Rub & Buzz and Other Irregular Loudspeaker ... - Klippel GmbH

Rub & Buzz and Other Irregular Loudspeaker Distortion 

Product Design Session PD11 

133rd AES Convention 2012 

by Wolfgang Klippel, 

KLIPPEL GmbH 

Institute of Acoustics and Speech Communication 

Dresden University of Technology 

Klippel, Rub & Buzz and Other Irregular Loudspeaker Distortion, 1

Abstract 

Loudspeaker defects caused by manufacturing, aging, overload or climate 

impact generate special kind of irregular distortion commonly known as 

rub & buzz which are highly audible and intolerable for the human ear. 

Contrary to regular loudspeaker distortions defined in the design process 

the irregular distortions are hardly predictable and are generated by an 

independent process triggered by the input signal. Traditional distortion 

measurements such as THD fail in the reliable detection of those defects. 

The Tutorial discusses the most important defect classes, new 

measurement techniques, audibility and the impact on perceived sound 

quality. 


Problems addressed here: 

• What is Rub & Buzz ? 

• What kinds of symptoms are generated by irregular 

loudspeaker defects ? 

• Why do straightforward measurements fail in 

detecting those defects ? 

• Do we need a measurement more sensitive than the 

human ear ? 

• How to cope with ambient production noise ? 

• How to find the root cause and how to fix the 

problem ? 


Desired and Undesired Components ? 

Generation of Signal Distortion in an Audio System 

Input 

Signal 

Desired Small 

Signal Performance 

Desired Large 

Signal Performance 

Stimulus 

Linear Model 

Nonlinear 

Model 

Unpredictable 

Dynamics 

Regular linear 

distortion 

Regular nonlinear 

distortion 

Undesired Loudspeaker Defects 

• Rubbing coils , buzzing parts 

• Wire beat, coil bottoming 

• Loose particles, air leak noise 

• Parasitic vibration of other components 

Excessive nonlinear 

distortion 

Measured 

Signal 

Noise 

Klippel, Sound Quality of Audio Systems, Part 1 Introduction , 4 

Output 

Signal

Loudspeaker Defect: Buzz problem 

Most defects behave as a nonlinear oscillator 

• active above a critical amplitude 

• new mode of vibration 

• powered and synchronized by stimulus 

• constant output power 

vibration 

distortion signal 

parasitic resonator 

one period 

Externally excited 

mass 

spring 

Klippel, Sound Quality of Audio Systems, Diagnostics on Irregular Loudspeaker Defects, 5 

Loose joint 

(Nonlinearity) 

time

Loudspeaker Defect: Voice Coil Bottoming 

Voice coil Voice coil 

Voice coil 

backplate backplate backplate backplate 


Short impulse, deterministic symptom 

one period 

Voice coil 


time

Loudspeaker Defect: Voice Coil Rubbing 

• signal contains reproducible and 

stochastic components 

gap 

Voice coil 

voice coil rubbing 


Cause: rocking mode at 328 Hz 

one period 


time

Loudspeaker Defect: Air Noise 

• stochastic signal 

• air pressure is changed by coil displacement 

• synchronized with stimulus – signal envelope 

one period 

Air noise 


gap 

cone 

leakage 

dust cap 

time

Loudspeaker Defect: Loose Particles 

• random process 

• impulsive 

• particles are accelerated by cone displacement 

• not synchronized with stimulus 

• constant output power 

bouncing 


bouncing 

one period 

Voice 

coil 

former 


gap 

cone 

dust cap 

Loose Particle 

time

Audibility and Impact on Sound Quality 

Klippel, Sound Quality of Audio Systems, Diagnostics on Irregular Loudspeaker Defects, 10

Searching for a Critical Stimulus 

Audibility of Voice Coil Rubbing 

Signal Stimulus Output 

1V 

Music 

Multi-Tone 

20 Hz – 20 kHz 

Multi-Tone 

20 Hz – 1 kHz 

Sinussoidal Sweep 

1 s 

Most sensitive Stimulus 

Output 

2V 

Output 

3V 


Masking of Loudspeaker Defects 

Masking Threshold 

Hearing Threshold 

Passed Failed 

Air Leak 

• irregular distortion produce high frequency components 

• no masking from fundamental component at low frequencies 

• close to the hearing threshold 


Impact on Sound Quality 

Amplitude modulation 

(Envelope modulated by the bass signal 20…300 Hz) 

Spectrogram and time signal of leak noise 

time 

High-Frequency components 

(Increasing spectral power above 3 kHz) 

