Key takeaways
- Resolution and frame rate share one resource. Your camera can only move so much data through its cable. More detail per image or more images per second both consume that limited resource, so adding one can cost you the other.
- Size resolution from the feature and the field of view (FOV), not the megapixel count. Aim for three to five pixels across the smallest defect you need to catch, and 10 or more pixels for dimensional measurement.
- Don’t focus on a single aspect of a camera. The sensor, lens, lighting, and bandwidth all must work together, and pushing one to its limit strains the others. Define your application needs, the part, the speed, the smallest defect, and let that drive every component choice.
Why resolution and frame rate compete for the same bandwidth
Resolution and frame rate both draw from one shared, limited resource, sensor data throughput. In turn, data throughput is limited by camera’s interface, the cable and protocol that regulates how data is exchanged between devices. There is a limited amount of data that can pass through a camera’s interface, so it has to share that capacity between resolution and frame rate. The governing relationship is simple:
Cognex Expert Insight: How do you explain the tradeoff between sensor resolution and frame rate?
Grant Wilson, Consulting Applications Engineer: "Generally, the higher the resolution the lower the frame rate. You need to choose a suitable resolution for the specific application.
Finding the right resolution is straightforward. First, look at the smallest feature or defect size and then the size of the part and the amount of part-to-part positional variation. With this information, you can estimate field of view and what resolution you need.
It gets more challenging when the speed of the application, measured in parts inspected per second, increases. If your camera needs to process more frames per second (FPS) to keep up with high-speed operations, you might need a higher resolution senor more bandwidth capacity. This can increase vision controller system complexity. These days, there is usually a sensor solution available which can meet the applications requirements."
How does sensor resolution affect area scan camera performance?
Sensor resolution determines the smallest detail an area scan camera can identify across its field of view. The metric that matters is spatial resolution: the physical size each pixel represents on the part, not the raw megapixel count. A high-megapixel area scan vision system can still miss a defect if the field of view spreads those pixels too thin.
These formulas help you find spatial resolution and the required sensor resolution:
- Spatial resolution = smallest feature size ÷ pixels-per-feature
- Required sensor resolution = FOV ÷ spatial resolution
- Quick estimate of the minimum pixels needed = (field of view ÷ smallest feature) × 2
For example, to locate a 0.25 mm feature (smallest feature size) four pixels (pixels per feature) across inside a 40 mm FOV, you need a spatial resolution of 0.0625 mm per pixel. That works out to a minimum of 640 pixels across that axis:
An area scan camera with the same number of megapixels (MP) delivers different spatial resolution depending on the field of view. A 12 MP area scan system inspecting a 600 mm area resolves about 0.149 mm per pixel. That same vision system inspecting a 60 mm surface resolves at 0.0149 mm per pixel, 10 times finer, with no hardware change. The right question isn't "how many megapixels do I need," it's "what feature, at what field of view."
Resolution carries a hidden cost, too. Adding more pixels onto the same sensor forces a smaller pixel pitch, meaning each pixel is physically smaller and sits closer to its neighbors. Smaller pixels enable higher spatial resolution but collect fewer photons per pixel, which can reduce sensitivity, meaning you may need brighter lighting or a longer exposure to compensate. That longer exposure is exactly what a fast production line can't afford, which is how a resolution decision can quickly become a frame rate problem.
Cognex Expert Insight: When can sub-pixel processing improve an under-resolved image?
Saravanan Ramachandran, Supervisor, Applications Engineering: "Sub-pixel processing alone cannot enhance an image. The image itself needs to have enough resolution to display the required details for the application.
Sub-pixel processing can produce extremely sharp images, making it easier to see fine details. That capability makes sub-pixel processing particularly useful edge detection, measurement, gauging, and pattern-matching applications.
The pitfall is relying on sub-pixel processing as a crutch. You need to fine-tune your setup to get the most out of sub-pixel processing. That means the right working distance, lens and lighting configuration to minimize noise. You can’t correct those problems with sub-pixel processing."
How does frame rate affect area scan camera performance?
Frame rate impacts two outcomes for an area scan camera: throughput and motion blur. Throughput is the number of parts your inspection system processes in a specific period. Motion blur is one of the problems teams underestimate most. If the part moves too fast while the camera is still exposing the image, its edges blur and the inspection loses accuracy. A simple formula predicts how much blur you can expect:
Generally, anything past one pixel starts to cause problems, and precise measurement can fail even at those levels.
This tradeoff has consequences. Shorter exposure freezes motion but reduces the light reaching the sensor, so you compensate with brighter, well-controlled illumination. Higher frame rates capture faster lines but increase data and processing load. A high frame rate area scan camera without matching lighting and bandwidth will become a bottleneck.
Cognex Expert Insight: Can you share an example from the field about a failure involving motion blur?
Grant Wilson: "I recently worked on a project to read codes on food pots traveling up to five meters per second. The customer wanted continuous or "always on" lighting as strobing the light would require shielding to avoid operator exposure to potentially harmful flickering.
The brightest light available was used in continuous mode, however the exposure time required on the camera was almost the minimum the sensor would allow. The image was dark and contrast was poor, increasing gain setting also increased noise in the image.
