Correct DSP estimator edge cases: TSA time axis, echo detection, EQ validation (#356)
* Report the true 1/fs time axis from time_synchronous_average
The averaged samples sit on the 1/fs sampling grid (each block is aligned
back to it by the band-limited fractional delay before averaging), but the
result reported a time axis with T/M spacing. For a non-integer number of
samples per period the two grids drift apart across the period: with
100.37 samples per period the noiseless recovery error read against the
reported axis was 9.3e-2, versus 2.8e-6 on the true m/fs grid.
Return times = arange(M)/fs, document that the M samples cover the period
only approximately when fs*T is not an integer (McFadden 1987 revised
model: a block average on the sampling grid, no angular resampling), and
assert the non-integer-period tests on the true grid, with a regression
test pinning the 100.37-sample case on both grids. Also soften the comb
response [0, 1] claim (float cancellation beside a tooth can exceed 1 by
a few 1e-9) and mark the ~14 dB N = 32 figure as derived from Eq. 8
rather than printed in the paper.
* Detect inverting echoes and off-sample delays in echo_detection
echo_detection picked the maximum of the signed power cepstrum, so an
inverting reflection (a < 0), whose first rahmonic is a itself and hence
negative, was silently missed: with a = -0.4 at 313 samples the true
rahmonic c[313] = -0.42 was ignored and an unrelated 4739-sample peak
returned. Peak-pick on |cepstrum| within the searched band instead and
report the signed cepstrum value at the peak as the reflection
coefficient (the Mercator series ln(1 + a e^{-j theta}) holds for either
sign of a).
Also refine the reported delay by three-point parabolic interpolation of
|cepstrum| through the peak, so an off-sample delay (e.g. 313.5 samples)
is estimated to a fraction of a sample, and document the bin-splitting
caveat: at off-sample delays the peak-sample coefficient underestimates
|a| (about 65 % of it midway between bins). One-line convention note
distinguishing Milner's IDFT(ln|X|^2) power cepstrum from Bogert's
squared original. Tests cover a < 0 (exact impulse-echo closed form and
a noise source) and the interpolated half-sample delay.
* Validate parametric-EQ sections at the numerical extremes
Three validation gaps in the RBJ cookbook designer:
- A wide 'bw' at an f0 near Nyquist drives the w0/sin(w0) warping factor
into math.sinh overflow, escaping as a raw OverflowError; slightly less
extreme pairs produced a large but finite alpha whose rounded a2 landed
exactly at -1.0: poles on the unit circle designed silently. Wrap the
sinh with an informative ValueError and check the realized coefficients
against the stability triangle (|a2| < 1, |a1| < 1 + a2) after
normalization.
- A shelf slope beyond the gain-dependent bound (A + 1/A)/(A + 1/A - 2)
makes the cookbook's alpha complex and surfaced as a bare 'math domain
error'. Validate the bound at section construction with the limit in
the message (any slope is fine at 0 dB, where A = 1).
- Non-finite f0/gain_db/q/bw/slope passed validation silently (NaN
compares false against every bound) and produced NaN/inf coefficient
rows. Require finiteness explicitly.
Tests cover all three paths plus extreme-but-valid sections that must
keep designing stably.
* Scope the envelope-spectrum amplitude claim and add the band-pass front end
The docstring promised that a sinusoidal modulation 'reads out as a line
at its exact amplitude'; that holds only when the modulation frequency
falls on an analysis bin. Off-bin lines read low by the taper's
scalloping loss (about 1.42 dB, ~15 %, for the default Hann midway
between bins), like any single-record amplitude spectrum. State the
on-bin condition on the closed forms and the caveat in the module,
function and result docstrings, pinned by a midway-between-bins test.
Also add the optional band=(low, high) pre-filter to envelope_spectrum:
a zero-phase 4th-order Butterworth band-pass ahead of the detector, the
band-pass front end of Bendat & Piersol Figure 13.11 and the classical
bearing-envelope chain (isolate the excited resonance band, then
envelope it). The band travels with the result; a test shows it
recovering the exact modulation line under a strong out-of-band
interferer that swamps the raw envelope.
* Warn when a tone burst cannot close on the tone's zero crossing
IEC 60268-1 Clause A2.1 asks for an integral number of full periods, but
the gate closes after round(fs*cycles/frequency) samples: sample-exact
only when f and fs are commensurate. For 10 cycles of 997 Hz at 48 kHz
the burst spans 481.44 samples, gated at 481, and the waveform carries a
residual step at the gate edge. Document the condition and emit a
PhonometryWarning quantifying the realized residual (bounded by
amplitude*sin(pi*f/fs)) for incommensurate pairs; commensurate bursts
stay silent.
Also tighten the resampler's 'equiripple-bounded' wording: the Kaiser
window method bounds both bands by the same delta = 10^(-A/20), but its
ripple decays away from the band edges, unlike a true equiripple design.
* Document estimator conventions verified against their sources
Documentation and comment corrections, no behavior changes:
- random_data: keep the reverse-arrangement acceptance bounds on the
normal approximation deliberately and say why: B&P Table A.6 itself
follows it, and no percentage-point convention on the exact Mahonian
distribution reproduces the published entries (N = 40 and N = 100 at
a = 0.975 pin it down), so exact bounds would disagree with the table
users check against. The p-value stays exact; the one-count boundary
disagreement is now documented on TrendTestResult.
- time_frequency: the spectrogram energy-recovery statement was wrong as
written; integrating 'density' columns over frequency and summing over
time recovers the record energy only after multiplying by the hop
duration, up to the taper roll-off at the record edges (verified
0.988 vs 1 for Hann at 75 % overlap on white noise).
- spectra: note that the adaptive multitaper sigma^2 is mean(x^2) with no
mean removal (consistent with the documented no-detrending
calibration), and mark the 'P&W Eq. 369a' citation as pending
verification against the book.
- compliance + tests/reference_data: transcription note for the bare
'+3' at 20 Hz, Type 2, in ANSI S1.4-1983 Table V (read as upper-only;
+/-3 plausible; either way the verdict cannot change, the realized
deviation is -0.05 dB).
- room_ir: the Mueller & Massarani 'below 4 dB' crest-factor figure
assumes their slight magnitude smoothing, which shaped_sweep_signal
does not implement; measured sweeps here sit at 4.2-4.3 dB.
* Translate EQ section labels, complete the bibliography and changelog
- The per-section legend labels of the parametric-EQ plot printed the raw
English filter-type names in Spanish figures; route them through the
plot string table (peaking -> campana, shelves, band-pass, notch,
all-pass; lowpass/highpass were already translated for the lifter
plot). English output stays byte-identical.
- docs/references.md: add the Harris 1978 window paper and IEC
60268-1:1985, both cited by guide pages but missing from the collected
bibliography; link the IEC entry (and the one in test-signals.md) to
the IEC webstore page.
- Changelog entries for the DSP fixes; regenerate the API reference,
llms-full.txt and the cepstrum_echo figures (the detected-echo marker
moved by the sub-sample delay interpolation); the conformance report is
unchanged (419/419).
* Propagate the echo and envelope semantics to the site guides
Review follow-ups:
- The EN and ES site guides (cepstrum-echoes.mdx) still described the
old echo reading and the unqualified 'exact amplitude' claim; bring
them in step with docs/cepstrum-echoes.md: |cepstrum| peak-picking,
signed coefficient, interpolated delay, bin-splitting caveat, the
on-bin/scalloping condition and the new band= front end (the docs page
gains the same envelope-section text).
