See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/323399206

Correction/measurement of rooms in the low

frequency range

Presentation · January 2018

DOI: 10.13140/RG.2.2.28084.96649

CITATIONS READS

0 37

3 authors:

Hervé Lissek Etienne Rivet

École Polytechnique Fédérale de Lausanne École Polytechnique Fédérale de Lausanne
144 PUBLICATIONS 410 CITATIONS 30 PUBLICATIONS 39 CITATIONS

SEE PROFILE SEE PROFILE

Thach Pham Vu
École Polytechnique Fédérale de Lausanne
3 PUBLICATIONS 0 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

INTERACTS: Intelligent low-frequeNcy acousTic Equalization of Rooms using Active ConTrol

Subwoofers View project

Transmission-Line Acoustic Metamaterials View project

All content following this page was uploaded by Hervé Lissek on 26 February 2018.

The user has requested enhancement of the downloaded file.

Correction/measurement of rooms
in the low frequency range

Hervé Lissek
Etienne Rivet
Thach Pham Vu

Signal Processing Laboratory 2 (LTS2)

École polytechnique fédérale de Lausanne, Switzerland

January 18, 2018

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Listening Rooms at Low Frequencies (1)

Mic. #1 Mic. #2 Mic. #3 Mic. #4 Mic. #5

100

90

80
Magnitude (dB)

70

60

50

40
20 25 31.5 40 50 63 80 100 125
Frequency (Hz)

Room correction at low frequencies Hervé Lissek 2

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Listening Rooms at Low Frequencies (1)

Mic. #1 Mic. #2 Mic. #3 Mic. #4 Mic. #5

100
Helmholtz equation
Boundary conditions
90
∆p + k 2 p = 0, ρω
ω ~ ·n
∇p ~ = −j p
80 k= Zn
Magnitude (dB)

c
70
Pressure field in the frequency domain
60
∞
X Ψn (x, y , z)Ψn (x0 , y0 , z0 )
50 p(x, y , z, ω) = jρωc 2 q0
n=1
Kn (ω 2 − ωn2 + 2jδn ω)
40
Z Z Z
20 25 31.5 40 50 63 80 100 125
Frequency (Hz) with Kn = Ψn (x, y , z)dxdydz
z y x

Room correction at low frequencies Hervé Lissek 2

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Listening Rooms at Low Frequencies (1)

Mic. #1 Mic. #2 Mic. #3 Mic. #4 Mic. #5

100
Helmholtz equation
Boundary conditions
90
∆p + k 2 p = 0, ρω
ω ~ ·n
∇p ~ = −j p
80 k= Zn
Magnitude (dB)

c
70
In rectangular room and ideally rigid walls
60      
mx πx my πy mz πz
50 Ψn (x, y , z) = cos cos cos
s lx ly lz
mx 2 my 2 mz 2
40
    
20 25 31.5 40 50 63 80 100 125 c
Frequency (Hz) and fn = + +
2 lx ly lz
Room correction at low frequencies Hervé Lissek 2
Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Listening Rooms at Low Frequencies (1)

Mic. #1 Mic. #2 Mic. #3 Mic. #4 Mic. #5

100

90
Helmholtz equation
Boundary conditions
80
∆p + k 2 p = 0, ρω
Magnitude (dB)

ω ~ ·n
∇p ~ = −j p
70 k= Zn
c
60
In other cases
50
Eigenfunctions Ψn (x, y , z) not accessible by mea-
40
20 25 31.5 40 50 63 80 100 125 surement (only numerical model).
Frequency (Hz)

Room correction at low frequencies Hervé Lissek 2

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Listening Rooms at Low Frequencies (2)

What happens in the time domain?

Room correction at low frequencies Hervé Lissek 3

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Listening Rooms at Low Frequencies (2)

Quality factor of mode n

jω
Hn (ω) =
ω 2 + 2jδn ω − ωn2
ωn ωn τn
→ Qn = =
2δn 2

Mic. #1 Mic. #2 Mic. #3 Mic. #4 Mic. #5 MT60 = 1 s

100
TRANSITION REGION STEADY STATE FREE RESPONSE
1
90

Normalized Sound Pressure

80 0.5
Magnitude (dB)

70
0

60
-0.5

50
-1
40
20 25 31.5 40 50 63 80 100 125 0 0.5 1 t0 t 0+0.5 t 0+1
Frequency (Hz)
Time (s)

Room correction at low frequencies Hervé Lissek 3

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Listening Rooms at Low Frequencies (2)

Modal decay time of mode n:= decay of 60 dB

3 ln(10) Qn
MT60n =
π fn

Mic. #1 Mic. #2 Mic. #3 Mic. #4 Mic. #5 MT60 = 1 s

100
TRANSITION REGION STEADY STATE FREE RESPONSE
1
90

Normalized Sound Pressure

80 0.5
Magnitude (dB)

70
0

60
-0.5

50
-1
40
20 25 31.5 40 50 63 80 100 125 0 0.5 1 t0 t 0+0.5 t 0+1
Frequency (Hz)
Time (s)

→ 1. How to improve the listening experience in any room at low frequencies?

