A big fight among the five major lidar companies

Latest update time：2021-01-08

Reads：

Source: The content is reproduced from the public account " Car Heart ". The author of this article is Zhou Yanwu, a senior expert in the industry and a special author for Car Heart. Thank you.

2020 is a landmark year for the lidar industry.

Velodyne and Luminar have successively landed on the U.S. stock market, and Innoviz, Aeva and Ouster are on the way.

Among these five major lidar companies, the highest one, Luminar, has a market value of over 10 billion US dollars, and the lowest one, Innoviz, has a valuation of 1.4 billion US dollars.

The reason why these companies are so popular in the capital market is that major car companies have chosen to equip their mass-produced cars with laser radar.

Honda and Toyota have decided to use lidar on their L3 autonomous driving models; manufacturers such as Mercedes-Benz, Volvo, BMW, NIO and Xiaopeng are also preparing to use lidar on their mass-produced cars in 2021.

It can be said that the golden age of lidar is coming.

This article will analyze the technical routes of the five major LiDARs and take you to see where these hot LiDARs are heading.

This article will cover:

Velodyne Core Technology Analysis: MLA
Luminar: Highest power for highest performance
Innoviz: MEMS brings the lowest cost
Aeva: Sticking with FMCW
Ouster: A technology similar to Flash

Basic structure of LiDAR

It is worth noting that Velodyne, Luminar, Innoviz, Aeva and Ouster, these five companies, have all abandoned the traditional bearing motor mechanical rotation solution in their main products for the original equipment automotive market.

This is different from the main products of Huawei, Sagitar, Hesai, Leishen, Yijing and other domestic companies.

Before analyzing each company, we first briefly introduce the terminology of lidar performance evaluation.

LiDAR is divided into pulse ToF type and continuous wave type according to the ranging principle .

Most of our common products are pulse ToF type, which consists of four parts in hardware:

Laser emission
scanner
Reflected light reception
data processing

Continuous wave lidar is divided into two types: phase modulation and frequency modulation, among which frequency modulation, namely FMCW, is more common.

When it comes to the laser emission part, we usually divide it into three categories:

EEL type laser diodes usually come in two types: 905 nanometers and 1550 nanometers, and the materials include silicon, GaAs (gallium arsenide) , and InP (indium phosphide) .
VCSEL , vertical cavity surface emitting type, usually appears in array form.
Fiber laser tube .

The receiving part is usually divided into four categories:

PIN diode, without any gain.
APD, avalanche diode, has a certain degree of gain.
SPAD, or single photon array, has ultra-high gain.
MPPC or SiPM, similar to SPAD.

As a sensor, the most important indicator of LiDAR is the signal-to-noise ratio . However, this is also an indicator that LiDAR companies never disclose.

Technical analysis of the top five lidar companies

(1) Velodyne’s core technology: MLA

Velodyne is the pioneer in automotive lidar.

Velodyne began developing solid-state lidar in 2015, released Velarray in 2017, and basically completed the design in early 2020.

It was designed-in by Hyundai Motor in the same year, and it is said that GAC is also testing Velarray.

Velodyne has been working in the lidar field for the longest time and has accumulated many research and development results in mechanical lidar.

Previously, Velodyne stated that the core technology of Velarray is not MEMS.

On the Velarray, Velodyne miniaturized the mechanical lidar, which can adopt a resonant scanning method. It still uses multiple laser emitters, so it is indeed not MEMS.

Traditional mechanical lidar has three scanning modes:

The first type is prism , which has the disadvantage of introducing unnecessary size increase, bearing or bushing wear, and affecting life over time; the advantage is that the number of lines can be very high. A typical representative is Huawei .

The second type is the rotating mirror . Its disadvantage is that it cannot fully utilize the time domain, has a certain amount of volume waste, has a low line count, and is difficult to achieve high performance. Its advantages are long life and high reliability. A typical representative is the Valeo Scala .

The third type is MEMS galvanometer , which has the disadvantages of limited FOV, questionable reliability, low signal-to-noise ratio, and short effective distance. The advantage is low cost.

