Home · Articles on balancing and vibration
Balance quality grades

How to Choose a Balance Quality Grade: G6.3, G2.5, and Everything Else

The work order says "balance to G2.5," and not another word after that: no rotor mass, no weight-mounting radius, no speed. Without these numbers, a grade doesn't turn into grams or into an acceptance criterion. Let's break down what's behind the letter G, how a grade turns into a g·mm tolerance, and why tightening the grade "just in case" costs more than it looks like.

Updated 27 August 2026 · by AXILINE · Vila Nova de Gaia

In short: Grade G per ISO 21940-11 (formerly ISO 1940-1) sets the allowable residual unbalance of the rotor itself, not the vibration level at the bearing housing. The number in the grade designation equals the product of the allowable specific residual unbalance and the angular velocity, expressed in mm/s: for G6.3, that product equals 6.3 mm/s. For most on-site work on fans, pumps, impellers, and general-purpose electric motors, the baseline choice is G6.3; G2.5 is used for high-speed machines, turbines, and machine-tool drives; G1.0 and G0.4 are reserved for precision spindles. At the same grade, the allowable mass falls in inverse proportion to speed: spin twice as fast, and half as many grams are allowed.

What Grade G Is, and What It Doesn't Regulate

The balance quality grade G is set by ISO 21940-11 (before revision, this part was called ISO 1940-1). The number in the grade designation equals the product of the allowable specific residual unbalance and the angular velocity, and it has units of mm/s. The notation G6.3 reads like this: e_per (the allowable specific residual unbalance) times ω (the angular velocity of rotation) equals 6.3 mm/s.

Specific residual unbalance is the total unbalance divided by the rotor's mass, in g·mm/kg. This quantity has an intuitive meaning: it's numerically equal to the displacement of the rotor's center of mass from the rotation axis, in micrometers. 40 g·mm/kg is 40 µm of eccentricity — four hundredths of a millimeter. A sheet of paper is about twice that thick.

Now, about the main source of confusion. The mm/s unit in the grade designation has nothing to do with the vibration velocity you read off a sensor at the bearing housing. G6.3 doesn't mean 6.3 mm/s on the housing, and zone A on the overall vibration-level scale doesn't confirm a G6.3 grade. These are two independent criteria with different units and different control points. We covered them together with a third one — the target residual 1x (vibration at the rotor's rotation frequency) in the instrument's software — in a separate article on the three tolerances in balancing.

The series of grades steps up by roughly a factor of 2.5 each time: G0.4, G1, G2.5, G6.3, G16, G40, G100, G250, G630, G1600, G4000. The upper grades address slow-turning, rigidly mounted systems; the lower ones address precision rotors. On-site work in practice lives almost entirely within the G16…G1 range.

Sources: ISO 21940-11:2016

Why a Grade Has Units of Speed

The standard's logic is simple, and it explains why you can't set a single eccentricity limit for every machine. The centrifugal force from an unbalanced mass grows in proportion to the square of the angular velocity. A rotor with its center of mass offset by 40 µm at 1500 rpm loads the bearings four times less than the same rotor at 3000 rpm. The same residual unbalance is harmless on a slow machine and destructive on a fast one.

That's why the standard regulates not the eccentricity itself, but the product of e_per and ω. Physically, this is the linear velocity of the center-of-mass point as it orbits the axis at a radius of e_per. Hence the mm/s in the grade designation.

You'll feel the practical consequence the moment you start calculating grams. One grade is one curve, not one number. Double the speed, and the allowable mass drops by half. A rotor that meets G6.3 at 750 rpm with 32 grams of residual must fit within 8 grams at 3000 rpm.

Speed, rpmω, rad/sAllowable eccentricity, µmTolerance U_per, g·mmSame, in grams at a 250 mm radius
75078808,00032
1000105606,00024
1500157404,00016
3000314202,0008
6000628101,0004

The table is calculated for grade G6.3 and a 100 kg rotor, with numbers rounded for convenience. It shows the relationship, not a substitute for the calculation: check the applicable part and edition of the standard, and the method for splitting the tolerance between planes, for your specific machine.

Sources: ISO 21940-11:2016

How to Get From a Grade to Grams: The Calculation Logic

For a grade to become an acceptance criterion, you need four numbers. Without any one of them, talk of G2.5 stays just talk.

  1. 01

    Gather the input data

    Rotor mass M in kilograms (specifically the rotating part, not the whole machine), running speed n in rpm, grade G, and the actual weight-mounting radius r in millimeters for each correction plane (a rotor cross-section where correction weights are mounted). Take the mass from the nameplate or weigh it; measure the radius with a tape on site, don't estimate it by eye.

  2. 02

    Convert speed to angular velocity

    ω = 2π·n/60 rad/s. At 1500 rpm that's about 157 rad/s, at 3000 about 314, at 1450 about 152. Easy to remember: divide the speed by 9.55.