- State in the changelog and the result docstring that delay is no
longer exactly delay_samples/fs when the interpolated offset is
nonzero (delay_samples remains the whole-sample index the coefficient
is read at), and merge the duplicated Unreleased headings.
- Clearer stability message (poles 'do not lie strictly inside the unit
circle', full-precision coefficients) and an informative short-record
error for the envelope band-pass pre-filter instead of scipy's padlen
message.
- Sharper tone-burst warning test (asserts the quantified span and
residual) and a tighter interpolated-delay tolerance.
* Address review notes: cepstrum terminology, band-pass claims, edge validation
- Guides (docs and site, EN and ES): call the signed quantity what it is,
Milner's cepstrum of the log-power spectrum, and spell out that
Bogert's original 1963 power cepstrum squares once more and is
non-negative, so the signed rahmonics are the library's documented
convention; qualify the band-pass front end (out-of-band content is
strongly attenuated, not eliminated: finite Butterworth rejection and
small zero-phase edge transients).
- cepstrum: render reflection_coefficient as inline code in the
delay_samples docstring so the generated API page stops linking the
unrelated underwater seabed function.
- envelope: validate that 'band' is exactly two numeric edges before
indexing (a one-item tuple raised IndexError, extra entries were
ignored); the validated edges now travel to the result.
- equalizer: reject finite but unrepresentable gains through a shared
_gain_amplitude helper (10^(gain_db/40) overflowing, or underflowing
to zero and then dividing by zero in the designers); document the 0 dB
exception to the shelf-slope bound on the slope field.
- tone_burst: validate the repetition configuration before emitting the
incommensurate-frequency warning, so invalid setups raise even under
a warnings-as-errors filter.
- comb_filter_response: reject phase products that overflow to infinity
(sin(inf) returned NaN instead of the documented bounded response).
Tests for every new rejection path; API reference and llms regenerated.
* Test the representable gain amplitude as a positive finite bound
* Scope the first-rahmonic height claim to the source-free identity
Correct the framing and the guards of the perception and noise-control fiches (#355)
A review pass over the psychoacoustics, hearing, speech, environmental and
noise-control fiches. The numerics held up; what did not were the statements
printed around them, plus one wrong table lookup.
ISO 1999 validated domain. nipts()/htlan() applied Formula (2) wherever they
were asked to: L_EX,8h = 130 dB over 60 years at the 0,99 fractile returned a
357,9 dB threshold shift with nothing marking it as an extrapolation. Since the
formula is quadratic in (L_EX,8h - L0) that number is meaningless, not merely
uncertain. Both now warn (NoiseInducedHearingLossWarning) outside the stated
ranges while still computing: durations outside 1-40 years (clause 6.3.1
validates Formula (2) over 10-40 and Formula (3) over 1-10), fractiles in the
tails clause 6.3.2 says "should not be estimated" (Q below 5 % or above 95 %),
and levels above the 100 dB of Annex D that the Scope's NOTE 4 restricts
validity to. The fiches print the caveat when the conditions sit outside.
ISO 1999 Q. The fiches printed the library's fractile under ISO 1999's symbol,
inverting its meaning: in Formulae (4)/(5) Q is the percentage with worse
hearing, so the most-susceptible tenth is Q = 10 %, not the "Q = 0.9" the fiche
showed while citing 6.3.2. They now print ISO's Q as ISO defines it; the guides
note the difference from the fractile argument. Two further disclosures on the
same fiches: the age component H is evaluated from ISO 7029:2017 (the edition
ISO 1999 references undated in 6.2.2), which departs from the illustrative
Table A.3 selection by up to about 7 dB at the median, and the 2/3/4 kHz
average is the user's own choice, the Scope's NOTE 1 specifying no frequency
combination. The footer states a population scope rather than "the results
relate only to the tested specimen", which describes a laboratory specimen.
ISO 1999 Annex C is wired in as an independent oracle: its worked example (male
population aged 50, 90 dB for 30 years, 1/2/4 kHz at Q = 10 %) reproduces the
Table D.2 shifts it consumes (0, 9, 19 dB), the 13,3 dB left of the 4 kHz shift
after the Formula (1) compression (C.5) and the 31,1 dB combined threshold
(C.11). The composition it exercises is now exposed as combine_age_and_noise(),
which clause 6.2 wants available on its own: 6.2.3 recommends a database B
collected on a control population of the country under consideration, and this
applies Formula (1) to whichever database the user brings.
Verdict precision. The STI and SII fiches printed the requirement with one
decimal beside a measured value at full precision, so a verdict could contradict
itself: "STI = 0.50, required >= 0.5 -> FAIL" against a 0,52 requirement, and an
SII minimum of 0,75 printed as "0.8" (the committed example said so). Both now
print the requirement at the quantity's own precision. IEC 60268-16 Annex M
joins the conformance report as an end-to-end oracle: its printed MTF matrix,
speech spectrum and ambient noise give back its published MTI row and
STI = 0,76, exercising the whole A.5.3 to A.5.8 chain.
Noise-control prediction framing. The enclosure, silencer and HVAC fiches
compute design models but read as measurements: no prediction statement,
measurement headers, a footer scoped to a tested specimen, and committed
examples citing instrumentation that never existed ("Two-microphone
transfer-matrix bench"). They now follow the EN/ISO 12354 prediction exemplar
already in the repo, with a prediction-basis line, a statement naming what each
model does not represent, and a prediction footer. Their verdicts also decided
on the half-away-from-zero display rounding while printing Python's
round-half-to-even, so 0,25 printed 0.2 against a verdict taken on 0,3; both
now use the display rounding.
ISO/PAS 1996-3. With every onset rate at or below 10 dB/s the fiche withheld
the adjustment on the onset-rate gate (clauses 4.5/8) but justified it with "no
prominent impulse is present (governing P = 6.01 <= 5)", contradicting its own
boxed P. The note now states whichever gate governs. The assessment period is
read from the result instead of being hardcoded: 30 min is the Clause 5
*default*, now the default of the new assessment_period_min argument.
Smaller corrections:
- the HVAC elbow insertion loss took the wrong row of Bies Table 8.11 for a
W/lambda exactly on a bin edge; the rows read "a <= W/lambda < b", so an edge
opens its row (1 dB at 0,14, not 0 dB);
- the wind-turbine tonality fiche cited "subclauses 9.5.2-9.5.5" of
IEC 61400-11 for a chain running through 9.5.6 to 9.5.8 (Formulae 31-34);
- the ISO 1996-2 tonal fiche notes that Table J.1 keys on the mean audibility
of the J assessed spectra (ISO/PAS 20065 Clause 5.3.9), which a
single-spectrum assessment stands in for;
- the shared 10-90 % fractile band carried a hardcoded English legend in
Spanish figures, and the Spanish silencer and HVAC fiches printed the device
kind and duct-element label in English.
docs/ERRATA.md records two more source defects found on the way: ISO/PAS
1996-3:2022 Clause 5 swaps the 3.4/3.5 cross-references of the onset rate and
the level difference (its own units contradict the clause numbers printed with
them), and ISO 9613-2:1996 Table 2 prints 4,1 dB/km at 15 C / 80 % / 1 kHz
where ISO 9613-1 gives 4,1511, the neighbouring cells rounding correctly. The
library computes that coefficient from ISO 9613-1 directly, so it is unaffected.
