Magnification Is Not Resolution: What Actually Determines the Smallest Visible Feature?

Engineers often start a vision project with a magnification number: “I need 20X,” “I want 400X,” or “The defect is 5 µm, so give me 1000X.” Magnification is easy to say, but it is not the right engineering specification. Magnification only tells you how large an object appears on the sensor or monitor. It does not tell you whether the system can actually distinguish that 5 µm defect from its background, whether the contrast is high enough, or whether the camera pixels are fine enough to sample the optical image.

This article translates the request for “more magnification” into the parameters that really control what a machine vision system can see: field of view (FOV), sensor size, pixel size, optical resolution, numerical aperture (NA), contrast, lighting, and the size of the smallest feature. It uses published specifications from POMEAS motorized zoom lenses, APO objectives, and industrial cameras.

Why “I need 20X” is the wrong starting point

Magnification is the ratio between image size and object size. A 20X lens makes a 1 mm object look 20 mm long on the sensor. But if the sensor has very large pixels, the 20 mm image may still be sampled by only a handful of pixels. Conversely, a lower magnification lens with a high-NA objective and a small-pixel camera can resolve finer detail than a high-magnification system with poor optics.

A more useful conversation starts with the object:

• What is the physical size of the region you need to inspect (FOV)?
• What is the size of the smallest defect or feature you must detect?
• Do you only need to see it, or do you need to measure it to a tolerance?
• What kind of contrast does the feature have against its background?

Once these are known, you can work backward to optical magnification, sensor format, and pixel size instead of picking a magnification number first.

The real specification chain

Think of the optical chain as a sequence of limits. Each stage can be the bottleneck:

Stage	What it limits	Typical source
Optical resolution	The smallest resolvable detail set by the lens NA and wavelength	Lens datasheet
Sensor pixel size	How finely the image is sampled; pixel limited systems cannot recover lost detail	Camera datasheet
Magnification / FOV	How large the object appears and how many pixels span a given physical length	Lens + sensor combination
Lighting and contrast	Whether the feature produces enough signal difference to be detected	Application setup
Measurement accuracy	The practical uncertainty when converting pixels back to physical units	Calibration + environment

If any of these stages is weaker than the others, raising the rest does not help. A 50X magnification on a low-NA lens is still optically soft. A high-resolution lens on a 1.4 µm-pixel camera can out-resolve a 5 µm-pixel camera at the same magnification.

What optical resolution actually means

Optical resolution is the smallest separation between two points that the optics can distinguish. It is governed by the numerical aperture (NA) and the illumination wavelength. For visible light, a common approximation is:

Resolution (µm) ≈ 0.61 × λ (µm) / NA

At green light (λ ≈ 0.55 µm), a lens with NA 0.34 resolves approximately 1.0 µm. POMEAS publishes an object-side resolution of 0.8 µm for its 20X plan apochromatic objective (NA 0.34), which is consistent with this estimate. The 14X motorized zoom lens family lists a resolution down to 2.80 µm at the high-magnification end, while the 6.5X family reaches 3.95 µm. These numbers are real lens limits.

Lens / objective	Magnification	NA	Object-side resolution (published)	Working distance
POMEAS 20X plan apochromatic objective	20X	0.34	0.8 µm	29.5 mm
POMEAS 14X motorized zoom lens	0.55X – 7.6X	0.023 – 0.12	14.59 – 2.80 µm	99.7 ±2 mm
POMEAS 12.5X motorized zoom lens	0.58X – 7.5X	0.025 – 0.11	13 – 3.05 µm	77.4 ±2 mm
POMEAS 6.5X motorized zoom lens	0.7X – 4.5X	0.03 – 0.085	11.18 – 3.95 µm	87 ±2 mm

Notice that the 14X and 12.5X zoom lenses reach similar resolution numbers at their high-magnification ends even though their maximum magnifications differ. Magnification alone does not determine resolution; NA does. The APO objective reaches 0.8 µm at 20X because its NA is much higher.

Pixel size vs optical resolution

A camera samples the optical image. If the optical spot is smaller than one pixel, the sensor cannot record it as a separate point. A useful rule is that the object-side pixel size should be smaller than the optical resolution, ideally by a margin, so the optical detail is properly sampled.

Object-side pixel size is calculated as:

Object-side pixel (µm) = Camera pixel size (µm) / Optical magnification

For the POMEAS CCU-050E2ABSC-85 camera (5 MP, 2448 × 2048, 3.4 µm pixel, 2/3" sensor), the object-side pixel at different magnifications is:

Optical magnification	Object-side pixel (3.4 µm / β)	Approx. FOV on 2/3" sensor
0.7X	4.86 µm	15.7 × 12.6 mm
1.0X	3.40 µm	11.0 × 8.8 mm
4.5X	0.76 µm	2.44 × 1.96 mm
7.5X	0.45 µm	1.47 × 1.17 mm
20X (APO objective)	0.17 µm	0.55 × 0.46 mm

At 0.7X the pixel is larger than the 6.5X lens optical resolution at any zoom position, so the system is pixel-limited. At 4.5X the pixel is much smaller than the optical resolution, so the lens is the limiting factor. Raising the camera resolution at 0.7X would help; at 4.5X it would not.

