# ISO 9613-2:2024 §7.4: what changes in the diffraction calculation compared with 1996

> The second edition rewrites Dz and Kmet and introduces the z_min threshold. The formulas side by side, the reason stated by the standard itself, and how much the numbers move for low barriers and long distances.

Published: 2026-05-22
Updated: 2026-09-03
Category: regolatorio
Tag: iso-9613-2, standards, environmental-acoustics, diffraction, barriers, deep-dive

Page: <https://www.stline.it/en/blog/iso-9613-2-2024-diffrazione-barriere/>

---

In January 2024 ISO published the second edition of 9613-2 — *Attenuation of sound during propagation outdoors, Part 2: engineering method for the prediction of sound pressure levels*. It cancels and replaces the 1996 edition, and the rewrite concerns §7.4, that is barrier diffraction: the formulation of the diffraction loss D<sub>z</sub> changes, and so does the meteorological correction K<sub>met</sub>.

**Update, 3 September 2026.** A draft amendment — ISO 9613-2:2024/DAmd 1, out for DIS ballot — restores both formulas on this page to their 1996 form. The two writings side by side, how much the numbers move and what survives of the second edition are in [ISO 9613-2:2024/DAmd 1: diffraction back to 1996](/en/blog/iso-9613-2-amd1-cosa-cambia/).

The reason is stated in the standard's own *Foreword*: *"modified specification of the barrier attenuation Dz and the correction for meteorological effects Kmet to eliminate well known shortcomings with low barriers and large source-to-receiver distances"*. The two shortcomings named are exactly the two cases where the 1996 edition was furthest off, and the two modifications address them one by one.

## The new D<sub>z</sub>, and the threshold that was not there

```
1996:  Dz = 10·log₁₀(3 + (C₂/λ)·C₃·z·Kmet)

2024:  Dz = 10·log₁₀[1 + (2 + (C₂/λ)·C₃·z)·Kmet]   for z > z_min
       Dz = 0                                       for z ≤ z_min
       z_min = -2λ / (C₂·C₃)
```

The constant going from 3 to 1+2 is not a matter of notation: in the 2024 form the constant term sits inside the bracket multiplied by K<sub>met</sub>, so wherever K<sub>met</sub> collapses — low barriers, large distances — D<sub>z</sub> tends to 10·log₁₀(1) = 0 instead of 10·log₁₀(3) ≈ 4.8 dB. In the 1996 edition the formula returned a positive attenuation even for small or negative `z`, that is for geometries where the barrier does not screen at all — it grazes the line of sight, or sits below it. An obstacle that does not intercept the direct path cannot attenuate, and crediting it with a few decibels is an error that accumulates precisely where barriers are low.

The 2024 edition introduces `z_min` as an explicit boundary: below that path-length difference, D<sub>z</sub> is zero. The threshold depends on wavelength, so it differs band by band — consistent with the fact that a low barrier screens high frequencies and not low ones. On its own, though, it moves not one decibel: substituting z_min = −2λ/(C₂C₃) into the argument of the logarithm gives 3 − 2 = 1 in the 1996 form and 1 + (2−2)·K<sub>met</sub> = 1 in the 2024 one, that is 0 dB in both. It is the boundary of the domain where the logarithm has an argument ≥ 1, written out explicitly — necessary in the 2024 form, which below that value would take the logarithm of a negative number.

## The new K<sub>met</sub>

```
1996:  Kmet = exp{-(1/2000)·√(d_SS·d_SR·d / (2z))}   [Kmet = 1 for z ≤ 0]

2024:  Kmet = exp{-(1/2000)·√[max(d_SS,d_SR)+e]·min(d_SS,d_SR)·d / [2(z-z_min)]}
```

Two changes. First, the source-to-edge and edge-to-receiver distances are treated **asymmetrically**, through `max` and `min`, whereas 1996 treated them as interchangeable inside a product. Second, the denominator becomes 2(z − z_min) instead of 2z, and with it goes the clause K<sub>met</sub> = 1 for z ≤ 0 — the meteorological correction now applies with a clear line of sight too, all the way down to the threshold. The direct distance `d` is in both editions: the canonical 1996 K<sub>met</sub> already carries it under the square root.