Committed example fiches, the conformance report, the API reference and the
llms files are regenerated.
Correct DSP estimator edge cases: TSA time axis, echo detection, EQ validation (#356)
* Report the true 1/fs time axis from time_synchronous_average
The averaged samples sit on the 1/fs sampling grid (each block is aligned
back to it by the band-limited fractional delay before averaging), but the
result reported a time axis with T/M spacing. For a non-integer number of
samples per period the two grids drift apart across the period: with
100.37 samples per period the noiseless recovery error read against the
reported axis was 9.3e-2, versus 2.8e-6 on the true m/fs grid.
Return times = arange(M)/fs, document that the M samples cover the period
only approximately when fs*T is not an integer (McFadden 1987 revised
model: a block average on the sampling grid, no angular resampling), and
assert the non-integer-period tests on the true grid, with a regression
test pinning the 100.37-sample case on both grids. Also soften the comb
response [0, 1] claim (float cancellation beside a tooth can exceed 1 by
a few 1e-9) and mark the ~14 dB N = 32 figure as derived from Eq. 8
rather than printed in the paper.
* Detect inverting echoes and off-sample delays in echo_detection
echo_detection picked the maximum of the signed power cepstrum, so an
inverting reflection (a < 0), whose first rahmonic is a itself and hence
negative, was silently missed: with a = -0.4 at 313 samples the true
rahmonic c[313] = -0.42 was ignored and an unrelated 4739-sample peak
returned. Peak-pick on |cepstrum| within the searched band instead and
report the signed cepstrum value at the peak as the reflection
coefficient (the Mercator series ln(1 + a e^{-j theta}) holds for either
sign of a).
Also refine the reported delay by three-point parabolic interpolation of
|cepstrum| through the peak, so an off-sample delay (e.g. 313.5 samples)
is estimated to a fraction of a sample, and document the bin-splitting
caveat: at off-sample delays the peak-sample coefficient underestimates
|a| (about 65 % of it midway between bins). One-line convention note
distinguishing Milner's IDFT(ln|X|^2) power cepstrum from Bogert's
squared original. Tests cover a < 0 (exact impulse-echo closed form and
a noise source) and the interpolated half-sample delay.
* Validate parametric-EQ sections at the numerical extremes
Three validation gaps in the RBJ cookbook designer:
- A wide 'bw' at an f0 near Nyquist drives the w0/sin(w0) warping factor
into math.sinh overflow, escaping as a raw OverflowError; slightly less
extreme pairs produced a large but finite alpha whose rounded a2 landed
exactly at -1.0: poles on the unit circle designed silently. Wrap the
sinh with an informative ValueError and check the realized coefficients
against the stability triangle (|a2| < 1, |a1| < 1 + a2) after
normalization.
- A shelf slope beyond the gain-dependent bound (A + 1/A)/(A + 1/A - 2)
makes the cookbook's alpha complex and surfaced as a bare 'math domain
error'. Validate the bound at section construction with the limit in
the message (any slope is fine at 0 dB, where A = 1).
- Non-finite f0/gain_db/q/bw/slope passed validation silently (NaN
compares false against every bound) and produced NaN/inf coefficient
rows. Require finiteness explicitly.
Tests cover all three paths plus extreme-but-valid sections that must
keep designing stably.
* Scope the envelope-spectrum amplitude claim and add the band-pass front end
The docstring promised that a sinusoidal modulation 'reads out as a line
at its exact amplitude'; that holds only when the modulation frequency
falls on an analysis bin. Off-bin lines read low by the taper's
scalloping loss (about 1.42 dB, ~15 %, for the default Hann midway
between bins), like any single-record amplitude spectrum. State the
on-bin condition on the closed forms and the caveat in the module,
function and result docstrings, pinned by a midway-between-bins test.
Also add the optional band=(low, high) pre-filter to envelope_spectrum:
a zero-phase 4th-order Butterworth band-pass ahead of the detector, the
band-pass front end of Bendat & Piersol Figure 13.11 and the classical
bearing-envelope chain (isolate the excited resonance band, then
envelope it). The band travels with the result; a test shows it
recovering the exact modulation line under a strong out-of-band
interferer that swamps the raw envelope.
* Warn when a tone burst cannot close on the tone's zero crossing
IEC 60268-1 Clause A2.1 asks for an integral number of full periods, but
the gate closes after round(fs*cycles/frequency) samples: sample-exact
only when f and fs are commensurate. For 10 cycles of 997 Hz at 48 kHz
the burst spans 481.44 samples, gated at 481, and the waveform carries a
residual step at the gate edge. Document the condition and emit a
PhonometryWarning quantifying the realized residual (bounded by
amplitude*sin(pi*f/fs)) for incommensurate pairs; commensurate bursts
stay silent.
Also tighten the resampler's 'equiripple-bounded' wording: the Kaiser
window method bounds both bands by the same delta = 10^(-A/20), but its
ripple decays away from the band edges, unlike a true equiripple design.
* Document estimator conventions verified against their sources
Documentation and comment corrections, no behavior changes:
- random_data: keep the reverse-arrangement acceptance bounds on the
normal approximation deliberately and say why: B&P Table A.6 itself
follows it, and no percentage-point convention on the exact Mahonian
distribution reproduces the published entries (N = 40 and N = 100 at
a = 0.975 pin it down), so exact bounds would disagree with the table
users check against. The p-value stays exact; the one-count boundary
disagreement is now documented on TrendTestResult.
- time_frequency: the spectrogram energy-recovery statement was wrong as
written; integrating 'density' columns over frequency and summing over
time recovers the record energy only after multiplying by the hop
duration, up to the taper roll-off at the record edges (verified
0.988 vs 1 for Hann at 75 % overlap on white noise).
- spectra: note that the adaptive multitaper sigma^2 is mean(x^2) with no
mean removal (consistent with the documented no-detrending
calibration), and mark the 'P&W Eq. 369a' citation as pending
verification against the book.
- compliance + tests/reference_data: transcription note for the bare
'+3' at 20 Hz, Type 2, in ANSI S1.4-1983 Table V (read as upper-only;
+/-3 plausible; either way the verdict cannot change, the realized
deviation is -0.05 dB).
- room_ir: the Mueller & Massarani 'below 4 dB' crest-factor figure
assumes their slight magnitude smoothing, which shaped_sweep_signal
does not implement; measured sweeps here sit at 4.2-4.3 dB.
* Translate EQ section labels, complete the bibliography and changelog
- The per-section legend labels of the parametric-EQ plot printed the raw
English filter-type names in Spanish figures; route them through the
plot string table (peaking -> campana, shelves, band-pass, notch,
all-pass; lowpass/highpass were already translated for the lifter
plot). English output stays byte-identical.
- docs/references.md: add the Harris 1978 window paper and IEC
60268-1:1985, both cited by guide pages but missing from the collected
bibliography; link the IEC entry (and the one in test-signals.md) to
the IEC webstore page.
- Changelog entries for the DSP fixes; regenerate the API reference,
llms-full.txt and the cepstrum_echo figures (the detected-echo marker
moved by the sub-sample delay interpolation); the conformance report is
unchanged (419/419).
* Propagate the echo and envelope semantics to the site guides
Review follow-ups:
- The EN and ES site guides (cepstrum-echoes.mdx) still described the
old echo reading and the unqualified 'exact amplitude' claim; bring
them in step with docs/cepstrum-echoes.md: |cepstrum| peak-picking,
signed coefficient, interpolated delay, bin-splitting caveat, the
on-bin/scalloping condition and the new band= front end (the docs page
gains the same envelope-section text).