Increased Roughness Increased Sharpness 

Sensation: aggressive, unnatural, more noticeable 

Degradation of 

Sound Quality 

Signal spectrum of an intact and a defective speaker 

Klippel, Rub & Buzz and Other Irregular Loudspeaker Distortion, 13 

frequency

Auralization of Irregular Distortion (Rub & buzz) 

Example 1 : Headphone, sinusoidal sweep as stimulus 

35 

Distortion to 30 Mask Ratio = 45dB 

dB 

55 

50 

45 

40 

35 

Distortion to Mask 30 Ratio = 22dB 

dB 

55 

50 

45 

40 

Perceptual Attributes: 

Change of Sharpness 25 = 130 % 

Discolouration 20 =130 % 

Modulation = 1200 

15 

10 


5 

Change of Sharpness 25 = 27% 

Discolouration 20=16 

% 


15 

10 

5 

Distortion (Ref) Masking pattern Internal noise 

102 103 104 f in Hz 


102 103 104 f in Hz 

KLIPPEL 

S Dis =0dB 

KLIPPEL 

S Dis =-24dB 

20 

Discolouration =50 % 

15 


10 


dB 

dB 

55 

50 

45 

40 

35 

10 

5 

55 

50 

45 

40 

35 

Perceptiual Attributes: 

30 

Distortion to Mask Ratio = 32dB 

25 

Change of Sharpness = 55 % 

5 


102 103 104 f in Hz 


30 

Distortion to Mask Ratio = 7dB 

25 

Change of Sharpness = 7 % 

20 

Discolouration =4 % 

15 



102 103 104 f in Hz 

KLIPPEL 

S Dis =-12dB 

KLIPPEL 

S Dis =-36dB

Example 2: Automotive Audio System 

organ music in woofer channel, Irregular distortion 

Door Buzzing 

End-of-line testing Off-line 

Processing 

Test with sweep for loudspeaker 

defects and parasitic vibration 

Recording of reproduced music 

stimulus if defect is detected 

reference 

Engineering 

S DIS = 24 dB S DIS = 12 dB S DIS = -6 dB 

S DIS = 0 dB 

Measurement 

results 

recorded 

wave-file 

Auralization + 

Perceptive 

Evaluation 

Klippel, Sound Quality of Audio Systems, Part 13 Specification in R&D and QC, 15 

S DIS = -12 dB 

Marketing 

Management

ated sound quality 

10 = „high“ 

0 = „low“ 

Sensations of Irregular Nonlinearities 

while increasing the distortion signal 

audibilty 

100% 

75% 

„improves bass 

sensation“ 

audibilty 

threshold 


psychometric 

function of 

audibility 

5 = „medium“ 50% 

Kms(x) 

distortion 

voice coil rubbing 

distortion ratio

Engineering 

Subjective and Objective Evaluation 

in Loudspeaker Development 

Objective 

Evaluation 

Physical Data 

• Distortion, Maximal Output 

• Displacement, Temperature 

• Evaluation of Design Choices 

• Clues for Improvements 

Listening Test + Auralization 

Perceptive Modeling 

Performance/cost ratio 

Subjective 

Evaluation 

S DIS 

Audibility of distortion 

Perference, 


Marketing 

Management 

• Defining target specification 

• Tuning to the market

Are loudspeaker defects are acceptable 

which are inaudible for a human ear in a 

production environment ? 

NO ! 

• some loudspeaker defects become worse over time 

(after break-in of the suspension) 

• some loudspeaker defects generate random symptoms 

(loose particles) 

• the production noise masks the symptoms of the defect 

(air leakage) 

• the end user may use a more critical stimulus (low 

frequency bass) 

• Significant loss of sound quality if symptoms detected by 

customer 


Consequences of Subjective Testing at 

the Assembling Line 

• Need for trained operators more sensitive than the end 

user 

• Avoiding fatigue of the human ears (regular breaks, low 

SPL) 

• Operating the loudspeaker under similar condition as in 

the final application (listening distance, amplitude, load) 

• Sufficient time for human inspection (flexible and long 

cycle times) 

Human testing is expensive, not reliable and highly subjective ! 