→ 2. How to measure the room acoustic characteristics at low frequencies?
Room correction at low frequencies Hervé Lissek 3
Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Correction Strategies

• Passive means

Passive absorbers
Optimal room design / configuration

Room correction at low frequencies Hervé Lissek 4

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Correction Strategies

• Passive means

Passive absorbers
Optimal room design / configuration

• Active equalisation

-G

Single input Multiple inputs

Plane wave propagation

Room correction at low frequencies Hervé Lissek 4

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Correction Strategies

• Active absorption

Olson and May, J. Acoust. Am., 1953.

Room correction at low frequencies Hervé Lissek 5

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Correction Strategies

• Active absorption

Olson and May, J. Acoust. Am., 1953.

→ Electroacoustic absorbers
Direct

S d pr i
Sd p f +

v
u

Controller

Nicholson, PhD thesis, 1994

Room correction at low frequencies Hervé Lissek 5

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Correction Strategies

• Active absorption

Olson and May, J. Acoust. Am., 1953.

→ Electroacoustic absorbers
Direct Shunt-based

S d pr i S d pr i
Sd p f + Sd p f +
v v Rl
u u Ll
Cl
- -

Controller

Nicholson, PhD thesis, 1994 Boulandet, PhD thesis, 2012

Room correction at low frequencies Hervé Lissek 5

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Correction Strategies

• Active absorption

Olson and May, J. Acoust. Am., 1953.

→ Electroacoustic absorbers
Direct Shunt-based Self-sensing

S d pr i S d pr i S d pr
Sd p f + Sd p f + Sd p f + i
v v Rl v
u u Ll u
Controller
Cl
- - -

Controller

Boulandet et al., Acta

Nicholson, PhD thesis, 1994 Boulandet, PhD thesis, 2012 Acustica united with
Acustica, 2016

Room correction at low frequencies Hervé Lissek 5

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Correction Strategies

• Active absorption

Olson and May, J. Acoust. Am., 1953.

→ Electroacoustic absorbers
Hybrid passive/active
Direct Shunt-based Self-sensing

S d pr i S d pr i S d pr S d pr
Sd p f + Sd p f + Sd p f + i Sd p f + i
v v Rl v v
u u Ll u u
Controller
Cl
- - - -

Controller Controller

Boulandet et al., Acta Galland et al., Applied

Nicholson, PhD thesis, 1994 Boulandet, PhD thesis, 2012 Acustica united with Acoustics, 2005
Acustica, 2016

Room correction at low frequencies Hervé Lissek 5

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Correction Strategies

• Active absorption

Olson and May, J. Acoust. Am., 1953.

→ Electroacoustic absorbers
Self-sensing Hybrid sensor-/shunt-based
Direct Shunt-based

Controller

S d pr S d pr S d pr Sd p f
Sd p f + i Sd p f + i Sd p f + i i
+ u/i
v v Rl v
u u Ll u v
u
Controller
Cl
- - - -

Controller

Boulandet et al., Acta

Rivet et al., IEEE CST,
Nicholson, PhD thesis, 1994 Boulandet, PhD thesis, 2012 Acustica united with
2017
Acustica, 2016

Room correction at low frequencies Hervé Lissek 5

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

Overview of the presentation

• New concepts allow efficient sound absorption in the low-frequency range

→ Presentation of the Electroacoustic Absorber (EA) concept
• General performance can be measured with standard facilities (1D)
→ Example of experimental assessment of EA (for design purpose)
→ Highlight why these standard measurement are not transposable to real room
applications
• Some 3D performance can be derived from simulations/in situ measurements
→ Present optimization procedures based on time decay with room simulations
→ Present some experimental results (time decay) measured in real rooms
• Still, limited tools to measure their performance in situ
→ Present preliminary results (space distribution) obtained through COMSOL
simulations

Room correction at low frequencies Hervé Lissek 6

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Introduction

1 Electroacoustic absorbers design

Electroacoustic absorbers: general principle
Electroacoustic absorbers: hybrid sensor-/shunt-based control
Performance assessment in 1D Case

2 From 1D to 3D
Identifying target Impedance in 3D
Optimisation of Multiple Degree-of-Freedom Target Impedance

3 In situ performance Evaluation

Frequency responses
Modal decay times
Spatial distribution of pressure

Room correction at low frequencies Hervé Lissek 7

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: general principle

General description

Mobile equipment: Mms , Rms , Cms

Moving coil: Re , Le
Electrodynamic transduction: B`
Enclosure: Vb
Diaphram area: Sd

Room correction at low frequencies Hervé Lissek 8

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: general principle

General description

Newton’s law on the mobile equipment

dv (t)
Mms =Sd (pfront (t) − prear (t)) − B`i(t)
dt Z
1
− Rms v (t) − v (t)dt
Cms
Mesh law on the electric circuit
 
di(t)
u(t) = Le + Re i(t) − B`v (t)
Mobile equipment: Mms , Rms , Cms dt
Moving coil: Re , Le
Electrodynamic transduction: B`
Enclosure: Vb
Diaphram area: Sd

Room correction at low frequencies Hervé Lissek 8

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: general principle

General description

Newton’s law on the mobile equipment

(Laplace domain: s = jω)

Zms (s)V (s) = Sd Pfront (s) − B`I (s)

1
where Zms (s) = sMms + Rms +
sCmc
Cms Vb
and Cmc =
Vb + ρc 2 Sd2 Cms
Mesh law on the electric circuit
(Laplace domain: s = jω)

