ISO 25178 Areal Filter Cutoff Selection for Tooling Surface Replication Qualification
Lock S-filter and L-filter cutoffs to feature pitch and optical resolution under ISO 25178 before tooling sign-off to prevent false replication approval.

Steel
Cavity inserts machined from hardened alloys define the topography transferred to each injected shot. Toolrooms shaping mould cores by sinker electrical discharge machining, picosecond laser ablation, or ultra-precision diamond milling impart characteristic spatial features across several decades of scale. Sinker discharge leaves overlapping thermal craters, laser passes deposit periodic ripples alongside re-cast slag, and diamond fly-cutting cuts kinematic cusps governed by tool nose radius and feed per revolution.
In each manufacturing route, the resulting steel surface combines high-frequency micro-roughness, mid-frequency process waviness, and low-frequency form errors.
Tool qualification protocols evaluate whether polymer mouldings faithfully duplicate these machined topographies. Replicating micro-fluidic channels, optical diffractive gratings, light-guide textures, or tactile automotive surfaces demands exact physical reproduction of tool asperities. During injection, the melt confronts steep thermal and mechanical gradients.
As advancing resin strikes mould steel held below its glass transition or crystallization threshold, an instantaneous skin layer freezes. Viscosity spikes under shear cooling, and molten polymer cannot reach the acute valley bottoms of the tool unless cavity pressure stays high during the narrow millisecond window before skin solidification locks the micro-flow.
A surface measurement parameter separated from its transmission band is a meaningless number on an inspection cert.
Topography characterization under ISO 25178 sets the spatial bandwidth evaluated during tool-to-part replication checks. When quality engineers inspect the cavity face and the ejected moulding with areal optical profilers, the instruments gather coordinate clouds containing millions of elevation points. Areal filtering then decomposes the raw data into separate scale components.
Selecting the filter cutoffs determines whether an apparent replication failure reflects actual polymer flow stagnation or an arbitrary mathematical boundary: trimming real asperities with an overly aggressive short-wavelength cutoff masks incomplete cavity fill, while an expansive cutoff that swallows bulk part warpage buries replication fidelity under shrinkage noise.
Mismatched filter boundaries produce false replication acceptance certificates, releasing unfillable cavity tolerances into production contracts that generate six-figure scrap write-offs during high-volume ramp.

Bandwidth
Spatial frequency transmission boundaries dictate which physical features enter an areal surface characterization. Under ISO 25178-2 and ISO 25178-3, scale-limiting operators parse the raw instrument extraction ~ the primary surface, which bundles form errors, mid-range waviness, target roughness, and sensor noise ~ into distinct spectral bands. The S-filter removes high-frequency micro-roughness and optical noise using a nested cutoff index, λs.
The F-operator subtracts nominal part geometry like cylindrical curvature or tilt by fitting polynomials across the data matrix. The L-filter then strips long-wavelength components such as sink marks and bulk part warpage using the cutoff index λc.
The scale-limited surface resulting from both S-filtering and L-filtering forms the S-L surface, and parameters such as arithmetic mean height, root mean square gradient, and developed interfacial area ratio derive entirely from this matrix. Comparing tool steel to a moulded part without locking identical S-filter and L-filter cutoffs renders any derived replication ratio physically meaningless, particularly since volumetric polymer shrinkage shifts the lateral pitch of periodic features and distorts areal heights unevenly.