Need for Objective Measurements 

which 

• are faster 

• can be integrated in an automated line 

• give reproducible, reliable results 

• are more sensitive than the human ear 

• are comparable with R&D, target performance 

• are immune against ambient noise 

• show the root cause of the defect 

• are the basis for process control 


Generation 

of Stimulus 

Basic Requirements to detect 

Irregular Loudspeaker Defects 

Measurement of 

State Variables 

Analysis 

Symptoms 

• High displacement x and/or velocity v is required 

Stimulus with sufficient low frequency content 

• Defects produce only acoustical symptoms 

Sensitive microphone required 

• Defects produce high frequency components 

Low-pass filtered stimulus and high-pass filtered microphone 

signal 

• Defects are similar to ambient noise 

Microphone is located close to the source (near-field 

measurement) 


[V] 

5,0 

2,5 

0,0 

-2,5 

-5,0 

Optimal Stimulus 

Stimulus (t) 

Excitation of Irregular Loudspeaker Defect 

Stimulus (t) vs time 

0 100 200 300 400 500 

Time [ms] 

600 700 800 900 1000 

Continous sweep with 

speed profile 

Why using continuous sweep? 

• Defects may be resonators with high Q-factor 

• Missed, if not excited exactly 

• Critical: Stepped excitation with low resolution 


0,2 

0,1 

0,0 

-0,1 

-0,2 

-0,3 

[V] 

0,50 

0,25 

-0,00 

-0,25 

-0,50 

Y1 (t) 

Response 

0 500 1000 1500 

Time [ms] 

2000 2500 

Fund. Buffer 

Response Detail 

6*101 8*101 102 2*102 Frequency [Hz] 

Defect

Stimulus: 

Sound Pressure [dB] 

120 

110 

100 

90 

80 

70 

60 

50 

40 

30 

20 

10 

Voice Coil Rubbing 

multi-tone complex 20 Hz – 20 kHz 

Noise Floor 

Fundamental 3V 

Multitone Distortion 

KLIPPEL 

10 2 10 3 10 4 

Frequency [Hz] 

Regular distortion masked higher-order harmonics 

Multi-tone distortion is not sensitive for coil rubbing and other defects 


2V 

1V 

3V


Measurement of Incoherence with a music stimulus 

Percent 

100 

10 

1 

Defect speaker 

Good speaker 

KLIPPEL 

10 2 10 3 10 4 


Distortion from regular nonlinearities mask symptoms from rub and buzz 

Incoherence measurement is not sensitive for impulsive distortion 



130 

120 

110 

100 

90 

80 

70 

60 



130 

120 

110 

100 

90 

80 

70 

60 

50 

140 

130 

120 

110 

100 

90 

80 

70 

60 

Frequency Response 

2 

10 

THD Max 


THD Max 

3 

10 


THD 

THD 


Stimulus: sinusoidal sweep 

KLIPPEL 

210 310 Frequency [Hz] 

410 

KLIPPEL 

4 

10 

1V 

2V 


THD Max 

100 1000 10000 


•THD measures the rms value 

of all harmonics 

•Regular distortion are 

dominant 

•Distortion of defects have not 

much energy 

THD is not sensitive for coil 

rubbing and other irregular 

defects !! 

THD 

KLIPPEL 


3V

Sound Pressure Level of Signal Components 

Voltage at 

Terminals 

Input 

Signal 

Measured 

Signal 

130 dB 

Linear System SPL at Mic 

Regular 

Harmonics 

Irregular 

Harmonics 

(Defects) 

90 dB 

50 dB 

Noise 

40 - 60 dB 


Symptoms of a Loudspeaker Defect 

loudspeaker with and without rubbing coil 

- 80 dB 

Masked by 

noise and 

regular 

distortion 

IN1 [V] (rms) 

10-1 

10-2 

10-3 

10-4 

10 

-5 

10-6 

10-7 

with failure without failure Fundamental 

KLIPPEL 

-0 500 1000 1500 


2000 2500 3000 

Fundamental 

50 Hz 

Regular 

nonlinear 

distortion 

Loudspeaker 

Defect 


Frequency Domain Analysis 

60 Hz Tone reproduced by a good and bad speaker 

fundamental 

High Pass Filtering 

2nd 3rd 

required to suppress 

fundamental and low order 

harmonics 

ambient noise 

Failed 

Passed 


Simple Approach 

exploiting amplitude of higher-order harmonics only 

SPL 

fundam ental 

Regular 

distortion higher order 

harmonics 

RMS value of 

Higherharmonics 

amplitude 

spectrum 

frequency 

PROBLEMS: 

• considering deterministic defects only 

• each harmonic is close to the noise level 

• insensitive to loose particles and air leakage noise 


SPL 

SPL 

360 

180 

0 

RMS value 

(MHD) 

fundamental 

fundamental 

Time Domain Analysis 

Regular 

Regular 

distortion 

distortion higher order 

harmonics 

FFT -1 

peak-value 

rms-value 

Crest factor 

amplitude 

spectrum 

frequency 

phase 

spectrum 

time 

Peak value 

(PHD) 

Solution back to the time domain 

exploiting amplitude and phase of 

higher-harmonics and all non-harmonic 

components 

peak value reveals small transients 

(clicks) 

