# Errata in EN 15610

> One page of the phonometry documentation, as Markdown. Index: https://docs.jmrp.io/phonometry/llms.txt

Canonical: https://docs.jmrp.io/phonometry/reference/errata/en-15610/
Language: en
Alternate: https://docs.jmrp.io/phonometry/es/reference/errata/en-15610/index.md
Version: phonometry 4.0.0
Install: pip install "phonometry[full]"
Author: José Manuel Requena Plens (https://orcid.org/0000-0003-1250-6212)
Published: 2026-10-28
Updated: 2026-11-16T21:36:14Z
License: MIT (https://github.com/jmrplens/phonometry/blob/main/LICENSE)
Summary: Errata in EN 15610: 2 defects in the printed text, with the evidence and what phonometry does about each.

---

These are the 2 entries of the [errata registry](https://docs.jmrp.io/phonometry/reference/errata/) on EN 15610:2009, in the order the registry lists them. The registry's introduction says how each claim is checked against the printed page and what the status of an entry means.

## EN 15610:2009, Annex B, B.9.2 (a spike-removal loop that asks for no change of sign and never ends on a bend)

- **Location:** Annex B (informative), B.9.2, the spike-removal section of the
  listing of RoughProcess.m, printed pp. 23 and 24; clause 7.2, printed
  pp. 12 and 13.
- **The print:** the listing repeats its sweep under
  `while min(d2rdx2) <-10^7`, and inside it treats a sample as a spike only
  `if (y < -10^7) && (d1 ~= d2)`, the second derivative and a change of sign
  of the first, and interpolates across it only `if height > w^2/3`. Clause
  7.2 c) identifies a spike "by the criteria $\mathrm{d}^2r/\mathrm{d}x^2 <
  -10^7\ \mathrm{\mu m/m^2}$ and a change of sign for $\mathrm{d}r/\mathrm{d}x$",
  removes it when $h > w^2/a$ with $a = 3$ m, and ends: "The spike removal
  procedure shall be repeated until no further spike is detected."
- **The problem:** the loop condition leaves out the change of sign that
  7.2 c) and the sweep inside the loop both ask of a spike. Wherever the
  record bends sharply without a maximum or a minimum, its second derivative
  falls below $-10^7\ \mathrm{\mu m/m^2}$ with no spike there by 7.2 c), no
  sweep changes the record, and the loop does not terminate. A rail rising
  20 µm per millimetre and then 5 µm per millimetre, sampled every
  millimetre, is such a bend: $\mathrm{d}^2r/\mathrm{d}x^2 = -1.5 \times
  10^7\ \mathrm{\mu m/m^2}$ at the bend, and $\mathrm{d}r/\mathrm{d}x$ is
  positive on both sides of it. That is where the listing departs from the
  clause it illustrates. On a peak that 7.2 c) does detect and that the
  height test keeps as roughness a wheel feels, the listing is no worse than
  the clause read literally: a ridge 200 µm high, sampled every millimetre,
  whose edges the listing's slope rule places 42 mm apart, has
  $\mathrm{d}^2r/\mathrm{d}x^2 = -2 \times 10^7\ \mathrm{\mu m/m^2}$ at its
  tip and $h = 2 \times 10^{-4}$ m below $w^2/3 = 5.9 \times 10^{-4}$ m; it
  is detected on every pass and kept, so neither the listing nor "until no
  further spike is detected" ever stops on it.
- **Evidence:** the listing read on PDF pages 25 and 26 (printed pp. 23 and
  24), and clause 7.2 on PDF pages 14 and 15 (printed pp. 12 and 13) of
  BS EN 15610:2009, the UK implementation of EN 15610:2009 (approved by CEN on
  16 April 2009). The non-termination is a run of the listing's loop,
  transcribed line for line, on the bend and on the ridge above.
- **Library behaviour:**
  [`remove_roughness_spikes`](https://github.com/jmrplens/phonometry/blob/main/src/phonometry/environment/sources/acoustic_roughness.py)
  detects a spike by both criteria of 7.2 c) and repeats its sweep until a
  sweep removes nothing, which is how it reads "until no further spike is
  detected" for a peak that is detected and kept. It returns the bend and the
  ridge above unchanged; `test_a_bend_without_an_extremum_is_no_spike` and
  `test_a_broad_ridge_is_kept_and_the_sweep_ends` in
  [`tests/environment/sources/test_acoustic_roughness.py`](https://github.com/jmrplens/phonometry/blob/main/tests/environment/sources/test_acoustic_roughness.py)
  hold it on both.
- **Status:** unreported.

