FMCW lidar science popularization (part 1): basic concepts, technical roadmap, advantages, controversies, and misunderstandings.

Definition of FMCW LiDAR

For quite some time, our discussions on the “technical route” of LiDAR have been focused on the classification of ToF LiDAR based on scanning architecture, but have ignored a higher-dimensional classification: based on the ranging method, LiDAR can be divided into two major categories: ToF and FMCW.

Apart from the common optical lens, these are two completely different categories. It can be said that the classification of ToF and FMCW is far more important than the classification of “mechanical rotation, MEMS, mirror, prism, and flash”.

From public information, there are roughly two forces involved in the development of FMCW LiDAR companies:

  1. Companies that focus on the FCMW route from the beginning, such as Aeva, Mobileye, Blackmore, LightIC, Strobe, Analog Photonics, Baraja, Scantinel Photonics, Bridger Photonics, Insight LiDAR, SiLC, OURS, LightSpoon Technology, LightPower Technology, LoWe Technology, and Microsource Photonics.
  2. Companies currently focused on the ToF route, such as Waymo, Hesai, and Huawei.

Interestingly, almost none of the companies established after 2017 have developed ToF products; they have all followed the FMCW route from the beginning. To these companies, the ToF route has already been matured by previous companies, leaving them with less opportunities to innovate. Additionally, the inherent limitations of the ToF technology route can only be “compensated” by a new technology route.

Despite the large number of players involved in this industry, according to Yao Jian, the CEO of LightSpoon Technology, in a speech several months ago, there are only a few companies that can provide FMCW LiDAR samples globally, such as Aeva, Blackmore, and LightSpoon.

The concept of FMCW may give people an “unfamiliar and intimidating” feeling, but except for a few practitioners, the outside world knows very little about FMCW LiDAR.

In September, “Jiuzhang Auto” read over 100,000 words of public materials about FMCW LiDAR, sorted out dozens of questions based on these materials, and had in-depth exchanges with six industry experts, including Yao Jian, CEO of LightSpoon Technology, Andy.Sun CTO of LoWe Technology, and Wu Lei, Director of Zhigan Photonics IC Design. The result is this “popular science article” series about FMCW LiDAR.

During the interview process, “Jiuzhang Auto” found an interesting phenomenon: practitioners have not yet reached a consensus on some basic issues regarding FMCW.The full name of FMCW is Frequency Modulated Continuous Wave, which is translated into Chinese as “调频连续波”. It is a kind of laser ranging principle compared to TOF. Therefore, before discussing FMCW, let’s briefly introduce TOF.

The ranging principle of TOF (Time of Flight, Chinese name “飞行时间法”) is: to use the flight time of light pulses between the target and the Lidar multiplied by the speed of light to calculate the distance. TOF Lidar adopts pulse amplitude modulation technology (AM), and is also called AM Lidar.

Unlike the TOF route, FMCW mainly sends and receives continuous laser beams, interferes with the reflected light and local light, and uses heterodyne detection technology to measure the frequency difference between the transmitted and received signals. The distance of the target object can be calculated by converting the frequency difference.

Specifically, the laser beam hits the target object and is reflected, which affects the frequency of light. If the target object is approaching the vehicle, the frequency will increase; if the target object is moving in the same direction as the vehicle, the frequency will decrease. When the reflected light returns to the detector and is compared with the frequency at the time of transmission, the difference between the two frequencies can be measured, and the distance information of the object can be calculated.

In short, TOF uses time to measure distance, while FMCW uses frequency to measure distance.

The following figure uses triangular wave FMCW as an example to illustrate its ranging principle: the blue represents the frequency of the transmitted signal, and the red represents the frequency of the received signal. The emitted laser beam is repeatedly modulated, and the signal frequency keeps changing.

Image from the official account "Meituan unmanned delivery" on June 28, 2021, in the article "Classification and Principles of mainstream Lidar"

In addition, according to Mr. Xi, the CEO of a certain FMCW Lidar company, there is still an obvious difference between FMCW and TOF: in order to reduce the interference of ambient light, TOF focuses on filtering, that is, blocking the light outside the working wavelength outside the Lidar receiver; while FMCW only interferes with the laser it emits and is not affected by other light sources. In short, TOF “rejects dissidents”, while FMCW “attracts the same kind”.