Sensitive for all loudspeaker defects 

most loudspeaker defects generate symptoms with 

high crest factor 

but 

regular loudspeaker distortions, electrical and 

microphone noise have lower crest factor 


Single tone with varying 

frequency f (chirp) 

Sinusoidal 

Generator 

~ 


Peak value Contra rms-Value 

Amplifier 

Drive 

Unit 

Microphone 

instantaneous frequency f 

130 

120 

110 

100 

90 

80 

70 

60 

peak value (PHD) 

rms value (MHD) 


High-pass 

cut-off frequency fc > 10f 

Tracking 

Filter 

100 1000 


10000 

squarer 

integrator 

peak 

detector 

High-pass filtered 

Distortion 

rms value 

(MHD) 

peak value 

(PHD) 

Peak value (PHD) is a sensitive 

measure for most irregular defects 

such as „rub and buzz“, loose 

particles !! 


[V] 

0,03 

0,02 

0,01 

0,00 

-0,01 

-0,02 

-0,03 

Distortion Signal in the Time Domain 

output of the high-pass filter f c=20f 

ZOOM 

Total sound pressure 

50 

Model error 


Period 


• Regular distortion have high energy 

• Disturbances have low energy 

• Disturbances are concentrated in time during a period 

impulsive distortion (high crest factor) 

Active compensation is useful 


KLIPPEL


130 

120 

110 

100 

90 

80 

70 

60 

50 

3V 

Frequency Response Response Max Response Min THD 

THD Max Rub+Buzz Rub+Buzz Max 

2 

10 


140 

130 

120 

110 

100 

90 

80 

70 

60 

3 

10 


KLIPPEL 

4 

10 

PHD 


Stimulus: sinusoidal sweep 

1V 

PHD Max 


130 

120 

110 

100 

90 

80 

70 

60 

Frequency Response Response Max Response Min THD 

THD Max Rub+Buzz Rub+Buzz Max 


KLIPPEL 

102 103 104 Frequency [Hz] 


KLIPPEL 

210 310 410 


2V 

Peak value of highpass 

filtered 

distortion exceeds 

limit by 40 dB !!

Sinusoidal 

Stimulus 

Defect: Loose Particle 


130 

120 

110 

100 

90 

80 

70 

60 

PHD 

Peak Value of high-pass 

50 

filtered distortion 

exceeds limit by 40 dB 

!! 


PHD Max 

Response Max 

Response Min 

THD 

THD Max 


KLIPPEL 

102 103 104 Frequency [Hz]

Sinusoidal 

Stimulus 

Very Small Loose Particle 


130 

120 

110 

100 

90 

80 

70 

60 

Peak Value of high-pass 

filtered distortion50 exceeds 

limit by 10 dB !! 

one grain of fine salt 

PHD Max 

Response Max 

Response Min 

PHD 


THD Max 


KLIPPEL 


MHD

Crest factor of high-pass filtered distortion (CHD) 

Stimulus: Sinusoidal sweep 

[dB] 

20,0 

17,5 

15,0 

12,5 

10,0 

7,5 

5,0 

2,5 

CHD 

CHD ( f ) 

rub & buzz, other disturbances 

50 100 200 500 


PHD 

MHD 

peak-value within one period 

Rms-value averaged over one period 

regular distortion 

KLIPPEL 

CHD can be interpreted on an absolute scale ! 

CHD expoits the phase information of all high frequency components 


12 dB

ICHD 

Instantaneous crest factor of 

high-pass filtered distortion ICHD(f,x) 

good 12 dB bad 

Frequency of sine sweep 

time 


Loudspeaker Defect or Noise ? 

dB 

130 

120 

110 

100 

90 

80 

70 

60 

50 

40 

Fund. mean 

3rd 

THD 

2nd 

PHD 

CHD 

MHD 

Fundamental 

PHD limit 

(-40dB) 

10 20 50 100 200 500 1k 2k 5k 10k 


Symptoms of a significant defect: 

1. Sufficient amplitude: peak value PHD > FUND mean - 40dB 

2. Impulsive: crest factor CHD > 12dB 

Check for coincidence ! 

(microphone noise is not 

impulsive) 


is not an 

audibility 

threshold 

!

