Audio System Evaluation with Music Signals - aes.org · Audio System Evaluation with Music Signals,...

Audio System Evaluation with Music Signals, 1

Audio System Evaluation

with

Music Signals

Stefan Irrgang,

Wolfgang Klippel

KLIPPEL GmbH


Motivation

• Field rejects are $$$

• Reproduce + analyse the problem before repair


Audio System Evaluation

over the product life cycle

DefinitionTarget

Performance

Target Application Condition

DefinitionTarget

Performance

ProductSpecification

Standard Measurement Condition


DefinitionTarget

Performance


Development(components,

system)

Physical + PerceptualAssessment

PhysicalEvaluation



DefinitionTarget

Performance



system)


PhysicalEvaluation

EOLTesting



Production

DefinitionTarget

Performance



system)


PhysicalEvaluation

EOLTesting

FieldMonitoring

Service



Production


After Sales / Field Measurements

Acoustical measurements using (available) microphones reveal

- Frequency Response, SPL

- Overall transfer function

- Coherence

- Residual Distortion

Electrical measurements based on voltage and current monitoring

(control technology) reveal

- Voice coil and magnet temperature

- Peak displacement and offset in coil rest

position

- Change of suspension stiffness indicating

fatigue

- anticipation of a motor or suspension defect

DSP~audio

signal

amplifier

u, i

microfone

wooferparasitic vibration

u, i

amplifier

internal

sound

sourcesambient

noise

Challenges - audio signal used

- influence of other sound sources

- use of compromised test elements

- measurement shall be realized at low

cost

Objectives - evaluation of the performance as seen

by the end user

- long term testing under real condition

- Prevention of a total failure

- Service for customer complaint


What is a critical defect?

• Related to customer complaints

• Observable in in-situ condition Impulsive distortion (panel buzzing, loose

particles, loose electrical connection) Significant air noise caused by a leakage of

the enclosure (Subwoofer) Excessive nonlinear distortion caused by

motor instability and severe asymmetries


Input

Signal

Output

Signal

Linear Model

H(s)-1

Nonlinear

Model

Noise

u(t) p(t)

dn(t)

n(t)

dl(t)

Linear

Distortion

Nonlinear

Distortion Large Signal

Performance

(motor, suspension)

Undesired Defects • Rubbing coil, Buzzing parts

• Wire beat

• Loose particles, air leak noise

• Parasitic vibration of other components

Desired Small

Signal Performance

Reproduced Sound Quality Generation of Signal Distortion in an Audio System

Irregular

Distortion

Residual

Distortion

StimulusMeasured

Signal


Properties of Music Stimulus

• Dense stimulus / Most complex

• Non persistant excitation

• Non stationary excitation

(Unknown time structure)

• Defects occur quasi randomly

How to separate defect from

accepted response?


State of the art Correlation Technique

Stimulus

FFT

DUT

Y(f,n)

Syy(f,n)

),(ˆ nfSyy

FFT

MA MA

()*

H(f,n)IC(f,n)

Sxy(f,n)

Calculation

MA

Sxx(f,n)

X(f,n)

()*

)(ty

),(ˆ nfSxx ),(ˆ nfSxy

HWHW

Y*(f,n)

)(tx

Linear System Identification:

• Correlation Technique

• Requires Stationary Signals

),(ˆ

),(ˆ),(

,

,

nfS

nfSnfH

xx

yx

Incoherence:

• Power metric

• Good for Noise problems

• Moderate for regular non-linearities

• Poor for time varying TRF

• Poor for impulsive distortion

%100),(ˆ),(ˆ

),(ˆ

),(

,,

,

nfSnfS

nfSnf

yyxx

yx

dB%100

),(1log10),(

nfnfICdB


State of the art Correlation Technique

Incoherence:

• Requires long (statistical!) time for accurate estimation (see paper) for low

energy symptoms.

• T is defined by required resolution (Woofer 1 Hz / Midrange 5 Hz)

• K (number of processed blocks) = f(error)

• Example: df = 1Hz, total time =5s

error: 30%, Incoherence too low by 1.5dB

• Assumption: stationary excitation

• When using non-stationary signals: Symptoms may be not stable!

