A telecentric image can look slightly blurred while still supporting a useful measurement. It can also look sharp while producing a biased result. For dimensional inspection, the practical question is whether the complete measurement stays within its error allowance as the part moves through the expected height range.
This procedure is for teams qualifying a gauging station before production. You will define an acceptance rule, sweep a calibrated reference through height, and separate repeatability from systematic error. The result is a tested operating window, not a depth claim based on appearance alone.
Separate telecentricity from usable focus
Object-side telecentric optics reduce changes in magnification when an object moves along the optical axis. That does not make their depth of field unlimited. Defocus still spreads an edge transition over additional pixels, reducing the information available to an edge detector.
For an isolated straight edge with symmetric blur and uniform illumination, the transition midpoint can remain stable as blur increases. Real parts are less cooperative. Nearby edges, reflections, chamfers, saturation and uneven backgrounds can shift the measured boundary.
Residual telecentricity error can also change scale with height. The explanation of telecentric distortion and telecentricity separates those effects from ordinary focus loss.
Treat a depth-of-field illustration as a concept, not a measurement specification. Your usable depth range depends on the feature, lighting, algorithm and allowable error. A range qualified for a wide silhouette may not work for a narrow slot or a small rounded corner.
A focus score remains useful for setup and monitoring. However, the guide to choosing a reliable sharpness metric should support, rather than replace, dimensional validation.
Define the measurand and acceptance rule
Write down exactly what the station measures: an outside diameter, distance between fitted lines, slot width or feature position. Define which physical boundary represents that dimension. A silhouette around a chamfer may not correspond to the drawing's specified edge.
Also distinguish size from position. Both edges can shift together while their separation remains nearly unchanged. A station that measures width successfully may still fail a positioning task.
Establish these requirements before changing focus:
- The permitted measurement error, separate from the part's manufacturing tolerance.
- The expected height interval, including fixture variation and part geometry.
- The field locations and orientations that must work.
- The reference value and its calibration uncertainty.
- The allowed failed-detection rate and cycle time.
For a preliminary engineering screen, use a conservative rule such as:
Here, s(z) is the sample standard deviation of repeated measurements. Uref is the stated expanded uncertainty of the reference dimension, and Eallow is your allocated measurement-error limit.
This is a screening rule, not a formal combined uncertainty calculation or guaranteed coverage interval. A release decision may require a fuller uncertainty budget, measurement-system study and documented decision rule.
Assemble a representative optical station
Use a stable fixture, controlled illumination and the intended acquisition settings. A height stage should move the reference along the optical axis without introducing appreciable tilt or lateral motion. Otherwise, your height test becomes a mixture of several effects.
Browse telecentric lenses by magnification, sensor coverage and working distance, then use the lens finder to narrow the mechanical options. Candidate optics include LS-TCL-0651 and LS-TCL-2118. Their nominal specifications help shortlist parts; they do not establish a qualified measurement depth.
- Sensor format
- 1"
- Mount
- C-mount
- Sensor format
- 2/3"
- Mount
- C-mount
For opaque profiles, backlights provide a useful starting point. Compare diffuse and collimated illumination when edge shape changes with height. The backlight comparison for edge gauging explains why illumination direction can affect the measured silhouette.
LS-BKL-0830 is a collimated backlight candidate. Confirm coverage and installation geometry using the lighting finder, then validate its effect with your actual parts.
- Size
- 100-200 mm
- Type
- Collimated
- Voltage
- 24 V
Choose area scan cameras for adequate sampling, acquisition timing and sensor compatibility. The camera finder helps compare options, but sensor format labels alone do not establish optical compatibility. Check active sensor dimensions, lens image coverage, mount and any required spacing.
Lock exposure, gain, image processing and illumination during the baseline sweep. Keep the bright background below saturation and retain the original image data.
Check sampling before interpreting subpixel measurements
Pixel sampling is not measurement accuracy, but it establishes the scale of the problem. Calculate the object distance represented by each pixel along the measurement direction.