Characteristics for Diagnostics 

Single-valued parameter derived from PHD and CHD 

Batch of TRF measurements 

6V 

8V 

10V 

12V 

voltage 

dB dB - - [V] [V] (rms) 

(rms) 

Fundamental Fundamental + + Harmonic Harmonic distortion distortion components components (SHAPED (SHAPED STIMULUS) 

STIMULUS) 

130 

130 

120 

120 

120 

110 

110 

110 

100 

100 

100 

90 

90 

90 

80 

80 

80 80 

70 

70 

70 70 

60 60 

60 60 

50 50 

CHD > 12dB 

impulsive 

Signal Signal 

Signal at at 

IN1 

IN1 

Fundamental Fundamental 

Fundamental THD THD 

THD 2nd 

2nd 

2nd Harmonic Harmonic 

Harmonic 3rd 3rd 

Harmonic 

Harmonic 

Absolute Absolute 

Absolute PHD 

PHD 

PHD Fund. Fund. 

Fund. mean 

mean 

mean (100 

(100 

(100 to to 

to 500 500 

500 Hz) Hz) 

PHD PHD 

limit limit 

(-40dB) 

(-40dB) 

f r&b 

-40 dB 

KLIPPEL KLIPPEL 

50 40 

50 40 

10 10 20 20 50 50 100 100 200 200 500 500 500 

1k 1k 

2k 2k 2k 

5k 5k 5k 

10k 10k 

10k 

Frequency Frequency 

Frequency [Hz] [Hz] 

[Hz] 

Significant 

defect (R&B) 

PHD > FUNDmean - 40dB 

high amplitude 

12V 

Conditions: 

Rms Voltage U r&b = 12 V 

Peak displacement X r&b= 4mm 

Frequency f r&b = 65 Hz 

Inst. Displacement x in, r&b = 3.8 mm 


Diagnostics on Irregular Defect (1 st example) 

Conditions: 

• Positive peak displacement at 0.5 mm below f s 

• above a terminal voltage of 0.9 V 

Root Cause: Bottoming 


Diagnostics on Irregular Defect (2 nd example) 

Conditions: 

• Negative turning point of voice coil excursion above f s 

• maximal acceleration Tilting of voice coil former 

• independent of displacement 

Root Cause: Voice coil Rubbing 


Clues for Diagnostics of Defects 

derived from ICHD(f,u) measured versus frequency f and voltage u 

• Coil rubbing initiated by a tilting of the voice coil former occurs at maximal acceleration 

corresponding with the positive and negative maxima of the displacement. This occurs 

usually at frequencies above resonance where the acceleration is maximal while the 

amplitude of the displacement decreases. When the voltage is increased the coil rubbing will 

also occur at the turning points of the voice coil at higher displacement values. 

• Buzzing part generates high ICHD for particular (reproducible) values of frequencies and 

displacement because it behaves like a coupled nonlinear resonator. Above a critical voltage 

when it starts this defect is almost independent on the peak displacement. 

• Coil bottoming at the rear plate generates high ICHD at a clearly defined negative 

displacement independent of frequency. Increasing the voltage of the excitation the maximal 

negative displacement is constant but the loudspeaker generates a dc displacement shifting 

the coil in the opposite direction. 

• Limiting surround generates high ICHD at maximal positive or negative displacement 

independent of the excitation frequency. The instantaneous displacement where ICHD > 12 

dB raises by increasing the voltage because the suspension behaves not as a hard limiting 

nonlinearity like coil bottoming. 

• Tensile slap generates a high value of ICHD at reproducible values of displacement and 

frequency because the wire behaves as a coupled mechanical system having its own 

vibration mode and natural frequency. 

• Air leakage in a dust cap generates a high ICHD for frequencies below resonance f< fs 

where the sound pressure of the enclosed air is high. 

• Loose particles generate ICHD > 12 dB at random values of frequencies and instantaneous 

displacement. 


Basic Classification of Loudspeaker Defects 

Coil hitting 

backplate 

Waveform is 

completely 

reproducible 

Buzzing loose 

joint 

vibration 

Rubbing 

voice coil 

Envelope is 

reproducible 

(Waveform is not) 

Flow noise at air 

leak 

Loose particle 

hitting 

membrane 

Semi‐random 

Deterministic Random 

(mixed characteristic) 

Waveform is not 

reproducible 


Loose particle

Exploiting all Symptoms of the Defect 

SPL 

fundamental 

2nd 

3rd 

Regular 

distortion 

4th 

High-pass filtered 

symptoms and noise 

SPL 

f 

Symptoms of 

defects 

noise level 

frequency 

High-pass Filter 

cut-off frequency f c > 10f 

SPL 

Comb Filter 

Order n > 10th 

frequency 

Deterministic part 

(higher-order harmonics) 


SPL 

Inverse Comb Filter 

frequency 

Random part 

(non-harmonics)

Example: 

tensile slap, bottoming 

Deterministic Distortion 

Results of three measurements 

Symptoms: 

• Reproducible, repeatable 

• Related with stimulus 

• impulsive distortion 

• Deterministic amplitude and 

phase of higher-order ha FAIL 

high-pass 

PASS 


SPL 

Defects Masked by Regular Nonlinearities 

@ 5 Volt stimulus 

fundamental 

Regular 

distortion 

order > 10 

Symptoms of 

defects 

frequency 

SPL 

fundamental 

Where are the limits of a human tester ? 