## EN 15610:2009, Annex B, B.9.2 (Method B filters run forward and back, outside every class of EN 61260)

- **Location:** Annex B (informative), B.9.2, the digital filtering section of
  the listing of RoughProcess.m, printed p. 25, described in B.6, printed
  p. 20; clause 7.4.3, printed p. 15, and clause 2, printed p. 5.
- **The print:** the listing sets the band edges `fsmin=fsc/(2^(1/6))` and
  `fsmax=fsc*(2^(1/6))` about the centres `fsc=1./wl_d`, designs each band
  with `order=3;` and `[b,a]=butter(order,Wn);`, filters with
  `rf=filtfilt(b,a,rraw);` and reports `specdf(q)=std(rf);` of the record
  left once 2 m are cut from either end. Clause 7.4.3 reads: "The digital
  filters shall comply with EN 61260", and clause 2 dates that reference as
  EN 61260 (IEC 61260:1995).
- **The problem:** `butter(3,Wn)` designs a band-pass filter 3 dB down at
  `fsmin` and `fsmax`, and `filtfilt` runs it forward and then backward, so
  the filter the record actually goes through has the square of its
  magnitude response: 6.02 dB down at its own band edges. Taking the
  listing's `fsc` as the exact midband frequency and the base-two ratio its
  edges are built on, Table 1 of EN 61260:1995 allows a relative attenuation
  just inside a band edge of at most 4.5 dB, 5.0 dB and 5.5 dB for classes 0,
  1 and 2, and 4.5.3 limits the filter integrated response
  $\Delta B = 10 \lg (B_\mathrm{e}/B_\mathrm{r})$, with $B_\mathrm{e}$ of
  equation (14) and $B_\mathrm{r} = G^{1/(2b)} - G^{-1/(2b)}$ of equation (9),
  to $\pm 0.15$ dB, $\pm 0.3$ dB and $\pm 0.5$ dB. The squared
  response gives $\Delta B = -0.59$ dB in every band of `wl_d` from 0.5 m
  to 4 mm at 1 mm sampling, and $-0.58$ dB in the 3.15 mm and 2.5 mm bands,
  where the sampling bends the response. The filters
  of the listing therefore meet no class of EN 61260 in any band, on either
  requirement, which is what 7.4.3 asks of them, and a band whose spectrum is
  flat across it reads about 0.6 dB low. The same filter run forward once
  would be 3.01 dB down at the edges, with $\Delta B$ from $+0.15$ dB to
  $+0.20$ dB, inside class 1. A comment of the same section, "used later for fsc(q)<50 ie
  lambda >0.02", disagrees with the test the code makes, `if fsc(q)>=20` and
  `elseif fsc(q)<20`, which sends the bands longer than 0.05 m to the
  decimated record.
- **Evidence:** the listing read on PDF page 27 (printed p. 25), B.6 on PDF
  page 22 (printed p. 20), clause 7.4.3 on PDF page 17 (printed p. 15) and
  clause 2 on PDF page 7 (printed p. 5) of BS EN 15610:2009, the UK
  implementation of EN 15610:2009; Table 1 on PDF page 14 (printed p. 10),
  equation (9) on PDF page 10 (printed p. 6), equations (13) and (14) on PDF
  page 15 (printed p. 11) and 4.5.3 on PDF page 16 (printed p. 12) of
  BS EN 61260:1996, which is IEC 61260:1995. The attenuations and $\Delta B$
  are a recomputation: the listing's filters designed with
  `scipy.signal.butter`, the same design as the MATLAB function, at 1 mm
  sampling (and on the record decimated by 10 for `fsc<20`), their response
  squared, read at the band edges and integrated by equation (14).
- **Library behaviour:**
  [`filtered_roughness_spectrum`](https://github.com/jmrplens/phonometry/blob/main/src/phonometry/environment/sources/acoustic_roughness.py)
  runs each band forward once through the order 4 Butterworth sections of
  [`roughness_filter_bank`](https://github.com/jmrplens/phonometry/blob/main/src/phonometry/environment/sources/acoustic_roughness.py),
  at the record's own rate, a bank that is class 0 of EN 61260:1995 on
  Table 1 up to the Nyquist wavenumber of the record, on 4.5.3 and on 4.9;
  `test_the_bank_is_class_0_on_table_1_of_en_61260_1995`,
  `test_the_bank_attenuates_class_0_beyond_g4_on_a_record`,
  `test_the_integrated_response_of_every_band_is_within_class_0` and
  `test_the_summation_of_outputs_is_within_class_0` in
  [`tests/environment/sources/test_roughness_method_b.py`](https://github.com/jmrplens/phonometry/blob/main/tests/environment/sources/test_roughness_method_b.py)
  hold it there, and the conformance checks "EN 15610:2009 7.4.3 and
  EN 61260:1995 Table 1" and "EN 15610:2009 7.4.3 and EN 61260:1995 4.8 and
  Table 1" grade it band by band and on tones run through it.
- **Status:** unreported.