• No Auralization of separated distortion (power measure!)

KfIC

fICfeIC

%100

)(

)()(

T

KTtot

2


State of the art Correlation Technique / Measurement Results

KLIPPEL

10

10 2 10 3 10 4

Incoherence

Frequency [Hz]

Defect SetupNormal

Setup

1

Frequency [Hz]Frequency [Hz]

IC [%

]

100

Distortion from regular nonlinearities mask symptoms from rattling

Incoherence measurement is not sensitive for impulsive distortion


Acoustical In-Situ Measurement

Linear Distortion

Audio Source

DUT

p

Audio Source

DUT

-

Physical Characteristics

AdaptiveLinear

Modelling plin

p

Linear Distortion

Adaptive Identification of Linear System

Properties:

• Adaptive identification

• Fast initial learning

• Copes with non

stationary input

• Follows changes of

linear system

Improved separation

• Typical ANC method


Short Term Response (1 s)

Compression of SPL Output

Long term response

measured after

applying the

sinusoidal stimulus

for 1 min

Sound Pressure Response

20 50 200 500 2k

Frequency [Hz]

80

85

90

95

100

105

110

115

120

125

130

dB

Long Term Response (1 min)

Short term response

measured within 1 s

(without voice coil

heating)


KLIPPEL

50

60

70

80

90

100

110

120

130

140

150

160

170

0 500 1000 1500 2000 2500 3000 3500

resonance frequency fs (t) at rest position X=0

Kms [N

/mm]

t [sec]

fs (X=0)

Hz

Influence of Ambient Conditions

More than one octave shift

KLIPPEL

0

25

50

75

100

125

150

0

1

2

3

4

5

6

7

8

9

10

11

12

0 500 1000 1500 2000 2500 3000 3500

Delta

Tv [K

]

P [W]

t [sec]

Delta Tv Tambient P real P Re Pnom

Winter

Sommer

Ambient temperature

-30° C

80° C

Influence to the Loudspeaker System Alignment

Properties of the mechanical suspension depend on humidity, temperature

voice coil displacement cannot be predicted by time-invariant parameters

adaptive learning process required


Acoustical In-Situ Measurement Residual Distortion

Audio Source

DUT

-

dr


Adaptive

Linear

Modelling

Residual

Distortion

plin

p

Linear

Distortion

Properties:

• Model identifies linear

system only Separation

of residual from linear

distortion

• Exploit Residual:

• Non-linear Distortion

• Defect Distortion

• Noise


Assessing Regular Nonlinear and Irregular

Distortion

Impact on sound quality:

• Parasitic resonances are excited by

sufficient (kinetic) energy

• Defect occurs rarely

• Depends on

• Level (not highest!)

• Phase relationship

• Distortion require high state (e.g. Bl(x))

KLIPPEL

0.00

0.25

0.50

0.75

1.00

1.25

-3 -2 -1 0 1 2 3

PD

F [

1/m

m]

Displacement [mm]

Proabilitiy densitiy function (pdf)

sine tone

music

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

-5 -4 -3 -2 -1 -0 1 2 3 4 5

Bl [N

/A]

<< Coil in X [mm] coil out >>

Nonlinear Force Factor

More comprehensive

then sine sweep!


Audio system defect: Parasitic Vibration problem

Most defects behave as a nonlinear oscillator

• active above a critical amplitude

• new mode of vibration

• powered and synchronized by stimulus

Loose joint

(Nonlinearity)

parasitic resonator

Externally excited

spring

mass

Loose joint

(Nonlinearity)

parasitic resonator

Externally excited

spring

mass

time one period

vibration

distortion signal


Residuum of the Linear Modeling (defect in car)

KLIPPEL

-0.10

-0.05

0.00

0.05

0.10

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Time [s]

Mic

. V

oltag

e

[V]

tambourine measured signal p(t)

modelled signal plin(t)

residuum e(t)