Regular distortion 

Symptoms 

of defects 

order > 50 

frequency 

• Symptoms of some defects have almost constant energy 

• Distortion of regular nonlinearities rise with amplitude 

• Defects are masked (become inaudible at higher amplitudes) 

• Increasing high-pass frequency less energy noise problems 


Stimulus 

Solution: Active Compensation 

DUT 

Model 

p(t) 

p’(t) 

Isolated 

Defect 

Distortion 

classificator 

Fail/Pass 

Meta-Hearing Technology 

• Regular distortion are deterministic and predictable 

• Modeling of regular distortion (adaptive learning) 

• Masking by regular distortion can be removed actively 

active compensation 


SPL 

Meta-Hearing Technology 


fundamental 

Regular 

distortion 

residual 

Symptoms of defects 

frequency 

• Exploiting symptoms masked by regular distortion 

• Test of driver at maximal amplitude becomes possible 

• Detection of defects with low energy (loose particles) 

• Detect failures even if they are inaudible 

(getting worse in final application) 

More energy can be used ! 


110 

100 

90 

dB 

80 

70 

60 

50 

40 

30 

2*10 1 20 

Benefits of Active Compensation 

Without compensation 

regular 

PHD 

Limit 

Loose particle not detected 

50 100 


200 

400 

90 

dB 

80 

Simple definition of PASS / FAIL thresholds 

Measurement below the hearing threshold 

110 

100 

70 

60 

50 

40 

30 

2*10 1 20 

With meta-hearing technology 

Limit 

Meta-hearing 

PHD 

Meta-hearing 

regular 

Loose particle detected 

50 100 



200 

400

Example: 

Turbulent air noise generated at 

leaks, coil rubbing 

Symptoms: 

• Distortion are NOT reproducible 

• Distortion occur at particular 

times 

• Dense spectrum (cover audio 

band and beyond) 

Semi-Random Distortion 

high-pass 


Averaging of Air Leakage Noise 

Multiple realizations 

Averaged realizations 

Air noise 

• High pressure generates high air velocity at the leakage 

• Turbulences generate a noise signal modulated by the air pressure 

• Noise is almost random 

Problem: Averaging suppresses the random noise signal 


time 

time

Generation of Turbulent Air Distortion 

u(t) 

voltage 

x(t) 

pbox(t) 

q(t) 

Generation of High Volume velocity q(t) 

Voltage displacement sound pressure volume velocity 

Linear System 

Generation of Turbulent Noise 

qmod(t,rs) 

Nonlinear System 

(Modulation Process) 

Linear sound propagation 

Linear System 


p(t) 

sound pressure

u(t) 

voltage 

Modeling by a Signal Flow Chart 

Hpre(j? ) 

Low-pass filter 

Linear System 

pbox(t) 

Linear 

System 

Regular 

Nonlinearities 

(motor, suspension) 

Transducer Defects 

(rub&buzz) 

N(pbox) 

Static nonlinearity 

Noise 

Source 

n(t) 

q(t,rs) 

Nonlinear System 

(modulation process) 

pa(t) 

Ambient noise 

pmod(t) 

qmod(t,rs) 

Hpost(j? ) 

Sound propagation 


) 

p(t) 

sound pressure 

Linear System

Why Are Traditional Measurements 

Not Sensitive for Air Leaks ? 

Deterministic 

第一次测量 

1st measurement 

1st measurement 

2nd measurement 

第二次测量 

2nd measurement 128次结果平均 

Averaged 120 times 

Averaged 128 times 

Noise is attenuated by averaging !! 

Deterministic 


Envelope of the Modulated Noise 

Multiple realizations 

Averaged Envelope 

Air noise 

• Envelope of the modulated noise is deterministic 

• Averaging of the envelope increases signal to noise ratio 


time 

time

Single tone 

p(t) 

Comb-Filter 

Demodulation 

p r(t) 

e(t) 

p f(t) 

Comb-Filter 

How to Calculate the Envelope ? 

