Areal Surface Texture Filtering Selection in High-Precision Micro Injection Moulding
Selecting areal surface texture filter cutoffs under ISO 25178 matches transmission bandwidth to moulding replication limits, preventing false tooling rejects.
Filter
Areal surface metrology under ISO 25178 relies on multi-scale spatial decomposition to separate functional topography from high-frequency measurement noise and long-wavelength form errors. In high-precision micro-injection moulding, surface texture defines crucial end-use characteristics including optical reflectance, fluidic boundary layer friction, hydrophobic contact angles, and mechanical tribology. A micrometer-scale surface detail transferred from a nickel electroformed tool insert or a femtosecond laser-ablated core to a polymer replica undergoes complex spatial distortion.
Plastic melt elasticity, frozen skin layer formation, thermal shrinkage, and ejection shear alter high spatial frequencies differently than long-wavelength undulations.
Evaluating surface texture without explicitly designated areal filter cutoffs yields arbitrary parameter values. Standard profile parameters like Ra or Rz condense three-dimensional topography into single-line traces, missing spatial anisotropy and micro-cavity filling defaults. ISO 25178 areal surface texture analysis replaces single profiles with surface operators that isolate micro-roughness, waviness, and underlying geometry.
Steel preserves spatial truth.
Areal Transmission Characteristics and Nesting Indices
Separating roughness from waviness across micro-moulded features requires precise mathematical boundaries that prevent spectral leakage. The S-filter removes spatial wavelengths shorter than a specified limit, effectively suppressing optical sensor noise or stylus tip tip-radius convolution artifacts. The L-filter removes long spatial wavelengths, isolating functional roughness from form errors caused by mould deflection, thermal warping, or tool plate misalignment.
The F-operator removes nominal geometric form, such as cylindrical or spherical curvature on micro-moulded lenses, using polynomial baseline fitting.
Combining these mathematical operators produces nested surface definitions. Applying an S-filter and an F-operator yields an S-F surface, containing all spatial scales between the high-frequency cutoff and the part geometry. Applying an additional L-filter to the S-F surface yields an S-L surface, representing the standard areal roughness scale.
The bandwidth between the S-filter cutoff spatial wavelength and the L-filter cutoff spatial wavelength defines the transmission band. Specifying surface roughness parameters without defining this exact transmission band allows measurement variation up to four hundred percent on the identical micro-injection moulded part.
Specifying an areal surface parameter under ISO 25178-2 without declaring the S-filter and L-filter nesting indices renders the measurement legally unenforceable during tooling acceptance trials.
Bandpass Boundaries for Micro-Moulded Geometries
High-precision micro-cavities exhibit spatial periodicities spanning from sub-micron tool polishing marks up to millimeter-scale runner shrinkage undulations. Matching the measurement bandwidth to the physical process window prevents false non-conformance reports. For micro-moulded polymer components with functional micro-structures between five micrometers and five hundred micrometers, standard macro-moulding filter cutoffs distort surface parameter calculations.
ISO 16610 defines linear and non-linear filter types suitable for areal processing. Linear Gaussian filters offer standard transmission profiles but suffer from distortion at edge boundaries and steep micro-structure steps. Spline filters reduce edge distortion on bounded micro-fluidic channels.
Morphological filters, using spherical or disk structuring elements, simulate physical surface contact for tribological sealing assessment without smoothing out functional micro-peaks.
| Feature Scale Domain | Primary Application | S-Filter Cutoff (λs) | L-Filter Cutoff (λc) | Recommended Filter Class |
|---|---|---|---|---|
| Sub-micron Optical | Diffractive optical elements, anti-reflective structures | 0.25 µm | 0.08 mm | Gaussian or Spline (ISO 16610-21) |
| Micro-Fluidic Channels | Capillary bio-chips, hydrophobic fluid guiding | 0.80 µm | 0.25 mm | Robust Gaussian or Morphological |
| Precision Mechanical | Micro-gears, watch movements, medical pump seals | 2.50 µm | 0.80 mm | Linear Gaussian (ISO 16610-21) |
| Functional Textures | Laser-textured tactile surfaces, light-diffusing panels | 8.00 µm | 2.50 mm | Areal Spline or Wavelet |
Selecting the spatial wavelength limits depends directly on the moulding technique. Micro-injection moulding processes operating under variothermal mould control replicate high spatial frequencies down to fifty nanometers. Conventional cold-mould cycles freeze the polymer skin before melt pressure can drive high spatial frequencies into micro-cavity corners.
