Three tolerances in balancing: what “within norm” actually means

The software says “in tolerance,” but the customer's reliability department says the machine is vibrating above the ISO norm. Both statements can be true at the same time, because they're talking about different quantities. Here are the three tolerances that get called by one word, and what to write in the report.

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

Three tolerances that get called by one word

A fan starts humming after the impeller is cleaned. You arrive, fit sensors to the bearing housings, balance in two planes, and the software reports “in tolerance.” A week later an email arrives: vibration is above the ISO norm, redo the job. Nobody is lying. You were simply talking about different tolerances.

These three quantities have different units, different measurement points, and different standards. Mixing them into one sentence means arguing about nothing.

One row of this table doesn't stand in for the other two. “In tolerance” on 1x doesn't mean zone A, and zone A doesn't mean grade G 6.3.

Tolerance #1: target residual 1x in the instrument's software

When you add a new rotor to the Balanset-1A archive, the software asks for two tolerances: one for vibration, one for residual imbalance. The first field is tolerance #1. It's your own target value for the residual running-speed component, in mm/s.

The instrument compares only 1x against this number. Not overall vibration, not the spectrum, not the harmonics. Just the amplitude at the rotation frequency — exactly where imbalance lives.

The direct consequence follows from that. You set the number. Set it to 4.5 mm/s, and the software will report “in tolerance” at 4.4. Set it to 1.8, and you'll end up doing a trim — adding small weights to the ones already fitted. Set a value tied to an actual requirement for the machine, not one that's just convenient to reach in two runs.

A practical guide: take tolerance #1 from the zone you've committed to bringing the machine into under ISO 20816, and leave a margin. If the target for overall vibration is 2.8 mm/s, a target 1x of around 1.5 mm/s leaves room for the other sources.

“In tolerance” in the software means exactly one thing: the residual 1x is below the number you entered. It doesn't confirm an ISO 20816 zone or a grade G under ISO 21940-11.

Tolerance #2: overall vibration under ISO 20816

The customer and the chief mechanic care about something else: how the machine vibrates as a whole. That's what ISO 20816 (formerly ISO 10816) is for. The standard describes assessing broadband vibration (across the whole frequency band at once) on the non-rotating parts — meaning bearing housings and supports, not the shaft.

Three things in this kind of assessment are fixed, and you can't change them to suit yourself.

• Parameter: vibration velocity, root-mean-square (RMS) value, in mm/s.
• Frequency band: 10-1000 Hz for general-purpose machines. Use a narrower or wider band, and the numbers stop being comparable.
• Point and direction: the bearing housing or bearing support, sensor on a rigid mount (a stud or a magnet on a clean surface), radial direction, kept the same from reading to reading.
• Mode: steady-state operating speed and operating load.

The numbers in the table are given in mm/s RMS for the 10-1000 Hz band and come from the classification previously used in ISO 10816-3. Use them as a working guide. Before contractual acceptance, open the applicable part of ISO 20816 in its current edition and check the scope: ISO 20816-3 addresses industrial machines above 15 kW at 120-30,000 rpm and contains exceptions by machine type. The standard itself divides foundation types more carefully than the simplified table does. For small fans, screw conveyors, mulchers, and belt-driven units, check applicability separately.

Sources: ISO 20816-1:2016

Rigid or flexible foundation: the row that gets mixed up most often

The last two rows of the table differ by one word, and the figures come out roughly double. This is exactly where online sources get it wrong, constantly.

The logic runs like this. A flexible (compliant) foundation deforms at the rotation frequency and absorbs part of the vibration energy. The bearing housing moves more noticeably as a result, but the dynamic load on the bearing itself comes out lower. A rigid foundation behaves the opposite way: the housing barely moves, but the full force goes into the bearing and the foundation. That's why the norm for a flexible foundation is higher, not lower.

• Formal indicator: compare the rotation frequency with the system's first natural frequency in the measurement direction. Above the operating speed means a rigid foundation, below means flexible.
• On-site check: measure vibration on the bearing housing and, alongside it, on the frame or foundation. If the frame vibrates comparably to the housing, the foundation is behaving as compliant.
• Separately, check for “soft foot” and anchor bolt torque. Loose fasteners turn a formally rigid foundation into a compliant one, and the tolerance doesn't get more lenient because of it: that's a defect, not a design choice.

If a table you've found shows a stricter norm for a flexible foundation than for a rigid one, the table is wrong. Don't sign off a report against it, and don't put it into a technical specification.

Tolerance #3: residual imbalance and grades G under ISO 21940-11

The third tolerance isn't about the vibration of the housing. It's about the rotor itself: how much unbalanced mass is left in it after the work. Units: g·mm for the whole rotor, and g·mm/kg in specific form — that is, per kilogram of rotor mass.

This tolerance is set by the balance quality grade G under ISO 21940-11 (formerly ISO 1940-1). The number in the grade designation equals the product of the allowable specific unbalance and the angular velocity, expressed in mm/s. From that come the working formulas for a rotor in the rigid state: specific tolerance e_per = G / ω, total tolerance U_per = 1000 · G · M / ω in g·mm, where M is in kilograms and ω is in rad/s.

Let's work through a typical example. A fan impeller with a mass of 80 kg, 1450 rpm, grade G 6.3, weight-mounting radius 300 mm. Angular velocity ω = 2π · 1450 / 60 ≈ 152 rad/s. Total tolerance U_per = 1000 · 6.3 · 80 / 152 ≈ 3300 g·mm. For a rotor symmetrically mounted between bearings, that's split in half: around 1650 g·mm per plane. At a 300 mm radius, that's roughly 5.5 g of residual unbalanced mass in each plane.