Velodyne develops Resonant Mirror technology.

This technology is only one letter different from MEMS galvanometer, but the actual difference is not small.

Resonant scanning does not have the disadvantages of the above three scanning mirrors, but it requires the laser radar to be miniaturized and needs to be matched with a concave mirror to form an arc shape. In addition, the cost will increase significantly.

Velodyne has applied for a patent for the Resonant Mirror . The patent is shown above, with 163 and 164 being the core.

In the field of optical communications, optical resonance is a basic element. The optical resonant cavity is one of the three elements that make up a laser amplifier. It can be divided into open cavity and closed cavity according to the presence or absence of a reflective surface.

The main function of the resonant cavity is to allow the gain medium to achieve population inversion, so that it can be used as an optical amplifier (Gain amplifier) . The amplified signal can be collected through the resonant cavity to form an oscillator (oscillator) .

Laser resonant cavities can be divided into three main categories:

The first type is a parallel plane cavity .

It consists of two parallel plane mirrors, optically known as a Fabry-Perot resonator (FP cavity for short), and is mostly used in solid-state laser systems.

The second type is double concave cavity .

It is composed of two concave mirrors. One special and commonly used form is the confocal cavity, which is composed of two concave mirrors with the same curvature radius, and the distance between the two mirrors is equal to the curvature radius. The two mirror surfaces coincide with the focal point. The confocal cavity has small diffraction loss and is easy to adjust.

The third type is the flat concave cavity .

It consists of a plane mirror and a concave mirror. A special and commonly used form is the semi-confocal cavity, which is equivalent to half of the confocal cavity.

Judging from Velodyne’s patent, the MLA array has a slight curvature, which should be used in conjunction with a concave reflector.

Detail of the MLA, showing the 8-channel module

Velodyne's newly appointed CTO Mathew Rekow comes from the field of optical communications and is very familiar with resonant cavities.

The main research and development work of Velarray is led by Mathew Rekow.

One of his tasks is to miniaturize and modularize the lidar to improve the mass production efficiency of Velarray and reduce costs.

The difficulty of this work lies in miniaturizing the module while ensuring high performance, especially since laser diode emission requires relatively large current, and traditional power devices cannot meet the requirements.

To address this problem, Velodyne began working with EPC in 2016 .

EPC excels in GaN power device technology, a wide-bandgap semiconductor material that enables field-effect transistors to switch more than 10 times faster than conventional transistors.

Velarray uses GaN field-effect transistors, which Velodyne calls a custom ASIC.

It is extremely small, measuring only 2 to 4 square millimeters. As it shrinks, its performance also increases.

There is a simple formula in LiDAR. The Z-axis resolution of LiDAR depends on the pulse width.

The Velarray, which uses GaN field-effect transistor ASIC, has a pulse width of up to 5 nanoseconds, which is the highest performance ASIC chip except for SPAD.

Most solid-state lidars are generally 50-150 nanoseconds, while SPAD can easily achieve 1 nanosecond or even tens of picoseconds.

Currently, Velarray's main product is an 8-channel module. High-performance products can use 4 to 16 modules, while low-performance products only require 1 module.

(2) Luminar: Highest power for highest performance

The simplest and most effective way to improve the performance of lidar is to increase the laser emission power.

When increasing power, the eye safety of the product must be considered.

Silicon photodetectors at 905 nanometers and InGaAs at 1550 nanometers are 100,000 times safer, allowing the power of lasers to be increased without any risk.

Luminar features the use of 1550nm InGaAs.

The laser power it uses is 40 times that of traditional silicon photoelectric systems. It not only has a high signal-to-noise ratio, but also reduces the pulse width to less than 20 nanoseconds, the pulse repetition frequency to less than 100MHz, and the duty cycle to less than 1%; this also increases the effective distance.

In rainy, snowy and foggy weather, the reflectivity of objects will decrease, which will shorten the effective detection distance of the lidar. However, increasing the power can solve this problem. This is what Luminar did.

