On this page
A microscope image can pass separate horizontal and vertical length checks yet still distort a diagonal distance. The missing question is whether the two image-coordinate directions remain orthogonal in the physical specimen plane. A two-dimensional calibration grid can reveal the cross-axis term that two scalar pixel sizes omit. This matters when ceramic-circuit acceptance depends on diagonal clearance, pad position or hole-to-feature geometry.
Measurement purpose
Determine whether microscope image coordinates require cross-axis calibration before evaluating diagonal ceramic-feature dimensions.
Specimens and conditions
- Planar geometry
- Defined specimen working plane and the field locations relevant to the dimensional decision
- Reference artifact
- Known two-dimensional coordinate relationships with suitable uncertainty and independent check positions
Equipment and records required
- Imaging chain: Controlled objective, camera mode and final measurement-image representation
- Coordinate analysis: Verified affine or other justified mapping with residual and independent-check evaluation
Method sequence
- Reference capture
Acquire distributed grid positions through the production image path
Record: Artifact images and configuration
- Calibration
Fit coordinate mapping and inspect cross-axis terms
Record: Coefficients, units and residual patterns
- Validation
Test independent diagonals and field locations
Record: Supported measurement envelope and exceptions
Decision and uncertainty
Use diagonal dimensions only when the two-dimensional coordinate relationship, not just individual axis scales, is supported for the actual image path.
Artifact geometry, localization, cross-axis coupling, nonlinear residuals and specimen-plane differences contribute separately.
The dimensional-method owner approves the coordinate model; the drawing owner defines the geometric requirement and tolerance.
Traceable outputs
| Record | Required contents |
|---|---|
| Calibration package | Reference coordinates, mapping, image transforms and independent checks |
| Measurement boundary | Validated field and height conditions, residual limits and change triggers |
Method review decisions
- Distinguish axis scale from orthogonality and cross-axis coupling.
- Validate the coordinate map with independent grid positions and diagonals.
- Do not use an affine correction to conceal unresolved nonlinear distortion or missing image detail.
Decide whether the result uses one direction or a two-dimensional displacement
A width along one known image axis and a distance between two arbitrary pad centres do not exercise the same coordinate model. The second combines horizontal and vertical displacement, so their physical relationship matters. Define the coordinates and geometric output before accepting a calibration represented by one micrometre-per-pixel value.
Keep image geometry separate from optical visibility. A coordinate map can improve the interpretation of resolved features but cannot restore a blurred edge or recover a gap lost to saturation. The selected points must already correspond to identifiable physical boundaries on the ceramic circuit.
Use a reference that supplies two-dimensional information
A stage micrometer provides known intervals along its scale. A suitable grid or other calibrated two-dimensional artifact also supplies relationships between positions in different directions. Record the certified coordinates or dimensional properties actually available; a visually square printed pattern is not automatically a calibrated orthogonality reference.
Acquire the artifact using the objective, camera mode, working plane and export path intended for the specimen. A calibration taken before an image is stretched or transformed may no longer describe the pixels measured later. Preserve both the native image and the measurement-file transformation so a coordinate change remains reproducible.
Write the image-to-specimen map with its cross-axis terms
For a locally adequate affine model, represent a pixel displacement by a two-component vector. Multiplying it by a two-by-two calibration matrix gives the corresponding physical displacement. Translation affects the coordinate origin but cancels from a difference between two points. The off-diagonal matrix entries allow cross-axis coupling.
The squared physical length is the pixel displacement multiplied through the metric matrix formed from the calibration matrix transpose and itself. If the metric has a nonzero cross term, separate horizontal and vertical scale factors are insufficient for arbitrary diagonals. The calculation should be implemented and checked with explicit units and coordinate conventions.
Δr = A Δp; L² = Δpᵀ AᵀA Δp
- Δp is the pixel-coordinate displacement between two resolved points.
- A maps pixel displacements to physical coordinates in the specimen plane.