  3. 03

    Calculate the specific tolerance

    e_per = G/ω. For G6.3 at 157 rad/s that works out to 0.040 mm, or 40 µm, or 40 g·mm/kg. That's the allowable displacement of the rotor's center of mass from the rotation axis after balancing.

  4. 04

    Move to the full tolerance

    U_per = e_per·M, or in a convenient form, U_per = 1000·G·M/ω in g·mm. For a 100 kg rotor at 1500 rpm and G6.3, that comes out to about 4,000 g·mm for the whole rotor.

  5. 05

    Convert to grams at your own radius

    m = U_per/r. At a 250 mm radius that's 16 g; at a 500 mm radius, just 8 g. The farther from the axis you mount the weights, the fewer grams correspond to the same tolerance. That's exactly why you can't just guess at the radius.

  6. 06

    Split the tolerance between planes

    For a symmetric between-bearings rotor, the total tolerance is usually split in half: 2,000 g·mm and about 8 g per plane. For an overhung or asymmetric rotor, the split is calculated based on the position of the center of mass and the distances to the bearings, and you can no longer just divide by two.

The numbers in the examples are rounded so the logic is easy to follow. The formulas apply to a rotor in the rigid state. Take the exact calculation, the grade-to-machine-type mapping, and the rule for splitting the tolerance between planes from the applicable edition of ISO 21940-11 and from the documentation for your specific rotor.

Sources: ISO 21940-11:2016

Guidelines: Which Grade for Which Machine

The standard gives recommendations by machine type, and on site they add up to a short list. The numbers in the last two columns are calculated for the same 100 kg rotor at 1500 rpm with a 250 mm correction radius, so you can see the cost of tightening the grade, in grams.

Why G6.3 is the working reference point

A fan, a pump, a blower, and a standard electric motor at 1000…3000 rpm all land right here. A tolerance of 16 grams at a 250 mm radius is realistically achievable with two trial runs and one trim step, and the machine's construction doesn't call for more than that. If the documentation doesn't name a grade, this is a reasonable starting assumption to agree on with the customer before the site visit.

When G16 is enough

Slow-turning, coarse drives with clearances, a belt drive, and a flexible frame. Tightening the grade there makes no sense: residual unbalance stops being the main source of vibration well before you reach the G6.3 boundary.

When you need G2.5 or tighter

High speed, bearings with tight clearance, requirements on machining quality or service life. A grinding-machine spindle feels the difference between G2.5 and G1.0 in the workpiece's surface quality. A roof fan won't notice it at all.

GradeTypical applicationTolerance U_per, g·mmGrams at a 250 mm radius
G16Driveshafts, crushers, mulchers, screw conveyors, agricultural machinery, coarse drives with low requirements10,20041
G6.3Fans, induced-draft fans, pump impellers, blowers, pulleys, drums, general-purpose electric motors. The baseline choice for most on-site work4,00016
G2.5Mid-range turbines and compressors, machine-tool drives, high-speed pumps, special-purpose electric motors1,6006.4
G1Machine-tool spindles, grinding wheels, high-precision drives6402.6
G0.4Precision grinding spindles, gyroscopes, special rotors2551.0

This table is a working guideline, not an excerpt from the standard. Before contractual acceptance, open the applicable edition of ISO 21940-11 and check the grade for your rotor type: the standard breaks things down in more detail, and separate recommendations exist for some machine types.

Sources: ISO 21940-11:2016

Five Questions That Decide the Grade Choice

The fifth question rules out more unrealistic requirements than the first four combined. Let's look at it separately, because this is exactly where the idea of "let's just ask for a tighter grade, just in case" falls apart.

The number of correction planes has nothing to do with grade selection: that's a separate decision based on the L/D ratio (rotor length to diameter) and speed, and we have our own article on it. But the number of planes does directly affect how the tolerance is split.

Why You Shouldn't Ask for G1.0 "Just in Case"

The honest way to work is this. Set the grade that matches the machine — usually G6.3 for fans, pumps, and general-purpose electric motors. Reach it, look at the spectrum (the breakdown of vibration by frequency), and compare overall vibration against the 1x running-speed component. If 1x still dominates and that's not enough for the customer, tighten the grade deliberately: by then you already know how many runs that costs and whether your weight resolution can handle it. But if something other than 1x dominates the overall vibration, tightening the grade won't achieve anything at all, and that's where vibration diagnostics starts, not balancing. We've covered the cases where balancing doesn't help at all in a separate article.