Luminar emphasizes:

Even for objects with a reflectivity of 10%, the effective detection distance of its products can reach 200 meters.

Luminar has also applied for a patent on laser power amplification.

Its patent is to use a two-stage large mode field erbium-doped fiber (EDFA) amplifier to modulate a seed source laser into a pulse laser system with a pulse width of less than 20 nanoseconds, a pulse repetition frequency of less than 100MHz, and a duty cycle of less than 1%.

Luminar's patented technology consists of a seed source laser and an erbium-doped fiber amplifier.

The above picture shows the internal structure of Luminar's seed source laser

The picture above shows the internal structure of Luminar's amplifier

In terms of scanners, Luminar does not have much innovation and still uses traditional MEMS dual-axis galvanometer scanning.

Generally speaking, traditional MEMS lidars have a low signal-to-noise ratio, but Luminar's amazing power density completely eliminates this shortcoming.

Due to the introduction of fiber lasers, Luminar lidar is slightly larger.

In addition, the use of 1550 nm InGaAs lasers also keeps the product cost high.

Although Luminar has repeatedly emphasized its ability to reduce costs, fiber lasers have been used for more than 20 years and have long since lost their performance potential.

Therefore, the industry has always had doubts about Luminar's cost control capabilities.

(3) Innoviz: MEMS route brings the lowest cost

MEMS is currently the fastest solution to be implemented.

Compared with mechanical laser radar, it has three advantages:

First of all, MEMS micro-vibration mirrors help LiDAR get rid of bulky motors, prisms and other mechanical motion devices. The millimeter-sized micro-vibration mirrors greatly reduce the size of LiDAR and improve its reliability.

Schematic diagram of Innoluce MEMS LiDAR acquired by Infineon

The second is cost. The introduction of MEMS micro-vibration mirrors can reduce the number of lasers and detectors, greatly reducing costs.

The number of transmission modules and receiving modules required for traditional mechanical lidars is determined by the number of wiring harnesses.

The use of a two-dimensional MEMS micro-vibration mirror only requires a laser light source to reflect the laser beam through a MEMS micro-vibration mirror.

The two work together at a frequency of microseconds, and after being received by the detector, they achieve the purpose of 3D scanning of the target object.

Compared with mechanical lidar structures with multiple transmit/receive chipsets, MEMS lidar requires significantly fewer lasers and detectors.

From a cost perspective, N-line mechanical LiDAR requires N sets of IC chipsets:

Transimpedance Amplifier (TIA)
Low Noise Amplifier (LNA)
Comparator
Analog-to-digital converter (ADC), etc.

If imported lasers (typically Excelitas' LD) and detectors (typically Hamamatsu's PD) are used , the cost of each line of lidar at 1K quantity is about US$200 . Domestic lasers, such as the commonly used Changchun Institute of Optics, Fine Mechanics and Physics lasers, can be cheaper.

MEMS can theoretically achieve 1/16 of the cost.

Finally, the resolution is that the MEMS galvanometer can precisely control the deflection angle, unlike mechanical lidar which can only adjust the motor speed.

For example:

Velarray has 2 million single echo points per second.

Velodyne's 128-line lidar has only 2.4 million lines, and Velarray is almost equivalent to a 106-line mechanical lidar.

What are the disadvantages of MEMS?

The disadvantages are low signal-to-noise ratio, short effective distance, and too narrow FOV.

Because MEMS uses only one set of transmitting laser and receiving device, the signal light power must be much lower than that of mechanical lidar.

At the same time, the light receiving aperture of the MEMS lidar receiver is very small, much smaller than that of the mechanical lidar, and the peak power of light reception is proportional to the aperture area of the receiver, which leads to a further decrease in power.

The above means that the signal-to-noise ratio is reduced and the effective detection distance is shortened.

The resolution of the scanning system is determined by the product of the mirror size and the maximum deflection angle.

The larger the mirror size, the smaller the deflection angle.

The larger the mirror size, the higher the resolution.

Finally, the cost and size of the MEMS galvanometer are directly proportional.

