Target scenarios: High-speed production line inspection · Low-light dark-field imaging · Precision defect detection · Multi-camera synchronized systems
Core challenge: Frame rate and exposure time compete for the camera's per-frame "time budget" — imbalance causes "captured but unclear" or "clear but missed" results.
1. The Core Conflict: Frame Rate vs Exposure Time
In any machine vision inspection system, frame rate (fps) and exposure time (μs/ms) are fundamentally interdependent parameters — they share the camera's per-frame "time budget."
1.1 Key Definitions
| Parameter | Definition | Primary Function | Typical Unit |
|---|---|---|---|
| Frame rate (fps) | Number of images captured per second | "Keep up with the target" — no missed captures, no motion blur | fps |
| Exposure time | Duration the sensor collects light per frame | "Capture clear details" — sufficient illumination, visible gray-level contrast | μs or ms |
- Insufficient frame rate: Missed parts on moving conveyors, or image tearing/blur on fast-moving targets. Frame rate directly determines inspection completeness.
- Insufficient exposure: Dark images with elevated noise, where defect gray-level contrast (scratches, coating gaps) is masked — causing sharp increases in algorithm misclassification rates.
1.2 The Time Conflict
Per-frame total time = Exposure time + Data transfer time + Sensor readout time
This total must be ≤ Frame interval (i.e., 1/frame rate), otherwise frame drops or stuttering occur.
Frame interval = 1/fps → Exposure ≤ Frame interval − Transfer time − Readout time
Example: At 30fps, the frame interval is ~33ms. Set exposure to 20ms → only 13ms remains for transfer + readout. Force exposure to 30ms → just 3ms left for transfer + readout — with insufficient interface bandwidth, frame drops are inevitable.
This is the push-pull relationship: higher frame rate → shorter per-frame time → exposure constrained; longer exposure → frame rate harder to increase.
2. Three Consequences of Imbalance
Approximately 90% of industrial vision inspection failures trace back to frame rate–exposure imbalance. The consequences fall into three categories:
| Imbalance Type | Symptom | Typical Scenario | Outcome |
|---|---|---|---|
| ⚡ High fps, insufficient exposure | "Captured, but unclear" | Forcing max frame rate on fast lines (e.g., 500fps), exposure compressed below 10μs | Dark/black images, blurred defects, misclassification spikes |
| 🔍 Long exposure, low fps | "Clear, but missed" | Low-light environments extending exposure to 50ms+, fps drops below 20 | High miss rate, severe motion blur |
| 🔧 Ignoring transfer/readout | "Looks good on paper, stutters in practice" | Calculations satisfy frame interval, but data transfer time unaccounted | Frequent frame drops (e.g., 1 out of 10 lost) |
Real-World Examples
- 📦 Battery cell inspection: 200fps with 3μs exposure → coating gap gray-level contrast drops dramatically, algorithm misclassification rate spikes
- 🚗 Auto parts inspection: 30ms exposure highlights dent shadows, but insufficient frame rate increases miss rate
- 🖥️ High-pixel transfer stutter: 15fps (66ms interval), 30ms exposure, appears to leave 36ms — but 20MP image transfer requires 40ms → 1 frame lost every 10
3. The 3-Step Balancing Method
Set frame rate floor → Calculate exposure ceiling → Address hardware gaps
No complex math required — follow the steps.
Step 1: Set the Frame Rate Floor — Prioritize No Missed Captures
Frame rate must capture target motion. Calculate minimum frame rate first, then add safety margin.
Minimum fps = Conveyor speed (m/s) ÷ Shortest part dimension (m)
Line scan cameras reference line frequency — higher line frequency supports faster motion.
| Speed Range | Margin Recommendation | Priority Selection |
|---|---|---|
| High speed (>2m/s) | ≥ 30% margin | Global shutter camera (eliminate motion blur) |
| Medium speed (0.5–2m/s) | 20–30% margin | Choose shutter type based on precision needs |
| Low speed (<0.5m/s) | Can reduce margin | Reserve more time for exposure |
Example: Conveyor 0.8m/s, shortest part edge 0.08m → min fps = 10fps → with 20% margin → set 12–13fps.
Step 2: Calculate the Exposure Ceiling — Then Ensure Clear Images
Exposure cannot exceed frame interval minus transfer and margin time.
