A typical phone call starts like this: “we can feel vibration in the house, can you measure it?”. The useful answer is not a quote, it is a question: vibration with respect to what? Because underneath that sentence there are four different problems, and measuring the wrong one well produces a report that is correct and unusable.
The four questions, and why none follows from another
Are the people inside disturbed? Is the building being damaged? Are workers overexposed? Does the sensitive equipment still work? Each has a different standard, a different quantity and a different threshold.
The thresholds sit on scales far apart. The human perception threshold for vibration is orders of magnitude below the damage threshold for masonry: a level that does not scratch the building can make a house unliveable. In the opposite direction, a level nobody notices can push an electron microscope or a coordinate measuring machine out of tolerance.
The Italian standard on disturbance, UNI 9614:2017, says so itself: it expressly excludes from its scope seismic vibration, structural damage, occupational hygiene and damage to equipment. It is built to answer one question, and it declares as much. For damage to the building the reference is another one, UNI 9916:2014, which covers structural response and architectural integrity. For workers it is statute, not a voluntary standard: Legislative Decree 81/08, Title VIII Chapter III, with action and limit values on daily exposure A(8).
Asking which question is being posed changes the instrument, it changes where the sensor goes, it changes the quantity computed and it changes the number the report concludes with.
The regulatory gap, which has to be filled in the contract
One fact is worth knowing before signing an engagement: in Italy there is no statutory limit for vibration in residential premises. Law 447/95 is the framework law on noise pollution; DPCM 14/11/1997 sets limit values for sound sources. Neither deals with vibration introduced into a building.
What exists is a voluntary technical standard. It becomes binding when a contract invokes it, when a local regulation adopts it, or when a court takes it as the criterion in assessing an emission under Article 844 of the Civil Code — normal tolerability.
The practical consequence: the criterion must be stated in the report, with its edition and the basis on which it applies. A report that compares a value to a threshold without saying where that threshold comes from is contestable before the number is even discussed. And whoever commissions it has an interest in that point being written rather than implied, because it is the first thing an opposing party attacks.
A detail of the 2017 edition worth a dispute
The 2017 edition of UNI 9614 is not a cosmetic update of its predecessor: it introduces a different assessment approach, taken from the Norwegian standard NS 8176, and refers to ISO 2631-2:2003 for measurement methods.
This means a 2010 report and a 2020 one are not comparable by default. It is the same problem as the transition between the two editions of ISO 9613-2 in acoustic calculation, which we have written about elsewhere: if earlier work has to be re-read, the edition it was produced under must be stated. Omitting that step delivers a comparison that does not survive scrutiny.
A second point of the same nature: road traffic, railways and construction sites do not produce stationary vibration, they produce events. An r.m.s. value averaged over an hour dilutes the passage of a train in the minutes of silence around it, and returns a low number that does not describe the experience of the people living there. This is why assessing repeated events does not use a simple average: it uses descriptors that represent the set of events.
Where the numbers are actually formed
The standards say what to assess. The value being assessed, however, comes out of an acquisition chain, and the defects of that chain do not show in the result: they show as a plausible number.
The most insidious case concerns sampling. Disturbance is assessed on frequency-weighted acceleration, and the weighting depends on direction: the body’s response along the vertical axis differs from that along the horizontal axes. A triaxial accelerometer is therefore needed, and the three axes must be in phase with each other, because composing the three components is a vector operation.
If the three channels are read in turn through a multiplexer, each axis is sampled at a different instant: the composition is computed on data that do not belong to the same moment. The skew is small in absolute terms, but its effect grows with frequency — and at 80 Hz, the upper limit of the band of interest, a delay negligible at 1 Hz becomes a non-negligible fraction of a period. It is a hardware requirement: it cannot be corrected afterwards.
It is the reason the four-channel front-end for IEPE sensors we develop samples its channels simultaneously rather than in turn, with 24-bit conversion over 20 kHz of bandwidth, on-board excitation and cable-break and short-circuit detection — the last of which is what saves an overnight run, where a cable pulled at the second minute is discovered eight hours later. The board is on the electronics page, in 3D.
What to ask, if you need a measurement
Three questions to put to whoever performs it, ourselves included:
- Which of the four questions you are answering, and under which standard in its edition in force.
- Against what criterion you compare the result, given that no statutory threshold exists for residential premises, and on what basis that criterion applies to the case.
- What the uncertainty of the value is, and what temporal representativeness it rests on: two trains or twenty are not the same measurement.
The technical detail — quantities, weightings, measurement points, acquisition chain, uncertainty budget — is in the wiki: vibration in buildings.