The asymmetry matters because a geometry with the source close to the barrier and the receiver far away is not equivalent to the opposite one, however equal the product of the distances: the curvature of sound rays caused by wind and temperature gradients acts along the whole path, and the long leg does not weigh like the short one. Over short urban scenarios the difference between the two formulations is small; over long-distance propagation it becomes significant, which is the second shortcoming named in the *Foreword*.

The upper bound is unchanged between editions: 20 dB for a single barrier, 25 dB for multiple diffraction (§7.4.4).

## Lateral diffraction (§7.4.3)

The 2024 edition also covers the paths that go around the **vertical edges** of the barrier, not only the one over the top edge. These are the paths that become dominant near the ends of a finite barrier, where sound passes around the side rather than over the top: modelling the top edge alone gives an attenuation that is too high at the extremities.

The three paths — over and around both sides — are combined per formula (25) of §7.4.4. The combination is only meaningful when the barrier is the screen governing the result: if on the same source-receiver pair a building shields more, the geometry of the barrier edges no longer describes the relevant path, and the lateral contribution must be discarded.

## How much the numbers move

The order of magnitude is what tells you whether the transition affects your own case:

- **short urban scenarios**, ordinary barrier height, source and receiver at comparable distances: the two editions differ by less than a decibel;
- **extreme cases** — low barriers, or large source-to-receiver distances: the difference reaches 4.4 dB per band, and these are the two cases the 2024 edition states it aims to correct. The per-geometry figures, computed on all three formulations, are in the [post on the amendment](/en/blog/iso-9613-2-amd1-cosa-cambia/).

Below a decibel you are inside the uncertainty of any field measurement, so the transition does not invalidate earlier urban work. In the extreme cases the difference is above the threshold at which a number changes a conclusion, and there the formulation used has to be declared.

## Why the 1996 edition is still needed

The transition between editions across the environmental noise prediction software in common use (CadnaA, SoundPLAN, NoiseModelling) did not happen simultaneously everywhere. A technician who has to compare a result with a study produced some years earlier, under the previous formulation, needs to reproduce that formulation — not an explanation of why the numbers do not match.

This is why a tool does well to declare which edition produced a result, instead of silently swapping it inside a version bump. The cost is one more selector; the return is that a result stays reconstructible years later, which is the precondition for comparing it with anything.

## The Italian regulatory context

9613-2 is a calculation method, not an obligation: what makes a result usable in Italy is the regulatory chain around it.

- **ISO 9613-1:1993** — atmospheric absorption, the term that feeds 9613-2;
- **ISO 9613-2:2024** — current edition of the engineering method;
- **Maekawa, Z. (1968)** — *Applied Acoustics* 1(3), 157-173, [DOI 10.1016/0003-682X(68)90020-0](https://doi.org/10.1016/0003-682X(68)90020-0): the experimental Fresnel-number curve from which the analytic approximations still in use derive;
- **Law 447 of 26/10/1995** — framework law on noise pollution;
- **DPCM 14/11/1997** — limit values for sound sources (Tables A, B, C, D);
- **DM 16/03/1998** — measurement and survey techniques;
- **DM 29/11/2000** — functional requirements for calculation models;
- **Legislative Decree 194 of 19/8/2005** — transposition of Directive 2002/49/EC.

On this point, to be explicit: "certified software" is not a category Italian regulation provides for, and no product is mandated. DM 29/11/2000 sets functional requirements the model must meet, and responsibility for the work rests with a competent acoustic technician on the ENTECA register. The two do not substitute for one another, and no program substitutes for the second.

## Where to try it

The comparison between formulations is best seen by running it: the [barrier calculation](/en/tools/barrier-calculator/) exposes the 2024 edition as the default method, the previous formulation as a comparison mode and Maekawa as a third option, on the same scenario. The model perimeter and its assumptions are in the [acmap](/en/wiki/acmap/) and [noise-barrier-calc](/en/wiki/noise-barrier-calc/) wikis; the code is public under the Apache 2.0 licence.