+10dB (!)

next


Time Frequency Analysis of the Residuum

KLIPPEL

210

310

410

0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fre

qu

en

cy

[Hz]

Time [s]

-74

-71

-68

-65

-62

-59

-56

-53

-50

-47

-44

-41

-38

-35

0 0.5 1.0 1.5Time [s]

Fre

q. [H

z]

100

1k

10k-35

-38

-41

-44

-47

-50

-53

-65

-68

<-68

-56

-59

-62

dBFS KLIPPEL

210

310

410

0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fre

qu

en

cy

[Hz]

Time [s]

-74

-71

-68

-65

-62

-59

-56

-53

-50

-47

-44

-41

-38

-35

0 0.5 1.0 1.5Time [s]

Fre

q. [H

z]

100

1k

10k-35

-38

-41

-44

-47

-50

-53

-65

-68

<-68

-56

-59

-62

dBFS

Defect

defective car (with tambourine)

Normal car (without tamborine)

KLIPPEL

0.00

60 62 64 66 68 70

V

Time [ms]

Normal car

Defective car

-0.02

0.01

0.02

-0.01

residuum in time domain

next


Standard EOL test of defect

KLIPPEL

210

310

410

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Fre

qu

en

cy

[Hz]

Time [s]

-95

-92

-89

-86

-83

-80

-77

-74

-71

-68

-65

-62

-59

-56

-53

-50

-47

-44

-41

-38

-35

0 0.5 1.0Time [s]

Fre

q. [H

z]

100

1k

10k

dBFS-35

-41

-47

-53

-65

-71

-59

-77

-83

-89

KLIPPEL

210

310

410

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Fre

qu

en

cy

[Hz]

Time [s]

-95

-92

-89

-86

-83

-80

-77

-74

-71

-68

-65

-62

-59

-56

-53

-50

-47

-44

-41

-38

-35

0 0.5 1.0Time [s]

Fre

q. [H

z]100

1k

10k-35

-41

-47

-53

-65

-71

-59

dBFS

-77

-83

-89

defective car (with tambourine)

Normal car (without tamborine)

So

und

Pre

ssure

[d

B]

Frequency [Hz]

KLIPPEL

30

40

50

60

70

80

90

100

110

101

102

103

104

Defect

Pass/Fail Limit

Fundamental

Rub&Buzz

Using log sweep


Audio Source

DUT

-


AdaptiveLinear

Modelling plin

p

Linear Distortion

Time-Variant

Distortion

Adaptive Identifcation of slowly changing

linear system

Acoustic In-Situ Measurements Auralization

pAAudio Source

DUT

- dr

Sdis


Listening

Test

Perceptual Characteristics

Auralization

plin

p

Linear

Distortion

Perceptional

Modeling

Adaptive

Linear

Modelling

Residual

Distortion

Adaptive Identification

provides:

• Separated linear

response

• Purified from time variant

linear distortion

• Residual Distortion to be

scaled up and down

Benefit:

• Objective Criteria

• Rate importance

• Is this defect acceptable?

• Action required?


Finding most critical section

KLIPPEL

-30

-25

-20

-15

-10

-5

0

5

50 75 10025 150

Distortion ratio

dB

re

l.

Time s

mean

distortion

peak

distortion

RMS

125

Challenge:

• Find highest distortion in a

long recording

• Non-Stationary excitation

Solution:

• Check Crest factor of

Residual:

High Peak and Low Mean

distortion indicate

impulsive distortion

(defect symptom)

Select Best Section


Summary

• Simple Setup – Wave file / Web based solutions

• Comprehensive System Evaluation – Linear Response + Residual distortion

• Using any stimulus – Customer specific audio material

• Fast Identification of defects – Automatic selection of critical section

– Time Domain Analysis

– Auralization

Audio System Evaluation with Music Signals - aes.org · Audio System Evaluation with Music Signals,...

Documents

Transcript of Audio System Evaluation with Music Signals - aes.org · Audio System Evaluation with Music Signals,...