The core technology of FMCW Lidar comes mostly from the field of optical communication. Zhu Xiaoqi, chairman and general manager of Micro Source Photonics, said that the modulation-demodulation algorithm for receiving and transmitting signals of FMCW Lidar is very similar to that of optical communication products. Yao Jian, CEO of Light Spoon Technology, said that “putting the transmission end and the receiving end of optical communication products together, and folding the optical fiber, is an FMCW Lidar”.In the field of optical communications, silicon photonics technology is widely used, and it’s also needed in the receiving and scanning of FMCW lidar. The so-called silicon photonic chip is a type of chip that controls many optical paths on a CMOS wafer, including active control, modulation, and demodulation. Simply put, compared with ordinary silicon chips, silicon photonic chips can conduct both electricity and light.

However, optical communication products were not originally based on silicon photonics technology. Early optical communication products also used many discrete components, which were characterized by large size and high cost. With the maturity and introduction of silicon photonics technology, optical communication products began to develop towards integration, and their application scale began to increase significantly.

Given that the development of FMCW lidar companies is highly dependent on the maturity of the silicon photonics industry chain, the growth rate of the FMCW lidar industry is also largely restricted by the maturity of the silicon photonics industry chain.

The founder of a certain lidar company told “Jiuzhang Autonomous Driving”: “Why are FMCW lidar companies under DARPA basically established after 2018? Because only at this time, the process of silicon photonics technology can be considered mature. Silicon photonics communication has only become popular in recent two or three years. Before this, because it was impossible to rely on the integration of silicon photonics to reduce costs, the cost of making FMCW was certainly high.”

The Technological Roadmap of FMCW Lidar

Classified by Coherent Method

According to the introduction of Yao Jian, CEO of Guangshao Technology, FMCW can be divided into two types: frequency modulation and phase modulation based on the coherent way of light waves. They are essentially the same. (The derivative of frequency modulation is phase modulation, and the integral of phase modulation is frequency modulation.) But the implementation methods are somewhat different:

  • Frequency modulation is like adding the signal of the light wave to a “spring”, and then compressing or lifting the spring;
  • Phase modulation is to add a random encoding modulation to the signal. Essentially, it is a coding technology.

According to Andy Sun, CTO of Livox, phase and frequency are correlated, and phase modulation can be seen as a nonlinear or coded form of frequency modulation.

Linear frequency-modulated FMCW has the advantage of providing an extremely high signal-to-noise ratio through FFT signal processing technology, and the chips and IPs are very mature. However, the linearity requirement for laser tuning is very high, and the prices of lasers currently on the market are relatively high. But in recent years, with the continuous influx of suppliers using low-cost optical communication laser technology, future costs will be greatly reduced. The representative companies of linear frequency-modulated technology are Blackmore and SILC.

Encoded phase modulation or nonlinear frequency modulation has lower requirements for laser modulation and can use more easily available, high-output power fiber lasers. However, the disadvantage is that the sampling rate of ADC is very high, special DSP algorithms are needed, and the signal-to-noise ratio is relatively lower than that of FFT. Companies that use this technology roadmap are Aeva, Guangshao Technology.Andy Sun believes that compared to non-linear frequency modulation or code phase modulation, linear frequency modulation can achieve higher signal-to-noise ratio, so it will be a better choice for long-distance FMCW lidar.

Classification of frequency modulation implementation

According to an unnamed lidar manufacturer, there are two ways to implement frequency modulation:

External modulation, which loads the RF signal onto the optical modulator to achieve frequency modulation;

Direct modulation, which directly changes the working current of the laser to achieve linear frequency modulation.

There is no difference in performance between the two methods, but there is a big difference in integration and cost:

  • External modulation has low integration and high cost;

  • Direct modulation saves the cost of the modulator and signal source, but the implementation difficulty is relatively high.

According to the above-mentioned person, currently, only Aeva has achieved direct modulation technology among lidar manufacturers.

Advantages of FMCW over TOF

What are the advantages of FMCW lidar compared to TOF? Below, we will compare and share them one by one in combination with the limitations of TOF lidar.

The light wave of TOF is easily interfered by ambient light, while FMCW has strong anti-interference ability

The key technology of TOF is the electrical signal (narrowing the pulse of light through modulating voltage, then converting the light into electrical signal, and then manipulating the electrical signal to resolve the pulse), and a major characteristic of electrical signal is that it is easily interfered by environmental noise.

TOF lidar detects directly, responding to all the light entering the detector. Therefore, when there are many lidar devices installed on a vehicle or multiple vehicles with lidar devices driving in the same area, how to avoid mutual interference of the light waves emitted by each lidar becomes a major challenge.

