A machinist hands you a finished shaft. It looks smooth. The dimension is within tolerance. But the customer returns it two weeks later — the rubber seal is leaking. You put the shaft under a profilometer and see it immediately: Ra = 0.4 µm, well within spec. But there is one deep scratch, 8 µm below the mean line, running straight through the sealing zone. Ra never caught it. Rz would have.
Surface roughness is not one number. It is a family of parameters, each measuring something slightly different. Knowing which one to specify — and why — is one of the marks of an experienced quality engineer.
What Surface Roughness Actually Measures
When a cutting tool moves across metal, it leaves a pattern of microscopic peaks and valleys. The wavelength and height of these irregularities depend on the tool geometry, feed rate, material, and cutting conditions. Surface roughness parameters quantify this pattern in different ways.
All roughness measurements start with a profile — a cross-sectional trace of the surface taken by a stylus (contact) or laser (non-contact) instrument. From this profile, a mean line is calculated. Every roughness parameter is then derived from how the actual profile deviates from that mean line, measured over a defined sampling length.
The sampling length matters. A surface that looks rough over 0.08 mm may look smooth over 0.8 mm, because shorter wavelengths get averaged out. JIS B 0601 specifies standard sampling lengths (λc cutoff values) that must match the expected roughness range. Measuring Ra = 1.6 µm with the wrong cutoff can give a reading of Ra = 0.4 µm — a passing result on a failing surface.
Ra — The Global Average That Hides Extremes
The arithmetic average of all absolute deviations from the mean line over the sampling length. The most widely used roughness parameter globally.
Ra is calculated by taking every point on the profile, measuring its absolute distance from the mean line, and averaging all those distances together. It is stable, repeatable, and easy to communicate. These properties explain its dominance in engineering specifications worldwide.
Its weakness is precisely what makes it stable: Ra treats a surface with one catastrophic scratch the same as a surface with uniform, gentle waviness — provided the total average is the same. In many applications, that is acceptable. In sealing surfaces, bearing journals, and fatigue-critical parts, it is not.
"Ra is the blood pressure of surface roughness. A normal reading reassures you. But it cannot tell you whether you had one massive spike an hour ago." — Metrology department training session, Shiga Prefecture, 2014
Rz — The Peak-to-Valley Parameter That Catches Defects
The average of the five largest peak-to-valley heights measured across five equal sections of the evaluation length. Sensitive to isolated defects that Ra will miss.
Rz divides the evaluation length into five equal sections. In each section, it finds the highest peak and the lowest valley and measures the distance between them. It then averages those five peak-to-valley heights. Because each section contributes its worst event — not an average — Rz is far more sensitive to scratches, pits, and tool marks than Ra.
The relationship between Ra and Rz is not fixed. For a uniform, sinusoidal profile, Rz ≈ 4 × Ra. But for a surface with isolated defects, Rz can be ten to twenty times Ra. When you see a large gap between Ra and Rz on the same surface, something unusual is happening — and it almost always warrants investigation.
Rzjis — Japan's Own Standard That Still Appears on Old Drawings
Here is where foreign engineers encounter their first trap with Japanese drawings. Until 1994, JIS B 0601 defined Rz differently from the ISO standard — as the average of the five highest peaks plus the five deepest valleys, not the five peak-to-valley heights within five sections.
When ISO 4287 was adopted, Japan created a new parameter — Rzjis — to preserve the old definition in the JIS standard. If you see "Rz" on a drawing issued before approximately 2000, it may mean Rzjis, not ISO Rz. The numerical values are similar but not identical, and the difference can be critical on tight specifications.
Always check the drawing standard referenced in the title block. If it cites JIS B 0601:1994, Rz means Rzjis. If it cites JIS B 0601:2001 or later, Rz follows the ISO definition. On legacy drawings with no explicit standard reference — ask. Do not assume.
The Triangle System — Reading Old Japanese Drawings
Before Ra and Rz became standard, Japanese drawings used a triangle symbol system to specify surface finish. You will still encounter this on drawings from the 1970s through the 1990s, particularly in automotive and machine tool supply chains where legacy drawings remain in use.
The triangle system communicated the process as much as the result. One triangle meant turning or coarse milling. Two triangles meant finish turning or fine milling. Three triangles implied grinding. Four triangles meant lapping or superfinishing. An experienced machinist reading the symbol knew immediately which machine and which operation would achieve it.
The system was replaced because it was imprecise — the Ra equivalent of three triangles depends on who you ask and which era's JIS edition you reference. But it remains embedded in the manufacturing culture. Older engineers in Japan still think in triangles, and will sometimes sketch them informally in meetings even when the official drawing uses Ra values.
Choosing the Right Parameter for the Application
| Parameter | Best Used For | Typical Values |
|---|---|---|
| Ra | General machined surfaces, comparative QC, process monitoring | 0.1 – 25 µm |
| Rz | Sealing surfaces, bearing journals, fatigue-critical areas, plated surfaces | 0.4 – 100 µm |
| Rzjis | Legacy Japanese drawings (pre-2000), automotive supplier drawings citing JIS 1994 | Similar to Rz |
| Rq (RMS) | Optical surfaces, electronics, where statistical distribution matters | ≈ 1.1 × Ra |
For sealing surfaces specifically, Rz is almost always the correct choice. An O-ring seating groove with Ra = 0.8 µm looks fine on paper. But if Rz = 12 µm, there are peaks tall enough to damage the O-ring under compression, or valleys deep enough to create a leak path. Ra cannot see either condition. Rz can.
How Japanese Drawings Specify Surface Roughness
Modern Japanese drawings following JIS B 0601:2013 (aligned with ISO 4287/4288) use a standardized symbol consisting of a tick mark with a horizontal bar, to which roughness values and additional requirements are attached. The general structure reads:
Ra 0.8 — Upper limit of arithmetic mean roughness: 0.8 µm
Rz 6.3 — Upper limit of maximum height: 6.3 µm
(U) — Upper limit only (default; "L" = lower limit, "U/L" = both)
When both Ra and Rz appear, the inspector must verify both. Passing one does not guarantee the other.
When a roughness value appears without a parameter name, and the drawing title block references JIS B 0601:2001 or later, Ra is assumed by default. On older drawings, Rz (meaning Rzjis) may be the default. This ambiguity is one reason modern drawings explicitly name the parameter.
Measuring Surface Roughness on the Shop Floor
The most common shop-floor instrument is the contact profilometer — a stylus with a diamond tip of defined radius (typically 2 or 5 µm) that drags across the surface while a transducer records vertical displacement. The raw signal is filtered to separate roughness (short wavelengths) from waviness (longer wavelengths) from form error (the longest wavelengths).
The cutoff wavelength λc defines this filter. For Ra = 0.8 µm, the standard cutoff is λc = 0.8 mm, and the evaluation length is 4.0 mm (five sampling lengths). Measuring with the wrong cutoff — say, λc = 0.25 mm on a surface intended for λc = 0.8 mm — will systematically underreport roughness. The instrument will pass parts that should fail.
"The most dangerous measurement error is not the one that fails a good part. It is the one that passes a bad part — quietly, repeatedly, without anyone noticing." — Quality audit finding, precision machining supplier, Osaka, 2019
For internal bores, small radii, and features inaccessible to a contact stylus, non-contact methods (white light interferometry, confocal microscopy, laser triangulation) provide the same data without physical contact. These instruments are increasingly common in Japanese precision parts manufacturing, though they require careful calibration and interpretation — a flat surface measured at an angle will report a higher roughness than the same surface measured normally.