How to Read a Spectrum Back to a Mechanical Cause

Why this matters

A spectrum peak is not a mystery, it is an arithmetic result. Most of what a machine radiates lands at frequencies you can compute in advance from shaft speed and a count: blades, teeth, poles, lobes, belt length. Bearings are the exception, and the worksheet below says so where it reaches them. Compute the candidates before you look at the data and the peaks name themselves. Skip that and you are staring at a picture, guessing, and eventually replacing the part that was easiest to order. This card is the arithmetic half. Once a peak has a name, the library's noise and vibration diagnosis family takes over on what the fault actually is and whether it is worth condemning the machine.

The worksheet

Fill this out before the meter comes off the truck. Each field is a frequency in Hz, and each one carries the condition it was derived under.

Field 1: shaft speed at the operating point, in Hz. Rev/min divided by 60. Measure it, do not take it off the nameplate. An induction motor's nameplate speed is its speed at RATED load, and slip rises with load, so a lightly loaded motor runs faster than nameplate, not slower. Use a non-contact optical tachometer with reflective tape applied to the shaft only after the drive is isolated, locked out and verified stopped under 29 CFR 1910.147, which covers the mechanical and stored-energy isolation this task needs.

Field 2: driven-shaft speed, in Hz. For a belt drive, driver rev/min multiplied by the driver sheave pitch diameter divided by the driven sheave pitch diameter, then divided by 60. For a direct drive, field 2 equals field 1.

Field 3: blade or vane pass, in Hz. Driven-shaft Hz multiplied by the number of blades or vanes. This holds the blade count constant, so a blade-pass peak moves in exact proportion to shaft speed and nothing else.

Field 4: twice line frequency, in Hz. 120 Hz on a 60 Hz supply, 100 Hz on 50 Hz. Magnetic forces in motors and transformers act twice per electrical cycle. This field is fixed by the supply and is the only entry on the sheet that does NOT move when you change speed, which is what makes it decisive.

Field 5: gear mesh, in Hz. Shaft Hz for the gear you are counting, multiplied by that gear's tooth count. Compute it on the shaft the teeth belong to, not the output shaft, or the answer is off by the ratio.

Field 6: belt passage, in Hz. Belt linear speed divided by belt length. Belt speed is pi multiplied by the sheave pitch diameter multiplied by revolutions per second on that sheave. Belt-related peaks appear at that frequency and its low multiples. Never measure a belt by touching it; take the sheave diameters and the belt length from the drive with the guard on and the machine locked out.

Field 7: bearing frequencies. These are NOT integer multiples of shaft speed. Cage, ball-pass-outer and ball-pass-inner frequencies depend on ball count, ball diameter, pitch diameter and contact angle, and they land at non-integer multiples typically somewhere between a fraction of shaft speed and several times it. Do not estimate them. Get the multipliers from the bearing manufacturer for that specific bearing, or leave the field blank and treat any peak you cannot account for from fields 1 to 6 or their harmonics as a bearing candidate to be confirmed by other means.

Field 8: the harmonics of fields 3, 4 and 5. Write the second and third multiple of each tonal field beside it. A blade-pass source at 176.8 Hz also radiates at 353.6 and 530.4 Hz, and twice line frequency at 120 Hz brings 240 and 360 Hz with it. Two things follow. A peak at a harmonic belongs to its fundamental and is not a separate fault, so do not name it twice. And harmonics collide: on a 60 Hz supply, a shaft turning near 20 Hz puts its sixth harmonic on top of the 120 Hz electrical line, and the only clean way to separate them is the speed test below. Fill in field 8 before you measure, not after, or you will find yourself explaining a peak you already predicted.

The filled-in worksheet

A tenant reports a whine from a rooftop supply fan serving an open office. Drive data taken with the unit locked out: motor nameplate 1750 rev/min at rated load on a 60 Hz supply, motor sheave pitch diameter 4.0 in, fan sheave pitch diameter 8.0 in, belt length 60 in, fan wheel with 12 blades.

Measured shaft speed with the unit running and the guard in place: 1768 rev/min. That is above nameplate, which is expected on a fan running below its rated load and is the first sanity check the sheet gives you.