Setting an S-filter cutoff at 0.25 micrometers on a part moulded without variothermal assistance captures background optical noise rather than true plastic topography.
Spatial resolution limits yield. Applying an overly aggressive L-filter removes authentic long-wavelength polymer sink marks caused by differential thermal contraction across micro-ribs. Selecting appropriate filter limits establishes an unambiguous metrology baseline that correlates directly with press-side moulding parameters.
The choice of spatial filtering boundary dictates the mathematical stability of modern quality control pipelines. Linear filters compute rapidly on automated shop-floor optical profilers, yet non-linear morphological operators better predict gas-tight sealing performance on elastomeric micro-valves. Whether non-linear morphological filters can replace linear Gaussian operators without introducing numerical instability during automated quality control on production shop floors remains an open empirical question.
Peak
Topographical height distribution across micro-injection moulded optical interfaces dictates light scattering, hydrophobic contact angle, and frictional performance. Areal field parameters derived under ISO 25178-2 quantify surface height variations, spatial periodicities, and volume characteristics. These metrics respond dynamically to changes in spatial cutoff boundaries.
Calculating areal parameters on raw unfiltered height data introduces severe measurement artifacts from instrument tilt, environmental vibration, and high-frequency sensor scatter.
Evaluating peak height distributions on precision micro-mouldings involves separating amplitude metrics from hybrid spatial properties. Amplitude parameters such as Arithmetical Mean Height (Sa), Root Mean Square Height (Sq), and Maximum Height (Sz) measure vertical feature scale. Skewness (Ssk) quantifies profile asymmetry, identifying whether a micro-textured surface possesses protruding peaks or recessed valleys.
Developed Interfacial Area Ratio (Sdr) measures the percentage of additional surface area contributed by micro-texture relative to a flat planar projection. Each metric changes magnitude when spatial filter cutoffs expand or contract.
Worked Mathematical Case for Areal Parameter Shift
Consider a micro-structured cyclic olefin polymer (COP) diffractive optical element moulded inside a nickel-cobalt electroformed insert with a nominal feature pitch of eight micrometers. The nickel insert core possesses an electro-deposited micro-texture with high spatial frequency sharpness. The COP replica is measured using a coherence scanning interferometric microscope with a 50x objective lens (numerical aperture 0.55, optical resolution 0.55 micrometers).
To evaluate the impact of S-filter cutoff selection on surface parameters, the identical raw height dataset gathered from the COP micro-moulding undergoes three distinct filtering routines under ISO 25178-3 rules:
- Unfiltered Raw Topography Baseline measurement dataset containing raw optical profiler point clouds including high-frequency electronic noise and baseline tilt.
- Narrow Spatial Bandpass Applied S-filter cutoff λs = 0.8 µm and L-filter cutoff λc = 0.08 mm to isolate pure micro-texture while stripping high-frequency optical scatter.
- Wide Spatial Bandpass Applied S-filter cutoff λs = 2.5 µm and L-filter cutoff λc = 0.25 mm, passing longer spatial wavelengths into the roughness calculation.
The calculated areal surface parameters derived from these three filtering configurations demonstrate the sensitivity of quality metrics to spatial cutoff decisions:
Under the raw unfiltered configuration, the measured Sa is 42.1 nanometers, Sz is 418.5 nanometers, Ssk is -0.12, and Sdr is 14.8 percent. High-frequency sensor shot noise inflates the surface area ratio Sdr by generating artificial point-to-point slope spikes across flat plateaus.
Applying the narrow spatial bandpass (λs = 0.8 µm, λc = 0.08 mm) strips sensor scatter while retaining authentic micro-moulded geometry. The resulting Sa drops to 31.4 nanometers, Sz falls to 245.2 nanometers, Ssk shifts to -0.68, and Sdr drops to 3.2 percent. The negative skewness (Ssk = -0.68) reveals that the polymer melt failed to fully penetrate narrow tool valleys, leaving recessed micro-voids across the replica while rounding top peaks.
