LiDAR is one of the most important technologies in the world of autonomous vehicles. It enables cars to view their surroundings in three dimensions (3D) so they can make safe driving decisions.
But not all LiDARs work the same way. Two main types are often compared:
- Frequency Modulated Continuous Wave (FMCW)
- Time-of-Flight (ToF)
FMCW is sometimes described as the “next-generation” of LiDAR, promising ultra-high accuracy and long range. Although Innoviz manufactures only ToF LiDARs, the company spent several years building prototype FMCW LiDARs and thoroughly investigating the technology to determine if it is indeed superior to ToF. After long and careful evaluation in the lab and on the road with real-world driving scenarios – such as rain, fog, high glare, and occlusion on the LiDAR window – and after considering the auto industry’s strict requirements for high resolution and low cost, Innoviz determined that ToF comes out on top.
This blog takes a close look at both technologies and explains why Innoviz is convinced that Time-of-Flight is the clear and proven choice for automotive LiDAR today and for many years to come.
Following are several take-ways from Innoviz’s real-world comparison of FMCW and ToF LiDARs.
1. FMCW’s Velocity and Accuracy Are Not a Real Advantage in Automotive Applications
Velocity first. ADAS needs the full velocity vector (direction + magnitude) of objects. FMCW can measure Doppler velocity only along the laser beam (the radial component). Sideways or diagonal motion doesn’t project onto the beam, so FMCW systems must still use frame-to-frame differentiation to estimate the orthogonal components. In practice, both FMCW and ToF pipelines compute velocity vectors by differentiating frames, so FMCW’s inherent radial velocity readout provides little net advantage. We have observed that the higher pixel rate and density of the ToF system result in improved speed estimation when differentiating frames.
Accuracy in context. Millimeter-level ranging precision sounds impressive and can be useful in some industrial or space applications. However, for automotive perception, feedback that Innoviz received from automotive Original Equipment Manufacturers (OEMs) indicates that centimeter-level accuracy is sufficient for safe lane keeping, cut-in detection, and collision avoidance.
Bottom line: ADAS requires reliable, vector-level motion estimates and robust centimeter-level ranging at scale. Both FMCW and ToF derive velocity vectors from frame differentiation. ToF already delivers the required accuracy more efficiently and affordably.
2. Range and Performance Trade-Offs
FMCW is often claimed to achieve higher ranges and strong resistance to sunlight interference. However, in practice, FMCW systems operate with an ultra-narrow optical bandwidth that leads to a significant speckle effect. This is a phenomenon where coherent laser light interferes with itself after reflecting off rough or irregular surfaces, producing random intensity patterns. This speckle noise severely degrades signal uniformity and reduces detection statistics, removing much, if not all, of the theoretical advantage in range and sensitivity. The following image illustrates the speckle effect:
FMCW systems can extend range by increasing integration time (the duration over which the sensor collects and averages reflected light energy to improve sensitivity). While longer integration times can enhance signal detection, this always comes at a steep trade-off with frame rate, pixel rate, spatial resolution, power consumption, and overall system cost.
Modern ToF LiDARs such as InnovizTwo have already achieved 450 meters detection range for dark or low reflectivity targets, meeting requirements for Level 3 highway autonomy, without the heavy trade-offs inherent to FMCW.
In short, ToF delivers the required range reliably, efficiently, and affordably.
3. Pixel Rate Limitations in FMCW
Detecting small or distant targets depends on both range and resolution. They must scale together. Having long-range without sufficient spatial resolution leads to blind gaps between sampling points, which allows small obstacles to go undetected.
FMCW systems are commonly limited in pixel rate because they require long coherent integration times per chirp to maintain sensitivity and velocity accuracy. This restricts how many points can be measured simultaneously compared to ToF systems, which can fire and process multiple independent pulses in parallel. The result is lower effective resolution and slower frame rates for FMCW.
Furthermore, the computational burden of FMCW signal processing is an order of magnitude higher than ToF, making it very hard for FMCW to scale to the high channel counts required for dense, real-time 3D mapping. As a result, current FMCW LiDARs cannot yet achieve the pixel density or resolution necessary to reliably detect small obstacles — such as road debris or narrow poles — at automotive ranges.
In contrast, ToF technology scales more efficiently in both range and resolution, enabling faster frame rates and higher fidelity point clouds that meet the stringent perception needs of ADAS and autonomous driving.
4. ToF is Better in All Weather
For a LiDAR to be truly effective, it must perform reliably in challenging conditions like rain, fog, and dust. FMCW systems use wavelengths (around 1550 nm) that are absorbed more by water, which means their performance drops in poor weather. ToF LiDARs use 905nm, which handles these conditions better and keeps working even when the sensor window is a bit dirty or partially blocked.
The following graph shows that FMCW performance in fog starts to degrade at roughly 25 meters, whereas ToF starts to degrade at around 125 meters. At 200 meters, FMCW degrades by about 17% more than ToF and continues to degrade to about 22% more than ToF at 400m.
5. Cost, Manufacturing, and Scale
ToF has another major advantage: ToF already is field-proven and mass-produced. There’s a fully automotive-qualified supply chain supporting ToF LiDAR at 905nm, and costs have dropped dramatically over the past few years from tens of thousands of dollars to just a few hundred dollars per-unit.
FMCW, as far as we know, still faces big challenges in scaling up. The components are complex, expensive, and not yet approved for automotive production. The five largest Silicon Photonics (SiPh) fabs used for FMCW chip fabrication are not certified for automotive reliability standards. The packaging process (which must combine optical, electronic, and thermal interfaces with extreme precision) is complex and costly, further slowing progress toward mass-production.
6. High FMCW Power Consumption
An additional challenge is due to FMCW’s use of coupling and switching high-power lasers. While SiPh platforms developed for telecommunications typically handle sources of only a few milliwatts, LiDAR applications require generating and switching laser power that is orders of magnitude higher. This introduces further hurdles related to device maturity, efficiency, thermal management, and size. The excessive heat dissipation makes integration in a vehicle very problematic. These drawbacks restrict practical large-scale implementation in automotive environments where power consumption, dissipation, and size are major considerations.
In summary: ToF is currently in volume production, while FMCW is being sampled for niche installations. Simply put, ToF is ready for production today; FMCW is still in sampling mode.
7. The Industry Has Spoken
Based on Innoviz’s experience with many automotive OEMs and public market research reports, the vast majority of LiDARs currently in production vehicles, from passenger cars to robotaxis, use Time-of-Flight technology. FMCW systems have gained little traction in automotive applications. To the best of Innoviz’s knowledge, most companies developing FMCW systems have shifted focus to non-automotive markets such as defense or industrial sensing.
This is a clear indication that the automotive industry has already made its choice.
The Bottom Line
While FMCW technology holds theoretical promise, its practical limitations remain significant. Real-world implementations face lower pixel throughput, speckle-induced image noise, and strong sensitivity to weather and optical contamination. Combined with complex SiPh manufacturing, narrow bandwidth constraints, and costly packaging, these factors prevent FMCW from achieving the scalability the automotive market demands.
By contrast, Time-of-Flight LiDAR is already proven, industrialized, and qualified for automotive deployment. Its high pixel rate, robust weather performance, and established 905nm supply chain enable cost-effective mass production. ToF continues to evolve rapidly, powering next-generation ADAS and autonomy platforms with dependable, high-fidelity perception.
In the race for automotive LiDAR, ToF isn’t just keeping up. It’s leading the way.