For a hypothetical setup covering a width of 20 mm across 4000 pixels:
These are application assumptions, not specifications of the candidate products. Use the sampling calculation from magnification when sensor pixel pitch and magnification are your starting inputs.
The resolution calculator helps establish the required pixel count; it does not predict the usable telecentric depth of field.
Subpixel fitting can improve repeatability when an edge has adequate contrast and a stable profile. It cannot remove a systematic edge shift caused by defocus or changing illumination. A reported coordinate with several decimal places is not evidence of corresponding accuracy.
If positioning is the final task, the discussion of vision-guided positioning accuracy limits explains why optical sampling is only part of the system budget.
Run a controlled height sweep
-
Establish the nominal plane. Place the calibrated reference at the intended operating height. Set focus and, where adjustable, aperture. Calibrate image scale using the documented station procedure, then freeze the calibration.
-
Record the baseline. Capture a repeated image sequence without moving the reference. Save measured dimensions, fitted edge coordinates, detection status and an edge-profile sample. This estimates short-term repeatability, not loading repeatability.
-
Move through the required height range. Start with a coarse sweep that includes both operating extremes. Record actual stage positions and approach them consistently. Do not refocus or recalibrate at each height unless that adjustment is part of the production process.
-
Repeat across the field. Test the center and representative peripheral locations. Preserve the reference's height and orientation while moving it laterally. A center-only result does not qualify the whole image.
-
Add realistic loading variation. Remove and reseat the reference, then repeat selected conditions. Test relevant orientations and representative production edge finishes. Keep stationary-image variation and reseating variation in separate records.
-
Refine the boundaries. Add intermediate height positions near the first failing conditions. Repeat the sweep in the opposite direction to expose drift, fixture movement or stage hysteresis.
As an initial study, 5 heights, 5 field locations and 30 images per condition require 5 × 5 × 30 = 750 images. That is a planning example, not a statistically sufficient sample for every application.
At each condition, inspect individual edge positions as well as their separation. For circular features, retain fitted diameter, center position and fit residuals. Rejecting poor fits without recording them can hide an unreliable operating region.
Work through an acceptance example
Assume a hypothetical calibrated reference measures 10.000 mm, with stated expanded uncertainty of 2 µm. The station has an allocated measurement-error limit of ±10 µm. Each table row summarizes 30 stationary captures at the field center.
The numbers below illustrate the screening calculation; they are not product performance claims.
| Height relative to nominal | Mean measurement error | Sample standard deviation | Screening value G | Outcome |
|---|---|---|---|---|
| −2 mm | +7 µm | 1.4 µm | 13.2 µm | Fail |
| −1 mm | +2 µm | 1.1 µm | 7.3 µm | Pass |
| 0 mm | +1 µm | 1.0 µm | 6.0 µm | Pass |
| +1 mm | +3 µm | 1.2 µm | 8.6 µm | Pass |
| +2 mm | +9 µm | 1.5 µm | 15.5 µm | Fail |
At the positive intermediate height, the arithmetic is:
The tested intermediate points pass, while the outer points fail. This does not yet qualify every height between −1 mm and +1 mm, nor any other field location.
Refine the sweep, for example with 0.25 mm steps near the proposed boundaries. Then include reseating, representative edges and other operating variation before declaring a continuous working interval.
A narrow passing margin is a reason to investigate drift and uncertainty, not an invitation to round the result down. If required, reduce the operating height range rather than claiming the entire visually acceptable focus region.