Mobile equipment: Mms , Rms , Cms U(s) = Ze (s)I (s) − B`V (s)
Moving coil: Re , Le
Electrodynamic transduction: B` where Ze (s) = sLe + Re
Enclosure: Vb
Diaphram area: Sd

Room correction at low frequencies Hervé Lissek 8

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: general principle

General description

Newton’s law on the mobile equipment

(Laplace domain: s = jω)

Zms (s)V (s) = Sd Pfront (s) − B`I (s)

If I (s) is controlled:
We can write the apparent acoustic impedance of
the diaphragm:

Pfront (s)
Zs (s) =
V (s)
Mobile equipment: Mms , Rms , Cms
Moving coil: Re , Le
Electrodynamic transduction: B`
Enclosure: Vb
Diaphram area: Sd

Room correction at low frequencies Hervé Lissek 8

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: general principle

Absorption performance (under normal incidence - 1D)

• Incident pressure pi
• Reflected pressure
pi
Zs (s) − Zc
pr = Γ(s)pi = pi
Zs (s) + Zc
(under normal incidence)
pr
• Total pressure pfront = pi + pr

Room correction at low frequencies Hervé Lissek 9

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: general principle

Absorption performance (under normal incidence - 1D)

Sound absorption coefficient
 Zs (f ) − Zc 2
 
Ia
α(f ) = = 1 −  
Ii Zs (f ) + Zc 
pi
Bandwidth of efficient sound absorption BW
pr 1
Frequency range over which Itot ≤ Iideal case
2
→ αth ' 0.83

Room correction at low frequencies Hervé Lissek 9

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: general principle

Absorption performance (under normal incidence - 1D)

pi
Example: open-circuit loudspeaker (I = 0)
pr
 
Zms (s) 1 1
Zs (s) = = sMms + Rms +
Sd Sd sCmc

Specific acoustic impedance Sound absorption coefficient

Room correction at low frequencies Hervé Lissek 9

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: hybrid sensor-/shunt-based control

Hybrid Sensor-/Shunt-Based Control

Constitutive laws

Zms (s)V (s) = Sd Pfront (s) − B`I (s)

I (s) = Θ(s)Pfront (s)

Controller

Acoustic impedance Sd p f
+ i u/i
Zms (s)
Zs (s) = v
u
Sd − B`Θ(s)
-

Room correction at low frequencies Hervé Lissek 10

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: hybrid sensor-/shunt-based control

Hybrid Sensor-/Shunt-Based Control

Transfer function
Controller
I (s) Sd Zst (s) − Zm (s) Sd p f
Θ(s) = =
Pfront (s) B` Zst (s) + i u/i

v
u
Target impedance
-
µMms µ
Zst (s) = s + Rst +
Sd sSd Cmc

Bruel&Kjaer Power
Amplifier 2706
Bruel&Kjaer Pulse
Multichannel Analyzer
ISO 10534-2
Output Input Output Inputs
Electroacoustic
Sound Source Absorber H12 − HI j2k.x1
Rf Γ̂12 = .e
R − H12
p1 p2 p3 Ri H
i Rs +
−jk(x −x2 )

uout
Rd
Rf
-
Rs Ri HI = e 1
pt Cf
HR = e jk(x1 −x2 )
Waveguide ADC FPGA DAC

Room correction at low frequencies Hervé Lissek 10

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Electroacoustic absorbers: hybrid sensor-/shunt-based control

Hybrid Sensor-/Shunt-Based Control

Transfer function
I (s) Sd Zst (s) − Zm (s) Case µ Rst
Θ(s) = =
Pfront (s) B` Zst (s) A 1 Rms /Sd
Target impedance B 0.15 ρc/8
C 0.15 ρc
µMms µ
Zst (s) = s + Rst +
Sd sSd Cmc
A B C A B C
1200 1
Magnitude (Pa.s.m-1 )

1000 0.9
800
0.8

Sound Absorption Coefficient

600
0.7
400
0.6
200
0 0.5
π 0.4
Phase (rad)

π/2 0.3

0 0.2

-π/2 0.1

-π 0
16 31.5 63 125 250 500 1000 16 31.5 63 125 250 500 1000
Frequency (Hz) Frequency (Hz)

Sound absorption coefficient Specific acoustic impedance

Rivet et al., IEEE Transactions on Control Systems Technology, 2017

Room correction at low frequencies Hervé Lissek 10

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Performance assessment in 1D Case

With Hard Wall

Signal
Generator
+
Amplifier

Room correction at low frequencies Hervé Lissek 11

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Performance assessment in 1D Case

With a Loudspeaker

Signal
Generator
+
Amplifier

Room correction at low frequencies Hervé Lissek 12

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Performance assessment in 1D Case

With an Electroacoustic Absorber

Controller

+
u/i
Signal
Generator
+
Amplifier

Room correction at low frequencies Hervé Lissek 13

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

1 Electroacoustic absorbers design

Electroacoustic absorbers: general principle
Electroacoustic absorbers: hybrid sensor-/shunt-based control
Performance assessment in 1D Case

