active noise control : study of the · pdf fileoutpass microphone signal micsig active output...
TRANSCRIPT
Event title (to be altered on title master)Presenter name – presenter job title (to be altered on title master)
Convergence speed, Latency time and Implementation on a boardActive Noise Control : Study of the algorithm
Paris – December, 15th 2009Marco Gallo – LMS International - [email protected]
Marco Gallo – Personal Presentation
� Born in Asti 26/06/1984
� B.Sc & M.Sc. in Automotive Engineering at Politecnico di Torino, 2008
� Several projects during M.Sc:
� Team member of Squadra Corse (2006-2008), Formula SAE team of Politecnico di Torino, as responsible of design and production of a dry sump system
� Internship @ University of Surrey, Guildford, Uk
� Marie Curie fellowship from July 2009 in the frame SMART STRUCTURES for LMS International (12 months)
� Main Research Topic: Active Car Noise Control
� Relational Skills: I enjoy team-working. I was 3 years responsible of the youth activities in San Damiano d’Asti, Asti, Italy. I played in many football team; my hobbies are football, reading, gardening, oenology.
2 - copyright LMS International - 2009
Agenda
2 Control Strategy - FxLMS
1 Introduction
3 FxLMS - Results
4
5
6
1 Introduction
Board - Processor
Control Strategy – NEX LMS
NEX LMS - Results
3 - copyright LMS International - 2009
0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0- 1
- 0 . 5
0
0 . 5
1
1 . 5
2
2 . 5
3
3 . 5
4
T i m e
Noi
se P
ress
ure
Am
plitu
de
U n w a n t e d N o is eA n t i - N o i s e S o u n dS u m m a t io n
• Sound reduction with destructive interference (Young’s principle)
• Obtained by the superposition of the original (primary) sound field with an artificially generated (secondary) sound field
Introduction – Young’s Principle
4 - copyright LMS International - 2009
Introduction – Lueg Patent, 1936
Problem of cancelling sound in a duct
Problem of cancelling sound in an area
Noise attenuation in an open space
5 - copyright LMS International - 2009
Agenda
2 Control Strategy - FxLMS
1 Introduction
3 FxLMS - Results
4
5
6
2 Control Strategy - FxLMS
Board - Processor
Control Strategy – NEX LMS
NEX LMS - Results
6 - copyright LMS International - 2009
Control Strategy - Algorithm
Adaptive Feed Forward Strategy ( Fx-LMS)
• Aim : Achieve a pre-defined order level vs RPM profile, achieve a desired sound quality
• Controller should work with varying engine speeds, disturbance• Controller should have a fast convergence speed• The adaptive algorithm adjust the coefficients of the digital filter
w(n) = [w0(n) w1(n)….. wL-1(n)]T
• Steepest descent method, to minimize a quadratic cost function
7 - copyright LMS International - 2009
W S-Function
FIR
Secondary PathModel
Sid
1
Secondary Path
y2'
1
Scope
S-Function
FXLMSReference
Signal
RefSig
Passive Output
OutPass
Microphone Signal
MicSig
Active Output
OutActDigital FilterSecondary Path
Adaptive Algorithm
Engine Sound
W S-Function
FIR
Secondary PathModel
Sid
1
Secondary Path
y2'
1
Scope
S-Function
FXLMSReference
Signal
RefSig
Passive Output
OutPass
Microphone Signal
MicSig
Active Output
OutAct
Adaptive Algorithm
Digital FilterSecondary Path
Control Strategy - Algorithm
)()()()1( nenxnwnw ⋅⋅+=+ µ
xx LPLP
1.001.0 << µ
• LMS algorithm for updating the coefficients of W:
• x(n): Reference Signal RPM
• e(n): Error Signal
• µ: Convergence Coefficient
8 - copyright LMS International - 2009
Engine Sound
Agenda
2 Control Strategy - FxLMS
1 Introduction
3 FxLMS - Results
4
5
6
3 FxLMS - Results
Board - Processor
Control Strategy – NEX LMS
NEX LMS - Results
9 - copyright LMS International - 2009
FxLMS - Results
• Input: constant amplitude sinusoid
• Secondary path model : two poles
• Fast convergence near the secondarypath resonances