The Balanset-1A calculates this field itself. You enter the rotor mass, speed, correction radius, and grade G, and the software outputs the tolerance in g·mm; after the verification run, it shows the residual imbalance actually achieved. That number is the third line in the report.

• G 16: driveshafts, crushers, agricultural machinery, drives with modest requirements.
• G 6.3: fans, pumps, impellers, wheels, screw conveyors, pulleys, standard electric motors. The most common grade in field work.
• G 2.5: mid-range turbines and compressors, machine-tool drives, special-purpose electric motors.
• G 1: machine-tool spindles, grinding wheels, high-precision drives.
• G 0.4: precision spindles and grinding shafts.

Take the grade-to-machine-type mapping and the split of tolerance between planes from the applicable edition of ISO 21940-11 and from the rotor's documentation. The formulas above work for a rotor in the rigid state. If the operating speed is close to or above the first critical frequency (the speed at which the shaft itself goes into resonance), the rotor behaves as flexible, and a one- or two-plane scheme no longer describes its behavior. That case is assessed under different documents.

Sources: ISO 21940-11:2016

How to close out all three tolerances in one visit

1. 01

   Take a baseline reading

   Before fitting any weights, measure the overall vibration and 1x at both bearing housings, in the radial and axial directions. Record the speed and the operating mode. Save the spectrum. This is your starting point under ISO 20816, and at the same time a check on whether balancing is even worth doing.

2. 02

   Test the imbalance hypothesis

   Compare 1x with the overall level. Imbalance gives a steady, dominant peak at the rotation frequency and a repeating phase. A strong 2x more often points to shaft misalignment or looseness, and that's fixed with shaft alignment and fasteners. Lots of harmonics and a raised noise floor mean looseness, rubbing, cracks. Peaks at high frequencies not related to running speed are usually about bearings.

3. 03

   Set both tolerances in the archive

   Enter the target residual 1x in mm/s and the tolerance on residual imbalance in g·mm. Calculate the second one from grade G, the rotor mass, and the speed. Both numbers will go into the report together with the result.

4. 04

   Balance

   Run 0, trial weight in plane 1, and — for two planes — a trial weight in plane 2, then correction, then a verification run. The trial weight has to change the 1x amplitude by 20-30% or the phase by 20-30°, or the influence coefficient comes out unreliable. The instrument will tell you whether the weight was adequate.

5. 05

   Check the result against all three tolerances separately

   Residual 1x against the target value. Overall vibration against the ISO 20816 zones, at the same points and in the same direction as before the work. Residual imbalance in g·mm against the calculated U_per. Three lines, three conclusions.

When you won't reach zone A or B

Balancing only reduces the running-speed component. If its contribution to the overall vibration is small, the overall figure will barely move no matter how many weights you fit. Here are the cases where the method doesn't work, and what to do instead.

• Overall 9 mm/s, but 1x only 2 mm/s. Balancing will only remove part of that. The cause is elsewhere: shaft misalignment, loose fasteners, bearings, rubbing. What's needed is vibration diagnostics and a mechanical repair.
• Resonance. The operating speed is close to the natural frequency of the rotor, the supports, or the frame. Amplitude jumps around, the 1x phase drifts, the result doesn't hold between runs. Address stiffness and fasteners first, and change the operating speed or rework the supports if needed.
• An aerodynamic or electromagnetic component. This kind of force grows in proportion to the angular velocity, while the centrifugal force from a correction weight grows with its square. You can only compensate it at the specific speed used for balancing; at other speeds a residual will appear. Narrow fan impellers behave this way regularly.
• Geometry and runout. An out-of-round shaft journal, runout in a fit, or a bent pulley all produce vibration exactly at the rotation frequency, but a weight won't remove it.
• A worn fit or play. The rotor moves around in the support, the influence coefficient changes from run to run, and the result can't be reproduced. Repair the fit first.

When imbalance really does dominate, the picture is the opposite. A typical situation: a fan with an overall of around 12 mm/s, almost all of it sitting in 1x. After two-plane balancing, it drops to around 1.5-2 mm/s — out of zone D and into zone A or B. Treat this as a sense of scale, not a promise: the result depends on exactly what's making noise in your machine.

The report, and where AXILINE can help

We close out all three tolerances on the visit. AXILINE comes to the machine, takes vibration readings at the bearing housings, separates 1x from the overall level, shows you the spectrum, and tells you plainly whether balancing will help or not. If it will, we balance the rotor in its own bearings, in one or two planes, and hand over a report with all three lines. If it won't, you get a diagnosis with what needs fixing before balancing.

The people doing the work are engineers who design and manufacture the Balanset instruments and use them ourselves on site visits. If you'd rather calculate grade-G tolerances yourself, the Balanset-1A is available to buy: two vibration sensors, a laser phase sensor, a two-channel USB module, and Windows software that keeps an archive and prints reports. Consulting support stays with us either way.

Tell us which machine is giving you trouble, what speed it runs at, and what figures you're already seeing. That's usually enough for us to tell you over the phone which of the three tolerances you're missing.

• The part and edition of the standard used for acceptance (for example, ISO 20816-3 with the year stated).
• Measurement points and directions, with a sketch or photo of the sensor placement.
• Parameter and band: mm/s RMS, 10-1000 Hz.
• The machine's operating mode during the reading: speed, load, temperature.
• Foundation type: rigid or flexible, with the reasoning for the table row chosen.
• Grade G, rotor mass, correction radius, the calculated U_per, and the residual imbalance actually achieved, in g·mm.
• Target and achieved residual 1x, in mm/s.
• Before-and-after values, so the trend is visible, not just the final figure.

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

Frequently asked questions

Related content