To solve these problems, various lidar manufacturers have to invest a lot of money to research special anti-interference technologies. The most common technology is to separately encode each pulse of the laser to prevent it from being interfered with by other lidar devices. However, encoding will cause a decrease in signal-to-noise ratio and sacrifice ranging capability.

In addition, TOF lidar is also easily affected by strong light. If it faces the sun during operation, it will be difficult to identify the target when the sunlight is too strong. Imagine what would happen if both the camera and the lidar malfunctioned due to strong sunlight?

To achieve longer range detection, companies such as Luminar even use 1550 nm laser, but Yao Jian and others believe that if 1550 nm laser is used for TOF, it will be more sensitive to sunlight and therefore cannot work properly under strong light.

Of course, this problem is not completely unsolvable. For example, various filtering films can be used to resist interference; the focal length of the system can also be made longer – a longer focal length means a smaller field of view for single point testing, and therefore less sunlight is encountered. Note that the cost of the latter method is a sacrifice of FOV.However, for FMCW lidar, “interference” is no longer a problem.

Firstly, FMCW is based on the coherent principle (interference between reflected light and local light), it can only receive the light it emits (with the same or close frequency), and therefore will not be interfered by various “ambient light” – here “ambient light” includes light from other lidars and environmental light such as sunlight.

Secondly, according to Ya Jian’s explanation, FMCW lidar is single photon, and the intensity of the built-in light source is at least three orders of magnitude higher than that of the reflected sunlight, so the impact of sunlight on it can be basically ignored.

In addition, FMCW lidar has strong anti-interference ability, and one reason is that the filter cover is very small.

As mentioned earlier, adding a filter cover is a means to resist environmental light interference, and the smaller the bandwidth of the filter cover, the stronger the anti-interference ability. Typically, the filter cover bandwidth of TOF lidar is 20-30 nanometers, while that of FMCW lidar is less than 0.01 nanometer.

TOF has low signal-to-noise ratio, while FMCW has high signal-to-noise ratio

Low signal-to-noise ratio is a pain point that TOF lidar is difficult to overcome. Specifically, a lidar claims a detection distance of 200 meters, but in practice, it may not be able to distinguish whether a given target is a real or false one.

The reason is that the reflection and diffuse reflection of the target on sunlight and other environmental light can cause unnecessary noise signals. These noise signals are then converted into electrical signals and amplified at the receiving end.

Ya Jian introduced that the noise encountered by TOF lidar in the industry is called “additive noise”. The so-called “additive noise” means that the detection end receives a signal, but it may be a “false target point”. Usually, TOF lidar needs to judge whether this point is a real target point or a “false target point” based on the reflection rate.

FMCW lidar does not need to worry about this problem. There are two reasons:

(1) The detection end of FMCW lidar can only receive its own emitted light. Therefore, if the returned light does not match the original transmitted time, frequency and wavelength, FMCW lidar knows to filter out that data point, which means more precise target detection can be achieved.

(2) In FMCW lidar, in addition to the signal light emitted by the laser, there is also a local oscillation light which passes through a beam splitter. The echo of the signal light is coupled with the local oscillation light and the received light signal power, and the local oscillation light power also competes with the background noise, resulting in noise suppression.Normally, the signal-to-noise ratio of FMCW lidar is more than ten times higher than that of TOF. Mr. Xi, CEO of a certain FMCW lidar company, even believes that FMCW lidar, which uses coherent detection, can achieve a signal-to-noise ratio three to four orders of magnitude higher than TOF, and “with the advancement of semiconductor technology, interference efficiency will be further improved,” theoretically achieving single-photon detection.

As Yao Jian puts it, the noise of FMCW lidar belongs to “multiplicative noise,” that is, “once the detection end receives a certain signal, the front target is real, not a ‘pseudo-target’.”

Long-range lidars often sacrifice FOV to pursue longer detection distances, which actually requires higher signal-to-noise ratio. Therefore, after the technology matures, FMCW will be a better choice for long-range sensing.

TOF has poor data quality in the velocity dimension, while FMCW can obtain velocity data for each pixel

In the PR of FMCW lidar manufacturers, “velocity data” is an unavoidable highlight. So, can TOF lidar not provide velocity data of the target object?

Not necessarily. To be precise, TOF only calculates distance by measuring the return time of the emitted laser pulse, so it “cannot directly obtain the velocity information of the target object”. In practice, TOF calculates the instantaneous velocity of the target object using deep learning techniques based on its displacement and time differences between frames in the lidar.