Currently, the largest size of MEMS galvanometer is Mirrorcle, which can reach 7.5mm and costs as much as US$1,199 .

The MEMS micro-vibration mirror developed by Xijing Technology, which was invested by Suteng, has a mirror diameter of 5mm and has entered the mass production stage.

The MEMS micro-vibration mirror used in Hesai Technology's PandarGT 3.0 was developed independently by the team.

There are two main solutions to MEMS shortcomings:

One is to use a laser with an emission wavelength of 1550 nanometers and further increase the power using erbium-doped amplifiers in the fiber optic field.

The eye safety threshold of lasers in the 1550-nanometer band is much higher than that of lasers in the 905-nanometer band. Therefore, the power of 1550-nanometer fiber lasers can be greatly increased within a safe range. A typical example is Luminar.

The downside is that 1550 nm lasers are extremely expensive.

Moreover, this is within the scope of the laser industry. LiDAR manufacturers have far less technical accumulation in this area than laser industry manufacturers, so it is almost impossible to lower costs.

The second is to use a SPAD or SiPM receiving array instead of a traditional APD array. The SPAD array is about 100,000 times more efficient than an APD.

However, SPAD arrays are not yet particularly mature and their prices are slightly high.

(4) Aeva: Sticking to FMCW

LiDAR, traditional cameras and millimeter-wave radar have something in common. Traditional ToF LiDAR can be regarded as a 3D camera, but the resolution is generally very low.

Traditional cameras are 2D imaging, while lidar is 3D.

Laser can also be regarded as a type of electromagnetic wave, which is very close to millimeter-wave radar.

FMCW LiDAR Schematic Diagram

Early cars also used electromagnetic waves to directly emit and reflect to measure distance. Later, it was discovered that this method had a low signal-to-noise ratio and high power consumption, just like the current ToF lidar.

Later, it was discovered that continuous wave frequency modulation coherent detection (FMCW) has a high signal-to-noise ratio and low power consumption, but the signal processing operation is large.

With the improvement of current chip computing power, this difficulty has been gradually overcome, and today's electromagnetic wave radars are all FMCW type.

In addition, early electromagnetic wave radars were also driven by motors for scanning, but were later converted to printed planar antenna arrays instead of mechanical scanning.

People seem to be able to conclude from the development history of automotive radar that lidar will eventually be FMCW, and arrays will be used instead of scanners.

ToF lidar has many interference factors or noise.

One is the influence of sunlight, which is more sensitive to 1550-nanometer lasers, while 905-nanometer ones are much better.

Second, the surface material and color of the object will also have an impact. Different colors and materials have different absorption rates for lasers. For example, the reflectivity of white and black is very different. The reflectivity is closely related to the effective distance. The lower the reflectivity, the shorter the effective distance.

Generally, when measuring the effective distance of a LiDAR, a test condition of 90% reflectivity must be added; if the reflectivity is 10%, in extreme cases, the effective distance may be shortened by 50%.

Black objects reflect a low number of point clouds and may not be sensed at long distances.

FMCW lidar uses phase interferometer beat frequency method for measurement, so these noises do not exist.

For FMCW lidar, the signal-to-noise ratio is proportional to the total number of photons emitted, not the peak laser power.

Because FMCW lidar has more than 10 times higher sensitivity, its transmitted average power can be 100 times lower than that of pulsed ToF lidar, which means low power consumption and higher eye safety levels.

The photonic circuitry of an FMCW lidar mixes a portion of the outgoing coherent laser light with the received light.

This provides a unique “unlock key” that effectively blocks any ambient radiation or other LiDAR interference.

The light source of FMCW lidar needs to modulate the frequency of the optical carrier in different forms according to the measurement purpose. The commonly used forms include triangular wave, sawtooth wave and sine wave.

The frequency of the transmitted signal varies periodically with time t around the optical carrier frequency fc. Each period T is called the signal repetition time, and the frequency variation range (f1-f2) is called the modulation bandwidth B.

The use of triangular wave shaped frequency modulation can easily demodulate the Doppler frequency of the target reflection signal, thereby achieving simultaneous distance and speed measurement.