2 From 1D to 3D
Identifying target Impedance in 3D
Optimisation of Multiple Degree-of-Freedom Target Impedance

3 In situ performance Evaluation

Frequency responses
Modal decay times
Spatial distribution of pressure

Room correction at low frequencies Hervé Lissek 14

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

1D Case under Normal Incidence

Modal decay time of mode n

3 ln(10)
MT60n =
δn

Duct dimensions: L = 1.7 m, W = H = 30 cm

Target acoustic resistance: Rs = ζρc

Room correction at low frequencies Hervé Lissek 15

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

1D Case under Normal Incidence

Rs = ρc Rs = 0.15ρc

rabs = 6 cm, Sabs = 113 cm2 (12.6 % of the wall area)

Sound pressure (color), active sound intensity (gray)

Room correction at low frequencies Hervé Lissek 16

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

1D Case under Grazing Incidence

Rs = ρc Rs = 0.32ρc

rabs = 10 cm, Sabs = 314 cm2

Sound pressure (color), active sound intensity (gray)

Room correction at low frequencies Hervé Lissek 17

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

In rooms?

• Design of demonstration prototypes

→ Value of target impedance?

Room correction at low frequencies Hervé Lissek 18

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

Optimization on Modal Decay Time

As seen before, the Modal Decay Times only depend on eigenvalues ωn + jδn
(invariant with position)
→ can we derive optimal values of acoustic impedance by minimizing MT60n for a
given set of room resonance frequencies?
Optimization with COMSOL: output δn for a given set of resonance frequencies
Medium room M1 Medium room M2
Small room S L = 7.02m,W = 5.10m,H = 2.70m
L = 7.87m,W = 6.36m,H = 3.48m
L = 5.33m,W = 3.76m,H = 2.13m

Modal decay time of mode n

3 ln(10)
MT60n =
δn

Room correction at low frequencies Hervé Lissek 19

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

Modal Decay Time

Small room S Medium room M1 Medium room M2

L = 5.33m,W = 3.76m,H = 2.13m L = 7.02m,W = 5.10m,H = 2.70 m L = 7.87m,W = 6.36m,H = 3.48m

1.2 120 1.2 120 1.2 120

1.1 110 1.1 110 1.1 110

1 1 1
100 100 100
0.9 0.9 0.9

Modal Decay Time (s)

Modal Decay Time (s)

Modal Decay Time (s)

90 90 90
0.8 0.8 0.8

Frequency (Hz)
Frequency (Hz)

Frequency (Hz)
0.7 80 80 0.7 80
0.7
0.6 70 0.6 70 0.6 70

0.5 60 0.5 60 0.5 60

0.4 0.4 0.4
50 50 50
0.3 0.3 0.3
40 40 40
0.2 0.2 0.2
0.1 30 30 0.1 30
0.1
0 20 0 20 0 20
16 31.5 63 125 250 500 16 31.5 63 125 250 500 16 31.5 63 125 250 500
Specific Acoustic Resistance (Pa.s.m -1 ) Specific Acoustic Resistance (Pa.s.m -1 ) Specific Acoustic Resistance (Pa.s.m -1 )

Room correction at low frequencies Hervé Lissek 20

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

Modal Decay Time

Small room S Medium room M1 Medium room M2

1.2 120 1.2 120 1.2 120

1.1 110 1.1 110 1.1 110

1 1 1
100 100 100
0.9 0.9 0.9
Modal Decay Time (s)

Modal Decay Time (s)

Modal Decay Time (s)

90 90 90
0.8 0.8 0.8

Frequency (Hz)

Frequency (Hz)

Frequency (Hz)
0.7 80 0.7 80 0.7 80

0.6 70 0.6 70 0.6 70

0.5 60 0.5 60 0.5 60

0.4 0.4 0.4
50 50 50
0.3 0.3 0.3
40 40 40
0.2 0.2 0.2
0.1 30 0.1 30 0.1 30

0 20 0 20 0 20
16 31.5 63 125 250 500 16 31.5 63 125 250 500 16 31.5 63 125 250 500
Specific Acoustic Resistance (Pa.s.m -1 ) Specific Acoustic Resistance (Pa.s.m -1 ) Specific Acoustic Resistance (Pa.s.m -1 )

Room correction at low frequencies Hervé Lissek 21

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

Modal Decay Time

Small room S Medium room M1 Medium room M2

1.2 120 1.2 120 1.2 120

1.1 110 1.1 110 1.1 110

1 1 1
100 100 100
0.9 0.9 0.9
Modal Decay Time (s)

Modal Decay Time (s)

Modal Decay Time (s)

90 90 90
0.8 0.8 0.8

Frequency (Hz)

Frequency (Hz)

Frequency (Hz)
0.7 80 0.7 80 0.7 80

0.6 70 0.6 70 0.6 70

0.5 60 0.5 60 0.5 60

0.4 0.4 0.4
50 50 50
0.3 0.3 0.3
40 40 40
0.2 0.2 0.2
0.1 30 0.1 30 0.1 30

0 20 0 20 0 20
16 31.5 63 125 250 500 16 31.5 63 125 250 500 16 31.5 63 125 250 500
Specific Acoustic Resistance (Pa.s.m -1 )
-1
Specific
Room Acoustic
S Resistance
Room M1 (Pa.s.m Room
) M2 Specific Acoustic Resistance (Pa.s.m -1 )
160

125

100

80
Frequency (Hz)

63

50

40

31.5

25

20

16
16 31.5 63 125 250 500
Specific Acoustic Resistance (Pa.s.m -1 )

