Range is a test result, not a property.

First preserve the manufacturer’s exact range word

A product page may publish ideal range, optimal range, detection range, ranging capability, instrumented range or maximum range. These labels are not synonyms. Instrumented range can describe what the sensor electronics can encode, while a reflectivity-conditioned ranging capability describes a target under a stated optical condition.

The first RCI sample makes the problem visible. RealSense calls 0.6–6 m an ideal range for D455. Orbbec publishes a broad depth range, a narrower optimal range and a separate theoretical maximum for Gemini 335. Livox publishes two MID-360 distances tied to 10% and 80% reflectivity under 100 klx. Ouster publishes a 10%-reflectivity range and a larger maximum range. Hesai separates instrument range from channel- and reflectivity-conditioned capability in its manual.

Stereo range depends on the scene and algorithm

Passive or active stereo estimates depth from image correspondence. Baseline, focal length, resolution and disparity precision create a geometric sensitivity that worsens with distance. Textureless surfaces, repeated patterns, motion blur, occlusion and lighting can prevent valid matches before a theoretical geometric limit is reached.

An accuracy claim therefore needs distance, image mode, region of interest, target, lighting, filtering and metric. A percentage error at one distance cannot be projected across the entire range without a model and evidence.

LiDAR range depends on the optical target and detection rule

A laser return changes with target reflectivity, incidence angle, beam divergence, atmosphere, ambient light and receiver threshold. Multi-return settings, channel layout and scanning pattern also affect usable point density. A distant detection on a bright target is not equivalent to reliable obstacle perception on a dark, small or oblique target.

For navigation, point density and vertical coverage may matter more than the farthest possible return. The mounting position, minimum range, blind zone and motion distortion also belong to the system record.

A fair comparison protocol

A cross-sensor test should publish enough detail for another lab to reproduce the scene.

• Target material, reflectivity, size, angle and distance grid.
• Ambient illumination, temperature, weather and airborne obscurants.
• Resolution, frame/scan rate, exposure, laser power, return mode and filtering.
• Confidence threshold, valid-pixel or valid-point criterion and error metric.
• Host, firmware, SDK, synchronization and calibration versions.
• Raw-data retention and the distinction between detection, accuracy and precision.