Max exposure = 1/fps − Transfer time − 5% margin
| Interface | 20MP Transfer Time | 10MP Transfer Time |
|---|---|---|
| USB 3.0 | ~5ms | ~3ms |
| GigE | ~8ms | ~6ms |
| CoaXPress | ~2ms | ~1ms |
Practical Adjustment Rules
| Lighting Condition | Exposure Setting | Notes |
|---|---|---|
| 🌞 Bright-field (e.g., PCB traces) | 70–80% of maximum | Avoid overexposure washing out details |
| 🌙 Dark-field (e.g., metal scratches) | Up to 90% of maximum | Ensure sufficient light for defect contrast |
| ⚠️ Image too dark | Add lighting first, don't just extend exposure | Avoid sacrificing frame rate for brightness |
Step 3: Address Hardware Gaps — Resolve Unresolvable Conflicts
If Steps 1 and 2 still produce insufficient exposure or frame rate, hardware optimization is needed:
| Core Conflict | 🔧 Hardware Solutions | Applicable Scenarios |
|---|---|---|
| Insufficient exposure (dark images) | 1️⃣ Add high-brightness lighting (bar light / pulsed light) 2️⃣ Use larger sensor format camera (more light collection) 3️⃣ Select low-noise sensor (better low-light performance) | Low-light environments, high-speed scenes |
| Insufficient frame rate (missed captures) | 1️⃣ Upgrade to faster interface (CoaXPress is 3× faster than GigE) 2️⃣ Reduce image resolution (less data to transfer) 3️⃣ Use global shutter camera (no blur, suited for fast motion) | Long-exposure needs, high-resolution scenes |
| Transfer time too long (stuttering) | 1️⃣ Enable image compression (JPEG reduces transfer by 50%) 2️⃣ Use PoE+ interface (power + data, reduced interference) 3️⃣ Optimize software processing (reduce data redundancy) | High-pixel cameras (≥20MP), multi-camera systems |
4. Real-World Case Studies
Conveyor: 3m/s · Defect size: 0.1mm
| Step | Calculation | Result |
|---|---|---|
| Set frame rate | Min fps = 3 ÷ 0.1 = 30fps; 30% margin → 30 × 1.3 | Set 40fps |
| Calculate exposure | USB3.0 transfer ~5ms; interval ≈ 25ms; max ≈ 18.75ms; set 70–80% | Set 15ms |
| Address gaps | Add high-brightness bar light (2× brightness increase) | 15ms exposure clearly reveals coating gaps ✅ |
Conveyor: 0.6m/s · Shadow emphasis needed
| Step | Calculation | Result |
|---|---|---|
| Set frame rate | Min fps = 0.6 ÷ 0.05 = 12fps; 20% margin → 12 × 1.2 | Set 15fps |
| Calculate exposure | GigE transfer ~6ms; interval ≈ 66ms; max ≈ 56.7ms; dark-field near ceiling | Set 50ms |
| Address gaps | Add dark-field light source + select low-noise sensor | Prevent long-exposure noise increase ✅ |
5. Top 4 Tuning Mistakes & Quick Reference
🚫 Four Most Common Tuning Errors
- Maxing out frame rate while ignoring exposure → Pitch-black images, defects invisible
- Extending exposure without accounting for transfer/readout → Stuttering, frame drops
- Relying on longer exposure instead of better lighting → Sacrificing frame rate for brightness — inefficient
- Using rolling shutter in high-speed applications → Image distortion and motion blur
⚡ Global shutter: All pixels expose simultaneously — ideal for fast motion (zero blur)
🔄 Rolling shutter: Row-by-row exposure — suited for static/low-speed scenes (lower cost)
✅ Quick Reference Formula
Set frame rate floor first, then exposure ceiling;
Add lighting before extending exposure; address hardware gaps promptly;
Global shutter for high speed, rolling shutter for low speed;
Test and fine-tune on-site — no blur, no misses, no stuttering.
6. MindVision Product Recommendations
MindVision's industrial camera lineup covers area scan, line scan, thermal, SWIR (short-wave infrared), and fiber-optic cameras. For frame rate–exposure balancing scenarios, these products offer distinct advantages:
| Application | Recommended Product Type | Key Advantage |
|---|---|---|
| ⚡ High-speed line inspection (>2m/s) | Global shutter area scan cameras | Simultaneous exposure, zero motion blur, high fps ensures no missed captures |
| 📏 Continuous motion scanning | Line scan cameras | High line frequency for fast continuous scanning, per-line exposure precision |
| 🌙 Low-light / dark-field defect detection | Low-noise area scan / thermal cameras | Low readout noise improves dark-field SNR; thermal imaging breaks visible-light limits |
| 🔬 Precision micro-defect identification | High-resolution area scan cameras | Large sensor format for maximum light collection, strong detail rendering |
| 🔗 Multi-camera synchronized systems | CoaXPress / fiber interface cameras | High bandwidth reduces stuttering; PoE+ integrates power and data |
Defining frame rate floor and exposure ceiling during project planning dramatically shortens on-site tuning time. Contact MindVision's technical team for tailored solutions.
7. Frequently Asked Questions
Published by MindVision · Content compiled from machine vision engineering practice.
Technical discussion and reasonable citation welcome. For reprints, please credit MindVision.
© 2026 MindVision. All rights reserved.