Room correction at low frequencies Hervé Lissek 21

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Identifying target Impedance in 3D

Active Sound Intensity

Mode (1,0,0) with Rsabs = ζρc Sound pressure (color), active sound intensity (gray)

Room correction at low frequencies Hervé Lissek 22

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Design

Method to prescribe frequency-dependent resistances on the EA

Multiple degree-of-freedom target impedance
1
Zstn−DOF (ω) = n
X 1
k=1
Zstk (ω, ν2k−1 , Rst2k−1 , ν2k )

where
Mms 1
Zstk (ω, ν2k−1 , Rst , ν2k ) = jω + Rst +
Sd ν2k−1 jω Sd ν2k Cmc

with conditions
n
X n−1
X
ν1 = νM1 − ν2k−1 and ν2n = νC1 − ν2k
k=2 k=1

Modified sound absorption coefficient

 Zs (f ) − Rs 2
 
α̃(Rs , f ) = 1 −  
Zs (f ) + Rs 

Room correction at low frequencies Hervé Lissek 23

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Performance Optimisation

Room S Room M1 Room M2

160

125

100

80
Frequency (Hz)

63

50

40

31.5

25

20

16
16 31.5 63 125 250 500
Specific Acoustic Resistance (Pa.s.m -1 )

Room correction at low frequencies Hervé Lissek 24

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Performance Optimisation

Room S Room M1 Room M2

160

125

100

80
Frequency (Hz)

63

50

40

31.5

25

20

16
16 31.5 63 125 250 500
Specific Acoustic Resistance (Pa.s.m -1 )

Weighting function of optimal resistances

2 !
Rs − g Rsp (f )
Rsopt (Rs , f ) = (a1 f + b1 ) exp −
2(a2 f + b2 )

Room correction at low frequencies Hervé Lissek 24

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Performance Optimisation

Room S Room M1 Room M2

160

125

100

80
Frequency (Hz)

63

50

40

31.5

25

20

16
16 31.5 63 125 250 500
Specific Acoustic Resistance (Pa.s.m -1 )

Weighting function of optimal resistances

2 !
Rs − g Rsp (f )
Rsopt (Rs , f ) = (a1 f + b1 ) exp −
2(a2 f + b2 )

Optimisation strategy
Z Z
A2 = max (α̃(Rs , f ) − αth , 0) Rsopt (Rs , f ) dRs df

Room correction at low frequencies Hervé Lissek 24

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Performance analysis
Z Z C2 Z Z Z C3
st1 st2 st1 st2 st3
500 500

Magnitude (Pa.s.m-1 )

Magnitude (Pa.s.m-1 )
400 400

300 300

200 200
Zst3−DOF =
Zst2−DOF = 100 100
Zst1 //Zst2
Zst1 //Zst2 0 0
π π //Zst3
(C2)
(C3)
Phase (rad)

Phase (rad)
π/2 π/2

0 0

-π/2 -π/2

-π -π
10 16 31.5 63 125 250 500 10 16 31.5 63 125 250 500
Frequency (Hz) Frequency (Hz)

Room correction at low frequencies Hervé Lissek 25

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Performance analysis
Z Z C2 Z Z Z C3
st1 st2 st1 st2 st3
500 500

Magnitude (Pa.s.m-1 )

Magnitude (Pa.s.m-1 )
400 400

300 300

200 200
Zst3−DOF =
Zst2−DOF = 100 100
Zst1 //Zst2
Zst1 //Zst2 0 0
π π //Zst3
(C2)
(C3)
Phase (rad)

Phase (rad)
π/2 π/2

0 0

-π/2 -π/2

-π -π
10 16 31.5 63 125 250 500 10 16 31.5 63 125 250 500
Frequency (Hz) Frequency (Hz)

C0 C1 C2 C3
500
Magnitude (Pa.s.m-1 )

400

300

200

100

0
π
Phase (rad)

π/2

-π/2

-π
10 16 31.5 63 125 250 500
Frequency (Hz)

Room correction at low frequencies Hervé Lissek 25

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Performance analysis
Z Z C2 Z Z Z C3
st1 st2 st1 st2 st3
500 500

Magnitude (Pa.s.m-1 )

Magnitude (Pa.s.m-1 )
400 400

300 300

200 200
Zst3−DOF =
Zst2−DOF = 100 100
Zst1 //Zst2
Zst1 //Zst2 0 0
π π //Zst3
(C2)
(C3)
Phase (rad)

Phase (rad)
π/2 π/2

0 0

-π/2 -π/2

-π -π
10 16 31.5 63 125 250 500 10 16 31.5 63 125 250 500
Frequency (Hz) Frequency (Hz)

C0 C1 C2 C3 C0 C1 C2 C3
500 100
Magnitude (Pa.s.m-1 )

Real Part (Pa.s.m-1 )

400 80

300 60

200 40

100 20

0 0
Imaginary Part (Pa.s.m-1 )

π 200

100
Phase (rad)

π/2

0 0

-π/2 -100

-π -200
10 16 31.5 63 125 250 500 10 16 31.5 63 125 250 500
Frequency (Hz) Frequency (Hz)

Room correction at low frequencies Hervé Lissek 25

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Modified Sound Absorption Coefficient

Basic
config.
(C0)

