Confocal Sensor Resolution: Why It Depends on Range

Resolution is a function of the model's range, not a fixed value

A chromatic confocal sensor splits white light into a spectrum along the optical axis and reads height from which wavelength returns strongest. Because the same spectral bandwidth has to be spread over the whole measuring range, a longer range means each wavelength corresponds to a larger height step. Resolution and range are therefore linked by the optical design, not chosen independently.

This is the single most important thing to understand when comparing confocal sensors. Two models from the same family can differ by orders of magnitude in resolution because one measures a few millimetres and the other tens of millimetres. Quoting a resolution figure without the range it applies to is meaningless.

Across the POMEAS confocal models that publish a resolution figure, the value spans 3 to 1550 um. That spread reflects the range each model is built for rather than a quality difference between them.

The practical consequence is that the range should be chosen from the part's height variation plus margin, and the resolution then follows. Choosing a wide range for convenience and expecting fine resolution is asking the optical design for something it cannot give.

Where a wide range is genuinely required, the usual answer is a dual-range configuration or a mechanical stage that repositions the sensor between measurements.

Surface condition changes the effective resolution

A confocal sensor measures by detecting the wavelength that returns most strongly from the surface. On a rough or diffuse surface the returned spectrum is broad and the peak is poorly defined, so the height reading scatters even though the sensor's nominal resolution is unchanged. The specification is met on a suitable surface, and the application has to provide one.

On a mirror-like surface the opposite problem appears. A strong specular return can saturate the detector, and a surface tilted away from normal returns too little light. Both degrade the reading in ways the resolution figure does not describe.

The practical range of measurable surfaces is therefore part of the selection, not an afterthought. Materials with mixed finish, such as a machined face with tool marks, need to be tested rather than assumed.

Averaging helps on rough surfaces because it reduces the random component of the scatter, but it does not remove a systematic bias caused by the surface texture interacting with the measurement spot. Distinguishing the two requires measuring a known artefact of similar finish.

Testing on the actual material, at the actual angle, is the only reliable way to confirm that the nominal resolution will be achieved.

Averaging time is part of the specification

Confocal sensors are usually specified at a particular averaging setting, often many samples per reading. Increasing the averaging reduces noise and improves repeatability, at the cost of measurement time. A resolution figure quoted at a long averaging time cannot be compared with one quoted at a single sample.

For inline inspection the cycle time sets the averaging budget. A station that must produce fifty readings per second has far less room for averaging than a laboratory measurement, and the achievable resolution differs accordingly.

This is worth clarifying before comparing models, because the difference between a single-sample and a hundred-sample figure can be larger than the difference between two models.

When a supplier quotes only one number, ask for the averaging setting. If the answer is not available, the figure cannot be placed on the same scale as the alternatives.

The useful specification for an inline application is resolution at the cycle time the machine will actually run, which is a measurement rather than a catalogue value.

What resolution means for a height measurement

For a confocal sensor, resolution usually refers to the smallest height change that produces a detectable change in the reading. It is not the same as accuracy, which includes calibration and linearity errors over the full range, and it is not the same as repeatability, which is the spread of repeated readings at one point.

An application that must detect a 2 micrometre step needs resolution well below that figure, plus repeatability that keeps the noise band from crossing the decision threshold. Meeting one of the three and not the others is a common source of failed validations.

The relationship between resolution and range also affects linearity. A sensor operating near the end of its range may show a different error profile than one operating in the middle, so the working point on the range matters.

For step height and roughness measurements the relevant figure is often the vertical repeatability on a certified artefact, measured over the same time window the machine uses. That single test resolves most of the ambiguity.

POMEAS publishes resolution per confocal model alongside its range, which is why the figure in this article is given as a spread across models rather than as a single capability claim.

Choosing range from the part, not from the catalogue

Start from the part's total height variation, including fixture error and any expected variation between batches. Add margin for the approach and for the sensor's own working distance tolerance. That sum sets the minimum range.

Then check the resolution available at that range against the smallest feature the measurement must resolve. If the resolution is insufficient, the options are to reduce the range by fixturing the part more precisely, or to accept a slower measurement with more averaging.

Selecting a wide range first and then discovering the resolution is inadequate is the most expensive version of this mistake, because it usually means changing the sensor rather than the fixture.

It is often cheaper to invest in fixturing that reduces height variation than to buy a sensor with a wide range and fine resolution, since the latter may not exist in a single unit.

Writing the height budget down, including every contributor, is what makes the range decision defensible.

Where confocal is the wrong choice

Confocal measurement excels on small spots, steep angles within its acceptance cone, and surfaces that return a usable spectrum. It is a poor fit for very rough surfaces, for transparent layers where multiple returns confuse the peak detection, and for measurements that need a large field in a single shot.

For transparent or multilayer parts the returned spectrum can contain more than one peak, and the sensor may lock onto the wrong interface. That is a measurement principle limitation rather than a resolution issue, and no averaging setting fixes it.

For large fields with modest resolution requirements, a flash measurement system or an image measuring machine usually delivers the same result faster. Using a point sensor to raster a large area is slow by construction.

Recognising these boundaries early prevents a specification that reads well on paper and cannot be commissioned.

The confocal range in this article is therefore presented with its resolution spread and its surface requirements together, because neither is meaningful alone.

Frequently asked questions

Why does a longer range confocal sensor have worse resolution?

Because the same spectral bandwidth is spread over a longer distance. Each wavelength corresponds to a larger height step, so the ability to distinguish small height changes falls. Range and resolution are linked by the optical design and cannot be optimised independently in a single sensor.

Can I improve resolution by averaging more samples?

Averaging reduces random noise and therefore improves repeatability, but it does not change the sensor's fundamental height resolution or its linearity. It also costs measurement time, which matters for inline use. Improvement from averaging saturates once the systematic error dominates.

How do I measure a transparent film with a confocal sensor?

With difficulty, because multiple interfaces return multiple spectral peaks. Some sensors can be configured to select one interface, and some can measure thickness from the separation between peaks. Both need testing on the actual material, since the result depends on layer thickness and refractive index.

What surface finish is needed for confocal measurement?

A surface that returns a usable spectrum: most machined, ground and moulded surfaces work, very rough or highly diffuse surfaces scatter too broadly, and mirror surfaces can saturate the detector. Testing on the real material at the real angle is the only reliable confirmation.

Is confocal resolution the same as accuracy?

No. Resolution is the smallest detectable height change, accuracy includes calibration and linearity error over the full range, and repeatability is the spread of repeated readings at one point. An application needs all three to be adequate, and meeting only one is a common cause of failed validation.

When should I use a confocal sensor instead of a laser triangulation sensor?

Confocal handles small spots, steep surfaces within its acceptance cone, and thin transparent layers better, and it measures along the optical axis rather than at an angle. Triangulation covers larger standoff distances more cheaply. The choice usually turns on surface angle and spot size rather than on resolution alone.