A LiDAR designer’s guide to making sense of sensors


Wafer batch before oxidation—a crucial step for high performance.

Multi-project wafer of APD-arrays for fast sampling.
Advanced sensors, such as light detection and ranging (LiDAR) scanners, are critical components in prototype systems for autonomous vehicles, as well as in systems for adaptive cruise control (ACC), collision avoidance, traffic-sign recognition, blind-spot detection, and lane-departure warning.
Mobility LiDAR systems must sense the environment quickly and reliably, assembling as detailed a picture as economically feasible of immediate surroundings and the road ahead. Mounted in fast-moving cars, they must “see” a minimum of 150 m—almost 500 ft—forward, and detect small objects down to 10 cm—about 4 in. This demands complementary but independent sensor systems, with guaranteed functional safety and environmental qualification.
Sensors must possess an optimum signal-to-noise ratio to “see” the signal through any distracting background. Since optical detectors must be prepared to deal with varying levels of environmental light, sensors should possess a wide dynamic range. Designers must also consider economics; automotive components should be maximally cost-efficient. Generally, the best cost/performance ratio trumps the best technology.
Currently, design engineers are applying several different sensor technologies for LiDAR mobility systems. A new white paper from First Sensor, “Making Sense of Sensors: A LiDAR designer’s guide to sensor technologies for automotive/mobility systems,” reviews the strengths of each of these LiDAR sensor technologies and helps design engineers select the best product for their application. The paper also provides tips on choosing a sensor supplier with the appropriate experience, customization capabilities, and automotive qualification expertise. You can download it from First Sensor, Inc.. What follows are some highlights from its authors.
Silicon PIN diode
These silicon-based detectors possess a structure featuring three semiconductor types layered together: P-type, Intrinsic, and N-type—hence, PIN.
They exhibit good dynamic range, with the ability to handle widely varying amounts of light. They can detect the reflection of a distant object, even when subjected to direct sunlight. They’re relatively inexpensive.
However, they cannot deliver the high levels of bandwidth or signal-to-noise performance most modern mobility LiDAR systems require. They’re neither very sensitive, nor very fast.
Silicon photomultiplier (SiPM) and single-photon avalanche diode (SPAD)
Originally developed for small, specialized scientific and medical applications, these solid-state, silicon-based sensors are being tried out in other automotive systems.
They function similarly to avalanche photodiodes (APDs), see below, but are optimized for very high internal amplification or gain, making them able to detect the smallest amounts of light. They’re very fast. They’re compatible with CMOS (complementary metal-oxide semiconductor) technology, and can be paired with associated electronics on the same chip.
The sensitivity of their single-photon counters is much lower than that of APDs and must rely on very high multiplication. Unfortunately, the multiplication process adds noise that often significantly degrades the signal/noise ratio. Their amplification mechanism is prone to false triggers caused by high temperatures.
Perhaps their most serious drawback: their high gain comes at the cost of saturation problems.
The sensors must deal with laser light reflected from objects ahead. Many LiDAR systems specify scanners with wide fields-of-view, placing quite a large amount of added light on an SIPM or SPAD sensor. Some phenomena routinely encountered in LiDAR mobility environments—such as bright sunlight, high-beam headlights, or other LiDAR systems—can saturate the sensor with higher light levels than it can handle, even with optical filters.
As work to offset their drawbacks continues, they’re often considered for various LiDAR applications. To date, their saturation issues and other problems keep them from becoming the detectors of choice for scanning long-range LiDAR.
Indium gallium arsenide (InGaAs) photodiode
Frequently used at small sizes in telecommunications glass fibers, they’re newcomers to LiDAR, except for specialized military or aerospace applications. The technology abandons conventional silicon-based construction for InGaAs material.
With laser systems specially built for its higher spectrum—1550 vs 905 nm for the other discussed sensors, this design should be more sensitive, and able to generate more power. It can enable automotive LiDAR with a longer range.
However, InGaAs detector performance can be significantly degraded by even slightly higher than normal ambient temperatures. They may well need an external cooling system, even in temperate climates.
In addition, its base material is significantly more expensive than widely used silicon substrates. Fabricating InGaAs sensors in large sizes for LiDAR use would require much more complex manufacturing than silicon designs. They have not yet been successfully made in high commercial volumes.
Since they’re new to automotive LiDAR, OEMs must be prepared to spend substantial time, effort, and revenue trying to develop LiDAR systems around any InGaAs detector.
Avalanche photodiode (APD)
Originally perfected for industrial and military applications, these silicon-based photodetectors work by enabling incoming photons to trigger a charge avalanche, multiplying gain by their internal amplification mechanism. Their absorption-optimized structure converts at least 80% of a laser’s reflected 905 nm light into photoelectric current. Result: greatly increased sensitivity.
Besides their notable sensitivity, APDs have an optimal signal-to-noise ratio, minimal saturation, and very good speed. They’re among the lowest-cost sensor technologies available.
Potential drawback: APDs use specialized bipolar technology not compatible with common-place CMOS fabrication. They can be sourced only from a small number of suppliers, and they cannot be paired on the same chip with their associated CMOS electronics.
However, experienced suppliers can fabricate packages with sensor and electronics on closely adjacent chips. Both can be optimized for best-in-class performance, with few compromises. For example, an APD sensor array can be complemented by specially designed transimpedance amplifiers (TIAs)—with customized gains and bandwidths—to convert the photocurrent to voltage, and to condition the signal going into the system for high gain. This can maximize performance, especially in low-light conditions.
APDs are produced by well-established, high-productivity commercial manufacturing processes, and are proven in a wide variety of systems on the road.
Done right, they combine proven performance with an attractive price. Currently the detectors of choice for automotive long-range LiDAR, APDs are critical components in a number of today’s most advanced mobility systems.
The right source
Once the right sensor technology is determined, designers still face the challenge of choosing the right sensor supplier. Candidates must be evaluated carefully. Do they have the technology, capacity, and know-how to adapt their sensors and systems to an OEM’s individual requirements? Will they work closely with the OEM team on design, manufacture, and scheduling to ensure a winning time to market?
Recommended strategies: insist on experience, assess for integrated manufacturing, check customization capabilities, investigate automotive qualification, and require future-proof support.
As LiDAR technologies continue to evolve, making the right sensor choices marks the path ahead.