C0
160

125

100
Frequency (Hz) 80

63

50

40

31.5

25

20

16
16 31.5 63 125 250 500
-1
Specific Acoustic Resistance (Pa.s.m )

Room correction at low frequencies Hervé Lissek 26

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Modified Sound Absorption Coefficient

Basic Zst1−DOF
config. (C1)
(C0)

C0 C1
160

125

100
Frequency (Hz) 80

63

50

40

31.5

25

20

16
16 31.5 63 125 250 500
-1
Specific Acoustic Resistance (Pa.s.m )

Room correction at low frequencies Hervé Lissek 26

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Modified Sound Absorption Coefficient

Basic
config. Zst1−DOF
(C0) (C1)

C0 C1 C2
160

125

100
Frequency (Hz) 80

63
Zst2−DOF
50
(C2) 40

31.5

25

20

16
16 31.5 63 125 250 500
-1
Specific Acoustic Resistance (Pa.s.m )

Room correction at low frequencies Hervé Lissek 26

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Optimisation of Multiple Degree-of-Freedom Target Impedance

Modified Sound Absorption Coefficient

Basic
config. Zst1−DOF
(C0) (C1)

C0 C1 C2 C3
160

125

100
Frequency (Hz) 80

63
Zst3−DOF
50
(C3) 40

31.5

25

20

16
16 31.5 63 125 250 500
-1
Specific Acoustic Resistance (Pa.s.m )

Room correction at low frequencies Hervé Lissek 26

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

1 Electroacoustic absorbers design

Electroacoustic absorbers: general principle
Electroacoustic absorbers: hybrid sensor-/shunt-based control
Performance assessment in 1D Case

2 From 1D to 3D
Identifying target Impedance in 3D
Optimisation of Multiple Degree-of-Freedom Target Impedance

3 In situ performance Evaluation

Frequency responses
Modal decay times
Spatial distribution of pressure

Room correction at low frequencies Hervé Lissek 27

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Frequency responses

Experimental Setup: Small Room

floor area = 20.04 m2 , volume = 42.69 m3

Room correction at low frequencies Hervé Lissek 28

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Frequency responses

Experimental Setup: Medium Room

floor area = 50.05 m2 , volume = 174.18 m3

Room correction at low frequencies Hervé Lissek 29

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Frequency responses

Modal Equalisation: Small Room

4 EA off 2 EA off - 2EA on 4 EA on 4 EA off 2 EA off - 2EA on 4 EA on

15 20

10 15
Magnitude (dB re. 1 Pa.s.m-1 )

Magnitude (dB re. 1 Pa.s.m-1 )

5 10

0 5

-5 0
Mic #1 -10 -5
Mic #2
-15 -10

-20 -15

-25 -20

-30 -25
20 40 60 80 100 120 140 20 40 60 80 100 120 140
Frequency (Hz) Frequency (Hz)

4 EA off 2 EA off - 2EA on 4 EA on 4 EA off 2 EA off - 2EA on 4 EA on

5 5

0 0

-5
Magnitude (dB re. 1 Pa.s.m-1 )

Magnitude (dB re. 1 Pa.s.m-1 )

-5
-10
-10
Mic #3 -15
-15 Mic #4
-20
-20
-25
-25
-30
-30 -35
-35 -40

-40 -45
20 40 60 80 100 120 140 20 40 60 80 100 120 140
Frequency (Hz) Frequency (Hz)

Room correction at low frequencies Hervé Lissek 30

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Frequency responses

Modal Equalisation: Medium Room

4 EA off 2 EA off - 2EA on 4 EA on 4 EA off 2 EA off - 2EA on 4 EA on

15 20

10 15
Magnitude (dB re. 1 Pa.s.m-1 )

Magnitude (dB re. 1 Pa.s.m-1 )

5 10

0 5

-5 0
Mic #1 -10 -5
Mic #2
-15 -10

-20 -15

-25 -20

-30 -25
20 40 60 80 100 120 140 20 40 60 80 100 120 140
Frequency (Hz) Frequency (Hz)

4 EA off 2 EA off - 2EA on 4 EA on 4 EA off 2 EA off - 2EA on 4 EA on

15 10

10 5
Magnitude (dB re. 1 Pa.s.m-1 )

Magnitude (dB re. 1 Pa.s.m-1 )

5 0

0 -5
Mic #3 -5 -10
Mic #4
-10 -15

-15 -20

-20 -25

-25 -30

-30 -35
20 40 60 80 100 120 140 20 40 60 80 100 120 140
Frequency (Hz) Frequency (Hz)

Room correction at low frequencies Hervé Lissek 31

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Modal decay times

Evaluation of Modal Decay Times with RFP

Could be directly measured with sine excitations at resonance frequencies
(provided they have been carefully identified)
Reverberation chamber:

Room correction at low frequencies Hervé Lissek 32

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Modal decay times

Evaluation of Modal Decay Times with RFP

Could be directly measured with sine excitations at resonance frequencies

(provided they have been carefully identified)
Reverberation chamber:

Room correction at low frequencies Hervé Lissek 32

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Modal decay times

Evaluation of Modal Decay Times with RFP

Modal decay time of mode n
3 ln(10) Qn
MT60n =
π f0n
But in real rooms...

