What Does 0.5 mm/m Accuracy Mean on a Spirit Level?

Two readings of the figure circulate. One takes 0.5 mm as a fixed error, whatever the length. The other sees 0.0005 in/in on an American level and takes it for a tighter class than a European 0.5 mm/m. Both are wrong. The figure is a slope, and the two units are the same number. A third mix-up comes from machinist levels, where mm/m means something else. Below, six makers' specs go on one scale, worked over 4 ft and over a 20 ft run.

What the accuracy figure on a level measures

A spirit level is a straight bar with sealed glass or acrylic tubes, called vials, set into it. Each vial holds liquid and an air bubble. When the bar is level, the bubble sits between the two lines on the vial. The bottom edge of the bar, the measuring surface, is what rests on the work.

The accuracy figure says how far the vial can disagree with the measuring surface. A level rated 0.5 mm/m can show a centered bubble while the surface it sits on rises or falls by up to 0.5 mm in each meter. It is a tilt, so it scales with length. Over 2 m the same level can be out by 1.0 mm. Over 10 m it can be out by 5 mm.

That makes it a ratio, not a distance. The mm on top and the m below cancel to a plain number: 0.5 divided by 1,000, or 0.0005.

The same unit, mm per meter, shows up on other tools with a different meaning. On a tape measure it is an error in length, as with Class II's 0.2 mm per meter of tape, covered in what Class II means on a tape measure. On a level it is an error in angle. The two cannot be added.

0.5 mm/m and 0.0005 in/in are the same number

Write each one as a ratio:

0.5 mm / 1,000 mm = 0.0005

0.0005 in / 1 in = 0.0005

They are equal. Any length unit works on both sides, as long as it is the same one. So 0.0005 in/in is also 0.0005 ft/ft, and it is 0.006 in per foot, since a foot is 12 in.

As an angle, the ratio is the tangent of the tilt:

arctan(0.0005) = 0.0286 degrees, which rounds to 0.029.

That is why Milwaukee prints its REDSTICK box level as "Accurate to .029° ( .0005"/in, 0,5mm/M)". Johnson prints its JLX box level as "0.029° or 0.5mm/m". Stabila prints its 96-2 as "0.029° = 0.5 mm/m". Each maker is giving one figure in two or three units.

The overhead figure works the same way. arctan(0.00075) = 0.0430 degrees, which matches Stabila's 0.043 degrees for 0.75 mm/m.

Six levels on one scale

The makers publish in degrees, mm/m or in/in. Converted to all three, with the error at 48 in and over 20 ft:

The bubble levels all publish one class. Five product lines from five makers sit at 0.5 mm/m. On their spec pages, the units differ and the figure does not. A 0.0005 in/in level is not tighter than a 0.5 mm/m one.

The difference is the overhead line. Only Stabila prints a separate, looser figure for overhead use, 50 percent more than its normal figure. Empire says its figure holds "in all working positions", Milwaukee "in all working positions", Johnson "in all positions", and Kapro "when used in the reverse position". None of the pages explains how the figure was checked in each position. So the table can show only what each maker claims.

The digital level is the loosest figure here. Johnson's 1760-4800 is rated "±0.1° for 0°and 90°, ±0.2° for other angles". tan(0.1°) = 0.00175, so 1.75 mm/m, three and a half times the bubble class. Its display also shows mm/m, but at a resolution of 1 mm/m. A number on a screen reads as precise. Its published error is still 0.1 degree.

Over 4 ft and over a 20 ft run

Take a job where the figure matters. Say you are hanging a run of wall cabinets on a 20 ft ledger, with a 48 in level rated 0.5 mm/m. Here is how the published figure plays out.

One placement. The level spans 48 in. At 0.0005 in/in, one end of the ledger under it can be up to 48 x 0.0005 = 0.024 in high or low while the bubble reads centered. That is 0.61 mm, a bit over 1/64 in and under 1/32 in.

Stepping along the run. A 48 in level covers 20 ft in five placements. The vial's error is a fixed tilt of the vial against the bar. Move the level along without turning it, and the same tilt repeats in the same direction each time. In the worst case the five add up:

5 x 0.024 in = 0.12 in (3.05 mm)

That is the same as the slope times the whole run, 240 in x 0.0005 = 0.12 in. Stepping does not change the slope. It also adds what the spec does not cover: how each placement is seated and how each bubble is read.

Against a published tolerance. M S International's guidelines for its Q quartz say: "The tops of the cabinets must be level within 1/8" over a 118" length." Over 118 in, a 0.5 mm/m level can be off by 118 x 0.0005 = 0.059 in. That uses 47 percent of the 1/8 in before any shim goes in. A 0.75 mm/m overhead figure would use 71 percent.

If the job sets 1/8 in end to end over the whole 20 ft, the level's own 0.12 in uses 96 percent of it. There is nothing left for the shims. A 0.5 mm/m level reaches 1/8 in at 0.125 / 0.0005 = 250 in, about 20 ft 10 in. Used overhead at 0.75 mm/m, it reaches 1/8 in at 167 in, about 13 ft 11 in.

Against a laser. A laser rated 1/8 in at 30 ft is 0.35 mm/m, as worked out in what 1/8 in at 30 ft means on a laser level. Over the same 20 ft that is 0.083 in, against the level's 0.12 in. The laser also puts one line on the whole wall, with no stepping. On a long run, the 1/8 in class laser is the tighter tool by its published figures. Over a single 48 in span, the two are close: 0.017 in for the laser and 0.024 in for the level.

Once the ledger is set, what screws to use for garage wall cabinets covers what holds it to the studs.

