Evaluating Next-Gen Depth Sensors

V2 sensor selection is approaching. We have samples from four vendors. Time for rigorous evaluation.

Evaluation Framework

Datasheets tell one story. Real performance is another. Our evaluation covers:

Accuracy

• Depth error vs range: Measure at 0.5m, 1m, 2m, 3m, 4m, 5m
• Depth error vs angle: Center vs corners
• Systematic bias: Is error consistent and correctable?
• Random noise: Per-frame variation at static target

Resolution

• Spatial resolution: Can we resolve 1cm features at 2m?
• Depth resolution: What's the minimum detectable depth difference?
• Edge quality: Sharpness of depth discontinuities

Robustness

• Sunlight immunity: Performance at 10K, 50K, 100K lux
• Multi-path handling: Corner geometry test
• Dynamic range: Black to white surfaces
• Temperature stability: -10°C to 50°C operating

System Integration

• Power consumption: Under various operating modes
• Latency: Trigger to data ready
• Interface: USB, MIPI, custom
• Calibration: Factory vs field requirements

Test Infrastructure

Built a dedicated sensor evaluation lab:

┌──────────────────────────────────────────────────────────┐
│                  Controlled Environment                   │
│                                                          │
│  ┌────────────┐      ┌─────────────┐    ┌────────────┐  │
│  │ Light      │      │ Reference   │    │ Sensor     │  │
│  │ Sources    │      │ Targets     │    │ Under Test │  │
│  │ (variable) │      │ (calibrated)│    │            │  │
│  └────────────┘      └─────────────┘    └────────────┘  │
│                                                          │
│  ┌────────────────────────────────────────────────────┐ │
│  │ Motion Stage (sub-mm repeatability)                │ │
│  └────────────────────────────────────────────────────┘ │
│                                                          │
│  Temperature Chamber: -20°C to 60°C                     │
└──────────────────────────────────────────────────────────┘

Reference targets:

• Flat plates at calibrated distances
• Depth wedge for resolution testing
• Multi-reflectance chart (5% to 95%)
• Corner geometry for multi-path

Preliminary Results

Sensor A (ToF, indirect):

• Best accuracy (±1% at 2m)
• Struggles with sunlight above 30K lux
• Moderate power (400mW)
• Strong multi-path artifacts

Sensor B (ToF, direct):

• Good sunlight immunity (works at 80K lux)
• Higher noise floor
• High power (700mW)
• Minimal multi-path

Sensor C (Structured light):

• Best resolution at close range
• Fails above 10K lux
• Low power (250mW)
• Limited range (3m max)

Sensor D (Stereo):

• Passive (zero depth-specific power)
• Resolution depends on texture
• Struggles with uniform surfaces
• Good outdoor performance

Trade-off Analysis

No sensor wins on all dimensions. Priority ranking:

1. Must have: Outdoor operation (>50K lux survival)
2. Must have: Range to 4m (room-scale)
3. Should have: Low power (under 500mW)
4. Should have: Resolution for hands (under 5mm at 60cm)
5. Nice to have: Multi-path robustness

Current leader: Sensor B, despite power concerns.

Continued evaluation through August. Final decision in September.