Applying the wide spatial bandpass (λs = 2.5 µm, λc = 0.25 mm) filters out true sub-micron replication details. The calculated Sa rises to 38.6 nanometers due to the inclusion of long-wavelength thermal sink undulations. Sz increases to 312.0 nanometers, Ssk shifts toward neutral at -0.22, and Sdr drops to 0.8 percent.
Increasing the S-filter cutoff from 0.8 micrometers to 2.5 micrometers attenuates sixty percent of the genuine high-frequency replication detail achieved by the variothermal moulding process.
A variothermal micro-moulding process holding cavity wall temperature at 155°C during injection achieves an Sdr parameter of 3.2 percent on COP optical replicas, whereas a standard tool at 95°C yields an Sdr of only 0.6 percent under identical 0.8 micrometer S-filter conditions.
Melt contact dictates replication. When molten polymer contacts cold cavity steel, an amorphous frozen layer forms within milliseconds. This solid skin resists cavity pressure, bridging across fine tool scratch features and suppressing high spatial frequency transmission.
Height and Spatial Texture Metrics across Scale Boundaries
Micro-injection moulding surface defects present unique spatial spectral signatures. Distinguishing tool wear from polymer processing failure requires identifying which spatial frequencies carry the defect signal. Shear-induced melt fracture alters mid-range spatial wavelengths, whereas tool polishing marks manifest as highly directional high-frequency spatial periodicities.
- Shear Induced Skin Tearing High melt shear rates through micro-gates cause polymer chain scission, producing micro-scale surface roughness oriented parallel to flow.
- Micro Cavity Underfilling Insufficient packing pressure or fast gate seal leaves tool corners unreached, reducing Ssk negativity and blunting functional peak sharpness.
- Tool Polish Mark Transfer Diamond paste scratch patterns on electroformed or CNC-milled steel insert surfaces replicate directly into the polymer matrix as high-frequency anisotropic ridges.
- Variothermal Thermal Hesitation Temperature fluctuations across the mould surface create localized boundary lines that appear as long-wavelength waviness under wide L-filter cutoffs.
Unfiltered data corrupts acceptance. Instrument noise skews parameters. Setting the short-wavelength cutoff below the physical tip radius of the stylus or the optical spatial resolution threshold always transforms instrument noise into non-existent surface micro-texture.
Window
Processing conditions during micro-injection moulding govern how accurately the molten polymer replicates high-frequency tool asperities. The press-side setting engineer holds direct control over cavity melt pressure, variothermal temperature boundaries, injection velocity profiles, and cooling dwell times. Each parameter moves the spatial replication bandwidth of the polymer skin layer.
Establishing a robust moulding process window requires tracking surface texture transmission as a function of processing variables.
How Does Melt Temperature Shift Surface Spatial Frequencies?
Thermal energy in the polymer melt directly alters the viscoelastic relaxation time and skin layer thickness during cavity filling. Higher melt temperatures reduce bulk viscosity, allowing molten polymer chains to flow into high-aspect-ratio sub-micron cavity features before thermal vitrification locks the surface structure. Lower melt temperatures increase frozen layer thickness, preventing the melt front from contacting the deepest troughs of electroformed micro-inserts.
Variothermal tool heating elevates cavity steel temperatures above the polymer glass transition temperature (Tg) during the injection stroke. Maintaining tool steel at 160°C for polyetheretherketone (PEEK) or cyclic olefin polymers allows cavity pressure to force the viscous melt front into complete contact with micro-machined structures. Rapid variothermal cooling then drops tool temperature below Tg prior to mold opening, preventing ejection deformation on fine surface ribs.
Advanced optical metrology used in semiconductor photolithography stepper lenses applies identical spatial bandpass concepts to evaluate sub-nanometer wave-front distortion. Surface roughness on micro-injection tool cores obeys these same wave-propagation mechanics, where spatial cutoff selection separates mechanical machine vibration from chemical polishing boundaries.
Quantifying surface transmission across a process study requires measuring the identical surface region on both the steel tool insert and the resulting plastic replica. The replication fidelity ratio (Rfr) for a given surface parameter is expressed as the percentage ratio of the plastic parameter value to the tool core parameter value under identical ISO 25178 filter conditions.
Viscosity limits filling velocity. Pressure locks the surface. Thermal gradient freezes flow.
Achieving ninety-five percent replication fidelity on features under two micrometers pitch requires peak cavity pressures exceeding eight hundred bar combined with variothermal mould temperatures matching or exceeding polymer Tg during injection.
