Light-detection and ranging (LiDAR) instruments are key components of automotive designs that implement advanced driver-assistance systems (ADAS), ranging from level three control to full autonomy.
LiDAR operates by emitting pulses of infrared light generated by a laser in a sweeping pattern. Detection circuitry calculates the time for each pulse of light reflected off objects in the LiDAR’s sweep to reach its photodetectors. Software uses the delay to estimate the distance to the object and may use other properties, such as pulse strength and shape, to obtain more information about each detected object.
Using the angle of the original emitted pulse in the horizontal and vertical plane and the distance to the object, the software can build up a three-dimensional model of the space in the sweep range in spherical coordinates. In this way, the software can use LiDAR readings to map the environment in the instrument’s sweep field.
LiDAR is a core component of higher-level ADAS because of the advantages it brings to the real-time detection of hazards facing a vehicle. Though it will often operate in conjunction with radar, LiDAR has superior performance in its ability to estimate accurately the distance to potential obstructions and their shape and size.
The automotive industry deems two wavelengths of infrared light suitable for ADAS. One is the 905 nm wavelength, which lies in the infrared region close to that of visible light. However, the use of 905 nm light is associated with several issues. One is interference from natural sunlight and other ambient light sources, as the wavelength lies not far outside the visible-light region of the spectrum.
Another issue with 905 nm light is that of eye safety. Because this wavelength suffers from water vapor and dust absorption, longer-range use calls for high laser power. But this increased power comes with a risk of damage to the eyes of road users and pedestrians because of the wavelength’s proximity to the visible-light range. For this reason, 905 nm is now considered more suitable for short-range LiDAR applications on the sides and corners of the vehicle where laser power can remain low.
The 1550 nm wavelength is far enough from the visible spectrum to allow lasers that are considerably more powerful than is possible with 905 nm lasers and still be considered eye-safe to the Class 1 standard. This, combined with a greater ability to penetrate fog, haze, and dust, makes the wavelength more suitable for front-facing LiDAR systems that detect potential obstructions over 100 meters from the vehicle: an essential requirement for automated driving at motorway speeds.
There are additional advantages associated with the use of the 1550 nm wavelength. These include reduced interference from sunlight compared to 905 nm and better compatibility with the fibreoptic components used in the telecom market. LiDAR systems based on 1550 nm lasers can benefit from mature fibreoptic technology and the components and infrastructure manufactured for the telecom industry, which helps to reduce costs and improve reliability.
Detection sensitivity is also different than the 905 nm. It’s easier for sensors to detect light reflected from objects that have high absorbance, such as a rubber tyre lying on the road. However, detecting such objects at long distances in time for an ADAS to react calls for high sensitivity in the sensor array, even under higher laser power levels.
Absorbance and reflections from oblique surfaces and atmospheric conditions make detection of the returning light challenging to achieve, particularly in an environment where ambient noise is likely to be prevalent. This places a great deal of emphasis on the performance of the sensor elements.
Whereas silicon photodiodes are commonly used for 905 nm-wavelength applications, compound semiconductors such as indium gallium arsenide (InGaAs) offer higher sensitivity at the longer 1550 nm wavelength. The avalanche photodiode (APD) based on InGaAs effectively amplifies the low-level light signals expected in LiDAR applications.
When photons strike the absorption layer of an APD, they generate pairs of electrons and holes, with the electrons promoted to the conduction band by the energy released. A reverse bias voltage applied vertically across the APD accelerates the conduction-band electrons toward the positively charged anode.
As the electrons move into the avalanche layer and collide with atoms inside it, the kinetic energy they release triggers the generation of many more electron-hole pairs. The avalanche effect results in a higher photocurrent than possible from the initial photon interaction. The higher output current that results from the avalanche effect improves the ability of the readout circuitry to register the signal.
LiDAR systems typically use a transimpedance amplifier (TIA) to convert the current pulses generated by an APD to voltage pulses. These signals can more easily be processed by a time-to-digital converter that provides the information the detection algorithm needs. Higher gain applied to the APD provides stronger output levels for each pulse, but background noise becomes a larger issue. This noise results from the statistical nature of the avalanche process: the number of collisions triggered by each incident photon can vary widely. Designing a diode for gain that is too high leads to excessive background noise in the TIA output that can obscure smaller pulses from remote, low-reflectivity objects.
The excess noise factor depends heavily on the ratio of the impact ionization coefficients of holes to electrons, which ideally should be below 1. As this ratio approaches unity, the resulting excess noise factor limits useful gain in conventional InGaAs diodes to below 40, though this noise factor is superior to germanium-based devices. The significantly lower ionization ratio of silicon allows the APDs used in 905 nm LiDAR circuitry to be operated with a gain of several hundred or more.
Researchers from the University of Sheffield, UK, showed that adding an antimony alloy to the InGaAs manufacturing process can suppress the excess noise factor encountered with conventional devices.
There are now APDs based on these insights and years of R&D that operate at gains of 120 or more before noise becomes problematic. These devices also exhibit faster overload recovery compared to traditional InGaAs devices. This allows weaker secondary pulses that closely follow a large pulse to be detected reliably. A further benefit is a temperature drift that is ten times lower than parts that do not use the antinomy alloy, plus more stable high-temperature performance.
“Noiseless” technology makes it possible to extend the effective range of LiDAR instruments by up to 50% — the range being an important factor for ADAS Level 3 and beyond. This higher sensitivity enhances the system’s ability to detect small, distant objects on or near the road when traveling at high speed. It also enables small objects with low reflectivity that might be missed by conventional APD sensors to be detected. Alternatively, designers may use the improved gain to trade off range against laser power to save energy.
Lower laser power simplifies thermal management and eases the demands on optical components. The increased thermal stability of Noiseless InGaAs APDs results in more accurate measurements across varying environmental conditions.
Thanks to the material and manufacturing changes made by long-term research into InGaAs structure, LiDAR designers can take full advantage of APDs with increased gain to save power and improve accuracy without making substantial changes to the surrounding circuitry. By changing a single diode, the Noiseless APD infrared sensor technology can enable a new generation of ADAS for self-driving vehicles.
You may also like:
Filed Under: FAQs, Sensors