Quick Answer
Ra, Rz and RMS are three different scores for the same surface, and they are never equal. On ground and abraded metal, Rz typically runs 4 to 7 times higher than Ra, while RMS (written Rq on modern drawings) runs about 1.11 times Ra (KB: Surface Finish Ra). One micron is 39.37 microinches, commonly rounded to 40, so an Ra of 0.8 µm is the familiar 32 µin callout (KB: Surface Roughness Parameters). Treat those ratios as sanity checks, not certifications: if the drawing specs Rz, measure Rz.
Three numbers, one surface
A finish spec is almost never a single fact. A stylus profilometer drags a fine diamond tip across the part, records a height profile, splits roughness from waviness with a cutoff filter, then computes several statistics from the same trace. Which statistic the drawing quotes changes what the number protects against (KB: Surface Roughness Parameters).
- Ra, arithmetic average roughness. The mean of the absolute height deviations from the mean line. The headline number on most drawings and the most forgiving: averaging dilutes isolated deep scratches.
- Rq / RMS, root-mean-square roughness. The RMS of the same deviations; it weights big excursions more heavily, so it always reads above Ra. The old US default, which is why legacy American prints quote RMS in microinches.
- Rz, maximum height of profile. Under ISO practice, the largest single peak plus the deepest valley within a sampling length, averaged over the default five. It keys off the extremes Ra smooths over.
- Rmax / Rt, worst-case peak-to-valley. Rmax is the largest peak-to-valley of the worst single sampling length; Rt spans the whole evaluation length. Specified when a single flaw, a leak path or a fatigue-crack starter, means a reject.
- CLA. Centre-line average, on old UK drawings: simply Ra under an earlier name (KB: Surface Finish Ra).
The ordering is fixed: Ra is always smallest, Rq sits just above it, and the peak-to-valley family stacks higher still, with Rt ≥ Rmax ≥ Rz (KB: Surface Roughness Parameters). Comparing an Rz reading against an Ra limit rejects good parts; the reverse ships bad ones.
The conversion multipliers
These are the working ratios most shops use to sanity-check a spec written in an unfamiliar parameter (Engineers Edge and FIRGELLI conversion references; KB: Surface Finish Grades and Ra Targets by Application):
| Parameter | What it measures | Rule of thumb vs Ra |
|---|---|---|
| Ra | arithmetic mean of profile deviations | reference (1×) |
| RMS / Rq | root-mean-square of the same deviations; old US default | ≈ 1.11 × Ra |
| Rz | mean of the five largest peak-to-valley heights (ISO) | ≈ 4–7 × Ra (grinding ≈ 5.5×, polishing ≈ 4×) |
| Rmax / Ry | single largest peak-to-valley in any sampling length | ≈ 5 × Ra |
| Rt | total peak-to-valley over the full evaluation length | ≈ 6 × Ra |
Some references put typical ground surfaces nearer 7.2 times Ra, which shows how widely the real ratio scatters with the process (Fractory, cited in KB: Surface Roughness Parameters). The ratios only hold for similar surfaces made by similar processes: you cannot back-convert a measured Rz into a guaranteed Ra, and if a customer specs Rz, you measure Rz (KB: Surface Finish Ra).
The chart: Ra, microinches, RMS and ISO N-grades in one table
The decode table to keep by the profilometer: Ra in both units, the RMS an older US drawing might carry, the legacy ISO N-grade, the cutoff, and the process that typically lands each band (Engineers Edge and Astro Pak conversion tables; KB: Surface Finish Ra):
| Ra (µm) | Ra (µin) | RMS (µin) | N-grade | Cutoff (mm) | Typical process / abrasive |
|---|---|---|---|---|---|
| 50 | 2000 | 2200 | N12 | 8.0 | Hot-rolled / sand-cast (as received) |
| 25 | 1000 | 1100 | N11 | 2.5 | Rough machining, heavy grinding |
| 12.5 | 500 | 550 | N10 | 2.5 | Heavy turning/milling; coarse 24–36 grit |
| 6.3 | 250 | 275 | N9 | 2.5 | Standard machining; 40–60 grit grinding |
| 3.2 | 125 | 137 | N8 | 2.5 | As-machined CNC; 80 grit |
| 1.6 | 63 | 69 | N7 | 0.8 | Fine machining; 120 grit |
| 0.8 | 32 | 36 | N6 | 0.8 | #4 brushed / sanitary finish; 180–240 grit |
| 0.4 | 16 | 18 | N5 | 0.25 | Honing / fine polishing; 320–400 grit |
| 0.2 | 8 | 9 | N4 | 0.25 | Lapping / fine polish; 600 grit |
| 0.1 | 4 | 4.4 | N3 | 0.25 | Super-finishing; 800–1000 grit |
| 0.05 | 2 | 2.2 | N2 | 0.25 | Fine lapping |
| 0.025 | 1 | 1.1 | N1 | 0.08 | Mirror / optical finish |
The bolded row is the single most useful anchor in the trade: Ra 0.8 µm, 32 µin, N6. It is the No. 4 architectural stainless grade, the 3-A sanitary ceiling and the ASME BPE SF3 limit all at once, and it is typically reached with a 180-grit ladder (KB: Surface Finish Grades and Ra Targets by Application). Each N-grade step is roughly a doubling of Ra. For the companion table mapping each grit band to the Ra it leaves, see our grit-to-Ra surface finish chart.
Microns to microinches
ISO drawings quote Ra in micrometres (µm); North American prints quote microinches (µin). The conversion is 1 µm = 39.37 µin, commonly rounded to 40, and 1 µin = 0.0254 µm (Astro Pak conversion, cited in KB: Surface Roughness Parameters). If a number on a print looks forty times too big or too small, you are looking at the other unit.