Discrete correction eats the tolerance

Take a fan with six fixed mounting positions, 250 mm radius. A 16 g weight is exactly the G6.3 tolerance from the table above. Move that weight over by one position — 60° — and the unbalance vector changes by 2·16·sin30° = 16 g, meaning the full 4,000 g·mm. The position grid itself introduces an error the size of the entire G6.3 tolerance. For G1.0 the tolerance is 640 g·mm, and there's simply nothing on fixed positions fine enough to hit it: you'd need a free rim, fine mass tuning, and drilling.

Every trim step means a machine stoppage

The instrument calculates an add-on weight in seconds, but you spend an hour: stop the unit, wait for it to coast down, open the guard, reach the correction plane, weigh and mount the weight, close up, spin it back up, take the reading. Going from G6.3 to G1.0 usually means not one cycle like this but three or four, all on a live production line.

Measurement error becomes comparable to the tolerance

The instrument doesn't measure the achieved residual unbalance directly. It calculates it from the residual running-speed component and the influence coefficients (the machine's response to a trial weight) obtained during trial runs. When residual 1x drops down to the level of noise and run-to-run scatter, the result stops being repeatable. You're no longer chasing unbalance — you're chasing your own measurement error.

The other sources of vibration haven't gone anywhere

Shaft misalignment, loose fasteners and soft foot, bearing wear, fit runout, aerodynamic unevenness at the wheel outlet, closeness to resonance. Not one of these causes cares whether you reached G2.5 or G1.0. On a real machine, whatever you gain from over-tightening the grade drowns completely in these other causes.

How a Grade Turns Into a Number on the Instrument's Screen

In the Balanset-1A software, the tolerance on residual unbalance lives in the "Balancing tolerance" field and is expressed in g·mm. You can type it in by hand if the number has already been calculated for you, or open a separate calculation window and get the tolerance using the ISO 1940-1 method — that is, the currently applicable ISO 21940-11. You set the grade and the input data, and the software outputs g·mm.

The second field the whole calculation depends on is "Mass mount radius" for each plane. The software uses the radius to convert the initial and residual unbalance from the measured vibration and weights, and to check whether it's within tolerance. Get the radius wrong by 20%, and the tolerance in grams drifts by that same 20%.

After the verification run, the software shows the achieved residual unbalance right next to the tolerance you set. That pair of numbers is what goes into the report from the archive: the calculated U_per and the actual residual. The archive separately stores a second, independent tolerance in mm/s of vibration. Don't confuse them: the first field is about the rotor; the second is about how the machine vibrates.

Sources: ISO 21940-11:2016 · Balanset-1A operation manual

What to Put in the Contract and the Report

The phrase "balance to G2.5" isn't a technical requirement. It can't be verified and can't be disputed. For a grade to become an acceptance criterion, the contract needs all the numbers that turn it into grams, plus a method of verification.

There's no direct conversion from grade G to mm/s at the bearing housing. The same residual unbalance on a massive foundation and on a flexible steel frame will produce different housing vibration, because that vibration also depends on the mass and stiffness of the supports, damping, and closeness to resonance. That's why the two criteria are checked separately and both go into the report.

Sources: ISO 21940-11:2016 · ISO 20816-1:2016

How AXILINE Can Help

Choosing the grade is half the job, and usually the most contentious half. We do it together with the customer before the site visit: we look at the machine type, speed, rotor mass, and where a weight can physically be mounted, calculate the tolerance in g·mm and in grams, and say plainly whether it's achievable on this machine or not. If a requirement from the spec is unrealistic given the correction resolution, you find out in advance, not at the end of the shift.

From there we come to the machine, mount sensors on the bearing housings, separate the running-speed component from the overall level, and look at the spectrum. If unbalance really does dominate, we balance the rotor in its own bearing housings, in one or two planes, and hand over a report where the calculated tolerance and the achieved residual unbalance sit side by side, with the overall-vibration criterion on its own line. If unbalance isn't the main cause, you get vibration diagnostics telling you what to fix before balancing.

The work is carried out by engineers who design and manufacture Balanset instruments and do their own on-site balancing with them. If you want to calculate grade-G tolerances and keep your own reports, the Balanset-1A can be bought: two vibration sensors, a laser phase sensor, a two-channel USB module, and Windows software with tolerance calculation, an archive, and a polar diagram. Consulting support stays with us either way.

Tell us the machine type, speed, rotor mass, and the grade you're being asked for. That's usually enough for us to tell you how many grams of residual you're allowed and whether hitting that is realistic.

Sources: Balanset-1A manufacturer specification

Frequently asked questions

G6.3 or G2.5 for a fan at 1500 rpm?

For a general-purpose fan impeller, the baseline choice is G6.3. At a wheel mass of 100 kg and a weight-mounting radius of 250 mm, that gives about 4,000 g·mm for the whole rotor — roughly 16 g of residual unbalanced mass — split in half between planes for a symmetric between-bearings rotor. G2.5 on the same rotor tightens the tolerance to 1,600 g·mm, about 6.4 g, and will require extra trim steps. It's worth going with G2.5 if it's stated in the machine's documentation, if the speed is above 3000, or if the fan sits in a vibration-sensitive system.