0 1 2 3 4 5 6 7 8 9 10-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time [s]
Am
plitu
de
66 Hz
83 Hz40 Hz
0 1 2 3 4 5 6 7 8 9 10-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time [s]
Am
plitu
de
100 Hz
116 Hz133 Hz
0 1 2 3 4 5 6 7 8 9 10-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time [s]
Am
plitu
de
150 Hz
166 Hz200 Hz
FxLMS - Results
11 - copyright LMS International - 2009
40
60
80
100
120
140
160
180
200
0
1
2
3
4
5
6
7
8
9
10
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time [s]
Freq [Hz]
Am
plitu
de
Agenda
2 Control Strategy - FxLMS
1 Introduction
3 FxLMS - Results
4
5
6
4 Control Strategy – NEX LMS
Board - Processor
Control Strategy – NEX LMS
NEX LMS - Results
12 - copyright LMS International - 2009
Control Strategy – NEX LMS
Reference Signal(Engine rpm)
Look up tableNormalization Filter
Engine sound
Equalization
13 - copyright LMS International - 2009
Agenda
2 Control Strategy - FxLMS
1 Introduction
3 FxLMS - Results
4
5
6
5 NEX LMS - Results
Board - Processor
Control Strategy – NEX LMS
NEX LMS - Results
14 - copyright LMS International - 2009
NEX LMS - Results
0 1 2 3 4 5 6 7 8 9 10-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Time [s]
Am
plitu
de
80 Hz
166 Hz
133 Hz
0 1 2 3 4 5 6 7 8 9 10-1.5
-1
-0.5
0
0.5
1
Time [s]
Am
plitu
de
200 Hz
233 Hz
266 Hz
0 1 2 3 4 5 6 7 8 9 10-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time [s]
Am
plitu
de
266 Hz
333 Hz
400 Hz
Fast convergence for each frequencies!!
15 - copyright LMS International - 2009
NEX LMS - Results
50
100
150
200
250
300
350
400
0
1
2
3
4
5
6
7
8
9
10
-3
-2
-1
0
1
2
3
Time [s]Freq [Hz]
Am
plitu
de
16 - copyright LMS International - 2009
NEX LMS - Results
• For β=0 the residual noise is the unchanged primary disturbance
• For β=1: maximum reduction
• For β=+0.5: half reduction
• For β<0 the system amplifies the original disturbance.
)()()()()( * nyndnyndne β−=−=
0 1 2 3 4 5 6 7 8 9 10-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Time [s]
Am
plitu
de
Beta = -1Beta = -0.5
Beta = 0
Beta = +0.5Beta = +1
)()1()( ndne β−≈
)()(
)()()(
wewd
wewdw
r
d
−−=β
)(
)(1)(
wd
wew d−≈β
17 - copyright LMS International - 2009
Agenda
2 Control Strategy - FxLMS
1 Introduction
3 FxLMS - Results
4
5
6
6 Board - Processor
Board - Processor
Control Strategy – NEX LMS
NEX LMS - Results
18 - copyright LMS International - 2009
Latency time study
19 - copyright LMS International - 2009
Latency time study
European Passenger CarDriver Front Passenger
Average Latency [ms] 16.50 16.70
Max Value [ms] 21.52 21.98
Min Value [ms] 9.89 10.64
Monovolume Car
Driver
Average Latency [ms] 12.64
Max Value [ms] 18.71
Min Value [ms] 9.17
American Passenger CarDriver Rear Passenger
Average Latency [ms] 11.87 11.79
Max Value [ms] 18.10 16.41
Min Value [ms] 7.05 8.73
20 - copyright LMS International - 2009
Board for audio processing
PNX1005 Series- power and cost optimized- tm3282 350-400Mhz VLIW core- integrated USB2.0 OTG- 2x quad-multiplexed video input- HD 1080p/60 video in and out
Applications- 8 channel CIF H264 DVR
- 2xD1 H264 or 720p H264 encode
- H264 SD IP Camera with analytics
- Picture optimization
21 - copyright LMS International - 2009
aNext Step
• Implementation of the algorithm (WIP)
• Test in the LMS research car
• MIMO implementation
• Implement adaptive feedforward control system in VCS
• Secondment @ VTT, Finland
• Based on :L. De Oliveira, 2009, “Active sound quality control: design tools and automotive applications”, Ph.D. thesis, K.U. Leuven, Fac. Eng., Dept. Mechanical Eng.
22 - copyright LMS International - 2009
Event title (to be altered on title master)Presenter name – presenter job title (to be altered on title master)
Thank you
Paris – December, 15th 2009Marco Gallo – LMS International - [email protected]