Camera and Lens Selection Guide for Machine Vision Applications

Camera and Lens Selection Guide for Machine Vision Applications

This technical guide helps automation engineers select cameras, lenses, and supporting optics for machine vision systems. It consolidates sensor and optics fundamentals, practical calculation methods, interface/standard requirements, product-class examples, and field-proven implementation strategies. The goal is to enable reliable inspection, barcode/OCR, robotic guidance, and high-speed capture by matching sensor format, pixel pitch, resolution, field of view (FOV), working distance (WD), depth of field (DoF), and spectral response to the target task.

Key Concepts

Sensor Formats, Pixel Pitch, and Resolution

Start with the camera sensor because the sensor defines the pixel count, pixel size (pitch), and physical sensor dimensions that drive lens selection. Common optical-format sensor sizes include 1/3", 1/2", 2/3", and 1" formats; these are nominal industry names where the quoted "inch" is an optical format, not a mechanical measurement. For example, a typical 2/3" CMOS sensor (Sony IMX250 class) measures roughly 8.8 mm × 6.6 mm with a diagonal near 11 mm and pixel pitches commonly in the 3–4 µm range for 5 MP parts cameras. Smaller pixels increase the demand for higher resolving power from the lens (higher MTF) to avoid limiting system resolution. According to Resolve Optics and Edmund Optics, designers should match lens resolution (MTF) to pixel size: small pixels (<3 µm) often require lenses with MTF >120 lp/mm, whereas older low-resolution systems (<1 MP, large pixels) can achieve acceptable performance with lenses rated 70–90 lp/mm.

Reference: EMVA 1288 provides standardized metrics for comparing camera performance (quantum efficiency, dynamic range, noise) and should be used when benchmarking camera candidates for demanding applications.

Field of View (FOV), Working Distance (WD), and Focal Length

Optical selection centers on the relationship between sensor size, desired FOV, WD, and required focal length. Use the thin-lens approximation and the practical focal-length equation:

f = (WD × sensor dimension) / FOV

Here f is the focal length, sensor dimension is the relevant physical sensor width (for horizontal FOV) or height (for vertical FOV), WD is the distance from the lens principal plane to the object plane, and FOV is the physical size of the scene you must image in the same axis. For example, imaging a 175 mm horizontal FOV at 500 mm WD on a 2/3" sensor (~8.8 mm width) gives:

f ≈ (500 mm × 8.8 mm) / 175 mm ≈ 25.1 mm

This matches practical catalog increments such as 25 mm. In practice, select the next available catalog focal length (3.5, 4.5, 6, 8, 12, 16, 25, 35, 50, 75, 100 mm) and verify full-sensor illumination and minimal vignetting at the intended WD.

See Edmund Optics' imaging guides for worked examples of these calculations and field plots used to choose lenses and mounting geometry.

Lens Mounts and Mechanical Interfaces

The most common industrial mounts are C-mount and CS-mount. C-mount has a thread diameter of 25.4 mm, 32 TPI (0.794 mm pitch), and a flange focal distance of 17.526 mm. CS-mount uses the same thread diameter and pitch but a shorter flange focal distance of 12.5 mm; CS lenses can be adapted to C-mount by adding a 5.026 mm spacer, but C-lenses will not focus correctly on CS cameras without modification. Choose mounts that support the sensor format (some larger 1" lenses require larger mount/bayonet arrangements) and maintain mechanical stability to withstand vibration and alignment challenges on production lines.

Modulation Transfer Function (MTF), Distortion, and Vignetting

MTF characterizes lens contrast as a function of spatial frequency (line pairs per millimeter). For accurate measurement tasks (edge detection, OCR), ensure the lens MTF at the Nyquist frequency of your camera remains adequate. A rule of thumb from industry white papers is to specify lens MTF at least 20–50% higher than required so the optics do not become the limiting factor; many designers over-spec optics by 20–50% to future-proof systems. Also factor distortion (pincushion/barrel) for dimensional measurement and vignetting for even illumination across the sensor.

Spectral Response: Visible, NIR, and SWIR

Select optics that transmit the wavelengths used in your illumination. Standard visible-range machine vision lenses work well for 400–700 nm. For near-infrared (NIR) or short-wave infrared (SWIR) imaging, choose lenses with coatings and glass optimized for those bands; some lenses designated for NIR have different MTF and distortion behaviors at longer wavelengths. Consult manufacturer spectral transmission curves when matching camera sensors (silicon vs. InGaAs) and light sources.

Implementation Guide

Stepwise Selection and Design Workflow

• Define the task: Identify measurement accuracy, smallest feature size, object speed, and nominal WD and FOV.
• Choose the camera first: Select sensor format, resolution (pixels across FOV), pixel pitch, dynamic range, frame rate, and interface (GigE Vision, USB3 Vision, Camera Link, CoaXPress). As Basler and other vendors recommend, use EMVA 1288 specs to compare performance objectively.
• Compute focal length: Use f = (WD × sensor_dim) / FOV to find a target focal length, then pick the closest catalog focal length. Verify required DoF and MTF at that focal length.
• Select the lens: Choose a lens with full-sensor coverage for the sensor format, adequate MTF at the pixel Nyquist frequency, acceptable distortion, and required spectral transmission. Prioritize fixed focal lenses for static WD inspection and varifocal/zoom for flexible architectures.
• Design lighting and aperture: Determine required exposure using lens f-number and available illumination; close aperture to increase DoF but compensate with more light or higher camera gain as needed.
• Prototype and test: Validate system on a physical sample with the exact camera/lens/lighting combination to check resolution, vignetting, contrast, and DoF. Adjust focal length, aperture, or WD if issues appear.