However, according to a certain founder of FMCW lidar company, in many cases, once the target object’s distance from a TOF lidar claiming to have a detection range of 300 meters exceeds some 100 meters, it is difficult to calculate its velocity due to too few lines of laser hitting it.

Andy Sun, CTO of Luminar, mentioned another reason: the sensitivity to noise is very high when calculating velocity by using the difference in distance between two frames to derive the time difference, which can cause a very large error or even make it impossible. In practice, it is usually necessary to use data from more than two frames to smooth out the noise interference, and the calculated velocity will be more reliable, but this will also cause a significant delay.

But FMCW lidar does not have to worry about this problem. Because the reflected light frequency of FMCW will change according to the speed of the moving object in front due to the Doppler effect, the velocity data of each pixel of the target object can be directly calculated.

What is the use of velocity data?

Mr. Xi, CEO of a certain FMCW lidar company, mentioned two scenarios:

(1) A self-driving car is traveling at 120 kilometers per hour, and suddenly another car traveling at 125 kilometers per hour cuts in at close range. What will happen?When driving a car, the driver usually realizes sensitively that the gap between the car behind and mine will become bigger after the overtaking car successfully merges in, so I don’t need to take any action. But for an autonomous driving system, this is a very difficult task- due to the inability to obtain accurate velocity information, both cameras and TOF LiDAR cannot help the decision-making system make the judgment that “the gap between the car behind and me will become bigger”. In fact, emergency braking is a more common measure for autonomous driving systems in this situation, which is easy to cause a rear-end collision.

However, if the TOF LiDAR is replaced with FMCW, the problem will be solved easily. Because FMCW LiDAR can not only accurately detect the position of the overtaking car, but also precisely detect the vector velocity of the overtaking car, helping the decision-making system make the conclusion that “I just need not to overtake to avoid a collision”. Therefore, unnecessary braking is avoided.

On rainy days, the front tires of the vehicle will splash water mist. After the TOF LiDAR shoots onto the water mist, the formed point cloud is also lumped, which looks no different from the vehicle or other obstacles. This information will cause a lot of trouble for the decision-making system- to brake or not? If false braking happens frequently due to water mist, not only the riding experience is poor, but also safety hazards (rear-end collision) are hidden.

However, if FMCW is used instead of TOF, the problem will be easily solved. Because the water mist splashed by the front tires all have obvious ascending and descending trajectories, FMCW LiDAR can help the decision-making system to judge that they are “water mist” through these vector velocity information and filter them out in the decision-making algorithm.

Yao Jian gave the following examples:

(1) A motorcycle is covered by a white car, leaving only a small angle exposed. TOF cannot recognize such a target, but FMCW can identify it because it can monitor the velocity information of this “small angle”.

(2) Two cars are very close to each other, and TOF LiDAR finds it difficult to distinguish whether it is one car or two cars. But FMCW can easily distinguish between two cars because it can obtain the velocity information of each pixel of the cars.

(3) When monitoring pedestrians, TOF LiDAR can only roughly determine “there is a person there”, while FMCW can clearly see which side is this person’s left arm, and which side is the right arm.

Yao Jian explained: “Because there is velocity information, even if the laser only reflects one point (due to reasons such as the low reflectivity or far distance of the target), it will not greatly affect the monitoring result.”According to Yao Jian, the detection distance of FMCW LiDAR can be up to 500-600 meters. This is a key reason why Aurora decided to acquire Blackmore, a FMCW LiDAR manufacturer, after turning to the mainline logistics scene – trucks need to see far away, and the laser spots are sparse in the distance, but if there is speed dimension data, the sparse spots are not a serious problem.

Yao Jian said that in addition to extending the effective detection distance, speed dimension data can also bring the following two benefits:

  1. Because objects with low reflectivity can also be detected with only one laser spot, the reflectivity is not so important. Therefore, users do not need to specifically measure the reflectivity of various targets.
  2. The sensor used to output the speed of each pixel of the target object, reducing the computational power required for backend processing. Moreover, the algorithm architecture of the sensor fusion is also easier to implement.

Furthermore, speed dimension data can also compensate for the disadvantage of FMCW LiDAR in terms of point frequency.

A responsible person from a certain TOF LiDAR manufacturer said that unlike TOF ranging takes two microseconds at a time, FMCW ranging takes 20 microseconds at a time, so the point frequency of the latter will be lower. But overall, the low point frequency does not affect the detection effect, because each pixel includes speed dimension data, and the point frequency does not need to be high to achieve perception of moving objects at long distances.