Diagnose what fails before changing the lens
Use the pattern of failure to select the next test. Change one factor at a time, and preserve the original data for comparison.
| Observed behavior | Likely mechanism to investigate | Discriminating test |
|---|---|---|
| Mean dimension changes smoothly with height | Residual scale variation or lighting-dependent boundary shift | Compare multiple reference dimensions and inspect both edges |
| Scatter increases while the mean stays similar | Reduced edge slope, noise or unstable fitting | Compare edge profiles and repeatability under controlled exposure |
| Center position shifts but width stays stable | Common edge displacement, stage drift or asymmetry | Check a stationary fiducial and separate edge coordinates |
| Only peripheral measurements fail | Field-dependent optical behavior or calibration residuals | Repeat a field grid at several heights |
| Narrow features fail before wide features | Overlapping blurred transitions | Compare isolated edges with slots of different widths |
| Failures appear after unloading | Fixture variation, tilt or inconsistent seating | Separate stationary capture tests from reseating tests |
Threshold sensitivity is another useful diagnostic. If a small threshold adjustment causes a large dimension change, investigate the edge profile before accepting the measurement recipe. Keep the production algorithm fixed during the final qualification.
Stopping down an adjustable aperture may improve defocus tolerance, but it reduces available light and eventually increases diffraction blur. Repeat the measurement sweep after an aperture change rather than assuming a smaller opening always improves accuracy.
Check selection assumptions before ordering
Use this checklist to turn the experiment into a clear optical requirement:
- Magnification and field coverage: include location variation and every boundary used by the measurement tool.
- Sensor compatibility: verify active dimensions, image coverage, mount and mechanical spacing.
- Working distance: allow room for the part and fixture throughout the required height travel.
- Aperture configuration: confirm whether adjustment exists and how the final setting will be retained.
- Illumination geometry: test the same source arrangement, wavelength and exposure strategy intended for production.
- Reference suitability: use known dimensions and representative edge geometry, with adequate calibration uncertainty.
- Evidence: save raw images, settings, calibration records, height positions and individual measurements.
The usual recommendation to select telecentric optics is incomplete when the feature itself changes appearance with height. Transparent boundaries, overlapping edges and sloped surfaces may require a different lighting or viewing arrangement.
Likewise, telecentric optics do not remove foreshortening caused by part tilt. If the required quantity is an axial depth rather than a lateral dimension, a single silhouette measurement may not observe it directly.
Release a measured operating window
Document the accepted height interval together with field location, feature type, lighting, aperture, exposure and algorithm settings. Record the evidence supporting the boundary and the margin remaining at the worst tested condition.
Use a check reference at startup and after relevant changes. A sharpness trend can warn of focus drift, but dimensional checks must still verify measurement performance. Requalify after optical adjustments, illumination changes or measurement-software changes that affect edge location.
For a new station, start with the lens configuration step. If the usable height range remains uncertain, send the required tolerance, height travel, field of view and representative images through applications support before selecting the final combination.
Questions
Does a telecentric lens keep measurements accurate outside focus?
No. Telecentric optics reduce magnification changes with object distance, but defocus still broadens edges. The usable range ends where bias, repeatability or detection reliability no longer satisfies your measurement requirement. That boundary depends on the feature, illumination, camera sampling and measurement algorithm.
Should I choose the sharpest image as the measurement plane?
Not necessarily. An isolated edge can remain measurable with moderate symmetric blur, while a sharp-looking edge can carry systematic bias. Select focus using image quality, then qualify the measurement across the required height range using a calibrated reference and a fixed production recipe.
How closely should height positions be spaced during validation?
There is no universal step size. Begin with a coarse sweep covering the expected operating extremes, then add positions where measurement performance approaches the acceptance limit. Use finer spacing when the error changes rapidly, and test relevant field locations before declaring a continuous interval.
Should I recalibrate the camera at every test height?
Usually not for a fixed-calibration station, because doing so can hide height-dependent scale changes. Keep calibration fixed during qualification. If production deliberately refocuses or recalibrates, validate that complete adjustment sequence, including its reference accuracy, repeatability and cycle-time effects.
Will a collimated backlight increase the usable measurement depth?
Collimated illumination can reduce some silhouette errors caused by light arriving from different directions, but it does not remove lens defocus or every boundary effect. Compare lighting arrangements using the same reference, height sweep and measurement algorithm, then choose the arrangement with adequate demonstrated margin.