We used high-dynamic range technology on the camera to help boost the contrast but even then, it was lower than I would have liked. Sometimes the only solution is a more sensitive camera with larger pixels, better efficiency, or to override the light.
This is a situation that normally occurs when no feasibility testing has been performed prior to deploying an application. Proof of concept is a vital step in preventing these issues."
Do you need a global shutter for moving parts?
A rolling shutter exposes the sensor row by row, which can cause fast-moving objects to appear skewed or distorted. A global shutter exposes every pixel at once and freezes motion without that distortion. Moving parts don't automatically require a global shutter, but it’s a safe default for applications with fast motion.
Historically, global-shutter cameras had a greater distance between pixels, giving them a lower resolution compared to rolling-shutter vision systems. While global-shutter cameras are getting more powerful, there’s generally still a compromise between motion blur and resolution.
Global vs. rolling shutter machine vision systems
| Factor | Global shutter | Rolling shutter |
|---|---|---|
| Motion handling | Freezes fast motion, no skew | Skew and distortion on fast motion |
| Best fit | High-speed lines, moving parts, flash sync | Slow or indexed motion, static parts, cost-sensitive applications |
| Historic tradeoff | Larger pixels, resolution limited | Smaller pixels, higher resolution |
| Modern shift | Stacked BSI designs shrink pixels without the penalty | Still strong where speed isn't critical |
Cognex Expert Insight: Where does the machine vision industry over-specify global shutter?
Grant Wilson: "I’ve worked in the machine vision industry for some time, so I’m instinctively drawn to global shutters where there’s a lot of motion in the field of view. Global shutters require less strobing light, which also bolsters operator safety.
However, rolling shutter cameras are great for indexed applications where the product is momentarily stationary during image capture. Using a rolling shutter with a strobe may require shielding to prevent operator exposure to flashing light. Shielding also helps control the low level of ambient light required before and after the strobe, which in turn helps compensate for longer rolling shutter exposure times.
My recommendation is to do the math: work out the required exposure time as normal and then the readout time of the rolling shutter sensor. If this is going to be significant relative to the distance traveled, maybe global shutter is still preferable."
The interface bandwidth limit that decides everything
The camera interface, the combination of the physical cable and data protocol, sets the physical limit on how much resolution and frame rate you can run at once. Every interface limits data throughput, and once your pixels-times-frames product exceeds that limit, you must reduce resolution, reduce the frame rate, or upgrade the interface.
The interface, not the sensor alone, decides real-world area scan camera performance.
| Protocol | Approx. usable bandwidth | Max cable run | Frame grabber | Typical fit |
|---|---|---|---|---|
| GigE Vision (1 Gb) | ~125 MB/s | 100 m | No | Most standard area scan camera applications |
| USB3 Vision | ~400 MB/s usable | ~3 to 5 m | No | High frame rate at moderate resolution, short runs |
| 5GigE / 10GigE | ~600 MB/s to ~1.2 GB/s | Up to 100 m | 10GigE often needs a network card | High-resolution and high-speed without a grabber |
| CoaXPress (CXP-12) | Up to ~12.5 Gb/s per coax, scalable | 100+ m | Yes | Ultra-large area-scan camera, top-end speed |
| Camera Link | Up to ~850 MB/s | Short | Yes | Legacy high-end, now yielding to 5/10GigE |
Cognex Expert Insight: How often do area scan applications require more than GigE vision?
Saravanan Ramachandran: "For most everyday inspection, GigE Vision is enough. If I'm running a low-resolution sensor, around 2 MP, a GigE area scan camera moves the images comfortably, and that covers a large share of the work I see. But when the resolution is too high, GigE can't move the images fast enough, and I switch to a 5GigE or 10GigE interface."
Where do area scan cameras fit in traceability and industry 4.0 operations?
An area scan camera rarely works alone. It’s part of a broader ecosystem where scanners, sensors, and AI systems work together to capture inspection and part information and connect it to manufacturing systems. Understanding how resolution and frame rate affect image quality provides a foundation for the decisions that follow.
The same imaging fundamentals apply across the traceability chain. Barcode readers, for example, also depend on having sufficient resolution across the field of view and capturing moving codes clearly, creating a similar balance between resolution and frame rate.
AI-powered solutions can extract more information from each image, which can influence how much resolution and frame rate you actually need. When these systems work from consistent, well-resolved image data, they can provide the reliable information needed to build true end-to-end traceability across an Industry 4.0 operation.
Stop buying specifications and start budgeting data
Resolution and frame rate are not independent settings. They are two aspects that determine how a vision system operates and share one limited bandwidth, which is determined by the cable and data protocol. Determining the balance between resolution and frame rate comes down to a straightforward formula:
- Size resolution from the feature and the field of view.
- Set frame rate from line speed and motion blur.
- Confirm the interface can carry the data load.
The teams that get area scan camera performance right aren't the ones with the biggest sensor or the fastest shutter. They're the ones who define what the application needs, then spend their data budget to match. Start from the part, work outward to the interface, and let the numbers rather than the datasheet decide. That is how a camera specification turns into a reliable inspection system.