The sawtooth wave shaped frequency modulation form is often used when the Doppler frequency shift introduced by the relative speed of the detection target can be ignored, and the maximum detection distance can be achieved.

The generation of sinusoidal FM signals is relatively convenient, but the demodulation method is complicated and its accuracy is slightly worse than that of high FM linearity modulation forms.

Generally, a triangle wave is used, which can measure the speed of the target like FMCW millimeter wave radar.

Today's FMCW millimeter wave radar is very simple, and the main chips are transceivers and processors, which has the benefit of easy chip integration. This also means small size and low cost.

However, it took nearly 10 years for FMCW millimeter-wave radar to mature.

Today's FMCW lidar technology is very immature. Laser modulation, reception and data processing are all in their infancy and are far from comparable to ToF lidar.

Laser modulation, in particular, is extremely difficult, and there are only a handful of companies engaged in related research.

According to the relationship between the tuning device and the laser, the current methods for realizing laser optical carrier frequency modulation can be divided into two types: internal modulation technology and external modulation technology.

Internal modulation technology Refers to the modulation technology that is carried out simultaneously with the establishment of laser oscillation. The resonant parameters of the laser cavity are changed by modulation to achieve the change of the laser output frequency. It mainly includes modulating the optical length of the resonant cavity or changing the position of the gain loss spectrum in the cavity.

External modulation technology refers to the technology of using a modulator to modulate the light field parameters on the optical path of the laser after the laser oscillation is established.

Either way, it is still in the exploratory stage.

Most light sources with good tunability are not stable enough, and most stable light sources are not widely tunable.

From the perspective of modulation, internal modulation is relatively easy to obtain a large tuning range because it directly changes the parameters of the resonant cavity. However, due to the existence of laser build-up time, the instantaneous line width of the output frequency modulated light is relatively wide, resulting in a reduction in the coherence length of the light source. Or in order to establish a stable light field, the tuning rate must be limited.

The external modulation method can quickly change the instantaneous frequency of the light field while maintaining the excellent characteristics of the seed light through tuning mechanisms such as the electro-optic effect. However, due to the limited working bandwidth of the electro-optic effect itself, the increase in the tuning range of the light source is restricted, which limits the maximum resolution that can be achieved by the system.

Currently, the industry tends to prefer external modulation, but the disadvantages of this method are high cost and large size.

The disadvantages of FMCW are high cost and the need for ultra-high precision of all its components, since the tuning frequency is at the THz level, which requires measurement instrument-grade components.

There are very few suppliers of this type of components, and each component requires high-precision testing, resulting in a low yield.

Even if mass production is achieved in the future, the cost will remain high. All optical surfaces must be within a tighter tolerance range, such as λ (wavelength) /20.

FMCW requires an ADC conversion rate that is 2 to 4 times that of a ToF system, and requires higher accuracy.

The requirement for the FPGA is to be able to receive data and perform ultra-high-speed FFT conversions.

Even with an ASIC, an FMCW system requires ten times the processing system complexity (and cost) of a ToF system.

In addition to cost, although FMCW is free from interference from external factors, it will bring new interference.

Like millimeter wave radar, FMCW lidar needs to consider the interference of side lobes . The FMCW system relies on window-based side lobe suppression to solve self-interference (clutter) , which is far less robust than the ToF system without side lobes.

To provide context, a 10 microsecond FMCW pulse can propagate radially over 1.5 km.

Any object within this range will be caught in the Fast Fourier Transform (time) sidelobes. Even a shorter 1 microsecond FMCW pulse can be corrupted by high intensity clutter at 150 meters.

The sidelobes of the first rectangular window Fast Fourier Transform (FFT) are known to be -13dB, which is much higher than what is needed to obtain a good quality point cloud.

In addition, FMCW lidar has a slight delay problem, which is an inherent defect of coherent detection and cannot be changed.

Aeva's main partners are Audi and ZF.

Other companies using FMCW lidar include Strobe, which was acquired by General Motors in 2017. The company has not taken any action since the acquisition.