- State in the changelog and the result docstring that delay is no
longer exactly delay_samples/fs when the interpolated offset is
nonzero (delay_samples remains the whole-sample index the coefficient
is read at), and merge the duplicated Unreleased headings.
- Clearer stability message (poles 'do not lie strictly inside the unit
circle', full-precision coefficients) and an informative short-record
error for the envelope band-pass pre-filter instead of scipy's padlen
message.
- Sharper tone-burst warning test (asserts the quantified span and
residual) and a tighter interpolated-delay tolerance.
* Address review notes: cepstrum terminology, band-pass claims, edge validation
- Guides (docs and site, EN and ES): call the signed quantity what it is,
Milner's cepstrum of the log-power spectrum, and spell out that
Bogert's original 1963 power cepstrum squares once more and is
non-negative, so the signed rahmonics are the library's documented
convention; qualify the band-pass front end (out-of-band content is
strongly attenuated, not eliminated: finite Butterworth rejection and
small zero-phase edge transients).
- cepstrum: render reflection_coefficient as inline code in the
delay_samples docstring so the generated API page stops linking the
unrelated underwater seabed function.
- envelope: validate that 'band' is exactly two numeric edges before
indexing (a one-item tuple raised IndexError, extra entries were
ignored); the validated edges now travel to the result.
- equalizer: reject finite but unrepresentable gains through a shared
_gain_amplitude helper (10^(gain_db/40) overflowing, or underflowing
to zero and then dividing by zero in the designers); document the 0 dB
exception to the shelf-slope bound on the slope field.
- tone_burst: validate the repetition configuration before emitting the
incommensurate-frequency warning, so invalid setups raise even under
a warnings-as-errors filter.
- comb_filter_response: reject phase products that overflow to infinity
(sin(inf) returned NaN instead of the documented bounded response).
Tests for every new rejection path; API reference and llms regenerated.
* Test the representable gain amplitude as a positive finite bound
* Scope the first-rahmonic height claim to the source-free identity
Correct the framing and the guards of the perception and noise-control fiches (#355)
A review pass over the psychoacoustics, hearing, speech, environmental and
noise-control fiches. The numerics held up; what did not were the statements
printed around them, plus one wrong table lookup.
ISO 1999 validated domain. nipts()/htlan() applied Formula (2) wherever they
were asked to: L_EX,8h = 130 dB over 60 years at the 0,99 fractile returned a
357,9 dB threshold shift with nothing marking it as an extrapolation. Since the
formula is quadratic in (L_EX,8h - L0) that number is meaningless, not merely
uncertain. Both now warn (NoiseInducedHearingLossWarning) outside the stated
ranges while still computing: durations outside 1-40 years (clause 6.3.1
validates Formula (2) over 10-40 and Formula (3) over 1-10), fractiles in the
tails clause 6.3.2 says "should not be estimated" (Q below 5 % or above 95 %),
and levels above the 100 dB of Annex D that the Scope's NOTE 4 restricts
validity to. The fiches print the caveat when the conditions sit outside.
ISO 1999 Q. The fiches printed the library's fractile under ISO 1999's symbol,
inverting its meaning: in Formulae (4)/(5) Q is the percentage with worse
hearing, so the most-susceptible tenth is Q = 10 %, not the "Q = 0.9" the fiche
showed while citing 6.3.2. They now print ISO's Q as ISO defines it; the guides
note the difference from the fractile argument. Two further disclosures on the
same fiches: the age component H is evaluated from ISO 7029:2017 (the edition
ISO 1999 references undated in 6.2.2), which departs from the illustrative
Table A.3 selection by up to about 7 dB at the median, and the 2/3/4 kHz
average is the user's own choice, the Scope's NOTE 1 specifying no frequency
combination. The footer states a population scope rather than "the results
relate only to the tested specimen", which describes a laboratory specimen.
ISO 1999 Annex C is wired in as an independent oracle: its worked example (male
population aged 50, 90 dB for 30 years, 1/2/4 kHz at Q = 10 %) reproduces the
Table D.2 shifts it consumes (0, 9, 19 dB), the 13,3 dB left of the 4 kHz shift
after the Formula (1) compression (C.5) and the 31,1 dB combined threshold
(C.11). The composition it exercises is now exposed as combine_age_and_noise(),
which clause 6.2 wants available on its own: 6.2.3 recommends a database B
collected on a control population of the country under consideration, and this
applies Formula (1) to whichever database the user brings.
Verdict precision. The STI and SII fiches printed the requirement with one
decimal beside a measured value at full precision, so a verdict could contradict
itself: "STI = 0.50, required >= 0.5 -> FAIL" against a 0,52 requirement, and an
SII minimum of 0,75 printed as "0.8" (the committed example said so). Both now
print the requirement at the quantity's own precision. IEC 60268-16 Annex M
joins the conformance report as an end-to-end oracle: its printed MTF matrix,
speech spectrum and ambient noise give back its published MTI row and
STI = 0,76, exercising the whole A.5.3 to A.5.8 chain.
Noise-control prediction framing. The enclosure, silencer and HVAC fiches
compute design models but read as measurements: no prediction statement,
measurement headers, a footer scoped to a tested specimen, and committed
examples citing instrumentation that never existed ("Two-microphone
transfer-matrix bench"). They now follow the EN/ISO 12354 prediction exemplar
already in the repo, with a prediction-basis line, a statement naming what each
model does not represent, and a prediction footer. Their verdicts also decided
on the half-away-from-zero display rounding while printing Python's
round-half-to-even, so 0,25 printed 0.2 against a verdict taken on 0,3; both
now use the display rounding.
ISO/PAS 1996-3. With every onset rate at or below 10 dB/s the fiche withheld
the adjustment on the onset-rate gate (clauses 4.5/8) but justified it with "no
prominent impulse is present (governing P = 6.01 <= 5)", contradicting its own
boxed P. The note now states whichever gate governs. The assessment period is
read from the result instead of being hardcoded: 30 min is the Clause 5
*default*, now the default of the new assessment_period_min argument.
Smaller corrections:
- the HVAC elbow insertion loss took the wrong row of Bies Table 8.11 for a
W/lambda exactly on a bin edge; the rows read "a <= W/lambda < b", so an edge
opens its row (1 dB at 0,14, not 0 dB);
- the wind-turbine tonality fiche cited "subclauses 9.5.2-9.5.5" of
IEC 61400-11 for a chain running through 9.5.6 to 9.5.8 (Formulae 31-34);
- the ISO 1996-2 tonal fiche notes that Table J.1 keys on the mean audibility
of the J assessed spectra (ISO/PAS 20065 Clause 5.3.9), which a
single-spectrum assessment stands in for;
- the shared 10-90 % fractile band carried a hardcoded English legend in
Spanish figures, and the Spanish silencer and HVAC fiches printed the device
kind and duct-element label in English.
docs/ERRATA.md records two more source defects found on the way: ISO/PAS
1996-3:2022 Clause 5 swaps the 3.4/3.5 cross-references of the onset rate and
the level difference (its own units contradict the clause numbers printed with
them), and ISO 9613-2:1996 Table 2 prints 4,1 dB/km at 15 C / 80 % / 1 kHz
where ISO 9613-1 gives 4,1511, the neighbouring cells rounding correctly. The
library computes that coefficient from ISO 9613-1 directly, so it is unaffected.