What Cutoff Ratio Prevents Transmission Leakage?
A separation factor of five or greater between adjacent nesting indices maintains signal stability across the Gaussian weight function. ISO 25178-3 defines Gaussian areal filters with transmission characteristics that drop to fifty percent transmission at the specified cutoff wavelength. Setting the short-wavelength cutoff λs too close to the long-wavelength cutoff λc creates overlapping filter skirts, attenuating the intermediate frequencies that represent the cavity’s true functional texture.
Standardized filter selection couples the instrument lateral resolution directly to the spatial features of the manufacturing process:
- Areal S-filter eliminates high-frequency instrument noise and spatial artifacts below the optical transfer limit without blunting steep asperity sidewalls.
- Areal L-filter segregates long-range thermal shrinkage, sink marks, and gross moulding bow from micro-scale replication topography.
- Areal F-operator subtracts nominal macroscopic CAD geometry, tilt, and curvature by fitting second-order or third-order orthogonal polynomials across the raw data matrix.
- Nesting index designates the specific spatial wavelength cutoff separating functional surface textures from non-functional high or low frequency components.
Optical instruments evaluate surface coordinates through different physical principles. Coherence scanning interferometry delivers sub-nanometre vertical resolution across super-smooth diamond-turned surfaces, though steep slopes produce data dropouts. Focus variation microscopy relies on shallow optical depth of field to resolve high-slope features like electrical discharge machining pits, but carries a vertical noise floor near twenty nanometres.
The objective lens transfer function establishes the physical floor for λs selection: setting λs to less than three times the lateral pixel spacing produces severe spectral aliasing, turning optical artifacts into fictitious peaks.
| Tool Surface Generation Method | Dominant Feature Pitch (µm) | Optical Objective Magnification | S-Filter Cutoff λs (µm) | L-Filter Cutoff λc (mm) | Bandwidth Ratio (λc / λs) |
|---|---|---|---|---|---|
| Ultra-precision Diamond Turning | 0.2 to 2.0 | 50x to 100x CSI | 0.25 | 0.08 | 320:1 |
| Picosecond Laser Texturing | 2.0 to 15.0 | 20x to 50x Confocal | 0.80 | 0.25 | 312:1 |
| Precision Micro-Milling | 5.0 to 40.0 | 10x to 20x Focus Variation | 2.50 | 0.80 | 320:1 |
| Fine EDM Finish (CH 12-18) | 8.0 to 60.0 | 10x to 20x Focus Variation | 2.50 | 0.80 | 320:1 |
| Coarse EDM Finish (CH 24-36) | 30.0 to 200.0 | 5x to 10x Focus Variation | 8.00 | 2.50 | 312:1 |
| Chemical Acid Etch Texture | 20.0 to 150.0 | 10x Focus Variation | 8.00 | 2.50 | 312:1 |
| Lateral sampling interval must not exceed one-third of the S-filter cutoff λs to comply with Nyquist criteria and ISO 25178-3 recommendations. | |||||
Default filter tables from line-profile standards like ISO 4288 fail on areal data sets. Where stylus tip mechanics physically filter spatial frequencies during line profiling, numerical convolutions govern areal optical metrology. Qualifying diamond-turned optical inserts with an S-filter of 0.25 µm and an L-filter of 0.08 mm preserves nanometric turning cusps while excluding overall lens curvature.
For coarse chemical textures on injection moulds, shifting λs to 8.0 µm and λc to 2.5 mm retains multi-tier etching structures without allowing part shrinkage to distort areal height parameters.
Incorporating ISO 25178-3 clause 4.2 into tooling acceptance contracts binds suppliers to explicit nesting indices, eliminating the post-moulding filtering discretion that conceals replication defects.

