Key Design Considerations for LIDAR

Successful self-driving cars will inevitably use tightly integrated sensor systems to reach or even exceed human driving capabilities. Human drivers generally use their eyes, ears, and feedback from the vehicle's motion to drive the car. Our brains process all this information in real time and react intuitively from our database of driving experience. The sensors needed to replicate human driving capabilities include radar, laser radar (LIDAR), cameras, inertial measurement units (IMU), and ultrasonic sensors. Each system has its advantages and disadvantages. The accuracy and performance of a single sensor are not enough to replace all other sensors. Multi-sensor fusion is a major trend that complements each other. This article will introduce the main design considerations related to LIDAR, a sensor that provides a large amount of data for various autonomous driving solutions.

Figure 1. Spider diagram comparing vision, radar, and LIDAR.

In self-driving cars, LIDAR works closely with radar. Both technologies operate without producing visible light, which is essential for nighttime driving or low-light conditions. While radar is good for long-range detection and tracking, LIDAR provides higher angular resolution, which can identify objects and classify them. In other words, radar is good for detecting the presence of an object, while LIDAR is able to provide specific information about the object based on what radar has detected.

Figure 2. LIDAR perception for autonomous vehicles.

There are technical challenges when designing a LIDAR system, one of the main challenges is keeping the near-infrared wavelengths below eye safety limits. For these safety guidelines, please refer to IEC 60825-1. This is not to diminish the importance of eye safety, all aspects discussed in this article will ultimately affect eye safety. There are many different LIDAR system technologies, with varying degrees of design complexity, and each has its own advantages and disadvantages.

Importantly, all designs share the same fundamental concerns. Here we focus on other aspects that impact system design beyond eye safety, including: SNR maximization, minimum detectable requirements, field of view, thermal dissipation, power consumption, and dead reckoning.

Looking at the receive chain, the system's signal-to-noise ratio (SNR) impacts the ability to detect small targets at long ranges (100m to 300m). The ADC noise floor cannot exceed other noise sources in the receive path. If background light or shot noise contributions are lower than the ADC's noise floor or printed circuit board (PCB) noise, the system accuracy will be limited. Using direct time-of-flight (ToF) methods requires that the system can output short pulses (~1 ns to 5 ns) and detect these pulses using a high sampling rate ADC. Sampling rates of up to 1 GSPS meet the receive signal chain requirements. Also, note that the effective number of bits (ENOB) of the ADC must support the entire output range of the transimpedance amplifier (TIA) without clipping the signal.

Does the system need to detect a basketball 100 meters away? Determining the reflectivity, size, and distance of the object of interest will determine the SNR requirements of the TIA. The TIA needs to detect the same narrow pulses as the ADC. Because the system needs to handle a wide range of distances, reflectivities, and sizes of objects, the TIA must be able to recover quickly after saturation. Highly reflective objects (such as traffic signs) or close-range objects can reflect strong light and cause the TIA to saturate. These are common events, and the speed with which the system recovers (to minimize blinding time) is critical to safety.

Figure 3. LIDAR electrical architecture.

The field of view and angular resolution of a system also affect the ability to detect a basketball. Transmit and receive optics are the primary factors in determining field of view. Angular resolution determines whether you can detect and classify a basketball-sized object at a distance or just detect its presence.

For LIDAR system designers, dealing with power consumption and heat dissipation in these systems is a challenge. Of course, reducing the power consumption of the signal chain will correspondingly reduce the heat generated. Components can have large performance changes with temperature, and some of the more sensitive components may require temperature compensation. Using a thermoelectric controller is a good way to cool or heat ICs with high accuracy. If you are looking for precision, both the light emitting and sensing diodes need temperature compensation to maintain a stable operating wavelength and efficiency over the operating temperature range of the LIDAR system.

In some cases, avalanche photodiodes and lasers are biased to hundreds of volts (positive or negative). Generating these voltages efficiently and using as few components as possible is a best design practice. To provide an accurate reference voltage source, a precision digital-to-analog converter (DAC) is required to generate the bias point, current, and voltage. Along with the traditional 1.8 V to 12 V voltage domain, the voltage requirements of LIDAR systems have increased. Careful selection of power supply solutions can handle this, especially when an additional voltage is added to the solution. It is also important to select ICs and power supplies that have shutdown or low-power modes so that the system has the flexibility to poll multiple channels efficiently.

IMUs with integrated LIDAR sensors offer multiple advantages. IMU sensors intelligently fuse multi-axis gyroscopes and accelerometers to provide reliable position and motion identification for vibration and navigation applications. Precision micro-electromechanical systems (MEMS) IMUs provide the required accuracy even in complex operating environments and when faced with extreme motion dynamics.

The IMU provides dead reckoning, positioning, and stabilization capabilities to the autonomous driving system. In turn, these capabilities provide reliable data to the system when ADAS or GPS performance degrades or is unavailable. The IMU is able to effectively utilize high update rates (thousands of samples per second) and is not affected by external environmental changes. The more stable the IMU, the longer it can provide the system with critical and reliable track information.

IMUs can be integrated directly into LIDAR modules to detect, analyze, and correct for vibrations common in the vehicle's operating environment. For example, IMU output can assist in stitching together LIDAR point clouds that would otherwise be deviated as the vehicle travels over a pothole. In addition, IMUs can be used to detect bearing wear in rotating LIDAR systems to repair the LIDAR before actual failure occurs, improving safety.

in conclusion

During the initial product definition, the complexity of the LIDAR system needs to be considered to determine the acceptable SNR, detection requirements, field of view, thermal limitations, and power consumption. Understanding which components are the main contributors to each problem and carefully selecting the IC can greatly increase the chances of a successful design.

About the Author

Sarven Ipek joined Analog Devices in 2006. During his tenure at Analog Devices, Sarven has gained extensive experience in failure analysis, design, characterization, product engineering, project management, and program management. Sarven is currently the Marketing Manager for the LIDAR Division of the Autonomous Driving and Safety Products Group at Analog Devices in Wilmington, MA.

He holds a BS in Electrical and Computer Engineering and an MS in Electrical Engineering with a concentration in Communications Systems and Signal Processing from Northeastern University.