Surface Finish Chart: Ra Roughness Conversion Table (Free PDF)

Any number on any line is matched with a known value. If you see a 125 microinch drawing callout, you know that is an Ra 3.2 micrometre. It is also RMS 137.5 microinches and a grade N8 on the ISO scale.

For reference, a N1 would be a mirror lapped surface (Ra = .025 um); and a N12 would be a sand cast or torch cut surface (Ra = 50 um). The numbers is based off the preferred-number series used around the world on drawings; they double each time down a row, increasing roughness accordingly.

Ra is the average roughness or roughness average (also expressed as AA, for arithmetic average). In other words, this is the arithmetic mean deviation of the profile from its centre line.

This is the parameter most commonly specified on today's drawings. You'll note the value in the first column, which is given in micrometres (um) on Asian and European drawings. It's given in the second column in microinches (uin) on North American drawings. Note that one um = 39.37 uin, so the same value appear in both columns, e.g., Ra 0.8 um and 32 uin are equivalent values.

CLA is an older ISO and British designation referring to the same thing; it is the center line average (given in microinches). Because some older drawings still refer to it by this name, the fourth column merely duplicates the Ra microinch value under the CLA heading.

The RMS is the root mean square of the same profile deviations. Before approximately 1955, RMS was the only specification used for US drawings and can still be found on older prints and other vessel/pipe specs today.

Generally speaking, RMS is around 11 percent larger then Ra for a machined surface, so the column is calculated as Ra x 1.1. Consider any old RMS call-out to be a little more tight than if expressed in Ra. Thus, a surface with an RMS value of 32 is approximately Ra 29 uin rather than Ra 32.

Published tables vary by several percent on each cell of RMS due to some prints showing 32.5 or 35.6 while the formula yield 35.2. The table provide clarity on what formula it uses.

Rt is the overall height of ridges, measuring from the high point to the lowest point on the valley bottom along the evaluation length. No mathematically exact correlation exist between Rt and Ra.

These are approximate traditional chart values of about 8-16 x Ra, given here as a guide and not used to decide acceptance. Measure Rt if a print has a stated Rt or Rmax value.

N Grades, ISO N grades N1 to N12 come from ISO 1302:1992 and earlier editions. From 2002 onwards, ISO 1302 revised this and dropped the N numbers in favor of calling parameters out directly.

Nowadays, the ISO 21920 series are the standard for indicating drawing levels, but N grades is still common on older drawings and discussed in workshops. Because an N grade represent one fixed Ra value, it is important that you understand the mapping.

The Lc Cut-off is the cut-off sampling length that the stylus instrument needs to apply when measuring at that level of roughness under ISO 4288. For 2 to 10 um, it's 2.5 mm; over that it's 8 mm.

Also, there are five sampling length per measurement. This is by far the most frequent reason why two gauges will not agree on the same part, they have been used with different cut-offs.

There is no conversion from Ra to Rz because this value is only an approximation. A different table is provided lower down the page.

It also doesn't cover lay or waviness, which are other surface texture elements with their own symbols on the drawing. You should of checked it first.

Use the table both ways to find out what Ra (surface finish) a process will produce or which processes can reach a affordable Ra (surface finish) level.

The Typical Ra columns indicates the expected band for that finish. This represents average use for applications produced by that process using standard tooling and conditions during normal production.

The Full range column represent the coarsest less-frequent finish obtained from that process, through to the best finish the process obtains when setup rigidly and fed finely with new tools etc. The N grade column is the translation of these typical ranges into the ISO grades using the conversion table above.

Here’s a pattern that you’ll see repeated: each type of process has its “territory” on the roughness scale. Within about a decade (give or take) of scale, there are families of processes that owns it:

Primary shaping processes, such as hot rolling, flame cutting, and sand casting, are on the rough end, with Ra values ranging from 12.5 to 50 um.

Just below them, around Ra 0.8 to 6.3 um, are conventional chip-making process such as turning, milling and drilling. This is why the near-universal default for an as-machined CNC part is Ra 3.2 um.

Below that is where abrasive processes comes into play: grinding, honing, etc., and they take over below Ra 0.8 um. Only polishing, lapping, superfinishing, and electropolishing consistently gets below Ra 0.1 um.

If a drawing calls out something outside the band of your intended process, you need to do a secondary finishing operation, or modify the process itself, and either way this cost you money.

In these ranges, look for overlap traps. You can achieve Ra 0.4 um by turning if you have a rigid machine, a large nose radius, and fine feeds; otherwise, scrap rates will be high.

Normal die-sinking EDM has Ra 1.6 to 6.3 um, and while EDM processes can gets a surface lower than 1 um with low energy finishing passes, there’s still the recast layer, and it counts when the part is fatigue critical.

A typical die cast finish range from Ra 0.8 to 1.6 um, and while it appears superior to much machined finishes, it needs machining anyway on functional faces due to porosity, flash, and draft.

Investment casting outperforms sand casting by about a factor of ten, frequently the whole reason to use it.

The cast and formed values describe the as-produced skin. The machining removes this skin and brings the finish down to the machining band. As a result, a sand-cast housing with milled faces have two different finishes on one part. This should of been noted in its drawing.