Committed example fiches, the conformance report, the API reference and the
llms files are regenerated.
Correct DSP estimator edge cases: TSA time axis, echo detection, EQ validation (#356)
* Report the true 1/fs time axis from time_synchronous_average
The averaged samples sit on the 1/fs sampling grid (each block is aligned
back to it by the band-limited fractional delay before averaging), but the
result reported a time axis with T/M spacing. For a non-integer number of
samples per period the two grids drift apart across the period: with
100.37 samples per period the noiseless recovery error read against the
reported axis was 9.3e-2, versus 2.8e-6 on the true m/fs grid.
Return times = arange(M)/fs, document that the M samples cover the period
only approximately when fs*T is not an integer (McFadden 1987 revised
model: a block average on the sampling grid, no angular resampling), and
assert the non-integer-period tests on the true grid, with a regression
test pinning the 100.37-sample case on both grids. Also soften the comb
response [0, 1] claim (float cancellation beside a tooth can exceed 1 by
a few 1e-9) and mark the ~14 dB N = 32 figure as derived from Eq. 8
rather than printed in the paper.
* Detect inverting echoes and off-sample delays in echo_detection
echo_detection picked the maximum of the signed power cepstrum, so an
inverting reflection (a < 0), whose first rahmonic is a itself and hence
negative, was silently missed: with a = -0.4 at 313 samples the true
rahmonic c[313] = -0.42 was ignored and an unrelated 4739-sample peak
returned. Peak-pick on |cepstrum| within the searched band instead and
report the signed cepstrum value at the peak as the reflection
coefficient (the Mercator series ln(1 + a e^{-j theta}) holds for either
sign of a).
Also refine the reported delay by three-point parabolic interpolation of
|cepstrum| through the peak, so an off-sample delay (e.g. 313.5 samples)
is estimated to a fraction of a sample, and document the bin-splitting
caveat: at off-sample delays the peak-sample coefficient underestimates
|a| (about 65 % of it midway between bins). One-line convention note
distinguishing Milner's IDFT(ln|X|^2) power cepstrum from Bogert's
squared original. Tests cover a < 0 (exact impulse-echo closed form and
a noise source) and the interpolated half-sample delay.
* Validate parametric-EQ sections at the numerical extremes
Three validation gaps in the RBJ cookbook designer:
- A wide 'bw' at an f0 near Nyquist drives the w0/sin(w0) warping factor
into math.sinh overflow, escaping as a raw OverflowError; slightly less
extreme pairs produced a large but finite alpha whose rounded a2 landed
exactly at -1.0: poles on the unit circle designed silently. Wrap the
sinh with an informative ValueError and check the realized coefficients
against the stability triangle (|a2| < 1, |a1| < 1 + a2) after
normalization.
- A shelf slope beyond the gain-dependent bound (A + 1/A)/(A + 1/A - 2)
makes the cookbook's alpha complex and surfaced as a bare 'math domain
error'. Validate the bound at section construction with the limit in
the message (any slope is fine at 0 dB, where A = 1).
- Non-finite f0/gain_db/q/bw/slope passed validation silently (NaN
compares false against every bound) and produced NaN/inf coefficient
rows. Require finiteness explicitly.
Tests cover all three paths plus extreme-but-valid sections that must
keep designing stably.
* Scope the envelope-spectrum amplitude claim and add the band-pass front end
The docstring promised that a sinusoidal modulation 'reads out as a line
at its exact amplitude'; that holds only when the modulation frequency
falls on an analysis bin. Off-bin lines read low by the taper's
scalloping loss (about 1.42 dB, ~15 %, for the default Hann midway
between bins), like any single-record amplitude spectrum. State the
on-bin condition on the closed forms and the caveat in the module,
function and result docstrings, pinned by a midway-between-bins test.
Also add the optional band=(low, high) pre-filter to envelope_spectrum:
a zero-phase 4th-order Butterworth band-pass ahead of the detector, the
band-pass front end of Bendat & Piersol Figure 13.11 and the classical
bearing-envelope chain (isolate the excited resonance band, then
envelope it). The band travels with the result; a test shows it
recovering the exact modulation line under a strong out-of-band
interferer that swamps the raw envelope.
* Warn when a tone burst cannot close on the tone's zero crossing
IEC 60268-1 Clause A2.1 asks for an integral number of full periods, but
the gate closes after round(fs*cycles/frequency) samples: sample-exact
only when f and fs are commensurate. For 10 cycles of 997 Hz at 48 kHz
the burst spans 481.44 samples, gated at 481, and the waveform carries a
residual step at the gate edge. Document the condition and emit a
PhonometryWarning quantifying the realized residual (bounded by
amplitude*sin(pi*f/fs)) for incommensurate pairs; commensurate bursts
stay silent.
Also tighten the resampler's 'equiripple-bounded' wording: the Kaiser
window method bounds both bands by the same delta = 10^(-A/20), but its
ripple decays away from the band edges, unlike a true equiripple design.
* Document estimator conventions verified against their sources
Documentation and comment corrections, no behavior changes:
- random_data: keep the reverse-arrangement acceptance bounds on the
normal approximation deliberately and say why: B&P Table A.6 itself
follows it, and no percentage-point convention on the exact Mahonian
distribution reproduces the published entries (N = 40 and N = 100 at
a = 0.975 pin it down), so exact bounds would disagree with the table
users check against. The p-value stays exact; the one-count boundary
disagreement is now documented on TrendTestResult.
- time_frequency: the spectrogram energy-recovery statement was wrong as
written; integrating 'density' columns over frequency and summing over
time recovers the record energy only after multiplying by the hop
duration, up to the taper roll-off at the record edges (verified
0.988 vs 1 for Hann at 75 % overlap on white noise).
- spectra: note that the adaptive multitaper sigma^2 is mean(x^2) with no
mean removal (consistent with the documented no-detrending
calibration), and mark the 'P&W Eq. 369a' citation as pending
verification against the book.
- compliance + tests/reference_data: transcription note for the bare
'+3' at 20 Hz, Type 2, in ANSI S1.4-1983 Table V (read as upper-only;
+/-3 plausible; either way the verdict cannot change, the realized
deviation is -0.05 dB).
- room_ir: the Mueller & Massarani 'below 4 dB' crest-factor figure
assumes their slight magnitude smoothing, which shaped_sweep_signal
does not implement; measured sweeps here sit at 4.2-4.3 dB.
* Translate EQ section labels, complete the bibliography and changelog
- The per-section legend labels of the parametric-EQ plot printed the raw
English filter-type names in Spanish figures; route them through the
plot string table (peaking -> campana, shelves, band-pass, notch,
all-pass; lowpass/highpass were already translated for the lifter
plot). English output stays byte-identical.
- docs/references.md: add the Harris 1978 window paper and IEC
60268-1:1985, both cited by guide pages but missing from the collected
bibliography; link the IEC entry (and the one in test-signals.md) to
the IEC webstore page.
- Changelog entries for the DSP fixes; regenerate the API reference,
llms-full.txt and the cepstrum_echo figures (the detected-echo marker
moved by the sub-sample delay interpolation); the conformance report is
unchanged (419/419).