Now the measurement. The A-weighting adjustments applied below are the standard values at those one-third-octave centres: minus 39.4 dB at 31.5 Hz, minus 16.1 dB at 125 Hz and minus 13.4 dB at 160 Hz. Third-octave bands, unweighted, Leq over 1 minute, dB re 20 uPa, taken at a fixed position about 3 ft off the fan discharge with the guard in place, hearing protection worn because the plant area runs close to the 85 dB(A) eight-hour action level in 29 CFR 1910.95(c). Three bands stand above their neighbours; everything else sits below 58 dB unweighted and contributes about 55 dB(A) in total.

Every peak has a name before any part is touched. The A-weighted total across the three bands plus the 55 dB(A) of broadband is 66.1 dB(A), Leq over 1 minute, re 20 uPa. The blade-pass band alone is 65.6 of it, standing 13.7 dB above the twice-line-frequency band, so under the logarithmic summing rule the blade-pass tone owns the number and the other two are almost invisible in it.

The test that separates aerodynamic from electrical

Fields 1 through 3, 5 and 6 all scale with shaft speed, and so do their harmonics in field 8. Field 4 and its own harmonics do not. So change the speed and watch which peaks move.

Before touching the drive, name what the command does: reducing supply fan speed reduces airflow, which can drive a heating or cooling coil below its minimum airflow and freeze it, and can reverse a pressure relationship in any space that depends on one. Do not run this test on a unit serving a pressure-critical space, and run it with heating and cooling disabled, the space unoccupied, and the facility operator holding the return.

With the drive reduced 10 percent, to 1591 rev/min, the sheet says fan shaft falls to 13.3 Hz and blade pass to 159.1 Hz. Measured result:

• 160 Hz band fell from 79 to 71 dB unweighted, so 71 - 13.4 = 57.6 dB(A).
• 125 Hz band held at 68 dB unweighted, unchanged, so 51.9 dB(A).
• 31.5 Hz band held near 62, so 22.6 dB(A).
• Broadband contribution about 55 dB(A).

New total: 60.2 dB(A), Leq over 1 minute, re 20 uPa - a fall of 5.9 dB. The band that moved with speed is aerodynamic and confirms field 3. The band that did not move is electrical and confirms field 4. Third-octave resolution is too coarse to SEE a 10 percent frequency shift, since the 160 Hz band spans roughly 141 to 178 Hz and 159.1 Hz is still inside it, so the evidence here is the level change and the fixed band, not a peak visibly sliding. If you need to watch the peak itself move, that needs narrowband analysis with a tachometer reference, which is the point at which this becomes a vibration-analysis job rather than a sound-level job.

Read the ceiling in the same numbers. At reduced speed the twice-line-frequency band is now only 5.7 dB under the blade-pass band, so it has started to participate. Chasing the blade-pass tone further will hit that floor: even taking the 160 Hz band to nothing leaves about 56.7 dB(A) from the remaining content. Tell the customer that before the second control is bought, not after.

Sibling-rule check against the cards this one sits beside. Every level printed carries quantity, weighting, reference, bandwidth and time basis. Band levels are summed logarithmically, never averaged. The A-weighting values used are the standard values at those third-octave centers. Field 3's proportionality is stated with the blade count held constant, and the speed test is the only place shaft speed changes, so nothing derived at 1768 rev/min is reused at 1591. No band is credited with a reduction while sitting more than 10 dB under the peak.

Where this hands off

The worksheet names the part. It does not tell you the part's condition, and the two questions have different owners in this library.

• A peak at field 1 or field 2, the shaft fundamental, points at imbalance, a bent shaft, or misalignment. The signature family covers reading those apart.
• A peak at field 4 that dominates points at the motor's electrical side, and the level is often unremarkable while the tonal character is what the customer is reacting to.
• A peak you cannot match to any filled field, especially a non-integer multiple of shaft speed or a broad haystack of energy rather than a line, is a bearing candidate. That is where the bearing-versus-belt cards and the vibration-signature reference do the work, and they are better at it than any sound-level measurement is.
• A tone that everybody hates at a level nobody can justify is a tonality problem rather than a level problem, and it is judged by different rules.

References

• 29 CFR 1910.147, control of hazardous energy, for the mechanical isolation and stored-energy lockout required before applying tape to a shaft or taking drive dimensions
• 29 CFR 1910.95, occupational noise exposure, for the hearing-conservation action level that governs standing beside running plant to take readings
• Bearing manufacturer data for that specific bearing, which is the only reliable source of cage and ball-pass multipliers; do not estimate them
• See related: The Vibration Signature Reference; The Difference Between a Bearing Fault Sound and a Belt Fault Sound; What Octave Bands Tell You That an Overall Level Cannot