Also note, a surface is not described by roughness only. Two different processes can produce the same Ra. Directional turning grooves differ from non-directional EDM craters, and their fatigue, sliding, and sealing behaviors will differs.

This applies to metals, primarily steels, aluminium alloys, and cast irons. Other materials, such as plastics, ceramics, and additively manufactured parts, has their own rules of thumb.

For metal AM parts, the Ra values typically sits between 5 and 15 um; when you need a smoother finish, these are then rubbed using one or more of the processes in this table.

The boundaries of these bands aren't hard-and-fast guarantees but rather chart conventions. If something is finer than what's typical for that band, think of it as a negotiation with your machinist rather than an entitlement.

When selecting a finish for a drawing, use the table above. The Typical applications column shows the function of that finish. And then where do you say the corresponding Ra value? Resist the temptation to state a finer finish then required.

The Finish character column explain how it feels to touch and how it appears. At Ra 12.5 μm you can feel toolmarks with your fingernail. At Ra 3.2 μm it also feels smooth, but you see machining lines. At Ra < 0.4 μm your eye cannot detect a direction anymore.

What this says is that over-specifying is bad because it costs too much. The Relative cost column shows that.

Ra 3.2 um (normal) is natural output of a normal cut; that’s the base line. To get Ra 1.6 um requires a controlled finishing pass. Ra 0.8 um require either a trip to the grinder or some fine passes.

Each step down from there adds an extra abrasive operation, which is generally on a different machine with its own setup. Also, inspections is tighter. And machining suppliers charges you for all this stuff.

A part quoted Ra 0.8 um instead of Ra 3.2 um will cost noticeably more, sometimes just for the finish. In general, each step down the table below N8 adds cost faster than the step before.

So when you do a finish, you don’t do it to look good; you do it to fit the application. An example is sealing surfaces. An ideal Ra for a dynamic O-ring groove is about 0.4 to 0.8 um. A rougher surface leaks past the seal and tears it up, while a much smoother one can't hold the lubricant film the seal needs, so smoother is genuinly worse.

Bearing journals has an Ra of 0.2 to 0.8 um depending on load and speed.

If you’re going to paint or powder-coat or anodise the surface it makes no difference if it is fine (Ra < 3.2 um). The paint buries the finish. An exception is when surfaces is under fatigue. In that case, a finer finish is better because it polishes out the spots where cracks starts.

There are two caveats. An Ra by itself doesn’t completely characterize the working surface. For example, cylinder bores are intentionaly hatched across their diameters and honed at some controlled Ra; smoother isn’t better.  They need to hold oil.

Where’s the texture? Add lay when it matters, when it follows a certain direction.  Where’s the outlier? Add Rz limit when it matters.  Check out the table below of Ra to Rz.

And secondly, these applications are the conventions for general machinery on metal parts.  Other industries like aerospace, medical implant, semiconductor and food contact has their own surface requirements that outweigh this generalized advice. You should of looked at the moddern standards too.

Because drawings from different traditions mix two parameters, you use this table. The US and UK generally draws to Ra. Many European and some Japanese drawings calls out Rz. Engineering always requires an equivalent to work by.

When converting a preferred value series across the two conventions, you refer to the column labelled “Rz nominal”. It reflect the link between Ra and Rz preferred values that was made by the older national standards. Thus it matches Ra 0.8 um with Rz 3.2 um.

To estimate the actual measurements of something defined by Ra or Rz when tested with an instrument calibrated to the other, you consult the broader “typical measured” band. This is roughly 4-7 times Ra.

Roughness measurements are just that: two parameters measuring different things; there’s no exact conversion possible.

Ra is the arithmetic mean deviation of the whole profile. It hardly registers if you have one deep scratch.

Rz is measured according to ISO 4287 and ISO 21920 as the peak-to-valley height averaged over five sampling lengths. It is designed exacty to detect extremes.

Given equal Ra, two surfaces can be quite different in Rz based off whether they’re spiky or scratched. For example, a surface ground smoothly will have an Rz close to 4 times its Ra; a built-up-edge or torn machined surface will be 7 times. An isolated defect surface could reach 20 times. But it’s not constant, it’s a characteristic of each surface.

These variations have practical implications. Where there’s an Rz limit on a customer print, you’d better be measuring Rz. A good auditor won’t accept that you reported a converted Ra as complying.

Where you’re quoting against a spec which says Rz but you only have capability data in Ra, then divide the Rz limit by 7 and that will keep you safe. In other words, if they require Rz 6.3 um, a process capable of meeting that requirement would hold something like Ra 0.9 um or better comfortabley.

On the flip side, where a supplier gives you Rz for a surface which you called out in Ra, you should of anticipate that they’ll tell you a number which looks four to seven times worse than your callout. It could well be the same surface.

Two other pitfalls exist. On older drawings, specifically from Germany, you might see Rz(DIN) or Rmax. That's a form of the measurement which differ slightly from today's ISO Rz but sounds very similar. Both are available on instruments, but using the wrong setting change the result.

Worse, the finest rows (N1 to N3) are rarely specified in Rz at all. Single-point defect isolated from others control the readout and it's a crapshoot when reaching an Rz limit. Rt limits or even just surface defect counts does better at mirror finishes where isolated points rule.

For seal and leak-path jobs, use both parameters: Ra describes overall texture and Rz tops out the outliers.