Press-Side Protocol for Spatial Transmission Mapping
Establishing the exact spatial resolution limit of a given moulding machine requires systematic variations in cavity pressure and mold wall heating. A structured press-side procedure isolates process limits without introducing external measurement error.
- Mount the micro-textured tool insert into the mold base and verify alignment using an optical profiler to measure insert baseline surface parameters (Sa, Sz, Sdr) under ISO 25178-3 standards with λs = 0.8 µm and λc = 0.25 mm.
- Purge the moulding press barrel and establish baseline barrel temperatures according to resin manufacturer specifications for high-flow micro-moulding grades.
- Set variothermal mold temperature target above polymer glass transition temperature and initiate dry cycling to achieve thermal equilibrium across all core and cavity plates.
- Inject first-shot trial parts at twenty percent maximum injection velocity, holding packing pressure at fifty percent of maximum hydraulic capability for three seconds.
- Increment injection velocity in ten percent steps up to ninety-five percent machine limit, collecting five consecutive stable parts at each increment while maintaining constant variothermal thermal targets.
- Increment cavity packing pressure from four hundred bar to twelve hundred bar in two hundred bar steps, recording cavity pressure sensor peak values and screw cushion stability.
- Measure replicated polymer parts across all experimental steps using the identical optical profiler settings, computing the replication fidelity ratio for Sa, Sz, and Sdr parameters.
- Plot surface transmission curves against peak cavity pressure to locate the asymptotic replication limit where further pressure increases deliver no further spatial resolution gains.
Moulding trials demonstrate that increasing cavity packing pressure from 400 bar to 1000 bar improves the high-frequency replication fidelity of 1-micrometer laser textures from 42 percent to 91 percent under variothermal conditions.
Process window studies on micro-moulded polyetheretherketone (PEEK) parts demonstrate that holding cavity surface temperature at 175°C during injection yields an Sdr replication fidelity of 88 percent when evaluated with an S-filter cutoff of 0.8 micrometers. Dropping cavity surface temperature to 130°C causes Sdr replication fidelity to drop to 34 percent under identical pressure and velocity settings. This published relationship rests on a ten-shot sample average using a 15-tonne electric micro-moulding press equipped with a 14-millimeter injection screw.
Shift the screw diameter to 18 millimeters or remove variothermal oil heating, and the thermal skin growth rate doubles, shifting the required packing pressure upward by three hundred bar to achieve equivalent surface transmission.
Press-side evaluation of micro-fluidic sealing ridges exposes uncertainty when converting profilometer area metrics into production machine settings. Moulders frequently rely on a simplified rule of thumb stating that doubling hydraulic packing pressure halves un-replicated boundary gaps on sub-micron features. Field data indicates that above eight hundred bar cavity pressure, polymer melt compressional heating and shear thinning yield non-linear spatial fill jumps that defy simple proportional scaling.
A buyer navigating tool trials under this uncertainty specifies variothermal hold times based on cavity pressure sensor threshold decay rather than arbitrary time-based timer settings on the press control screen.
| Polymer Grade | Mould Heating Type | Cavity Pressure | Cutoff λs | Cutoff λc | Sa Replication Ratio | Sdr Replication Ratio |
|---|---|---|---|---|---|---|
| COP Optical Grade | Variothermal (150°C) | 950 bar | 0.8 µm | 0.25 mm | 96.4 % | 92.1 % |
| COP Optical Grade | Standard Water (90°C) | 950 bar | 0.8 µm | 0.25 mm | 68.2 % | 41.5 % |
| PMMA High-Flow | Variothermal (115°C) | 850 bar | 0.8 µm | 0.25 mm | 94.1 % | 88.7 % |
| PMMA High-Flow | Standard Water (75°C) | 850 bar | 0.8 µm | 0.25 mm | 61.0 % | 32.8 % |
| PEEK Medical Grade | Variothermal (185°C) | 1100 bar | 0.8 µm | 0.25 mm | 91.5 % | 84.2 % |
| PEEK Medical Grade | Standard Oil (140°C) | 1100 bar | 0.8 µm | 0.25 mm | 52.3 % | 21.9 % |
| Data compiled from variothermal press qualification trials using 0.5 mm nickel electroformed micro-fluidic tool core inserts. Transmission ratios calculated against baseline steel topography measured on a coherence scanning interferometric microscope. | ||||||
Selecting an inappropriate filter cutoff during process window qualification leads to mistaking poor variothermal mould heating for tool wear, resulting in unearned tool refurbishment costs and premature mold cavity rework.