Lighting and contrast decide whether the feature is visible at all

Resolution formulas assume ideal contrast. In production, a 5 µm crack on a matte black surface is harder to see than the same crack on a polished metal surface. Low-angle grazing light may make the crack cast a shadow; coaxial light may wash it out. The choice of illumination can change the smallest detectable feature more than a change in magnification.

POMEAS motorized zoom lenses are available with coaxial illumination options, and the MP-series auto-focus video microscope uses a multi-zone lighting system with coaxial, fill, and ring light segments that can be adjusted independently. Lighting is not a number on a datasheet; it is a variable that has to be matched to the surface and defect type.

Worked example: detecting a 5 µm defect

A project needs to detect a 5 µm dark spot on a reflective surface. Using a 5 MP camera with 3.4 µm pixels and a 12.5X motorized zoom lens at 5.0X magnification:

• Object-side pixel = 3.4 / 5.0 = 0.68 µm
• A 5 µm defect spans about 5 / 0.68 ≈ 7.4 pixels in the image
• The optical resolution of the 12.5X lens at 5.0X is listed as 3.05 µm, so a 5 µm feature is above the lens’s optical limit

With good coaxial or dark-field lighting, a 7-pixel feature is generally detectable by both classical blob analysis and modern AI segmentation. The same system at 1.0X magnification would give a 5 µm defect only about 1.5 pixels across, making detection unreliable even though the lens still works. Magnification changed the pixel coverage; it did not change the lens NA.

If the requirement is not just detection but measurement to ±1 µm, more pixels are not enough. The system needs calibration, a stable working distance, temperature control, and a measurement error budget. See the object-side resolution vs measurement accuracy FAQ for why resolution and measurement accuracy are not the same.

When higher magnification is not the answer

More magnification is useful only while the lens NA and the lighting support it. Beyond that, you are simply enlarging blur. Do not add magnification when:

• The lens NA is already the limiting factor.
• The required FOV becomes so small that parts move out of the view.
• Depth of field becomes too shallow to keep the whole feature in focus.
• The object-side pixel is already much smaller than the optical resolution.
• The measurement uncertainty is dominated by mechanical vibration or calibration, not by the image.

In these cases, the right fix may be a higher-NA objective, a smaller-pixel camera, better lighting, a telecentric lens, or a measurement-stage redesign—not more magnification.

Key takeaways

• Magnification is a scaling ratio, not a performance guarantee.
• The smallest visible feature is set by the combination of optical resolution (NA), pixel size, magnification, lighting, and contrast.
• Compare the object-side pixel size with the lens optical resolution to find the real bottleneck.
• For detection, plan on the feature covering at least 3–5 pixels, and ideally more, depending on contrast.
• For measurement, add calibration, working-distance stability, and an error budget on top of the optical chain.

Frequently asked questions

Does higher magnification always mean higher resolution?

No. Resolution is limited by the numerical aperture (NA) of the optics and the illumination wavelength. Magnification beyond what the NA supports only enlarges the image; it does not add detail.

How small a feature can a 5 MP camera detect?

It depends on the lens magnification and the optical resolution. A 5 MP camera with 3.4 µm pixels at 5X magnification gives a 0.68 µm object-side pixel. A feature of roughly 3–5 pixels, or about 2–3.5 µm at that magnification, can be detected under good contrast.

What is the difference between magnification and optical resolution?

Magnification is the ratio of image size to object size. Optical resolution is the smallest separation between two points that the lens can distinguish. A 20X APO objective with NA 0.34 resolves about 0.8 µm, while a lower-NA lens at the same magnification cannot.

How does pixel size affect microscope resolution?

The sensor samples the optical image. If the object-side pixel size is larger than the optical detail, the detail is lost. For adequate sampling, the object-side pixel should be smaller than the optical resolution.

Can a higher-megapixel camera fix a low-resolution lens?

No. If the lens is the bottleneck, adding smaller pixels or more pixels only records a sharper image of the blur. Upgrade the optics first.

When should I choose a telecentric lens instead of a zoom lens?

For high-precision dimensional measurement where magnification must stay constant across the depth of field, a telecentric lens is usually the better choice. A zoom lens is better when the same system must inspect parts of different sizes or locate a feature at low magnification and then zoom in.

What lighting is best for small defect detection?

There is no universal answer. Dark-field or low-angle grazing light emphasizes scratches and cracks. Coaxial light emphasizes flat, reflective surfaces. The right choice depends on the surface texture, defect type, and magnification.

Where can I find a practical lens selection guide?

The industrial lens selection guide walks through matching sensor size, FOV, working distance, and accuracy, and the programmable optics overview explains how motorized zoom lenses can switch between inspection magnifications.

Need help specifying a system for a specific defect size and FOV? Contact POMEAS with your part size, defect size, and throughput requirements.