However, Mr. Xi, the CEO of a certain FMCW LiDAR company, believes that the detection results in the case of sparse laser spots not only depend on the ability of the LiDAR itself, but also on how downstream customers’ algorithms are implemented – if the algorithm thinks that the LiDAR point cloud is too sparse and hits only three points on the vehicle far ahead, the “confidence level is not high”, and the detection results will be “ignored”.

Now there is a new saying that AI algorithms need to perceive surrounding objects and verify in real-time the accuracy of the perceived results, and how much deviation there is from the true value will showcase the role of LiDAR – generally speaking, measurement is objective and referable but the premise condition is that the measurement results provided by the LiDAR are precise, dense, and stable enough.

TOF is difficult to be compatible with OPA scanning structure, while FMCW is naturally more suitable for OPA.

As mentioned earlier, whether TOF or FMCW is mainly for the transmitter-receiver system. The scanning system is also critical. The scanning modes of TOF LiDAR include mechanical rotation, mirror rotation, prism, MEMS, Flash, and OPA, among which only Flash and OPA are purely solid-state scans, and the advantage of OPA scanning is more obvious.

Because Flash is a planar scanning, while OPA is a point scanning and the optical power is more concentrated, theoretically, the detection distance can be longer than Flash.However, in truth, the first several scanning solutions all had header lidar manufacturers involved, while OPA, although once popular, never made much noise. In fact, Quanergy, the first company to propose the use of OPA for TOF lidar, has been out of the automotive lidar market for three years.

Why? OPA scanning is typically based on silicon photonics chips, which cannot withstand the high peak power of TOF (usually 40-50 watts, or even up to 100 watts).

Of course, TOF power can be lowered to avoid damage to OPA chips. However, this also means a shorter detection range. Quanergy’s previous TOF+OPA combination had a detection range of less than 100 meters, which obviously cannot meet the requirements of autonomous vehicles for the main lidar.

In contrast, FMCW’s peak power is only at the “hundred milliwatt level,” which is four orders of magnitude lower than that of TOF. This is because:

  • TOF only requires 2 microseconds per measurement, while FMCW requires 20 microseconds, and although the total energy is not less, the peak power will be lower because the energy is spread over time;
  • The signal-to-noise ratio of TOF is lower, so if the power is too low, it will not obtain enough “effective signal”, while the signal-to-noise ratio of FMCW is very high, so even with very low power, it can obtain sufficient “effective signal”.

Therefore, if OPA scanning is combined with FMCW transceiver, there is no need to worry about peak power issues.

Conversely, it can also be said that OPA can only be applied to automotive lidar after FMCW matures.

FMCW can achieve a higher level of “chipification”

Currently, many TOF lidar manufacturers are trying to improve product integration and reduce costs through “chipification.” However, the modules that can be chipified are at most signal processing, lasers, detectors, etc. For optical lenses and scanning components (Flash is a matrix, with no independent scanning components), they cannot be chipified temporarily.

(Note: Optical lenses in TOF lidar cannot be integrated, which is the consensus of most manufacturers, but Mr. Xi believes that with the support of silicon single-crystal optical technology, optical lenses, even in TOF, can be integrated in the future.)

However, in FMCW lidar, in the most ideal situation, even optical lenses and scanning components can be chipified.

Of course, from a technical principle point of view, FMCW is not naturally chipified.According to Andy Sun, CTO of Luowei, early coherent optical communication products based on the same principle were “spliced” together from a bunch of discrete components. However, in recent years, there is a new trend, where giants like Cisco, Huawei, and ZTE have started to adopt integrated solutions based on silicon photonics chips.

The reason for this is that, although the performance of optical communication products based on discrete components can be guaranteed, due to the large number of components and the need for high-precision alignment coupling, the cost of each set can reach tens of thousands of dollars. In contrast, the cost of silicon photonics-based solutions has been reduced to one to two thousand dollars or even several hundred dollars per set, thanks to a relatively mature supply chain.

Similarly, early FMCW lidars (such as Blackmore’s previous products) were stacked with discrete components such as light sources, splitters, antennas, mixers, and detectors. However, not only are these solutions costly, with each set costing thousands of dollars, but they are also difficult to pass safety regulations. Therefore, similar to optical communication products, FMCW lidars ultimately have to be based on silicon photonics technology.