Worked example: an Rz spec lands on the same 180-grit ladder
Say a stainless drawing calls out "Rz ≤ 6 µm." By the 4–7× rule that corresponds to an Ra around 0.9 to 1.5 µm, and in practice the spec-translation table treats "Rz under 6 µm," "32 Ra," "N6," "SF1" and "No. 4" as pointing at the same roughly 180-grit finishing ladder (KB: Surface Finish Grades and Ra Targets by Application). The walkthrough for the finish itself is in our guide to getting a No. 4 or No. 8 finish on stainless. For sign-off the caveat still rules: a critical Rz spec is verified with an Rz measurement at the stated cutoff, not back-calculated from Ra.
The ISO version traps
Three different Rz values share one symbol
"Rz" has meant at least three things across eras: the legacy DIN ten-point Rz, which averaged the five highest peaks and five deepest valleys; the ISO 4287 Rz, the largest peak plus deepest valley per sampling length, averaged over five sampling lengths; and the ISO 21920 Rz, renamed "maximum height of the profile" and computed once over the whole evaluation length (KB: Surface Roughness Parameters). Same symbol, different arithmetic: two "Rz 6.3" callouts under different standards are not interchangeable, so confirm which standard the drawing references before comparing readings.
ISO 21920 replaced ISO 4287 and 4288 in December 2021
The ISO 21920 family, published December 2021, superseded ISO 4287, ISO 4288, ISO 13565 and ISO 1302 (Digital Surf metrology guide, cited in KB: Surface Roughness Parameters). It replaced the term "sampling length" with "section length," made the "max rule" the default acceptance basis instead of ISO 4287's 16% rule, and computes most parameters once over the evaluation length rather than averaging five sampling lengths. Even Ra is not immune: the definition is essentially unchanged, but the filtering order changed, so the same surface can return a slightly different Ra under the new document (KB: Surface Finish Ra). A huge installed base of drawings still calls out ISO 4287, so expect mixed callouts for years. The parallel North American standard is ASME B46.1; its 2019 edition is current and is the one most often referenced on US OEM prints (KB: Surface Finish Ra).
No cutoff, no number
Every roughness value depends on the cutoff filter (λc) that separates roughness from waviness, and the cutoff is chosen from the expected Ra, not picked freely (ISO 4288 / JIS B 0633 rules via the Mitutoyo roughness guide; KB: Surface Roughness Parameters). For Ra between 0.1 and 2 µm, which covers most abrasive-finished metal, the default is λc = 0.8 mm with five sampling lengths, a 4 mm evaluation length. Coarser work (Ra 2–10 µm) needs a 2.5 mm cutoff. A spec without a stated cutoff is only half a spec: quote λc with the number.
What this means at the grinder
Decode the spec into an Ra band, then pick the grit ladder that lands it: whatever the drawing says, Rz 6, 32 RMS, N6, translate it with the multiplier table and step through the grits without skipping more than one grade; a finer abrasive cannot reach the bottom of a coarser grit's scratch valleys (KB: Grit to Surface Finish Crosswalk). Sequencing discipline is covered in our grit progression guide. One more trap: FEPA-P and CAMI grit numbers agree only to about 220, then diverge, so confirm the scale before trusting a fine grit number; the details are in our FEPA vs CAMI comparison.
On the product side, the coarse-to-middle bands are flap disc and sanding belt territory, and the fine satin and pre-polish steps move to non-woven finishing wheels. We stock the abrasives that produce these numbers, not the gauges that measure them; the finish you achieve depends on pressure, speed, lubricant and wear as much as on grit, so verify a critical Ra with a profilometer rather than assuming it from the grit number (KB: Surface Roughness Parameters).
FAQ
How do I convert Rz to Ra for a ground surface?
Divide by roughly 4 to 7; ground surfaces typically sit near 5.5, so Rz 6.3 µm suggests an Ra around 1.1 µm (KB: Surface Finish Ra). It is a sanity check only. The ratio depends on the process, so a spec written in Rz should be verified with an Rz measurement, never certified by back-conversion from Ra.
Is RMS the same as Ra?
No. RMS (Rq) is the root-mean-square of the profile deviations, which weights large peaks and valleys more heavily, so it always reads higher than Ra, about 1.11 times Ra as a rule of thumb (KB: Surface Finish Ra). Older US drawings often quote RMS in microinches where a modern print would quote Ra.
How many microinches is 1 micron Ra?
1 µm = 39.37 µin, commonly rounded to 40 (KB: Surface Roughness Parameters). The useful anchors are 0.8 µm ≈ 32 µin, 1.6 µm ≈ 63 µin and 3.2 µm ≈ 125 µin.
Why does the same part read a different Rz under ISO 21920 than under ISO 4287?
ISO 4287 averaged Rz over five sampling lengths; ISO 21920 (December 2021) computes it once over the whole evaluation length and renamed it "maximum height of the profile" (KB: Surface Roughness Parameters). Different arithmetic on the same trace gives a different number, so state which standard governs before comparing readings.
Which is always bigger, Ra or Rz?
Rz, always. The fixed ordering is Ra smallest, then Rq (RMS), then the peak-to-valley family with Rt ≥ Rmax ≥ Rz (KB: Surface Roughness Parameters). Ra averages the whole trace while Rz keys off the extreme peaks and valleys, so Rz sits several times higher on the same surface.
What grit gets me under an Rz 6 micron spec on stainless?
In spec-translation terms, "Rz under 6 µm" points at the same roughly 180-grit ladder as a No. 4, 32 Ra, N6 or SF1 callout, an Ra near 0.8 µm (KB: Surface Finish Grades and Ra Targets by Application). Step to it through the grits without skipping more than one grade, and confirm the finished part by measuring Rz at the stated cutoff.