Point Spectral Confocal for Transparent Plastics: Benefits, Limits and Four Mistakes

Transparent plastic defeats measurement for a simple reason: being visible is not the same as being measurable. Conventional imaging decides where an edge is by comparing pixel intensities, and clear material lets whatever sits behind the part contribute to those intensities. The system loses its reference.

Point spectral confocal offers a different premise. It does not try to find the edge; it measures distance. That shift happens to line up well with several of the specific difficulties transparent parts present.

The principle: why transparency stops being the problem

The method exploits dispersion. In the optical system, different wavelengths come to focus at different axial positions. When broadband light reaches the surface, only the wavelength that focuses exactly on it returns efficiently; the rest is defocused.

Analysing the peak of the returned spectrum gives the axial position of the surface. Nothing in that depends on an image intensity edge, which is why the approach tolerates clear, semi-transparent and shiny surfaces. Additionally, on transparent material the upper and lower surfaces each produce their own identifiable return - which is the physical basis for thickness measurement on clear parts.

Four parameters that matter more than the numbers

These four come up in every selection, and understanding them beats memorising values.

Alongside those sit stated repeatability and linearity error. To be clear about what they mean: repeatability describes agreement between repeated readings and linearity describes departure from an ideal line. Neither is synonymous with the measurement uncertainty you will actually achieve, which also depends on environment, fixturing, target surface character and the evaluation method.

Model comparison and how to trade off

Under this configuration the equipment capability has a defined ceiling: for the integrated point-spectral confocal flash measurement setup used here, measurement accuracy reaches 2 μm. What actually matters in selection is not who claims the highest accuracy, but whether range, reference distance, spot size and sampling method match your part.

Taking the POMEAS SFS family as a worked example, the range-versus-precision trade off is unusually clear:

All figures above are published values from the respective product pages. None of these models is universally better than another; the choice follows from the height variation you must cover and the standoff you can hold. Temperature characteristic is stated below 0.03 percent of full scale per degree C and ingress rating IP40 across the family, subject to the individual model page.

The logic is direct: a larger range generally means looser stated repeatability. So establish the minimum range from the actual part variation first, then pick the tightest repeatability available within that tier, rather than starting from whichever model has the best headline figure.

Four mistakes

Assuming non-contact automatically handles transparent parts

Non-contact is necessary, not sufficient. Laser triangulation is also non-contact, yet light entering transparent material refracts and scatters internally, so the returned position may correspond to somewhere inside the material rather than its surface. Whichever single-point method you pick, confirm its behaviour on your specific material - preferably with samples.

Believing smaller spots are always better

A small spot reaches into narrow features and resolves fine detail, but it also becomes sensitive to local defects, texture and contamination, so readings may stop representing the surface. When characterising an area, spot size should match the scale at which you are evaluating, not simply be as small as available.

Ignoring tilt

Each model publishes a tolerated measurement angle, beyond which readings degrade or simply stop. Curved surfaces, inclined faces and areas that distort after assembly all need the angle between the surface normal and the sensor axis checked first.

Counting heads and forgetting the controller

In multi-point schemes the controller decides how many heads connect, what each channel can sample at, and whether you can synchronise to an encoder or line signal. POMEAS publishes 4, 8, 12 or 16 head configurations for the multi-channel unit, with sampling rate falling from 20 kHz on a single channel to 4 kHz when sixteen run together. Plan concurrency-aware throughput from the start, not after purchase.

Start from samples

However good the table looks, it comes back to parts. Provide several real pieces including known-good, borderline and known-bad; the zones and tolerances to be evaluated; and the target cycle time. The test should answer concrete questions: do readings hold steady on each zone, does range cover every sample's height variation, and can data be acquired at target pace.

This stage also surfaces everything a parameter table cannot: obstruction from some feature, an unusual reflection from a particular finish, or drift caused by how parts are stored. Finding them early costs an order of magnitude less than finding them later.

Frequently asked questions

Can spectral confocal measure transparent material thickness?

Yes. In principle the upper and lower surfaces each produce their own return, from which thickness follows. POMEAS publishes a minimum measurable thickness of 5 percent of full scale for the SFS family. Whether that satisfies a specific requirement has to be confirmed on actual samples.

How does it cope with glossy surfaces?

Because distance comes from spectral peak position rather than from an image intensity edge, mirrored and polished surfaces that would saturate an imaging approach are generally manageable. Very low reflectance or strongly absorbing materials still need verification.

How do reference distance and range relate?

Reference distance is the nominal mounting standoff; range is the usable interval either side of it. Install so that the usual part surface sits near the centre of range, leaving margin both ways for incoming dimensional variation.

How many heads can one controller drive?

It depends on the controller. POMEAS publishes 4, 8, 12 and 16 channel configurations for its multi-channel unit. Note that sampling rate falls as more channels run concurrently, so design cycle time against the number you will actually use.

Do temperature and vibration matter?

Both do. Published parameters include a temperature characteristic - below 0.03 percent of full scale per degree C for the SFS family - while vibration tends to show up as fixture stability and reading convergence, and belongs to the tooling design stage. Confirm the site's vibration environment during selection.

Single head or several?

It follows from area and cycle time. Small areas with relaxed timing favour one head on a stage. Large areas or short cycles favour parallel heads, at higher controller and calibration cost. For any particular part, evaluate both options on total cost rather than picking by habit.

Related reading

Overall route comparison is available elsewhere, as is the general selection procedure, and multi-point versus single-point trade-offs have their own article.

Related products and configuration notes

• SFS-D108 chromatic confocal sensor - +/- 0.05 mm range, stated repeatability 3 nm - tightest tier in the family
• SFS-D1220 chromatic confocal sensor - 20 mm reference distance, +/- 0.6 mm range; documented for transparent material thickness and warpage related work
• SFS-D606 compact chromatic confocal sensor - 20 mm by 110 mm body at about 73 g for space-constrained arrays
• 20 kHz multi-channel confocal controller - 4 / 8 / 12 / 16 head configuration with encoder and trigger support

Related reading

• Transparent Plastic Part Flatness Inspection: What Two Data Sets Actually Show
• Two Flatness Zones, Two Opposite Verdicts on the Same Part
• Flatness Out of Tolerance: A Troubleshooting Path That Actually Works
• What Flatness Really Means, and What It Is Often Confused With
• Flatness Measurement Methods Compared: Plate, CMM, Vision, Confocal
• Repeatability Is Not Accuracy: Reading Range, Sigma and Cpk Honestly
• How to Choose Flatness Inspection Equipment: A Five-Step Reverse Method
• Multi-Point or Single-Point Flatness: Choosing How to Collect the Points
• Where Warpage in Injection Moulded Parts Comes From
• Turning Inspection Data into Decisions: Four Stages of Refinement

Confocal Configurations for Transparent Parts

Transparent plastics put two surfaces in the same beam. These controllers cover single-point and multi-point arrangements; the head choice follows the material and the thickness range.

• Single-Channel Confocal Displacement Sensor Controller — one measuring head, the entry configuration for single-point thickness and position
• 32kHz Dual-Channel Confocal Displacement Sensor Controller — two heads on one controller for differential thickness or opposing surfaces
• 32kHz Four-Channel Confocal Displacement Sensor Controller — four heads for multi-point flatness and simultaneous thickness tracking

Comparing models side by side? compare channel count and controller bandwidth.

Send the material, the thickness range and whether both surfaces matter, and we will propose a head and channel count. Contact us.