The machine's documentation doesn't state a balance grade. What should you use?

Base it on the component type and speed, and confirm the choice with the customer in writing before work starts. For fans, induced-draft fans, pumps, blowers, pulleys, and general-purpose electric motors, use G6.3. For slow-turning coarse drives, crushers, and agricultural equipment, G16 is enough. For machine-tool drives and high-speed compressors, lean toward G2.5. Then check that the calculated tolerance in grams is larger than your weight resolution: the minimum washer mass and the spacing of fixed positions.

Why, at the same grade, is the tolerance at 3000 rpm half what it is at 1500?

Because the grade regulates the product of the allowable eccentricity and the angular velocity, not the eccentricity itself. Double the speed, and ω doubles too, so e_per = G/ω is cut in half. The physics behind it: centrifugal force grows as the square of speed, so a fast machine gets four times the bearing load from the same residual unbalance. If the machine runs off a VFD, calculate the tolerance at the maximum running speed and state that in the report.

How does the weight-mounting radius affect the tolerance?

The tolerance in g·mm doesn't depend on the radius, but the tolerance in grams depends on it directly and inversely. A 4,000 g·mm tolerance at a 250 mm radius is 16 g; at a 500 mm radius, just 8 g; at a 100 mm radius, a full 40 g. Two practical conclusions follow. First: the radius has to be measured on site and the actual value entered, or the whole tolerance and residual-unbalance calculation drifts off. Second: the farther from the axis you can mount the weight, the less mass it takes to cover the same unbalance, and the more precise the correction turns out.

The customer is demanding G1.0 for an ordinary fan. What do you tell them?

Show them the numbers. At a wheel mass of 100 kg, 1500 rpm, and a 250 mm radius, G1.0 gives 640 g·mm — that's 2.6 g of residual. On a wheel with six fixed positions, moving a weight over by one position changes the unbalance vector by about 4,000 g·mm, six times the entire tolerance. That means the requirement is only achievable on a free rim with fine mass tuning, over several extra machine stoppages, and the achieved result will be comparable to the measurement error. Add that housing vibration will barely change, because that also involves misalignment, fasteners, and bearings. Usually, after a conversation like this, the requirement turns into G6.3 plus a separate overall-vibration criterion.

Can grade G be used to predict how many mm/s will show up at the bearing housing?

No, there's no direct conversion. Grade G describes the rotor's condition, while housing vibration also depends on the mass and stiffness of the supports, the foundation type, fastener condition, damping, and how close the running speed is to resonance. The same residual unbalance on a concrete foundation and on a flexible steel frame will produce different mm/s. That's why a contract states two independent criteria: residual unbalance in g·mm per grade G, and overall vibration velocity in mm/s RMS in the 10–1000 Hz band per the applicable part of ISO 20816, with measurement points and machine condition specified.

Related content

How to Choose a Balancing Instrument: Vibrometer, Analyzer or Balancer

Name the task first, then choose the instrument class. For a route walk and trending, a single-number vibrometer is enough. To understand the cause of vibration, you need an analyzer with a spectrum and a time waveform. To balance, you need an instrument with a phase sensor and a correction calculation by influence coefficients, and for two planes, two vibration channels with simultaneous acquisition as well. You can't balance with a vibrometer: without phase, the instrument doesn't know at what angle to place the weight.

Open page

How to choose a balancing and vibration diagnostics contractor: questions, red flags, contract

Ask three things: what you measure with and how many channels, whether you measure phase and speed, and what you'll do if the cause isn't unbalance. A specialist will answer with a two-channel analyzer with a phase sensor keyed to a mark on the shaft, comparing overall vibration against the 1x running-speed component — the part of the vibration that repeats exactly once per shaft revolution — and that they won't balance if the 1x share is small. Someone with a plain vibrometer promises a result before the measurement and won't show you before-and-after data. A method's limits, stated out loud, are more reliable than any 'we'll fix everything' guarantee.

Open page

Conveyor Pulley Balancing: Drive, Take-Up, and Bend Pulleys

Yes, we balance conveyor pulleys on site, in their own bearing units, with weights on the end discs in two planes. An honest caveat: below roughly 120 rpm the forces from imbalance are negligible, and the running frequency drops toward the lower measurement limit. So we first check belt tension and tracking, lagging wear, buildup, the motor-gearbox unit, and the bearing units. Balancing is worthwhile on pulleys from ~120 rpm up, after a shell repair, when a factory weight has been lost, or when there's water or buildup inside the body.

Open page

Describe your equipment and the problem

We'll answer your questions, clarify the details, and let you know what's needed for an estimate and a visit.

Submit a Request WhatsApp Pricing