Sampling length in surface measurement
Separating legacy sampling lengths, ISO 21920 evaluation and section lengths, filter scales and digital point spacing.
Tags: inspection, metrology, surface-texture, profilometry
On this page
Distinguish legacy sampling length, digital point spacing and total traverse. ISO 21920-3:2021, published December 2021, specifies profile-method operators and supersedes ISO 4288:1996. ISO 21920-3:2021
In Mitutoyo’s ISO 21920-3 table, an Ra upper limit of 1.6 µm falls in Sc3, with an L-filter nesting index of 0.8 mm and an evaluation length of 4 mm. The five sections in that table apply only to section-length parameters. A measured Ra of 1.6 µm does not select the setting class. Mitutoyo quick guide, printed page 55
Distinguish the length quantities
Under the ISO 4287/4288 framework described in NPL's older stylus guide:
| Quantity | Meaning in that framework |
|---|---|
| Roughness sampling length, lr | Length used to characterize profile irregularities; numerically equal to the roughness filter cutoff wavelength λc. |
| Evaluation length, ln | Total assessed length, often containing several successive sampling lengths. |
| Traverse length | Full physical trace, including any additional travel needed for transients and filtering. |
| Sampling interval | Spacing between acquired data points; distinct from the lengths above. |
NPL explains that additional travel is normally required beyond the evaluated region. The guide describes legacy parameter calculations, including parameters averaged over sampling lengths and others evaluated over the full evaluation length. “Five sampling lengths” is consequently not a complete definition of every texture measurement. NPL Good Practice Guide 37, issue 2, legacy terminology
A filter cutoff is a spatial-scale setting, not a hard geometric border that cuts the trace into isolated pieces. Filter response, edge treatment and parameter algorithms also matter. The same surface can yield different numerical results when the retained spatial bandwidth changes. Record the filter type as well as its scale.
What changes under ISO 21920
Mahr's explanation of ISO 21920 distinguishes evaluation length le, section length lsc, the number of sections and filter nesting indices. Section length remains relevant to parameters calculated from profile sections, such as Rz, Rp and Rv; it is not a universal calculation unit for all parameters. Nesting-index terminology also accommodates filters for which “cutoff wavelength” is not an adequate general description. Mahr: improved standards for surface texture
Mahr further notes that the new framework removes the previous periodic-versus-aperiodic distinction and that measurement software can provide a choice of standard. Choosing the appropriate mode matters because the definition and evaluation of a familiar parameter can change. Keep the drawing's standard basis and software configuration together; identical parameter names do not establish equivalent evaluation. Mahr: standard-compliant measurement and MarWin
The change also affects how unspecified settings are selected. Legacy procedures can use preliminary measurements and different rules for periodic and non-periodic profiles. ISO 21920 settings are tied to the specification framework, including setting classes or explicitly specified filter settings. Do not repeatedly change the filter according to the measured result until a part passes. Resolve the controlling requirements before measuring production parts. Mahr: specification-based settings
A bounded setting-class example
Mitutoyo's Quick Guide to Precision Measuring Instruments, catalog E11003(7), reproduces an ISO 21920-3 upper-tolerance-limit table. In that table, an Ra upper limit of 1.6 µm falls in Sc3. The listed Sc3 settings include L-filter nesting index 0.8 mm, evaluation length 4 mm, S-filter nesting index 2.5 µm, maximum sampling distance 0.5 µm and maximum nominal tip radius 2 µm. These are the settings for that table and specification case, not a universal recipe for any surface whose measured Ra happens to be 1.6 µm. Mitutoyo quick guide, printed page 55
The same table gives section length 0.8 mm and five sections only for section-length parameters, identifying Rz, Rp and Rv as examples. It does not instruct the user to calculate every parameter as an average of five section values. Explicit drawing requirements and the applicable parameter rules still control. Mitutoyo: ISO 21920-3 setting-class table
Reconcile evaluation length, travel and point count
Original arithmetic example: assume an approved legacy procedure specifies lr = 0.8 mm and five successive sampling lengths. Then ln = 5 × 0.8 = 4.0 mm. Suppose that particular instrument program adds 0.4 mm before and 0.4 mm after the evaluated region. Its traverse is 4.8 mm. The two additional lengths are hypothetical program settings, not recommended defaults. Their sum explains why the drive can travel farther than the reported evaluation length.
At a digital interval of 0.5 µm, the 4.0 mm evaluated span contains 4000/0.5 = 8000 intervals, or 8001 points if both endpoints are retained. It does not contain 8000 sampling lengths. Acquisition and processing software may trim endpoints differently, so verify actual output rather than using this arithmetic to infer a file's row count.
The same numerical 0.8 mm and 4 mm may appear in both a legacy procedure and an ISO 21920 setting class. That coincidence does not establish identical filtering, parameter calculation or acceptance rules. Preserve the whole method, not just these two numbers.
Short surfaces and optical measurements
A narrow land may not provide enough usable length for the required evaluation plus the instrument's additional travel. NPL discusses traverse allowances and finite stylus geometry; reducing travel can change the measurement method. If the specified settings cannot be implemented, obtain an approved alternative and report it rather than silently using a shorter evaluation. NPL: stylus measurement and lengths
An optical field of view is not automatically the required evaluation length. Bruker's correlation guidance explains that contact and optical measurements need matched scan lengths, cutoffs and filters for meaningful comparison. Multiple optical fields may be stitched to capture the relevant structure, while lateral resolution and usable data still need checking. Extracting a profile from an areal map does not by itself make it equivalent to a specified stylus trace. Bruker application note 558
Inspection record and comparison limits
Record the parameter and standard edition, drawing revision, trace location and direction, evaluation length, relevant section settings, filter types and nesting indices or legacy cutoffs. Add traverse allowances, point spacing, probe geometry, form removal and any nondefault processing. Keep raw and processed profiles where practical so differences can be investigated.
When two reports disagree, first compare their complete settings and inspected regions. A difference may arise from a changed bandwidth, shorter evaluated area, different parameter algorithm or a genuinely different surface. A calibrated instrument does not compensate for choosing the wrong specification operator. Neither a convenient factory default nor the longest available trace is automatically the correct inspection method.
Sources were reviewed November 1, 2026. NPL Guide 37 explicitly states updated February 2014 and is used here for legacy context. Mahr's standards article is dated 06.04.2022. Mitutoyo's PDF identifies E11003(7) and print code 2408 (JP12); that code is retained without inventing a separate revision date. Bruker's web note did not display a revision date.