is worth noting that a new class of single-photon direct-detection photodetectors are being developed for sensitive pulsed detection (SPAD). These detectors operate near avalanche break down voltage and can provide similar gain to coherent detection. However, these detectors suffer from false detection due to thermal noise of the detector itself or photons from ambient light or other lidars that happen to be at the detector wavelength. Therefore, SPADs are often used in a statistical architecture where many SPADs used in parallel to measure time of flight of the transmitted beam and not fooled by random photon detection.
Thus coherent detection by far is the simplest and most sensitive technique for lidar receivers while being practically immune to interference from ambient light sources and other lidars. Despite all the advantages of the coherent detection method, it was not used in the first generation lidars being sold today primary because of inavailability of integrated coherent mixing components that enable this sensitive detection method. Initial demonstration of the technology by startups have relied on discrete components developed for telecom applications.
Why Coherent Lidar?
Coherent lidar devices use low phase noise laser sources (narrow linewidth lasers) and relatively low speed PIN receivers to achieve 200m-400m detection range that are already in high volume production in the telecom industry. In addition to providing long range distance measurements, being able simultaneously measure speed and distance at the same time has been advertised as the key advantages of the initial systems despite the size and cost. Finally, coherent lidar technology is inherently more accurate in measuring distance and using today's low speed electronics and lasers, can enable sub-millimeter accuracy where as pulsed lidar technology requires state-of-art high speed and high power electronics to reach to 2-10cm accuracy. This allows lidar functionality to be fine tuned for specific ranges and go beyond capabilities cameras for short range and provide improved object and gesture recognition under all light conditions.
The current pulsed technology provides say 100-130 meters of visibility, which when considering reaction time is not enough to assist an operator in stopping a car going 50 MPH on an icy road where breaking distance is ~8-10 x longer. Furthermore, traveling safely at freeway speeds requires long breaking distances, and with up to 400 meters of visibility provided by coherent lidar, the roads just got much safer.
Figure 2. Stopping distances of an average automobile for snow/ice road conditions.
The economics of sensor instrumentation in automobiles dictates that until sensor cost is few orders of magnitude lower than the vehicle cost, it will be not have mass deployment and its use will be limited to specialty applications. In addition to cost, sensor size is another key feature for high-level adoption.
Some of today's lidars cost as much as the vehicle itself and require substantial modification externally to the vehicle to accommodate bulky sensors.
Figure 3. The economics of automotive lidar devices require optical hardware to be smaller, low-cost and high-volume capable. This can only be achieved through chip-scale integration of coherent optical components.
As seen in many other areas such as radar, cost and size reduction can be achieved through chip scale sensors that can be manufactured in high volumes. Initial radars were based on discrete/expensive high speed components and radar instruments could only be placed on top of the vehicle or by extending the front bumper. Today one can make a single chip radar for few tens of dollars and due to their small size, can be better integrated into vehicle design without noticeable external modifications.
Chip-scale integration for lidar that includes lasers, receivers and possibly beam steering can only be achieved through photonic integration. This technology is used today for telecom applications and has reduced cost and size of optical transponders by orders of a magnitude in the last 10 years.
Chip-scale coherent lidar Photonic Integrated Circuits can provide similar cost and size impact while enabling > 2 x range improvement and provide additional information such as velocity that can be used for improved image segmentation and object classification.