* Propagate the echo and envelope semantics to the site guides
Review follow-ups:
- The EN and ES site guides (cepstrum-echoes.mdx) still described the
old echo reading and the unqualified 'exact amplitude' claim; bring
them in step with docs/cepstrum-echoes.md: |cepstrum| peak-picking,
signed coefficient, interpolated delay, bin-splitting caveat, the
on-bin/scalloping condition and the new band= front end (the docs page
gains the same envelope-section text).
- State in the changelog and the result docstring that delay is no
longer exactly delay_samples/fs when the interpolated offset is
nonzero (delay_samples remains the whole-sample index the coefficient
is read at), and merge the duplicated Unreleased headings.
- Clearer stability message (poles 'do not lie strictly inside the unit
circle', full-precision coefficients) and an informative short-record
error for the envelope band-pass pre-filter instead of scipy's padlen
message.
- Sharper tone-burst warning test (asserts the quantified span and
residual) and a tighter interpolated-delay tolerance.
* Address review notes: cepstrum terminology, band-pass claims, edge validation
- Guides (docs and site, EN and ES): call the signed quantity what it is,
Milner's cepstrum of the log-power spectrum, and spell out that
Bogert's original 1963 power cepstrum squares once more and is
non-negative, so the signed rahmonics are the library's documented
convention; qualify the band-pass front end (out-of-band content is
strongly attenuated, not eliminated: finite Butterworth rejection and
small zero-phase edge transients).
- cepstrum: render reflection_coefficient as inline code in the
delay_samples docstring so the generated API page stops linking the
unrelated underwater seabed function.
- envelope: validate that 'band' is exactly two numeric edges before
indexing (a one-item tuple raised IndexError, extra entries were
ignored); the validated edges now travel to the result.
- equalizer: reject finite but unrepresentable gains through a shared
_gain_amplitude helper (10^(gain_db/40) overflowing, or underflowing
to zero and then dividing by zero in the designers); document the 0 dB
exception to the shelf-slope bound on the slope field.
- tone_burst: validate the repetition configuration before emitting the
incommensurate-frequency warning, so invalid setups raise even under
a warnings-as-errors filter.
- comb_filter_response: reject phase products that overflow to infinity
(sin(inf) returned NaN instead of the documented bounded response).
Tests for every new rejection path; API reference and llms regenerated.
* Test the representable gain amplitude as a positive finite bound
* Scope the first-rahmonic height claim to the source-free identity
Noise-control performance reports via .report() (#331)
* Add noise-control performance reports via .report()
Add a one-page PDF .report() fiche to the three noise_control result
types, each laid out with a per-band table beside the result's own plot,
a boxed single-number performance figure and an optional PASS/FAIL
verdict:
- EnclosureResult: machine-enclosure insertion loss (Bies, Hansen &
Howard, section 7.4.2). The table lists the supplied panel transmission
loss R, the interior build-up correction C and the net insertion loss
IL = R - C; the boxed figure is the mean insertion loss with the
external and internal surface areas. verbose=True adds the interior
room constant column. A declared minimum passes when the mean meets it.
- ReactiveSilencerResult: reactive-silencer transmission loss (Munjal
Eq. (3.27); Bies sections 8.8-8.9). The table lists the transmission
loss TL and, when end impedances were given, the insertion loss IL; the
boxed figure is the mean transmission loss with the peak and the device
kind. A declared minimum passes when the mean meets it.
- HvacSpectrumResult: HVAC duct-noise spectrum (Bies Chapter 8; VDI
2081-1). A regenerated-noise spectrum boxes the A-weighted sound power
level with the overall total (lower is better); an attenuation spectrum
boxes the mean attenuation (more is better). verbose=True adds the
A-weighting correction and A-weighted band-level columns.
The three renderers share a two-panel skeleton in
_report/_noise_control_fiche.py and reuse the sound-power table builder,
band labels and header grid. Each accepts an optional metadata header,
states its method basis and renders in English or Spanish.
Register one committed example per fiche under .github/reports/, add
structural and clean-room number-presence tests (EN and ES), and update
the CHANGELOG and the regenerated API reference.
* Address SonarCloud findings on the noise-control renderers
Reduce render_noise_control_fiche below the parameter-count threshold by
fixing the two-panel split widths internally (the three renderers never
overrode them), and lift the HVAC verdict symbol/unit selection out of a
nested conditional into an explicit if/elif/else. No change to rendered
output; the committed example fiches are unaffected.
* Round the requirement to display precision and fit the verbose table
Compare the declared requirement at the same one-decimal precision as the
measured value in the noise-control verdict, so the printed comparison can
never contradict the verdict at the boundary. Also trim the verbose
enclosure table columns to sum to the 64 mm left panel width.
Perception, hearing and speech guides: result figures and when-to-use prose (#361)
* Perception, hearing and speech guides: result figures and when-to-use prose
Add ToneAssessment.plot(), which draws an ECMA-418-1 tone against the
prominence criterion of its own method (recovered from criterion_db) over the
whole range of interest, with the deciding margin marked, and a
view="levels" option on ToneAudibilityResult.plot() that exposes the
level-versus-frequency view the ISO 1996-2 fiche already embedded.
Add eleven single-concept figures to the perception, hearing and speech
guides, each a realistic result drawn by its own .plot(): the tone prominence
verdict, the ISO/PAS 20065 tone levels, the ISO/PAS 1996-3 onset detection,
the ISO 532-2 specific loudness, the ECMA-418-2 specific tonality, the
specific fluctuation strength, the SII band audibility under a hearing loss,
the ISO 7029 fractile band, the ISO 1999 NIPTS audiogram, the STOI per-band
correlation and the IEC 60268-16 per-band MTI. Every figure is embedded in
the English guide, the Spanish mirror and the docs mirror with the dual
snippet block, and every plottable snippet on those pages now names its
.plot().
Prose: the tone-prominence guide gains a comparison of the four tonality
assessments the library implements and when each applies, the
hearing-threshold guide explains how the ISO 7029 fractiles feed the ISO 1999
age component and gains a See-also section, and the objective-intelligibility
guide gains the "which measure, and when" section that existed only in the
docs mirror, extended with the STOI/ESTOI, STI and SII choice.
Also label the STOI band axis at whole-hertz resolution and let the shaded
critical band of the tone-audibility levels plot follow the figure
background, so it reads on the white report page and on dark figures alike.
* Correct two figure alt texts against the plotted values
The ISO 7029 fractile figure describes the median and the 90 % fractile with
the values it actually draws (10 to 50 dB and 22 to 74 dB), and the
ECMA-418-2 specific tonality peak is at 9 Bark_HMS.
Correct DSP estimator edge cases: TSA time axis, echo detection, EQ validation (#356)
* Report the true 1/fs time axis from time_synchronous_average
The averaged samples sit on the 1/fs sampling grid (each block is aligned
back to it by the band-limited fractional delay before averaging), but the
result reported a time axis with T/M spacing. For a non-integer number of
samples per period the two grids drift apart across the period: with
100.37 samples per period the noiseless recovery error read against the
reported axis was 9.3e-2, versus 2.8e-6 on the true m/fs grid.
Return times = arange(M)/fs, document that the M samples cover the period
only approximately when fs*T is not an integer (McFadden 1987 revised
model: a block average on the sampling grid, no angular resampling), and
assert the non-integer-period tests on the true grid, with a regression
test pinning the 100.37-sample case on both grids. Also soften the comb
response [0, 1] claim (float cancellation beside a tooth can exceed 1 by
a few 1e-9) and mark the ~14 dB N = 32 figure as derived from Eq. 8
rather than printed in the paper.