P f 

P f 

P r 

E 

fundamental 

regular 

Envelope 

Envelope 

frequency 

frequency 

frequency 

frequency 

p r(t) 


p(t) 

p f(t) 

e(t) 

time 

time 

time 

Envelope 

time

MOD 

abs 

fundamental 

Absolute Modulation Level 

Peak value of the 

envelope 

e 

10 lg 

 

2 p 

max 2 

0 

Harmonics 

 

 

 

 

Absolute hearing threshold p 0 

deterministic components 

• Shows the peak value of the envelope 

• is in dB referred to the absolute hearing 

threshold p0 • good for PASS/FAIL decisions 


e max

MOD 

Relative Modulation Level 

Peak value of deterministic part 

Rms value of random components 

deterministic components 

fundamental 

e 

10 lg 

r 

max 

rel ~ 2 

harmonics 

 

 

 

Random components 

• Considers the ratio of deterministic and 

random parts of the envelope 

• Shows significant modulation (MODrel > 0) 

• Is in dB 

• good for PASS/FAIL decisions 


e max

Influence of Excitation Voltage 

Box with leak 

Generation of 

turbulences 

Box without leak 

MOD rel= 0 dB 

indicates no 

modulation 

Box with leak 


Box without leak 

•Turbulent air leakage noise set in abruptly if the velocity of the air at the leak 

exceeds a critical value. 

• Search for the critical voltage of the stimulus !!

Influence of Measurement Time 

MOD abs 

MOD rel 

The relative 

modulation level 

MOD rel rises by 3dB 

for doubling the 

measurement time! 

A 1s stimulus of 100 Hz will increase the sensitivity by 20 dB for semi-random distortion ! 


How to Separate Port and Leak Noise ? 

Leak Noise 

30 dB 

Port Noise 

Inset of leak 

turbulences 

Inset of port 

turbulences 

• Close the port for leakage detection (if possible) 

• Select optimal excitation voltage (e.g. 1-2 Volt) between inset of leak and port 

turbulences 

• Shield test microphone from port noise 

• Choose optimal microphone position 

• Use directional microphone 


Leak Noise 

Port Noise

Combining Subjective and Objective Assessment 

Stethoscope 

Noise 

Detection 

Objectives: 

• Auralization of Irregular 

Distortion 

• Ear Protection 

• Signal Transformation 

• Defect Localization 

• Diagnostics 


Air 

Leak 

Localization of Loudspeaker Defects 

relative 

Modulation 

Good Bad 


A sensitive Measurement system is 

not sufficient ! 


Voltage at 

Terminals 

Coping with Ambient Noise 

Sound Pressure Level of Signal Components 

130 dB 

Linear System SPL at Mic 

Regular 

Nonlinearities 

Irregular 

Defects 

90 dB 

Noise 

50 dB 40 - 60 dB, Peaks 

Problems: 

• Symptoms of defects are very small (but still audible) 

• Ambient noise in a production environment has similar properties 


Professional Test Boxes 

Photos by courtesy of Phase Design 


Solution: Ambient Noise Microphone 

• Measure loudspeaker in the near field 

• Measure noise / vibration in the far field 

• Predict noise at test microphone 

• Calculate impact on measured characteristics 

• Store valid part of the measurement 

• Repeat measurement automatically 

Full noise immunity for random ambient noise 


Merging Technique 

repeating measurement automatically and accumulating valid parts 

Ambient noise generated by permanent hand clapping 

28 % valid 

62% accumulated 

85% accumulated 

100% valid 

Full immunity against random noise 


Sound Quality in the Car Interior 

ambient noise source 

Progress in Data Acquisition 

test microphone inside the car 

fundamental 

rub and buzz 


tambourine 

Simulation of Door Buzzing 


reliable detection 

of the defect

How to fix the problem? 


Root Cause Analysis of Loudspeaker Defects 

Bottoming 

Asymmetrical 

Suspension 

Buzzing 

vibration 

Glue dispenser 

Problem 

Coil Rubbing 

mass distribution 

on the cone 

Air Leak 

Missing screw 

Design 

Supplier 

Qualtiy Inspection 

of Incoming Parrts 

Loose particle 


dirty storage condition 

Manufacturing 

(Process Control)

Automatic Learning from Manufacturing 

1st step: Automatic clustering of the raw data 

2nd step: Prototype selection (best representative of the cluster) 

3rd step: Human inspection and tagging of the defect 

4th step: Updating the knowledge base for on-line diagnostics 


Clustering is a statistical 

process of separating 

devices under test (DUTs) 

into subgroups wherein 

all members of one 

cluster have similar 

properties but are very 

different to the members 

of other clusters. 