- Δr is physical displacement and L its Euclidean length.
A validated local affine mapping of a planar specimen. Residual nonlinear distortion, working-height changes and feature-localization uncertainty require separate evaluation.
See why two axis-length checks can miss a diagonal error
Use an illustrative mapping x equals two times u plus 0.2 times v, and y equals two times v, with physical coordinates in micrometres and u and v in pixels. A one-hundred-pixel horizontal displacement maps to two hundred micrometres. A one-hundred-pixel vertical displacement maps to the vector twenty, two hundred micrometres, whose length is approximately 201.00 micrometres.
If those two lengths are stored only as independent axis scales, a displacement of one hundred pixels in both directions appears to have length approximately 283.55 micrometres. The full mapping gives the vector two hundred twenty, two hundred micrometres and length approximately 297.32 micrometres. Reversing the horizontal direction gives approximately 269.07 micrometres, although the independent-scale calculation still gives 283.55. These assumed values isolate a geometric effect, not a real microscope error.
| Pixel displacement | Physical displacement in µm | Length from full map |
|---|---|---|
| (100, 0) | (200, 0) | 200.00 µm |
| (0, 100) | (20, 200) | 201.00 µm |
| (100, 100) | (220, 200) | 297.32 µm |
| (−100, 100) | (−180, 200) | 269.07 µm |
Fit enough positions and reserve an independent check
A general planar affine map has six coefficients including translation. Three noncollinear point correspondences provide a mathematical minimum, but an exact fit to that minimum supplies no redundancy for examining model inadequacy. Use an appropriate distributed set and retain independent positions or a separate acquisition to test the mapping.
Check both diagonal directions, locations near the intended features and the relevant field extent. A correction that removes mean scale error can still leave spatially structured residuals. Plot those residuals against position rather than reporting only one average magnitude that can conceal a corner-dependent problem.
Distinguish shear from nonlinear distortion and specimen tilt
An affine transformation preserves straight lines and parallelism, but it does not represent every optical distortion or perspective effect. Curved grid rows or residuals that vary strongly with distance from the field centre may require a different justified model or a restricted measurement region. Do not keep adding coefficients merely until the calibration data fit.
A tilted or warped specimen also changes the relation between the image and the desired physical plane. Establish the working-plane condition and challenge the permitted seating range. A correction fitted to a flat artifact at one height does not automatically validate a raised terminal, a curved glaze meniscus or an oblique ceramic section.
Carry the coordinate mapping through cropping and resampling
Cropping native pixels changes the origin but not their original spacing. Resampling, shear or unequal resizing changes the displacement map and can introduce interpolation effects. If software modifies the image geometry, either propagate the approved transformation mathematically or calibrate the final measurement representation with a suitable check.
A scale bar burned into a picture can be transformed along with the picture. Its presence is therefore not proof that every direction remains correct. Retain numerical calibration metadata, original dimensions and processing history instead of attempting to reconstruct a two-dimensional map from one displayed bar.
Release a geometry model with a defined measurement envelope
The calibration record should identify the artifact, acquisition configuration, mapping coefficients, units, fitted points, independent check points and residual limits. State which field region and specimen-plane conditions were supported. Reevaluate the affected geometry after a camera, objective, adapter, image-processing or export change.
ChipSimple can review dimensional questions and drawing coordinates for a project-specific inspection plan. The useful outcome is a coordinate transformation demonstrated to support the required pad or aperture measurement. It remains distinct from a defect-detection limit, a general microscope accuracy claim or proof that every visible edge represents the correct material boundary.
Define the image-coordinate measurement
Provide the two-dimensional feature relationship that determines acceptance.
- Drawing coordinates and diagonal or angular requirement
- Native and final measurement images
- Artifact calibration and working-plane setup
- Objective, camera mode and transformation history
- Independent grid or feature checks with residuals
The drawing-upload form loads as you reach this section.