* Detect inverting echoes and off-sample delays in echo_detection
echo_detection picked the maximum of the signed power cepstrum, so an
inverting reflection (a < 0), whose first rahmonic is a itself and hence
negative, was silently missed: with a = -0.4 at 313 samples the true
rahmonic c[313] = -0.42 was ignored and an unrelated 4739-sample peak
returned. Peak-pick on |cepstrum| within the searched band instead and
report the signed cepstrum value at the peak as the reflection
coefficient (the Mercator series ln(1 + a e^{-j theta}) holds for either
sign of a).
Also refine the reported delay by three-point parabolic interpolation of
|cepstrum| through the peak, so an off-sample delay (e.g. 313.5 samples)
is estimated to a fraction of a sample, and document the bin-splitting
caveat: at off-sample delays the peak-sample coefficient underestimates
|a| (about 65 % of it midway between bins). One-line convention note
distinguishing Milner's IDFT(ln|X|^2) power cepstrum from Bogert's
squared original. Tests cover a < 0 (exact impulse-echo closed form and
a noise source) and the interpolated half-sample delay.
* Validate parametric-EQ sections at the numerical extremes
Three validation gaps in the RBJ cookbook designer:
- A wide 'bw' at an f0 near Nyquist drives the w0/sin(w0) warping factor
into math.sinh overflow, escaping as a raw OverflowError; slightly less
extreme pairs produced a large but finite alpha whose rounded a2 landed
exactly at -1.0: poles on the unit circle designed silently. Wrap the
sinh with an informative ValueError and check the realized coefficients
against the stability triangle (|a2| < 1, |a1| < 1 + a2) after
normalization.
- A shelf slope beyond the gain-dependent bound (A + 1/A)/(A + 1/A - 2)
makes the cookbook's alpha complex and surfaced as a bare 'math domain
error'. Validate the bound at section construction with the limit in
the message (any slope is fine at 0 dB, where A = 1).
- Non-finite f0/gain_db/q/bw/slope passed validation silently (NaN
compares false against every bound) and produced NaN/inf coefficient
rows. Require finiteness explicitly.
Tests cover all three paths plus extreme-but-valid sections that must
keep designing stably.
* Scope the envelope-spectrum amplitude claim and add the band-pass front end
The docstring promised that a sinusoidal modulation 'reads out as a line
at its exact amplitude'; that holds only when the modulation frequency
falls on an analysis bin. Off-bin lines read low by the taper's
scalloping loss (about 1.42 dB, ~15 %, for the default Hann midway
between bins), like any single-record amplitude spectrum. State the
on-bin condition on the closed forms and the caveat in the module,
function and result docstrings, pinned by a midway-between-bins test.
Also add the optional band=(low, high) pre-filter to envelope_spectrum:
a zero-phase 4th-order Butterworth band-pass ahead of the detector, the
band-pass front end of Bendat & Piersol Figure 13.11 and the classical
bearing-envelope chain (isolate the excited resonance band, then
envelope it). The band travels with the result; a test shows it
recovering the exact modulation line under a strong out-of-band
interferer that swamps the raw envelope.
* Warn when a tone burst cannot close on the tone's zero crossing
IEC 60268-1 Clause A2.1 asks for an integral number of full periods, but
the gate closes after round(fs*cycles/frequency) samples: sample-exact
only when f and fs are commensurate. For 10 cycles of 997 Hz at 48 kHz
the burst spans 481.44 samples, gated at 481, and the waveform carries a
residual step at the gate edge. Document the condition and emit a
PhonometryWarning quantifying the realized residual (bounded by
amplitude*sin(pi*f/fs)) for incommensurate pairs; commensurate bursts
stay silent.
Also tighten the resampler's 'equiripple-bounded' wording: the Kaiser
window method bounds both bands by the same delta = 10^(-A/20), but its
ripple decays away from the band edges, unlike a true equiripple design.
* Document estimator conventions verified against their sources
Documentation and comment corrections, no behavior changes:
- random_data: keep the reverse-arrangement acceptance bounds on the
normal approximation deliberately and say why: B&P Table A.6 itself
follows it, and no percentage-point convention on the exact Mahonian
distribution reproduces the published entries (N = 40 and N = 100 at
a = 0.975 pin it down), so exact bounds would disagree with the table
users check against. The p-value stays exact; the one-count boundary
disagreement is now documented on TrendTestResult.
- time_frequency: the spectrogram energy-recovery statement was wrong as
written; integrating 'density' columns over frequency and summing over
time recovers the record energy only after multiplying by the hop
duration, up to the taper roll-off at the record edges (verified
0.988 vs 1 for Hann at 75 % overlap on white noise).
- spectra: note that the adaptive multitaper sigma^2 is mean(x^2) with no
mean removal (consistent with the documented no-detrending
calibration), and mark the 'P&W Eq. 369a' citation as pending
verification against the book.
- compliance + tests/reference_data: transcription note for the bare
'+3' at 20 Hz, Type 2, in ANSI S1.4-1983 Table V (read as upper-only;
+/-3 plausible; either way the verdict cannot change, the realized
deviation is -0.05 dB).
- room_ir: the Mueller & Massarani 'below 4 dB' crest-factor figure
assumes their slight magnitude smoothing, which shaped_sweep_signal
does not implement; measured sweeps here sit at 4.2-4.3 dB.
* Translate EQ section labels, complete the bibliography and changelog
- The per-section legend labels of the parametric-EQ plot printed the raw
English filter-type names in Spanish figures; route them through the
plot string table (peaking -> campana, shelves, band-pass, notch,
all-pass; lowpass/highpass were already translated for the lifter
plot). English output stays byte-identical.
- docs/references.md: add the Harris 1978 window paper and IEC
60268-1:1985, both cited by guide pages but missing from the collected
bibliography; link the IEC entry (and the one in test-signals.md) to
the IEC webstore page.
- Changelog entries for the DSP fixes; regenerate the API reference,
llms-full.txt and the cepstrum_echo figures (the detected-echo marker
moved by the sub-sample delay interpolation); the conformance report is
unchanged (419/419).
* Propagate the echo and envelope semantics to the site guides
Review follow-ups:
- The EN and ES site guides (cepstrum-echoes.mdx) still described the
old echo reading and the unqualified 'exact amplitude' claim; bring
them in step with docs/cepstrum-echoes.md: |cepstrum| peak-picking,
signed coefficient, interpolated delay, bin-splitting caveat, the
on-bin/scalloping condition and the new band= front end (the docs page
gains the same envelope-section text).
- State in the changelog and the result docstring that delay is no
longer exactly delay_samples/fs when the interpolated offset is
nonzero (delay_samples remains the whole-sample index the coefficient
is read at), and merge the duplicated Unreleased headings.
- Clearer stability message (poles 'do not lie strictly inside the unit
circle', full-precision coefficients) and an informative short-record
error for the envelope band-pass pre-filter instead of scipy's padlen
message.
- Sharper tone-burst warning test (asserts the quantified span and
residual) and a tighter interpolated-delay tolerance.
* Address review notes: cepstrum terminology, band-pass claims, edge validation
- Guides (docs and site, EN and ES): call the signed quantity what it is,
Milner's cepstrum of the log-power spectrum, and spell out that
Bogert's original 1963 power cepstrum squares once more and is
non-negative, so the signed rahmonics are the library's documented
convention; qualify the band-pass front end (out-of-band content is
strongly attenuated, not eliminated: finite Butterworth rejection and
small zero-phase edge transients).
