Small pixel camera selection: resolution versus usable detail

A smaller pixel can reveal finer defects, but only when the lens, illumination and acquisition chain preserve that extra information. For buyers replacing an existing camera, the useful question is not which sensor is newer. It is whether the replacement improves inspection at the required field of view, exposure and production rate. Start with the smallest feature you must distinguish, then check optical contrast, collected light and delivered frames before choosing hardware.

Decide what the replacement must improve

Small pixel camera selection should begin with an inspection requirement, not a sensor architecture. Write down the smallest relevant defect, required field of view, part speed and acceptable decision error. Also record the existing camera's exposure, image dimensions, lens aperture and processing time.

Smaller pixels provide finer sampling at unchanged magnification. They do not automatically produce sharper images, more reliable measurements or better low-light performance. A replacement can deliver more pixels while preserving the same blurred edges.

Backside illumination places the light-sensitive structure ahead of much of the wiring relative to incoming light. This can improve light collection, but the architecture alone does not establish quantum efficiency, noise or saturation capacity. Those need camera-specific data under comparable conditions.

Likewise, global shutter describes exposure timing across the image, not sensitivity or resolving power. It avoids the row-dependent timing of rolling shutter, but moving parts can still blur during exposure.

Use this decision table to identify the likely constraint before comparing area scan cameras.

Calculate required sampling before choosing pixel pitch

The two useful quantities are object-space sampling and pixels across the smallest feature. Pixel pitch only becomes meaningful after you account for magnification.

Use consistent units in each calculation. The object-space pixel size explanation shows how sensor pitch and magnification connect to the part.

Consider a hypothetical replacement with unchanged optics and a 120 mm field width. Both sensors have a 12 mm active width, but their pixel pitches differ:

• With 6 µm pixels, horizontal sampling is 12 mm / 0.006 mm = 2,000 pixels.
• With 3 µm pixels, horizontal sampling is 12 mm / 0.003 mm = 4,000 pixels.
• Object-space sampling changes from 120 mm / 2,000 = 0.060 mm per pixel to 120 mm / 4,000 = 0.030 mm per pixel.

A 0.15 mm feature therefore spans 2.5 pixels in the first image and 5 pixels in the second. That is a meaningful sampling improvement, provided the feature has sufficient contrast.

Five pixels across a defect is a starting assumption for this example, not a universal acceptance rule. Low-contrast flaws, dimensional tolerances and variable edge shapes can require different sampling. Establish the requirement with representative parts and the intended algorithm.

For the same example, the required width is 120 × 5 / 0.15 = 4,000 pixels. Use the calculator to explore alternative fields and feature sizes.

The dimensional inspection resolution guide explains why pixel count alone cannot establish measurement accuracy.

Check whether the lens delivers the added detail

Smaller pixels raise the spatial frequencies that the imaging system can sample. The lens must provide useful contrast at those frequencies across the required image area, not just at the center.

For a regular repeating pattern, the sensor sampling limit is its Nyquist frequency:

These are sampling limits, not guaranteed resolved detail. Contrast near the limit can be weak, and the result depends on pattern orientation and alignment with the pixel grid. Color sampling and image processing add further considerations.

Ask for lens modulation transfer function data at the intended aperture, working distance and image height. A lens resolution label is useful for screening, but it does not specify contrast at every frequency or operating condition.

The geometry also needs checking. Changing active sensor dimensions changes field coverage unless you alter magnification. Matching a nominal sensor format does not replace checking actual active width and height.

Stopping down can improve aberrations and depth coverage, but diffraction eventually reduces fine-detail contrast. For an ideal circular aperture, the diameter to the first dark ring of the diffraction pattern is approximately:

This is an optical estimate, not a pass/fail threshold. At close working distances, effective f-number can differ from the marked aperture. Test the actual lens setting rather than assuming a diffraction spot covering several pixels makes those pixels useless.

Compare machine vision lenses through the lens finder, then use the lens selection guide to check coverage, working distance and aperture together.

Compare the photon budget at the permitted exposure

At equal image-plane irradiance, exposure time and quantum efficiency, collected electrons scale approximately with pixel area. Halving square-pixel pitch reduces that area to one quarter.

Suppose a hypothetical larger pixel collects 4,000 electrons during the allowed exposure. With otherwise equal conditions, the smaller pixel collects about 1,000 electrons. Their shot-noise-limited signal-to-noise ratios are approximately √4,000 = 63 and √1,000 = 32.

That is a per-pixel comparison. Combining adjacent small pixels over the same image area can recover similar total photon statistics, although read noise and processing affect the result. It also gives up some of the finer spatial sampling that motivated the replacement.

Backside illumination may improve the quantum-efficiency term, but do not assume it completely offsets reduced area. Compare efficiency at the illumination wavelength, not only at the peak of a spectral curve.

Also compare saturation capacity, temporal read noise and dynamic range under matched settings. The EMVA 1288 camera comparison guide explains how to read those measurements without confusing gain or output bit depth with captured information.

Include motion and delivered frame rate in the calculation

A global shutter does not freeze motion by itself. Exposure must still be short enough for the permitted object-space blur.