Discrepancy
Instruments used for non-contact areal surface characterisation generate non-physical data when measuring high-aspect-ratio micro-moulded features. Optical profilers, including coherence scanning interferometers, confocal microscopes, and focus variation systems, interact differently with transparent or semi-crystalline polymer matrices than with polished tool steel cores. Selecting surface texture filtering parameters without accounting for optical artifact generation leads to invalid quality rejections.
Roughness masks tool wear. Draft angle restricts access. Optical noise alters parameters.
Transparent polymers like polymethyl methacrylate (PMMA) or polycarbonate (PC) allow partial internal light reflection below the physical surface layer, creating subsurface optical scattering that appears on the profiler map as artificial high-frequency height noise.
Optical Interrogation Limits on Micro-Moulded Polymers
Coherence scanning interferometry and confocal microscopy struggle with steep local slope angles exceeding twenty degrees on clear acrylic parts. When light rays strike a micro-structure sidewall at high angles, reflected light falls outside the objective lens numerical aperture, creating unmeasured data points known as dropout pixels. Internal profiler software algorithms frequently interpolate these missing points, generating synthetic high-frequency spatial frequencies that corrupt unfiltered Sa and Sq readings.
Filtering acts as the primary line of defense against optical measurement artifacts. Applying an S-filter removes single-pixel spikes caused by phase-unwrapping errors in interferometric data. Applying a robust Gaussian filter suppresses the influence of localized data dropouts without smearing the physical boundary of micro-fluidic channels or diffractive optical steps.
- Specular Reflection Dropout Steep micro-structure sidewalls reflect light away from the collector optics, leaving missing data points that distort baseline spatial calculations.
- Subsurface Internal Scatter Partial transparency in unpigmented polymers shifts the perceived optical focal plane beneath the physical part surface, adding artificial vertical variance.
- Numerical Aperture Clipping Lenses with numerical aperture below 0.55 lose spatial signals finer than 0.8 micrometers wavelength, acting as an uncalibrated hardware low-pass filter.
- Phase Unwrapping Jumps Coherence scanning interferometers measuring step heights greater than quarter-wavelength threshold generate monochromatic phase jumps, introducing fake vertical steps.
Specular reflection limits signal. Physical stylus profiling provides a mechanical alternative to optical measurement but introduces mechanical filtering effects. A stylus tip radius of two micrometers cannot physically penetrate a tool core valley narrower than four micrometers width.
The physical tip acts as a mechanical S-filter, smoothing out narrow troughs while faithfully recording peak heights. Calculating surface texture parameters from stylus data without adjusting digital S-filter cutoffs to match stylus tip geometry results in double-filtering, artificially suppressing real roughness values by up to thirty-five percent.
Applying a digital S-filter cutoff equal to the physical stylus tip radius avoids double-filtering distortion, preserving true surface height representation across micro-moulded inspection lots.
Stylus Radius Smoothing and Tip Filtering Dynamics
Physical contact mechanics set a hard boundary on stylus measurement resolution on soft polymer surfaces. When a diamond stylus with a 2-micrometer tip radius traverses a soft thermoplastic like high-density polyethylene under a 0.75-milliNewton contact force, the tip induces localized elastic and plastic deformation. This mechanical indentation smooths micro-peaks during the measurement sweep, effectively altering the high-frequency surface spatial spectrum before software filters process the trace signal.
Published metrology studies show that measuring micro-moulded polypropylene textures with a 2-micrometer stylus tip under 0.75 milliNewtons contact force reduces measured Sa values by 18 percent relative to non-contact optical measurements using a 0.55 numerical aperture objective lens. This reduction rests on ten repeat profile runs performed at 20°C ambient room temperature under ISO 3274 contact instrument calibration rules. Increasing contact force to 1.5 milliNewtons deforms the polymer micro-peaks further, increasing measured Sa deviation to 31 percent.
Digital filtering cannot recover physical height detail erased by mechanical stylus indentation during profiling sweeps.