Imprint
Thermoplastic melt forced against a temperature-controlled cavity face undergoes severe rheological throttling at the sub-micrometre boundary. Molten polymer chains carry a finite radius of gyration ~ typically between five and fifteen nanometres at rest, swelling under shear orientation. When the melt front meets cold cavity steel, heat conducts into the metal at rates exceeding one thousand degrees Celsius per second, forming an instantaneous solid boundary layer.
The molten core behind this frozen skin must transmit enough mechanical force under packing pressure to drive the semi-solid layer into the finest recesses of the tool.
Replication fidelity quantifies the proportion of steel surface geometry transferred into the cooled moulding. Calculating this value requires matched spatial filtering on both substrates. Simply dividing part roughness by tool roughness produces misleading ratios whenever the evaluation bandwidth captures form warpage or instrument noise instead of scale-limited functional features.
Under identical injection pressures of 90 MPa, cyclic olefin copolymer achieves 94 percent replication fidelity at a 2.5 micrometre cutoff while semi-crystalline polyamide 66 drops to 61 percent at the identical spatial frequency.
Polymer morphology dictates replication across different filter bandwidths. Amorphous resins with broad relaxation spectra and manageable glass transitions ~ such as polymethyl methacrylate, polycarbonate, and cyclic olefin copolymer ~ reproduce high-frequency tool details with fidelity exceeding ninety percent. By contrast, semi-crystalline resins like polyoxymethylene, polyamide, and polyphenylene sulfide undergo abrupt volumetric crystallization shrinkage at the solidification front.
As crystallites nucleate, the polymer pulls back from micro-cavity asperities; even under maximum machine packing pressure, the phase change breaks intimate mechanical contact with high-frequency steel cusps.
- Cavity Topography Baselines register the un-moulded steel insert under ISO 25178 areal confocal or coherence scanning profilers using calibrated optical objectives.
- Process Stabilization Runs establish continuous hydraulic repeatability over fifty production cycles, verifying steady-state cavity steel temperature via embedded thermocouples.
- Sample Replication Extraction strips residual processing oils from moulded components before mounting parts in rigid fixtures to prevent tilt artifacts.
- S-L Filter Normalization executes identical Gaussian nesting index operations across both tool and part datasets, removing macro-curvature via matching polynomial operators.
- Spectral Coherence Calculation correlates power spectral density functions across spatial frequencies, isolating the exact cutoff threshold where replication degrades below specification.
Replication fidelity drops off sharply once spatial feature sizes fall below ten micrometres. Features wider than one hundred micrometres replicate almost completely across most processing windows because bulk cavity pressure acts uniformly across broad geometries. Below ten micrometres, melt surface tension strongly resists the generation of new interfacial area, demanding local pressures above one hundred megapascals to force resin into acute corners.
| Polymer Matrix and Grade | Polymer Morphology | Replication Ratio (λ = 0.8 µm) | Replication Ratio (λ = 2.5 µm) | Replication Ratio (λ = 8.0 µm) | Replication Ratio (λ = 25.0 µm) |
|---|---|---|---|---|---|
| Cyclic Olefin Copolymer (Optical) | Amorphous | 88.4% | 94.2% | 98.7% | 99.8% |
| Polycarbonate (High Flow) | Amorphous | 82.1% | 89.6% | 96.5% | 99.2% |
| Polymethyl Methacrylate (Moulding) | Amorphous | 79.5% | 87.3% | 95.1% | 99.1% |
| Polyoxymethylene (Unfilled) | Semi-Crystalline | 54.2% | 68.7% | 84.3% | 96.4% |
| Polyamide 66 (Standard Flow) | Semi-Crystalline | 46.8% | 61.2% | 78.9% | 94.8% |
| Polyetheretherketone (High Temp) | Semi-Crystalline | 38.1% | 52.4% | 71.6% | 91.3% |
Consider an optical insert bearing diffractive ridges with an average height of 1.20 µm and a pitch of 4.0 µm, moulded in optical polycarbonate. An inspection protocol applying an S-filter λs of 0.8 µm and an L-filter λc of 25 µm yields an arithmetic mean height Sa of 1.07 µm on the part against 1.20 µm on the steel, establishing an 89.2 percent replication index. If the inspection setup instead widens the bandwidth to λs = 0.25 µm and λc = 0.8 mm, high-frequency optical noise inflates part Sa to 1.18 µm while gross component bow adds false amplitude, driving the calculated index to an apparent 98.3 percent.
The insert qualifies on paper, yet on the optical bench the lens scatters beam energy and fails because the acute groove roots never filled.
Variations in raw resin viscosity between pellet lots are frequently cited to account for replication discrepancies, yet cavity pressure decay and rapid frozen-skin formation dominate micro-texture height loss on moulded components.

Variance
Discrepancies between tool geometry and moulded topography concentrate within specific spatial bandwidths, where standard height parameters often fail to register functional losses. Because arithmetic mean height Sa averages absolute ordinate deviations across the entire evaluation field, it overlooks localized tip rounding. If a diamond-machined tool profile features acute 90-degree crests and the entering polymer freezes into rounded domes, the displaced volume remains nearly identical, leaving the Sa values between steel and polymer within four percent of each other.
Reliable qualification depends on the hybrid and functional parameters defined in ISO 25178-2. Root mean square gradient Sdq captures the mean slope of surface asperities, while the developed interfacial area ratio Sdr quantifies the additional surface area contributed by texture relative to a flat projected plane. When polymer melt fails to penetrate sharp crevices, Sdq and Sdr drop sharply, exposing incomplete replication even while Sa holds steady.