M. H. Richardson, D. L. Formenti, ”Global curve fitting of frequency response measurements using the rational
fraction polynomial method.” Proceedings of the Third International Modal Analysis Conference, 1985.
Room correction at low frequencies Hervé Lissek 32
Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Modal decay times

In situ evaluation of Modal Decay Times

4 EA off 2 EA off - 2 EA on 4 EA on
0.7 0.7

0.6 0.6

0.5 0.5
MT60 (s)

MT60 (s)
0.4 0.4

Small room 0.3 0.3

0.2 0.2

0.1 0.1

0 0
4 EA off 2 EA off - 2 EA on 4 EA on
.0

.4

.0

.6

.8

.1

.7

.2

.6

.5

7
2.
29

40

48

55

60

74

77

85

90

98
Configuration

10
Frequency (Hz)

4 EA off 2 EA off - 2EA on 4 EA on

1.8 1.8

1.6 1.6

1.4 1.4

1.2 1.2
Medium room
MT60 (s)

MT60 (s)
1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
4 EA off 2 EA off - 2 EA on 4 EA on
27 3
34 3
40 2
45 8
48 7
50 5
53 3
56 1
58 8
62 6
63 2
67 6
68 4
72 6
76 6
79 4
81 0
85 5
88 7
92 3
96 1
97 2
.9
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
21

Configuration
Frequency (Hz)

Room correction at low frequencies Hervé Lissek 33

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Is spatial sound pressure distribution accessible...?

• Target: Spatial evaluation of sound pressure field with and without absorbers
• Framework:
• Input: sound pressure measurement at a limited number of microphone
positions
• Output: visualize/render sound pressure distribution within the whole
room
• Challenges:
• Limited number of actual microphones
• If possible, limited computational requirements

Room correction at low frequencies Hervé Lissek 34

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

... through Compressed Sensing?

Compressed sensing1
• Reconstruction of a signals using far fewer samples than required by Nyquist
theorem, given that the signal is sparse
• In acoustics: Reconstruction of the whole sound field in room using a much less
number of measurements.
• Sparsity of sound field in room
• Modal summation
• Approximation of Modeshape function Ψn using plane waves summation

1
Mignot et al., Low frequency interpolation of room impulse responses using compressed sensing, IEEE/ACM
Trans. on Audio, Speech, and Language Processing, 22(1) (2014)
Room correction at low frequencies Hervé Lissek 35
Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
dj (t) = e jωi t .e −δi t
5 Project the measured signals on the dictionary
6 Select the column n1 (corresponding to mode ωn1 ,δn1 )
with the highest energy
7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
dj (t) = e jωi t .e −δi t
5 Project the measured signals on the dictionary
6 Select the column n1 (corresponding to mode ωn1 ,δn1 )
with the highest energy
7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
M =
dj (t) = e jωi t .e −δi t m1 (t1 ) m1 (t2 ) ... m1 (tNt )

 m2 (t1 ) m2 (t2 ) ... m2 (tNt ) 
5 Project the measured signals on the dictionary 
 . . . .


mM (t1 ) mM (t2 ) ... mM (tNt )
6 Select the column n1 (corresponding to mode ωn1 ,δn1 )
with the highest energy
7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
 D =
dj (t) = e jωi t .e −δi t d1 (t1 ) d2 (t1 ) ... dN (t1 )

 d1 (t1 ) d2 (t2 ) ... dN (t2 ) 
5 Project the measured signals on the dictionary 
 . . . .


d1 (tNt ) d2 (tNt ) ... dN (tNt )
6 Select the column n1 (corresponding to mode ωn1 ,δn1 )
with the highest energy
7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
M ·D =
dj (t) = e jωi t .e −δi t 
a11 a12 ... a1N

 a21 a22 ... a2N 
5 Project the measured signals on the dictionary 
 . . . .


6 Select the column n1 (corresponding to mode ωn1 ,δn1 ) aM1 aM2 ... aMN

with the highest energy

7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
dj (t) = e jωi t .e −δi t
 
a1n1
 a2n1 
5 Project the measured signals on the dictionary
 
 . 
aMn1
6 Select the column n1 (corresponding to mode ωn1 ,δn1 )
with the highest energy
7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
dj (t) = e jωi t .e −δi t
5 Project the measured signals on the dictionary
6 Select the column n1 (corresponding to mode ωn1 ,δn1 )
with the highest energy
7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
dj (t) = e jωi t .e −δi t
5 Project the measured signals on the dictionary
6 Select the column n1 (corresponding to mode ωn1 ,δn1 )
with the highest energy
7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: identifying the resonance frequencies

1 Measure a set of FRFs Hm (ω) (wrt a reference signal,

eg. source velocity) at M microphone positions
2 Get the RIRs through inverse Fourier Transforms
hm (t) = F −1 {Hm (ω)}
3 Build a matrix of M vectors of size Nt
4 Define a database of N eigenvalues (ωi ,δi ) within
limit values
Build a matrix (dictionary) of modes, such as
dj (t) = e jωi t .e −δi t
 
ωn 1 + jδn1
 ωn + jδn2 
2
5 Project the measured signals on the dictionary
 
 . 
ωnL + jδnL
6 Select the column n1 (corresponding to mode ωn1 ,δn1 )
with the highest energy
7 Subtract each RIR hm (t) with the projected vector
amn1 e jωn1 t .e −δn1 t
8 Repeat the operation until a set number of modes L
(should be lower than M the number of microphones)
9 The outcome of this phase is the identification of L
eigenfrequencies (ωni + jδni )