At 500 mm/s and 0.030 mm per pixel, a half-pixel blur allowance gives:

0.5 × 0.030 mm / 500 mm/s = 0.000030 s = 30 µs.

With 0.060 mm per pixel, the same half-pixel allowance permits 60 µs. Keeping blur below a fixed pixel fraction can therefore require a shorter exposure after increasing sampling density. Confirm that this stricter blur target actually benefits the inspection.

The moving-part exposure guide covers this trade-off in more detail. More illumination may be needed to maintain signal during the shorter exposure.

Larger images also increase transfer and processing demand. For hypothetical uncompressed monochrome images:

This excludes transport overhead and assumes one transmitted byte per pixel. Packed formats, unpacked higher-bit-depth formats and color conversion change the calculation.

At the same frame rate, a 2,000 × 1,500 image requires 90 MB/s under those assumptions. Doubling both image dimensions creates four times the payload.

Measure completed inspections per second, not just camera output. Include transfer, conversion, processing, storage where required, and decision delivery. If acquisition falls behind, follow the camera frame-rate bottleneck checks before changing the sensor again.

Shortlist parts without assuming unlisted sensor properties

The catalog examples below provide starting points for shutter, interface and optical compatibility checks. Their supplied catalog entries do not establish pixel pitch, backside illumination, quantum efficiency or lens MTF. Request those details before treating any example as a small-pixel upgrade.

The MV-CU020-90GC is listed as a global-shutter color camera with a GigE interface and C-mount. It is a candidate for interface and shutter screening, not proof of a particular pixel architecture.

The MV-CH120-11UC is listed as a 12 MP global-shutter color camera with a USB3 interface and C-mount. Verify its detailed timing, pixel pitch and signal measurements against the application requirements.

For an optical shortlist, the LS-FFL-0567 is listed as a 16 mm, 1.1", 20 MP C-mount machine vision lens. Its catalog resolution label does not prove adequate contrast for an unspecified pixel pitch.

These examples are not a prequalified camera-and-lens combination. Confirm actual sensor dimensions, image-circle coverage, required magnification, mount clearance and resolving performance before pairing them.

Use the camera finder to filter the required shutter and interface. Carry the resulting sensor dimensions into the lens selection rather than selecting each part independently.

Verify improvement with a controlled replacement test

Run an A/B evaluation using the same parts, field of view and production motion. Use a rigid mounting arrangement and record focus, aperture, illumination settings, exposure, gain and pixel format for each capture.

A useful initial set might contain 30 acceptable parts and 30 defective parts spanning known variation. That is a screening exercise, not statistical proof of production capability. Expand validation according to defect prevalence and the consequences of an incorrect decision.

Include the following tests:

1. Capture static parts with low gain to isolate sampling and optical contrast from motion.
2. Check the smallest defects at the center and corners, including the full part-height range.
3. Repeat at production speed using the permitted exposure and intended lighting.
4. Compare native images and images resampled to matching object-space sampling.
5. Measure false accepts, false rejects, missed frames and completed-decision latency.
6. Repeat after normal machine warm-up and across expected illumination variation.

Do not normalize brightness by clipping highlights or applying aggressive sharpening. Keep an unprocessed reference where the acquisition system permits it, and document every processing difference.

The lens and lighting test plan provides a useful structure for controlling these variables. A sharper-looking display is not sufficient evidence; the inspection decision must improve.

Know when larger pixels remain the better choice

Retain a larger-pixel camera when the existing sampling already supports the inspection and the real limit is collected light. This is especially relevant when exposure cannot increase and additional illumination is impractical.

Retain the existing camera when lens contrast, depth variation or motion blur prevents finer detail from reaching the sensor. Replacing the camera alone will only sample the same blur more densely.

A larger active sensor can sometimes provide both adequate pixel area and more sampling, but it may require a larger image circle or different optics. Evaluate the complete replacement cost and mechanical consequences, not only the camera body.

A smaller-pixel camera is more attractive when defects are undersampled, optics remain capable, and sufficient light reaches the sensor within the exposure limit. It also needs an acquisition chain that can deliver and process the additional data.

If color is not part of the decision, consider whether monochrome acquisition would simplify the photon and spatial-sampling requirements. Confirm this through application images rather than assuming a universal improvement.

Replacement checklist and next step

Before ordering, confirm the following against the actual inspection:

• Required field width and height: both dimensions must cover part position variation.
• Smallest feature and sampling target: state how many pixels it must span and why.
• Actual pixel pitch and sensor dimensions: nominal format alone is insufficient.
• Shutter and exposure behavior: verify motion performance at production timing.
• Lens coverage and contrast: check the intended aperture, working distance and image corners.
• Signal performance: compare wavelength-specific efficiency, read noise and saturation data.
• Data format and payload: calculate transfer demand using the transmitted representation.
• Sustained decision rate: test the full inspection path rather than an acquisition preview.
• Mechanical compatibility: check mounts, lens clearance and the available working-distance envelope.
• Acceptance evidence: retain representative images, settings and decision statistics.

If the replacement passes these checks, begin with the camera configurator. If pixel pitch, optical contrast or short-exposure signal remains uncertain, send your field size, defect size, part speed and representative images through applications support before selecting the final combination.