The direct answer
Can 3D LiDAR sensors be ranked by one maximum-range number? No. In this nine-record audit, all nine sources publish at least one target-reflectivity-qualified detection value, but only five records state ambient illumination beside that value and only three state a probability-of-detection or equivalent detection criterion. Three Ouster records also publish a much larger maximum representable range, which is a digital or mode boundary—not proof that a target will be detected there.
What is the minimum valid comparison? Match the hardware revision or orderable variant, target reflectivity, ambient light, detection probability and false-alarm criterion, channel or field-of-view region, return or signal mode and any thermal state. If one of those conditions differs or is absent, preserve the two claims as different observations instead of turning them into a leaderboard.
RCI extracted 21 target-qualified detection observations from nine records. Four records permit a within-sensor high-versus-low-reflectivity calculation at 100 klx; their published distance factors range from 1.50× to 2.89×. That spread is evidence that target reflectivity matters. It is not a cross-brand performance score.
The nine range records
| Hardware boundary | Qualified range statement | Separate non-detection boundary | Editorial limit |
|---|---|---|---|
| Livox MID-360 MID-360; first-return range table | 40 m @ 10%; 70 m @ 80% — both at 100 klx | The 0.1 m close-proximity boundary does not establish full accuracy from 0.1 to 0.2 m. | Effective range varies across the vertical field of view; the two distances are not a universal obstacle-detection guarantee. |
| Ouster OS1 Rev8 Rev8 OS1-080-xxx; excludes OS1 Max and Rev7 | 90 m at a 10% target; ~250 m maximum representable range is a separate boundary | ~250 m is labeled maximum representable range, not target-detection range. | Do not attach the OS1 Max 200 m at 10% target claim to this OS1 record, and do not reuse Rev7 point-rate or frame-rate values. |
| Hesai XT32 XT32 32-channel row inside the XT32M/32/16 family | 80 m @ 10% with PoD >90%; 0.05–120 m is the separate instrumented range | The 0.05–120 m instrumented range is not a guarantee of 10%-target detection at 120 m. | The selected product page does not publish an ambient-light value beside the 80 m / 10% / PoD result. |
| Livox Avia Avia; published 0 klx and 100 klx detection tables | 190/230/320 m at 100 klx and 190/260/450 m at 0 klx for 10/20/80% targets | None published in the selected source. | Return mode changes point rate, and the published range table does not make all scan patterns or targets equivalent. |
| Ouster OS0 Rev8 Rev8 OS0-080-xxx family | 35 m at a 10% target; ~250 m maximum representable range is a separate boundary | ~250 m is labeled maximum representable range, not target-detection range. | Channel count, beam configuration, frame mode and temperature state all remain material; zero minimum range is not zero-error range. |
| Livox Mid-70 Mid-70; 100 klx detection table | 90/130/260 m for 10/20/80% targets at 100 klx | The 0.05 m minimum range does not imply full published precision at that distance. | The independent Mid-70 registration result in RCI is a sensor-plus-algorithm metric and is not substituted for these manufacturer detection distances. |
| Ouster OS2 Rev7 Rev7 OS2; 32/64/128-channel family | 200 m at a 10% target; 404 m typical / 381 m minimum maximum representable range is firmware- and mode-bounded | 404 m typical / 381 m minimum maximum representable range at 0.25×, 0.5× or 1× signal multiplier on firmware 2.5.x; returns beyond the boundary may alias. | Representable range is not target detection, and signal multiplier, horizontal resolution and frame rate alter the operating boundary. |
| Hesai QT128C2X QT128C2X, Q03 manual dated 2026-04-10 | 20 m for channels 9–120; 15 m for outer channels — 10% target, 0–100 klx, PoD >90% | The 0.05–50 m instrumented range is not the 10%-target detection result. | The current product page and current manual disagree on both single-return point rate and mass; those conflicts remain separate C-state claims. |
| SICK MULS1AA-112211 multiScan136 Exact ordering code 1140133 | 10/12 m at 10% and 15/30 m at 90%, crossed with 100/10 klx | The 0.05–60 m working range is not equivalent to the reflectivity-qualified scanning range. | The row applies to ordering code 1140133 and a >99% detection criterion; other multiScan100 variants must not inherit it automatically. |
The table deliberately keeps families and exact ordering codes visible. “Ouster OS1” now means the Rev8 OS1-080-xxx family in the current catalog; OS1 Max is separate. “SICK multiScan136” means ordering code 1140133. QT128’s 20 m result applies to channels 9–120, while its outer sixteen channels are listed at 15 m under the same stated target and test envelope.
The same sensor can publish 1.5× to 2.89× different distances as target reflectivity changes
Four sources expose both a lower- and higher-reflectivity target under 100 klx. RCI divided the higher-reflectivity distance by the lower-reflectivity distance within the same record and light condition:
| Record | Lower-reflectivity claim | Higher-reflectivity claim | Light | Within-record factor |
|---|---|---|---|---|
| SICK multiScan136 | 10 m @ 10% | 15 m @ 90% | 100 klx | 1.50× |
| Livox Avia | 190 m @ 10% | 320 m @ 80% | 100 klx | 1.68× |
| Livox MID-360 | 40 m @ 10% | 70 m @ 80% | 100 klx | 1.75× |
| Livox Mid-70 | 90 m @ 10% | 260 m @ 80% | 100 klx | 2.89× |
The Mid-70’s 260 m at 80% target is almost 2.9 times its 90 m at 10% target under the same published 100 klx condition. That does not make Mid-70 “2.9 times better” at detecting robot obstacles. It means a high-reflectivity test target travels much farther in that manufacturer table. Material, angle, target area, beam incidence and a real perception stack still determine whether an application detects what matters.