Room correction at low frequencies Hervé Lissek 36

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: trying to recover the mode shapes

Making the assumption of plane wave decomposition:

~ ~
αni ,r e −kni ,r ·Xm e jkni ct
X
p(t, X~m ) ≈
ni ,r ∈N2
The second sparsity comes from the fact that each of the mode shapes can be
approximated by a finite sum of plane waves coming from different direction (but with
the same wave number ∀r , ||kn~i ,r || = kni ).
1 For each mode ni , we build a set of r eigenvectors
within a sphere of radius kni
2 Projecting all M microphone (with known positions
Xm ) measurements on all r wavevectors kni ,r
3 LMS minimization allows retrieving the most likely
wavevector kn~i ,r

Room correction at low frequencies Hervé Lissek 37

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: trying to recover the mode shapes

Making the assumption of plane wave decomposition:

~ ~
αni ,r e −kni ,r ·Xm e jkni ct
X
p(t, X~m ) ≈
ni ,r ∈N2
The second sparsity comes from the fact that each of the mode shapes can be
approximated by a finite sum of plane waves coming from different direction (but with
the same wave number ∀r , ||kn~i ,r || = kni ).

1 For each mode ni , we build a set of r eigenvectors

within a sphere of radius kni
2 Projecting all M microphone (with known positions
Xm ) measurements on all r wavevectors kni ,r
3 LMS minimization allows retrieving the most likely
wavevector kn~i ,r

Room correction at low frequencies Hervé Lissek 37

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: trying to recover the mode shapes

Making the assumption of plane wave decomposition:

~ ~
αni ,r e −kni ,r ·Xm e jkni ct
X
p(t, X~m ) ≈
ni ,r ∈N2
The second sparsity comes from the fact that each of the mode shapes can be
approximated by a finite sum of plane waves coming from different direction (but with
the same wave number ∀r , ||kn~i ,r || = kni ).

1 For each mode ni , we build a set of r eigenvectors

within a sphere of radius kni
2 Projecting all M microphone (with known positions
Xm ) measurements on all r wavevectors kni ,r
3 LMS minimization allows retrieving the most likely
wavevector kn~i ,r

Room correction at low frequencies Hervé Lissek 37

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing framework: trying to recover the mode shapes

Making the assumption of plane wave decomposition:

~ ~
αni ,r e −kni ,r ·Xm e jkni ct
X
p(t, X~m ) ≈
ni ,r ∈N2
The second sparsity comes from the fact that each of the mode shapes can be
approximated by a finite sum of plane waves coming from different direction (but with
the same wave number ∀r , ||kn~i ,r || = kni ).

1 For each mode ni , we build a set of r eigenvectors

within a sphere of radius kni
2 Projecting all M microphone (with known positions
Xm ) measurements on all r wavevectors kni ,r
3 LMS minimization allows retrieving the most likely
wavevector kn~i ,r

Room correction at low frequencies Hervé Lissek 37

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing reconstruction: results so far

• Goal: sound field reconstruction (in thereverberation chamber)

• Setup: 50 randomly placed microphones measurements in the room
• For now: COMSOL simulations (as ground truth) and validation of compressed
Sensing interpolation out of a limited number of microphones

FEM modeling (40.88 Hz)

CS result (40.88 Hz)

Room correction at low frequencies Hervé Lissek 38

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Spatial distribution of pressure

Compressed Sensing reconstruction: results so far

• Goal: sound field reconstruction (in thereverberation chamber)

• Setup: 50 randomly placed microphones measurements in the room
• For now: COMSOL simulations (as ground truth) and validation of compressed
Sensing interpolation out of a limited number of microphones

FEM modeling (66.09 Hz) CS result (66.09 Hz)

Room correction at low frequencies Hervé Lissek 38

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Conclusion

Summary of results

• Design guidelines for setting electroacoustic absorbers

• Impedance control through multiple degree-of-freedom resonators
• Hybrid sensor-/shunt-based impedance control

• Performance optimisation
• Definition of new performance metrics
• Methodology for performance improvement

• Performance evaluation for modal equalisation

• Development of demonstration prototypes
• Objective evaluation in real rooms
• Subjective evaluation of music playback in rooms (not presented here)

Room correction at low frequencies Hervé Lissek 39

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Conclusion

Perspectives

• Absorber design
• Ratio surface/mass
• Active panels controlled by electrodynamic inertial actuators
• Development of electroacoustic absorption-diffusion systems

• Performance optimisation
• Target impedance (reactive terms in simulations)
• Calibration of electroacoustic absorbers directly in the room

• Performance evaluation
• Spatial sound field reconstruction (compressive sensing)

Room correction at low frequencies Hervé Lissek 40

Introduction Electroacoustic absorbers design From 1D to 3D In situ performance Evaluation Conclusion

Conclusion

Thank you for your Attention

Thanks to: Etienne Rivet, Sami Karkar, Romain Boulandet,

and Thach Pham Vu

Most of the presented results are taken from

Etienne Rivet PhD Thesis No. 7166 (open access)

Open access paper:

E. Rivet et al, Acta Acustica, 103(6), pp. 1025-1036 (2017)

View publication stats

Room correction at low frequencies Hervé Lissek 41