Interfaces, Standards, and Performance Metrics

For industrial integration, choose cameras that support well-adopted machine vision protocols: GigE Vision for long cable runs over Ethernet and USB3 Vision for high throughput over short runs. Use EMVA 1288 data sheets to validate camera performance (sensitivity, noise, dynamic range) and consult manufacturer MTF and resolution charts for lenses. Although standards like IEC or IEEE do not define optics-specific rules, industry de facto standards (C-/CS-mount, GigE Vision, USB3 Vision, EMVA 1288) ensure interoperability across suppliers.

Practical Example: Calculating Focal Length and Verifying MTF

Design requirement: horizontally capture 175 mm of a conveyor at WD = 500 mm using a 5 MP camera with a 2/3" sensor (sensor width ≈ 8.8 mm) and pixel pitch ≈ 3.45 µm (Sony IMX250 class).

Calculation: f = (500 × 8.8) / 175 = 25.14 mm → select a 25 mm lens.

Next, estimate sampling frequency: pixel pitch 3.45 µm corresponds to Nyquist frequency ≈ 1 / (2 × 3.45 µm) ≈ 145 lp/mm (line pairs per mm on the sensor plane). Verify the chosen 25 mm lens MTF retains adequate contrast at that frequency; if not, increase focal length (reduces FOV) or select a lens with higher MTF rating (e.g., Thorlabs high-resolution series or Computar high-resolution lenses). For most inspection tasks where feature sizes exceed several pixels, lens MTFs in the 70–120 lp/mm range may suffice depending on pixel pitch and required accuracy. See Resolve Optics and Edmund Optics white papers for MTF testing methods and required design margins.

Best Practices

Sequence: Camera First, Lens Second

Always select the camera sensor and interface before the lens. The sensor size and pixel pitch determine the lens family and required coverage. Selecting the lens first risks incompatibility (e.g., wasting high-resolution optics on a large-pixel camera or starving a small-pixel camera with a low-MTF lens).

Choose Fixed Focal for Stability, Zoom for Flexibility

Fixed focal-length lenses deliver superior sharpness, lower distortion, and better MTF for a given price; they are the preferred choice for static WD inspection and high-precision measurement. Varifocal or zoom lenses suit development labs or applications where WD and FOV must change without swapping optics. Thorlabs and Computar publish product lines that include both choices; where repeatable settings are required, lock the focus and aperture to avoid drift.

Match Lens Resolution to Pixel Size and Over-spec When Practical

Match lens resolving power to camera Nyquist frequency and application accuracy. Many integrators choose a lens that exceeds the camera resolution by 20–50% to avoid optical limits becoming the failure mode and to accommodate future camera upgrades. For small pixels (<3 µm), select lenses whose MTF at the relevant spatial frequencies is high and maintain contrast across the image field; 1stVision and Resolve Optics recommend testing lenses on actual sensors to quantify system MTF.

Design Lighting Around Aperture and DoF

Balance aperture (f-number) and illumination. A smaller aperture increases DoF but reduces light; compensate with stronger, more uniform lighting or longer exposures in static applications. For moving objects, use strobe illumination synchronized with camera frames to freeze motion and allow smaller apertures. Consider NIR or polarized lighting to improve contrast for certain materials, and use diffusers and backlights where silhouette or edge detection is required. The Computar blog and Edmund Optics lighting notes provide practical patterns and examples.

Mechanical Stability, Calibration, and Environmental Protection

Mount lenses rigidly with vibration-tolerant brackets and anti-rotation features. Include calibration targets and procedures (dot grids or slanted-edge charts) to measure system MTF, geometric distortion, and lens centering. For industrial environments, protect optics with appropriate housings and lens seals; consider IP ratings, desiccants, and anti-reflective coatings to reduce contamination and glare. Use proper cable strain relief for high-speed interfaces like CoaXPress or Camera Link to maintain stable connections under motion.

Validation and Test Protocols

Establish repeatable validation tests: resolution charts (USAF 1951 or slanted-edge methods per ISO 12233), contrast and lighting uniformity mapping, and defect detection sensitivity runs on representative parts. Record EMVA 1288 and MTF data for each camera-lens combination to allow future comparisons and audits. As a practical field rule, if the system fails to detect features consistently, re-evaluate the least expensive change first: lighting geometry and intensity, then aperture, then lens, and finally camera sensor choice.

Product Comparison and Typical Specifications

Manufacturer	Product Line / Type	Typical Focal Lengths	Max Sensor Format Coverage	Notes / Compatibility
Thorlabs	Fixed focal length & high-res optics	3.5–100 mm; zoom 18–111 mm	Up to 1" formats; high MTF lenses for 10 MP	Good for laboratory & high-resolution factory setups. See Thorlabs catalog for MTF charts[3].
Tamron	M118 C-mount machine vision series	Typical ML focal options like 12, 25, 35, 50 mm	1/3"–1" depending on lens	Optimized for pixel sizes <5 µm; suitable for visible/NIR[2].
VST (generic series)	Standard machine vision lenses	3.5–100 mm	1/3"–1"	Low distortion, cost-effective for production lines[2].
Computar	Machine vision fixed/varifocal/zoom	Wide range; F1.4–F16 aperture options	1/3"–1"	Documented application notes on aperture vs. resolution trade-offs[6].

Summary

Effective machine vision begins with selecting the right camera sensor and then choosing a lens