- cepstrum: render reflection_coefficient as inline code in the
delay_samples docstring so the generated API page stops linking the
unrelated underwater seabed function.
- envelope: validate that 'band' is exactly two numeric edges before
indexing (a one-item tuple raised IndexError, extra entries were
ignored); the validated edges now travel to the result.
- equalizer: reject finite but unrepresentable gains through a shared
_gain_amplitude helper (10^(gain_db/40) overflowing, or underflowing
to zero and then dividing by zero in the designers); document the 0 dB
exception to the shelf-slope bound on the slope field.
- tone_burst: validate the repetition configuration before emitting the
incommensurate-frequency warning, so invalid setups raise even under
a warnings-as-errors filter.
- comb_filter_response: reject phase products that overflow to infinity
(sin(inf) returned NaN instead of the documented bounded response).
Tests for every new rejection path; API reference and llms regenerated.
* Test the representable gain amplitude as a positive finite bound
* Scope the first-rahmonic height claim to the source-free identity
Core metrology guides: result figures, ResampledSignalResult.plot() and real introductions (#364)
* Add ResampledSignalResult.plot(): the delivered anti-alias filter against its design spec
* Rewrite the multichannel guide intro: concrete use cases, per-channel guarantees, Bendat & Piersol citations
* Give the levels page a real introduction: descriptor families, when-to-use guidance and survey references
* Time weighting: framing intro with detector history, ANSI S1.4/Bies references and sibling links
* Calibration: proper introduction (SPL vs dBFS choice, workflow framing) with Bies reference and sibling links
* Add eleven core-metrology figures: resampling spec, cepstrum variants, liftering, correlation, IR alignment, Hilbert envelope, cross-spectra, Golay recovery, TSA noise law, runs test
* Test signals: anti-alias figure with dual snippet, res.plot() mention, See also, docs mirror sync
* Cepstrum guide: variants and liftering figures with dual snippets, complex-cepstrum plot mention
* Correlation guide: normalization, IR-alignment and Hilbert-envelope figures with dual snippets
* Spectral analysis: cross-spectral density and coherent-output-SNR figures with dual snippets
* System measurement: Golay exact-recovery figure and plot mentions for the inversion workflow
* Synchronous averaging: sqrt(N) figure, diagnostics-tool guidance, See also, recording naming
* Data qualification: runs-test figure with dual snippet and peak-statistics plot mention
* MISO coherence: conditioning-order snippet, practical pitfalls and See also across the three trees
* Time-frequency: Harris window reference and See also across the three trees
* GUM uncertainty: map the domain pages that consume the clause-5 machinery, expand See also
* Filter banks: compliance plot mentions and recording naming in the crossover snippet
* Changelog entry for the core-metrology documentation figures and prose
* Regenerate llms-full.txt for the core-metrology documentation
* Drop the duplicated phase-axis translation key
Base the whole-body A(8) on the Directive 2002/44/EC Part B dominant axis and tighten the vibration fiches (#352)
Directive 2002/44/EC bases the whole-body daily exposure A(8) on the
highest of the frequency-weighted axis values 1,4*a_wx, 1,4*a_wy, a_wz
(Annex, Part B, point 1), not on the ISO 2631-1 Eq. (10) vector total
a_v; the hand-arm side keeps the Part A vector total a_hv.
- Relabel the whole-body magnitude of the daily-exposure fiche as
a_w,max with a printed note stating the Part B basis (EN/ES); add
vibration.wbv_exposure_basis() returning the dominant-axis value and
document in daily_vibration_exposure which magnitude each kind must
be fed with; state the same basis in the human-vibration guide
(EN/ES) and the measurement-chain diagram.
- Derive the boxed exposure zone of the fiche from the same
displayed-rounded comparisons as the assessment rows and the verdict,
so an A(8) printing exactly at a threshold cannot show a zone one
step below its own Exceeded row.
- ISO 10848 Kij fiche: distinguish an empty single-number mean caused
by every in-range band being bracketed (M < 0,25) from a spectrum
with no bands in the Annex A range (EN/ES).
- Spanish ISO 2631-5 outputs: translate the subject sex in the
injury-probability plot title and the word Formula in the fiche's
clause references.
- Correct the running_rms comment: the linear window divides by the
full window length (zero-padded front), it does not average over the
available samples.
Add slow-sound slit and Helmholtz-resonator perfect absorbers (#346)
* Add slow-sound slit and Helmholtz-resonator perfect absorbers
Introduce materials.slow_sound_absorber, a transfer-matrix model of a
rigid panel of thin closed slits loaded by an array of Helmholtz
resonators. The resonators slow the sound inside the slit and pull its
resonance into the deep-subwavelength regime, and the visco-thermal
losses of the sub-millimetre slit and of the square resonator necks and
cavities make perfect absorption possible.
slit_helmholtz_absorber predicts the oblique-incidence absorption of the
panel, returning a SlitResonatorAbsorberResult with the surface
impedance, reflection factor, absorption, retrieved effective parameters
and the full chain matrix, plus a .plot() of alpha(f) with |R| overlaid.
critical_coupling_design solves the inverse problem: it tunes the cavity
length and the slit height so the reflection zero sits on the real
frequency axis, giving perfect absorption at a chosen frequency and
angle. The building blocks slit_effective_properties,
rectangular_duct_properties and helmholtz_resonator_impedance and the
HelmholtzResonator geometry are exported alongside.
The model is pinned to its exact analytic anchors: perfect absorption at
the design frequency, the Poiseuille resistivity limits 12 eta / h^2
(slit) and 28.454 eta / w^2 (square duct), and the loss-free
effective-parameter limits (rho -> rho0, kappa -> kappa0). Adds a
conformance section, an example figure, tests, an API-reference row set,
and EN/ES guide sections.
* Keep the Sonar gate clean for the slow-sound absorber
Exempt materials.slow_sound_absorber from S107 the same way the other
scientific APIs are: its functions carry the slit and resonator geometry
alongside the shared keyword-only air state, matching the porous-absorber
and JCA signatures, so the parameter count reflects the physics rather
than incidental complexity.
Extract a shared alpha(f) + |R| overlay renderer for the layered-absorber
and slit-resonator plots so the two oblique-incidence absorber renderers
no longer duplicate their body. Output is byte-identical.
* Make the slow-sound absorber files Ruff 0.16 clean
Use dict literals for the shared air-state mappings and drop the
redundant quoted type annotations (the module already imports
annotations from __future__). Keeps the new feature files clean under
the widened Ruff 0.16 default rule set without touching unrelated files.
* Modernize the slow-sound absorber additions for ruff 0.16 and regenerate references
* Silence the unused figure handle in the slow-sound graph helper
* Clarify slow-sound docstrings and cover the optional-correction branches
Fix the trailing space in a docstring code span, state the design warning is
emitted rather than raised, correct the deep-subwavelength depth ratio in the
figure docstring, and add a test exercising the end_correction and
slit_radiation False branches.
* Regenerate API reference and llms for the slow-sound docstring fixes
* Regenerate conformance and references after rebase
* Translate the slow-sound figure strings and require convergence in the check
Add the Spanish figure-string translations for the slow-sound absorber plot
(legend, design annotation, title and the panel-depth pattern) so the ES
variant is not left in English, and make the critical-coupling conformance
check fail unless the design solver converged.
* Regenerate conformance and references after rebase