The overhead figure only Stabila prints

"Overhead" means the level is held up against a surface above you, such as the underside of a shelf or a joist, and read from below. The vial is then upside down compared with normal use.

Stabila's listing for the 96-2 gives the normal position as "0.029° = 0.5 mm/m" and "Overhead Measurement 0.043° = 0.75mm/m". The overhead figure is half as large again.

Over a 48 in level, that is 0.036 in instead of 0.024 in. Over 20 ft, 0.18 in instead of 0.12 in, just under 3/16 in. If your level's maker gives only one figure, the table above is what it claims. If it gives two, use the overhead one for any reading taken from below.

Checking a level by turning it end for end

A level cannot check itself in one position, because its error is in every reading. Turn it end for end on the same spot and the error changes sign.

Here is the arithmetic. Say the vial reads centered when the bar is tilted by e. Set the level on a surface and shim it until the bubble centers. The surface now sits at e. Lift the level, turn it end for end, and set it on the same marks. The vial now wants the surface tilted e the other way, so the bubble reads as if the surface were off by 2e.

So a level that is exactly at its rating can show twice its rating in this check. For a 0.5 mm/m level, that is 1.0 mm/m. On a 48 in level, the shim at one end to center the bubble again would be:

2 x 0.024 in = 0.048 in (1.2 mm), about 3/64 in.

That figure is our derivation from the published rating, not a check procedure printed by any of these makers. Halve whatever gap you find before you compare it with the spec. True level sits halfway between the two readings.

The same doubling shows up in a laser's two-way field check, for the same reason. The laser article walks through it with DeWalt's and Milwaukee's own tables.

Turning the level also helps on a long run. If the vial tilt repeats each time you step the level forward, turning it end for end on every other placement flips the sign of that tilt. The repeated error then tends to cancel instead of adding.

Machinist levels: mm/m is a division, not an error

Here is where the third reading comes from. A search for spirit level accuracy turns up precision levels sold with figures like 0.02 mm/m or 0.05 mm/m. Beside a builder's 0.5 mm/m, those look 10 to 25 times better. They are a different quantity.

A machinist level, or precision level, has a vial ground to a curve and marked with a scale of short lines, called divisions. Its figure says how far the bar must tilt to move the bubble by one division. That is its sensitivity. It is not the most the level can be wrong.

Hoffmann Group's data sheet for its shaft level built to the German standard DIN 877 lists three classes:

The sheet lists the class under "Accuracy" and then gives "Scale divisions 0.02 mm/m" for the class 1 level. So the number on the label is the division. The sheet gives no separate maximum error figure.

The last column shows the gap. A builder's level's whole allowed error, 0.5 mm/m, would push the bubble of a class 1 vial 25 divisions. On Starrett's 199Z, at 0.042 mm/m a division, it is 12 divisions. The 199Z has "seven graduations on each side of the bubble", so that tilt runs off its scale.

The US version of the trap. Starrett states its 199Z as "one division equals 1/2 thousandth (0.0005) of an inch per foot, or 0.04mm per meter." Its 98-12 machinists level reads ".005" per foot" and "0.42mm per meter". Set beside Empire's 0.0005 in per inch:

0.0005 in per ft / 12 in per ft = 0.0000417 in/in = 0.042 mm/m

The same digits, 0.0005, sit on both levels. One is per foot and one is per inch, so they are 12 times apart. And they measure different things: one is a division, the other a maximum error. The Starrett 98 at 0.42 mm/m per division sits close to Hoffmann's class 4 at 0.4 mm/m.

So the rule is to read the unit and the noun. "Accuracy 0.5 mm/m" on a box level is the most it can be off. "0.4 mm/m" per division on a machinist level is how far the bubble moves. Neither number converts into the other.

Which figure a job needs

1. Cabinets and countertops up to about 10 ft. A 0.5 mm/m level uses about half of MSI's 1/8 in over 118 in. Check the level end for end first, and read the bubble the same way each time.
2. Runs over about 20 ft held to 1/8 in end to end. The level's own error nearly fills the tolerance. Use a laser in the 1/8 in at 30 ft class, or turn the level on alternate placements.
3. Reading from below. Use the maker's overhead figure if it prints one. For the Stabila 96-2 that is 0.75 mm/m.
4. Machine beds and setup. That is a machinist level's job. Its divisions show how much a bed is out. None of the box level pages above lists a vial division at all.

For anything else, weigh 0.006 in per foot, or 0.024 in per 48 in level, against the tolerance the job sets.

A level is a calibrated instrument. Like the click wrench in the torque wrench storage guide, it holds its number only inside its limits.

Where this answer stops applying

• A level that has been dropped or bent. Every figure here is the maker's rating for a level in its original state. Milwaukee guarantees its accuracy for life, but a guarantee is not a reading. Run the end-for-end check after any drop.
• Long levels without milled edges. Stabila's listing says the 96-2 has milled measuring surfaces only up to 120 cm, and Kapro mills the 985XL only up to 80 cm. None of these pages gives a flatness figure for the bar itself, so this page cannot say how much a bowed bar adds.
• Plumb vials. The figures here are as each maker prints them. Stabila's names a normal and an overhead position. The pages do not give a separate plumb figure, so none is worked out here.
• Digital levels at other angles. Johnson's 0.1 degree applies at 0 and 90 degrees. At other angles its figure is 0.2 degree, which is 3.49 mm/m.
• DIN 877 itself. The class figures come from Hoffmann's data sheet for one level built to the standard. This page does not quote the standard's text.