Why Do Optical Bandwidths Distort Shrinkage Ratios?
Bulk volumetric contraction pulls polymer boundaries inward across millimetre dimensions while surface tension stabilizes micro-scale peak spacing. Semi-crystalline thermoplastics show anisotropic mould shrinkage between 1.2 and 2.5 percent depending on flow orientation and local cooling. When an areal measurement window spans several millimetres under an expansive L-filter cutoff λc, differential shrinkage introduces low-frequency undulations into the dataset, which the instrument logs as surface waviness and distorts the resulting functional parameters.
Filtering decisions directly dictate the magnitude and stability of derived areal parameters:
- Arithmetic Mean Height Sa registers average surface amplitude but conceals localized peak flattening and valley rounding under standard replication conditions.
- Maximum Height Sz exhibits severe sensitivity to single dirt particles, optical measurement dropouts, and ejector pin witness marks.
- Root Mean Square Gradient Sdq quantifies the steepness of micro-flanks, serving as an early indicator of incomplete cavity filling under frozen skin layers.
- Developed Interfacial Area Sdr reflects functional micro-texture complexity, dropping toward zero when polymer melts fail to replicate high-frequency tool features.
- Core Roughness Depth Sk isolates the load-bearing plateau structure, providing robust replication data independent of outlier scratches or transient void defects.
Tool steel wear also shifts the reference baseline over long production campaigns. Abrasive mineral fillers like glass beads or chopped carbon fibers can erode micro-scale asperities within tens of thousands of cycles. When qualifying replacement inserts or balancing multi-cavity moulds, verification requires confirming that each cavity yields matching areal parameters under identical filter cutoffs.
| Measurement Target Substrate | S-Filter λs (µm) | L-Filter λc (mm) | Sa (µm) | Sz (µm) | Sdq (rad) | Sdr (%) |
|---|---|---|---|---|---|---|
| Tool Steel EDM Master Core | 2.50 | 0.80 | 1.42 | 12.85 | 0.68 | 21.4 |
| Moulded Part (Correct Bandwidth) | 2.50 | 0.80 | 1.18 | 9.42 | 0.41 | 9.8 |
| Moulded Part (Under-Filtered Noise) | 0.80 | 0.80 | 1.31 | 14.20 | 0.74 | 25.6 |
| Moulded Part (Over-Filtered Feature Loss) | 8.00 | 0.80 | 0.94 | 6.80 | 0.22 | 3.1 |
| Moulded Part (Contaminated by Bow) | 2.50 | 2.50 | 1.64 | 18.90 | 0.44 | 10.9 |
These shifts highlight how easily cutoff manipulation alters replication data. In the second row, correctly matched nesting indices expose the real shortfall: Sa drops from 1.42 µm to 1.18 µm, and Sdr falls from 21.4 percent to 9.8 percent, capturing the viscoelastic rounding of sharp EDM craters. In the third row, tightening λs to 0.80 µm lets high-frequency optical noise leak into the dataset, artificially rebounding Sdr to 25.6 percent and giving the false impression that the moulding exceeds the tool in texture complexity.
In the final row, opening λc to 2.50 mm lets bulk part warpage inflate Sa to 1.64 µm, hiding the replication defect entirely.
Metrology committees continue to debate whether a single fixed nesting index can ever capture both the volumetric shrinkage of semi-crystalline matrices and the localized shear-induced replication loss of high-aspect micro-ridges.

Acceptance
Signing off on hardened production tools without verifiable replication criteria creates substantial commercial exposure. Purchase agreements routinely fall back on legacy specifications like VDI 3400, SPI finishes, or two-dimensional Ra roughness, none of which transfer usefully to micro-structured cavities, optical surfaces, or micro-fluidic chips. Profile traces sample under one percent of the functional surface, missing localized unfill, weld-line gloss shifts, and directional texture loss.
A defensible tooling agreement relies on areal characterization under ISO 25178. Robust specifications fix five parameters before machining begins: the optical instrument modality, objective magnification and numerical aperture, lateral sampling interval, S-filter cutoff λs, and L-filter cutoff λc. Freezing these variables in advance prevents retrospective metrological adjustments, avoiding the heavy expense of sinker re-machining, chemical stripping, or remaking inserts after failure in the press.
A production tool approved on un-nested surface profiles transfers dimensional and cosmetic liability to the buyer the moment first shot samples pass receiving inspection.
Multi-cavity tooling amplifies qualification risk. Across an eight- or sixteen-cavity mould running micro-textured parts, thermal variance through the hot runner generates cavity-to-cavity viscosity differences. Outer drops see longer runner paths and higher heat loss, cooling the melt front.
Even under geometrically balanced cavity pressures, cooler resin in peripheral cavities skins over earlier and dulls micro-texture replication. Full qualification therefore requires evaluating every cavity under identical filter cutoffs: if cavity one delivers 92 percent Sdr replication while cavity eight reaches only 64 percent, the tool fails qualification regardless of whether part dimensions clear macroscopic tolerances under DIN 16742.
Tool sign-off demands physical validation tied directly to part performance. For micro-fluidic channels, qualification correlates areal void volume Vvv with liquid flow resistance; for anti-glare display bezels, core roughness Sk against gloss reflectivity; for tribological seals, areal material ratio curves against dynamic leakage rates. When these metrics hit specified replication targets within locked filter bandwidths, production release rests on defensible metrology.
A tool qualification procedure holds commercial authority only when the measurement protocol locks the spatial filter cutoffs before the toolroom cuts the cavity steel.