Toolmakers frequently attribute dimensional non-conformance on micro-textured inserts to measurement instrument discrepancy rather than incomplete cavity filling during the packing phase.
Invoice
Drawing callouts that fail to specify nested spatial cutoff boundaries create commercial disputes during tooling acceptance trials. Tooling purchase specifications frequently request a specific surface roughness metric, such as Sa = 0.05 micrometers, without identifying the required S-filter cutoff, L-filter cutoff, or filter type under ISO 25178-3. This ambiguity enables toolmakers to validate inserts using filtering parameters optimized to hide polishing defects, while the buyer inspects incoming moulded parts using tighter filtering windows that reveal replication flaws.
Aligning commercial tooling contracts with ISO 25178 metrology standards eliminates ambiguity during quality sign-off. A legally robust engineering drawing designates the complete surface texture filter chain alongside upper and lower parameter specification limits. Cutoff boundaries set the cost.
Unfiltered data corrupts acceptance.
Contractual Drawing Callouts and ISO 25178-1 Requirements
Engineering drawings for high-precision micro-mouldings designate explicit transmission bands rather than isolated numerical limits. A complete areal surface callout includes the standard designation, parameter symbol, target numeric limit, filter type, and nesting index limits. For example, a drawing callout stating ISO 25178-2 Sa 0.05 µm / Gaussian S 0.8 µm / Gaussian L 0.25 mm forces both tool manufacturer and moulder to measure parts under identical mathematical filter conditions.
Tooling sign-off agreements should establish explicit measurement protocols for both the polished steel cavity insert and first-article polymer replicas. Discrepancies between tool core topography and plastic part surface parameters trigger formal defect analysis. If the tool core meets the requested Sa parameter under specified filter cutoffs, but the plastic part fails to achieve ninety percent replication fidelity, commercial responsibility sits with the moulding injection press processor rather than the toolmaker.
| Drawing Specification Callout | Metrology Ambiguity | Toolmaker Verification Route | Buyer Incoming Quality Control | Commercial Dispute Consequence |
|---|---|---|---|---|
| Sa < 0.05 µm (No filter specified) | Undefined S-filter and L-filter spatial cutoffs | Applies wide λs (8 µm) to filter out fine polishing scratches | Applies narrow λs (0.8 µm) capturing fine scratch detail | Tool rejected at receiver dock; $25,000 tool rework dispute over non-conforming surface finish. |
| Ra < 0.1 µm (Profile callout on 3D surface) | Profile trace applied to 3D micro-texture geometry | Takes single trace along polishing lay direction | Takes cross-lay trace or areal optical scan | Batch rejection of 100,000 optical components; dispute over directional surface lay interpretation. |
| Sdr > 5 % (No instrumentation specified) | Optical profiler vs mechanical stylus method variance | Measures with optical profiler susceptible to noise spikes | Measures with mechanical stylus that smooths micro-peaks | Dispute over functional hydrophobic surface performance; contract delay pending third-party audit. |
| ISO 25178 Sa 0.04 µm S 0.8µm / L 0.25mm | Fully specified nesting indices and filter class | Executes standard calibration scan matching exact parameters | Executes identical digital bandpass scan on incoming parts | Zero ambiguity; pass/fail determined objectively; immediate sign-off of tooling milestone payment. |
Quality Management Integration for Replicated Micro-Textures
Integrating areal surface filtering standards into statistical process control protocols guarantees long-term production stability. Micro-injection moulding processes producing optical or micro-fluidic parts require routine monitoring of surface texture parameters across multi-cavity tools. Dimensional wear on micro-moulding tools manifests as a gradual loss of high spatial frequency replication, detectable only when tracking parameters isolated by narrow S-filter boundaries.
Tracking the Developed Interfacial Area Ratio (Sdr) under an S-filter cutoff of 0.8 micrometers serves as an early indicator of cavity contamination and micro-vent clogging. As volatile resin additives deposit micro-films inside tool cavity corners during long production runs, Sdr values degrade prior to any detectable drift in primary part dimensions or bulk weight. Identifying surface degradation early allows press setters to execute scheduled tool cleaning before producing out-of-specification scrap parts.
Invoking ISO 25178-3 section 5.2 in the tooling purchase agreement establishes that any surface texture parameter quoted without its associated S-filter and L-filter nesting indices is legally non-binding during tool handover.