Four Blackmore LiDARs are installed on the roof.

Costs up to $400,000

Then there is Blackmore, which was invested by BMW i Venture and acquired by Aurora in 2019.

(5) Ouster, which is similar to Flash

In a strict sense, Flash LiDAR refers to a LiDAR that performs imaging with a single flash (laser pulse) .

Borrowing the terminology from the camera industry, it is also called global shutter lidar.

Flash LiDAR in a broad sense refers to focal plane array imaging LiDAR, which does not necessarily require a global shutter but can also have a local shutter.

A typical representative of global shutter lidar products is ASC , which was acquired by Continental Automotive of Germany in 2016 .

Compared with scanning imaging LiDAR, Flash LiDAR has no moving parts and is an absolutely solid-state LiDAR that can meet the highest level of automotive regulations.

Scanning imaging needs to scan the entire workspace to provide an image (point cloud) , and the frame rate is usually 5-10Hz.

This means there is a delay of at least 100 milliseconds, which is unacceptable in high-speed scenarios.

If the scanning lidar wants to increase the frame rate, the horizontal angular resolution must be reduced, which are contradictory.

The reason is simple, the faster the scan, the lower the resolution will be.

But Flash doesn't. Theoretically, its pulse is only tens of nanoseconds to 1 nanosecond, which means that the frame rate can reach tens of kHz or even 1MHz.

Of course, considering the data processing capability, the current Flash LiDAR is still 30Hz, but it can be said to be delay-free.

HFL110 Flash LiDAR from Continental Automotive

It has been confirmed to be used in Toyota's L3 autonomous driving mass production vehicles

Although Toyota has invested in Luminar, it still uses the lidar from Germany's Continental Automotive.

The disadvantages of Flash LiDAR are obvious:

The power density is too low, resulting in its effective distance generally being difficult to exceed 50 meters, and the resolution is also relatively low. Using high-power VCSEL and SPAD can solve some of the problems, but the cost also increases rapidly.

Germany's Continental Motors has struck a balance between performance and cost, and its cost is estimated to be no more than $300 , and can be reduced by about $100 after mass production.

In order to solve the shortcomings of signal-to-noise ratio and short effective distance, some companies have made improvements to Flash LiDAR.

The improved design uses a VCSEL laser emitting array, which is manufactured using semiconductor chip technology. The current conduction of each small unit can be controlled, allowing the light-emitting units to be turned on and lit in a certain pattern, thereby achieving the effect of a scanner and accurately controlling the scanning shape.

For example, if the vehicle speed is high, the FOV will be reduced to improve the scanning accuracy.

When the vehicle speed is low, the FOV is increased and the detection range is expanded.

Ibeo and Ouster both have this design.

Ibeo considers this a scanning lidar.

Ouster believes it is a Flash LiDAR, but with a Multi-Segment added in front of it .

In fact, both are the same type of lidar.

Ibeo has been working in the field of LiDAR for more than 20 years, and its Flash LiDAR has excellent performance. Except for the number of pixels, which is slightly lower than Lumianr, most other indicators are comparable.

But its reliability is far superior to Luminar, and it is easier to pass vehicle regulations.

Why are these super giants optimistic about the Flash route?

I think the development direction of LiDAR is Flash, which can also be called depth camera .

The reason for this is that Flash LiDAR:

Easiest to pass strict vehicle regulations
Smallest size
Most flexible installation location
Full Chip
Lowest cost (unit price can easily be less than $100)
The greatest potential for performance development (depth cameras are similar to the CMOS image sensors that were just emerging at the time and eventually replaced CCDs)

The global technology industry's R&D investment in the global Flash field is far higher than that of other types of LiDARs, and all of them are super giants:

Broadcom (Tesla's partner) , Sony, Samsung, Apple, STMicroelectronics, Infineon, AMS, Lumentum, Toshiba, Panasonic, Canon, Hamamatsu, ON Semiconductor, Denso and Toyota are all developing Flash automotive lidar.