Results of the formal cluster analysis: 

• Total number of defect classes 

• Devices representing each class 

• Properties of the good units 

• Significant features 

• Related or similar defects 

tagging 

Clustering of the DUTs 

in the Feature Space 

Feature 1 (R (Re) e) 

P 3 

Similarity 

P 5 

(distance) 

Feature 2 (f (fs) s) 


P 2 

P 1 

P 4 

Volume 

(deviation) 

Prototype 

(mean value)



120 

110 

100 

Raw Measurement Data (300 DUTs) 

90 

80 

70 

60 

50 

40 

85 

80 

75 

70 

65 

60 

55 

50 

45 

40 

35 

30 

25 

20 

Curve Bulk 

Curve Bulk 



Rub+Buzz 


Clustering 

KLIPPEL 

Good 



120 

110 

100 

90 

80 

70 

60 

50 

40 

85 

80 

75 

70 

65 

60 

55 

50 

45 

40 

35 

30 

25 

20 

Results of the Clustering 


1: GOOD (SD) 2: Low SPL (SD) 3: not connected (SD) 

4: high THD @ 1.5 kHz (SD) 5: lower fs (SD) 6: not connected (SD) 

7: Rub+Buzz (SD) 

Good 

Low SPL 


Rub+Buzz 

1: GOOD (SD) 2: Low SPL (SD) 3: not connected (SD) 

4: high THD @ 1.5 kHz (SD) 5: lower fs (SD) 6: not connected (SD) 

7: Rub+Buzz (SD) 

Coil rubbing 

Low SPL 



Surround 

Resonance 

Coil rubbing 

Not Connected 

Surround 

Resonance 

Not Connected

Automatic Classification of Defects 

1st source: theoretical 

knowledge in engineering 

(static) 

2nd source: practical 

experience in 

manufacturing 

(dynamic) 


Irregular Loudspeaker Distortion 

Summary 

• are related with loudspeaker defects 

• not found in approved prototypes, golden reference unit 

• are caused by manufacturing, overload, ambient conditions 

• are not directly related to cost, size, weight 

• are difficult to model and not predictable 

• depend on the operation condition (e.g. orientation + loose 

particles) 

• are time variant (aging) and usually become worse over time 

• generate impulsive distortion with high crest factor but low energy 

• are inacceptable if detected by customer 


Conclusions 

• Defective loudspeakers with irregular distortion should not be 

shipped to the customer even if the symptoms are inaudible 

• Sensitive Measurement techniques are required more 

sensitive than the ear of the customer 

• Time domain analysis is required to consider transient 

properties of the symptoms 

• Reliable measurements require automatic detection of invalid 

tests corrupted by ambient noise in a production environment 

• Production limits depend on the target application 

• The human operator works on the diagnostic station and 

provides information for the knowledge base 

• Automatic defect classification shows the root cause of the 

defect 

• Process control ensures high yield rate in production 


Many thanks ! 


References 

• W. Klippel, Tutorial: Loudspeaker Nonlinearities - Causes, Parameters, 

Symptoms J. Audio Eng. Society 54, No. 10 pp. 907-939 (2006 Oct.). 

• W. Klippel, „Measurement of Impulsive Distortion, Rub and Buzz and 

other Disturbances”, presented at the 114th Convention of the Audio 

Engineering Society, 2003 March 22–25, Amsterdam, The Netherlands, 

preprint # 5734. 

• S. Temme, et. al., “Loose Particle Detection in Loudspeakers,” 

presented at the 115 th Convention of the Audio Eng. Soc., September 

2003, preprint 5883. 

• W. Klippel, et. al. , “Loudspeaker Testing at the Production Line,” 

presented at the 120 th Convention of the Audio Eng. Soc., Paris 

(France), September 2006, May 20-23, preprint 6845. 

• W. Klippel, J. Schlechter, „Fast Measurement of Motor and Suspension 

Nonlinearities in Loudspeaker Manufacturing,“ presented at the 127 th 

Convention of the Audio Eng. Soc., New York, USA, October 9-12 

2009, preprint 7903 


Application Notes and Templates 

Application Notes: 

AN 24: Measuring Telecommunication Drivers, Micro-speaker, Headphones 

AN 22 Rub & Buzz Detection without Golden Unit 

R&D Operation Templates available for Transfer Function Module 

TRF RUB + Buzz without Golden Unit Speaker 1 (woofer) 

TRF RUB + Buzz without Golden Unit Speaker 2 (headphone) 

Object Template: 

Diagnostics Rub + Buzz Distortion Sp1 (woofer) 

Diagnostics Rub + Buzz Distortion SP2 (micro-speaker, headphone) 

Further information under www.klippel.de

Rub & Buzz and Other Irregular Loudspeaker ... - Klippel GmbH

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?