In TOF lidars, the laser and detector use different chips, while in FMCW, in the ideal case, the laser and detector can be integrated onto the same SoC.

Chipification in TOF does not include optical lenses and scanning components, while in FMCW, it is possible to chipify the optical lens and use silicon photonics chips for scanning components. In the ideal case, the scanning module can be integrated onto the same chip as the transceiver module (laser + detector) (which Intel and Mobileye, as well as a company in Shanghai, China, are working on).

Therefore, many manufacturers believe that the degree of chipification of FMCW lidar can be more thorough, making it more likely to enjoy the benefits of Moore’s Law.

Some Controversies and Misunderstandings about FMCW Lidars

Does FMCW have longer detection range than TOF?

Many media outlets have mentioned the advantages of FMCW lidars in terms of detection range. This view is not entirely wrong but is also controversial.

According to Yao Jian, CEO of LiDAR Maker Optics, FMCW lidars can still detect obstacles even if there are few points hitting the target or few returning light spots, thanks to velocity information. Therefore, it is possible to detect vehicles 500-600 meters away.

ZhiGan Photonics IC Design Director Wu Lei holds a similar view.The CEO of a certain FMCW LiDAR company, Mr. Xi, even claimed that due to its strong detection capability for weak signals, the detection distance of FMCW LiDAR applied in the aerospace scenario can reach 12 kilometers, while the detection distance of FMCW LiDAR applied in the automotive scenario can also reach 2 kilometers.

However, a certain TOF LiDAR manufacturer’s executive said that compared with TOF, FMCW LiDAR does have an advantage in detection distance, but it is not a “crushing advantage.”

With the increasing popularity of single-photon detectors, the ranging ability of TOF LiDAR is constantly improving, while FMCW’s advantage lies in direct velocity measurement. However, its advantage in ranging ability compared to TOF is gradually shrinking.

Moreover, the detection distance must be considered in combination with the frequency. Only when the resolution is high enough, does talking about detection distance make sense. However, in reality, FMCW’s frequency is lower than TOF’s, so its claimed detection distance is limited in recognizing static objects.

Another founder of an FMCW LiDAR company believed that if TOF also uses a 1550nm laser, FMCW’s advantage in detection distance would disappear because “the loss of FMCW system is relatively high.”

Does FMCW LiDAR have a time delay?

There is a saying online that FMCW LiDAR “has a time delay.” With this question in mind, the author specifically consulted some industry experts.

According to the TOF LiDAR manufacturer’s executive, unlike TOF LiDAR which often requires only 2 microseconds for single ranging, FMCW LiDAR usually requires about 20 microseconds for single ranging. The reason is that the light emitted by FMCW LiDAR needs to do a beat frequency with its own signal after returning. If the beat frequency time is too short, power will be affected.

However, this 18-microsecond time difference is not enough to affect driving safety. Because “even if two cars are driving towards each other at 100 kilometers/hour, the relative speed is only 200 kilometers/hour. In less than 20 microseconds, the relative position of the two cars changes only by a few millimeters.”

Is the inability to provide lateral velocity a disadvantage for FMCW?

As mentioned earlier, FMCW can provide velocity dimension data, but it also has limitations – it can only provide radial velocity (the speed of the target object parallel to the direction of the vehicle’s travel when it is in motion), and cannot provide lateral velocity (the speed of the target object perpendicular to the direction of the vehicle’s travel when it is in motion). Some media even call it a “disadvantage” of FMCW LiDAR.Regarding the “disadvantage” of FMCW being said as the lack of horizontal velocity, some manufacturers are not convinced. For example, the founder of a certain FMCW lidar company asked, “TOF cannot even provide radial velocity, while FMCW can at least provide radial velocity. Why has it become a disadvantage?”

Having multiple radial velocities is better than having nothing, right?

In addition, according to the founder, usually when the target object “has a certain volume”, the lidar will detect many points on the object. By measuring the radial velocity at these points, and calculating with some formulas, the horizontal velocity can be obtained.

This person also said, “We can use our eyes as an analogy. Horizontal velocity does not need to be emphasized because the object’s horizontal motion is easily ‘detected’ by the human eye. However, radial velocity- whether the target object is moving away or getting closer, and at what speed- is difficult for the human eye to accurately evaluate, which is where FMCW lidar come into play.”

Next preview: FMCW lidar popularization (Part II): emission, reception and scanning. Don’t forget to follow “Jiuzhang Tech-Intelligence Driving” if you’re worried about missing out.

This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.