In the field of optoelectronics:

Whether it is the SPIE International Society for Optoelectronics Engineering, the OSA American Optical Society, the ISSCC International Solid-State Circuits Association, or the EPIC conference of the European Photonics Industry Association, almost all the papers are about SPAD or VCSEL, the key components of Flash LiDAR, and there are no traditional LiDAR papers at all.

Depth cameras can be used not only in the automotive field, but also in other solid-state 3D sensing fields, as well as AR/VR.

Broadcom, the world's second-largest IC design company, which is working with Tesla to develop next-generation chips, launched SPAD or SiPM array chips for automotive Flash lidar at the EPIC online conference in November 2020.

The upper and lower parts of the Apple iPhone 12 Pro's ultra-wide-angle lens make up the lidar.

This is no different from the Flash LiDAR used in cars, which is also a VCSEL+SPAD design, but with lower power and smaller size.

In fact, LiDAR has already been widely used in the mobile phone industry, but it is called ToF camera.

Apple is returning to its true name.

Apple has decided to build cars, so naturally it will also make use of its research and development results in the field of lidar, which can be fully used in the automotive field.

Apple iPad LiDAR disassembled, the sensor, also known as SPAD, is provided by Sony.

Sony presented a paper at the December 2020 ISSCC titled:

The paper A 189×600 Back-Illuminated Stacked SPAD Direct Time-of-Flight Depth Sensor for Automotive LiDAR Systems also directly targets automotive flash lidar.

Generally speaking, there are two factors that limit the performance of Flash LiDAR:

One is the VCSEL laser
The second is the received SPAD

VCSEL is small in size, low in cost, and easy to control, but has relatively low power.

Several major VCSEL manufacturers are working hard to develop high-power VCSEL arrays. The fastest progress is Lumentum, Apple's main supplier and the world's largest VCSEL manufacturer, with a market share of approximately 45%.

Currently, the maximum power of the experimental product can reach 10 watts, and 30 to 50 watts of power can be on par with non-Flash LiDAR.

In terms of automotive laser radar SPAD, it currently has only about 10,000 pixels, while 300,000 pixels are already mainstream in the mobile phone field.

Japan has accumulated rich experience in the CCD field and has an overwhelming advantage in the SPAD field.

Canon has developed a 1-megapixel SPAD that can easily crush the current highest-performing 128-line lidar, not to mention Luminar's MEMS lidar.

Samsung presented a paper titled:

The paper A 4-tap 3.5μm 1.2Mpixel Indirect Time-of-Flight CMOS Image Sensor with Peak Current Mitigation and Multi-User Interference Cancellation proposed a 1.2-megapixel ToF sensor (i.e. SPAD) .

Panasonic has developed a stacked SPAD that can achieve an effective range of 100 meters.

Toshiba is also developing chip-based SPADs.

The SPAD chip that Toshiba trial-produced in March 2018 has a resolution of 240x96.

MEMS lidar is just a transitional product, but it is difficult to judge how long this transition period will last.

I think it will take 3 years at the fastest and 6 years at the slowest. By then, LiDAR will be installed in the position of rearview mirror like traditional cameras today.

*Disclaimer: This article is originally written by the author. The content of the article is the author's personal opinion. Semiconductor Industry Observer reprints it only to convey a different point of view. It does not mean that Semiconductor Industry Observer agrees or supports this point of view. If you have any objections, please contact Semiconductor Industry Observer.

Today is the 2548th content shared by "Semiconductor Industry Observer" for you, welcome to follow.

Latest articles about

■TSMC's 2nm is too powerful, UMC is too miserable

■The United States has heavily funded this semiconductor technology

■Tesla is also snapping up HBM 4

■Intel's next-generation AI chip is exposed for the first time

■Nvidia releases its largest chip yet

■The Danish robot giant invites you to "do something" together

■Self-developed DPU released: Microsoft chip, full of firepower

■In the post-Moore era, optical computing chips have become the key to breakthrough, and domestic manufacturers have great potential!

■Open source software in crisis

■The global semiconductor equipment giants are all in trouble