Ambient light does not create one universal penalty
Only Avia and the selected multiScan136 row provide enough values to calculate a light contrast at more than one target reflectivity. Their own tables show that the light effect changes with the target:
| Record | Target | Brighter condition | Lower-light condition | Lower-light / bright |
|---|---|---|---|---|
| Livox Avia | 10% | 190 m @ 100 klx | 190 m @ 0 klx | 1.00× |
| Livox Avia | 20% | 230 m @ 100 klx | 260 m @ 0 klx | 1.13× |
| Livox Avia | 80% | 320 m @ 100 klx | 450 m @ 0 klx | 1.41× |
| SICK multiScan136 | 10% | 10 m @ 100 klx | 12 m @ 10 klx | 1.20× |
| SICK multiScan136 | 90% | 15 m @ 100 klx | 30 m @ 10 klx | 2.00× |
Avia publishes the same 190 m distance for a 10% target at 0 and 100 klx, but its 80% target changes from 320 m to 450 m. The SICK row changes from 10 to 12 m for a 10% target, and from 15 to 30 m for a 90% target. “Sunlight halves range” and “ambient light has no effect” would both overgeneralize these bounded tables.
Detection, instrumented and representable range answer different questions
Target-qualified detection range says a target with a stated reflectivity was detected at a distance, sometimes with a disclosed probability and false-alarm rule. Instrumented range describes a measurement interval or device boundary. Maximum representable range describes how far the encoded range can extend under a mode before aliasing or another representation limit becomes material.
The distinction changes procurement decisions:
- Hesai XT32: 0.05–120 m instrumented range sits beside an 80 m result for a 10% target with PoD >90%.
- Hesai QT128C2X: 0.05–50 m instrumented range sits beside 15 or 20 m at 10% reflectivity, depending on channel.
- Ouster OS0 Rev8: 35 m at a 10% target sits beside roughly 250 m maximum representable range.
- Ouster OS1 Rev8: 90 m at a 10% target sits beside roughly 250 m maximum representable range.
- Ouster OS2 Rev7: 200 m at a 10% target sits beside a 404 m typical / 381 m minimum representable boundary for the stated firmware and signal multipliers.
Dividing those Ouster pairs produces arithmetic, not a performance factor. The numerator and denominator are different quantities.
The audit found two forms of official-source drift
Ouster generation drift: the stable OS1 URL now describes Rev8. Earlier RCI data retained Rev7’s 5.2 million points/s and 20 Hz maximum. Catalog v0.4.2 updates the current record to the Rev8 family’s published maximum 10,485,760 points/s and available frame-rate list through 40 Hz, adds the ~250 m representable boundary, and explicitly excludes OS1 Max. Immutable v0.4.1 remains unchanged so a reader can reproduce what RCI previously published.
Hesai document conflict: the current QT128 product page states 1,152,000 points/s and 700 g. The Q03 manual dated 2026-04-10 states 864,000 points/s and 900 ± 15 g. RCI now publishes both disagreements as C-state claims. The gap may reflect a hardware or configuration change, but the two current sources do not name that boundary, so RCI does not invent one.
What to ask before selecting a LiDAR
- Exact identity: hardware revision, order code, channel count, connector option and firmware branch.
- Target: reflectivity, area, material, angle of incidence and whether the target fills the beam footprint.
- Environment: ambient illumination, temperature, precipitation, contamination and multi-LiDAR interference.
- Criterion: probability of detection, false-alarm rate, minimum valid returns, confidence threshold and integration time.
- Operating mode: frame rate, horizontal resolution, returns or echoes, signal multiplier and field-of-view region.
- System result: timestamp path, motion compensation, calibration, filtering and the perception algorithm that consumes the points.
A range table is a starting point for a test plan. It is not a substitute for testing the exact obstacles, mounting geometry, weather, speed and safety envelope of the robot.
Method and limitations
RCI selected the nine records already present in the 3D LiDAR comparison group, checked current manufacturer pages or manuals on 2026-08-10, and represented each published target/reflection/light/channel combination as one observation. RCI calculated only within-record ratios. No value was normalized across brands when its detection criterion, target or operating boundary differed.
This is an audit of public manufacturer evidence, not independent range testing and not a market census. The 21 observations remain M-state claims. The separate Mid-70 V-state mapping result in the component catalog measures a sensor-plus-algorithm registration experiment; it is not silently substituted for the manufacturer’s range table.
Download the observation-level data
Download the immutable research dataset as JSON or CSV. Stable current aliases are /data/research/lidar-range-conditions.json and /data/research/lidar-range-conditions.csv. The revised component records are in catalog v0.4.2 JSON and its claim CSV.
Suggested citation: Robot Component Index, “3D LiDAR Range Claims: 9 Sensors, 21 Conditioned Values,” RCI 014, research dataset v0.1.0 and component catalog v0.4.2, 2026-08-10.