Aachen Acoustics Colloquium 2015 135

The Porsche Wind Tunnel Microphone Array System Jörg Ocker1, Swen Tilgner2

1 Dr. Ing. h.c.F. Porsche AG, 71287 Weissach, Germany, Email: joerg.ocker@porsche.de 2 gfai tech GmbH, 12489 Berlin, Germany, Email: tilgner@gfaitech.de

Abstract The new Porsche aero-acoustic wind tunnel opens great options for the automotive development process at Porsche. A very low background noise and the homogeneous flow characteristics al-low the use of various measurement techniques and lead to an excellent quality and degree of re-producibility of acoustic measurements.

Especially in the early development stages a quick and reliable rating of aero-acoustic sound sources is important to fulfil the challenging requirements. For a localisation and detailed inves-tigation of these aero-acoustic sources, a complex measurement system consisting of three single microphone arrays, was developed and installed.

All arrays are fixed at a framework positioned out-of-flow, with the smallest possible distance to the test object. In order to prevent negative effects during aerodynamic testing, the whole system can be moved back to a parking position out of the test section area.

The basic evaluation of the recorded microphone data is known as “acoustic camera principle” where acoustic source maps computed by special beamforming algorithms are mapped to an op-tical image (2D) or the car surface (3D).

The quality of these basic source maps (“dirty maps”) - given by the dynamic of the localized sound levels and the spatial source resolution - is mainly defined by 4 hardware parameters:

Array dimensions 5 m x 3 m each

Number of microphones 192 microphones each

Microphone layout irregular with an optimized layout

Measurement distance automated positioning to 4 m

By the use of advanced post-processing techniques, e.g. algorithms like CLEAN SC [12], it is possible to improve the results significantly. However, the better the basic result the greater the success using these enhanced software tools.

Therefore great attention was paid to excellent properties of the mechanical installation as well as to a comfortable and fast data acquisition achieving an outstanding basic acoustic perfor-mance. Hence, further activities have been started and various investigations were scheduled to advance the data analysis.

The paper focuses on the first and terminated project phase: The development, design and instal-lation of the system. Based on the results of several simulations and a study with real sound sources (loudspeakers with and without flow); the capability of the system is verified and the quality of first results is shown by means of a typical example.

Introduction For a car manufacturer like Porsche, an outstanding sportive design and a great performance combined with a favourable passenger comfort and high efficiency are fundamentals to be re-garded at all times. Therefore, the new Porsche wind tunnel is a very important tool in the aero-dynamical and aero-acoustic optimization process, allowing evaluation of design- and material concepts as well as styling-driven shape decisions. The new Porsche wind tunnel was designed over several years and realized from 2011 to 2014 [14].

136 Aachen Acoustics Colloquium 2015

Now, aero-acoustic development can be performed over the full velocity range up to 300 km/h, and due to very low background noise levels, the localisation and precise investigations of sound sources are possible using microphone arrays. The Porsche wind tunnel microphone array system was developed and built in cooperation with GFaI, Berlin and installed in January 2015.

Basic Array Hardware The system consists of 4 data recorders, 3 microphone arrays, each equipped with 192 electret microphones and an integrated Full HD camera. Additionally 24 measuring channels are provid-ed to record reference signals and the test conditions. A high performance working station is dedicated to the data acquisition and analysis, operated by special controller software, adapted to the Porsche wind tunnel.

To achieve a high localisation accuracy of the acoustic sources, a signal sampling frequency up to 192 kHz is realized for each differential transmitted channel and a positional accuracy of ± 1 mm is ensured for all microphones. This precision is reached up to maximum speed by a high stability and a low weight of the arrays resulting from a special sandwich construction inte-grated in a lightweight reinforcing frame.

Beamforming in a Wind Tunnel The result generated by an array system is called acoustic source map; superposed to an optical image or a 3D geometry of the considered object. The calculation of the source map is carried out using beamforming algorithms.

A picture (2D) or model (3D), divided into a grid of points, is used to represent the object under investigation. For 2D applications, the grid is arranged in the scan plane with a known distance to the microphones in the array plane (Figure ). The evaluation of the runtime differences and amplitude decays of assumed sound sources between every grid point of the scan plane and eve-ry microphone of the array plane leads to the basic source map (“dirty map”). This elementary beamforming algorithm can be applied in the time or frequency domain whereat the time domain “delay and sum” algorithm is well known in acoustic signal processing for many years now.

In recent years more advanced frequency domain algorithms have been developed to improve the quality of the maps. In the frequency domain most algorithms base on the calculation of the cross spectral matrix (CSM). The CSM contains all cross spectra data of the microphone signals and therefore all level and phase information representing the runtime differences of the acoustic waves. By applying a so called steering vector to the CSM - to include the geometric information and distances - it is possible to calculate the source map. By the formulation of the steering vec-tor, the focus of the evaluation can be directed to calculate exact levels or to a good source local-isation in the source maps.

The main diagonal of the CSM contains the auto spectra data of the microphone signals which are often dominated by the background noise. Thus it is possible to improve the dynamic range of the source maps by removing the elements of the main diagonal from the CSM. However the energy content and thereby the absolute levels of the source maps are not correct in that case.

There are many publications in which the different methods and algorithms with their pros and cons are discussed in detail [2], [6], [7], [9], [13].

In a wind tunnel a shear layer appears between the flow region and the area in the plenum with-out a major flow where the arrays are positioned. The flow around the test object generates sound sources radiating acoustic waves. These waves get strongly affected when passing the shear layer (Figure ). The shear layer causes diffractions of the acoustic waves and lead thereby to displacements and distortions in the source maps. For the application of beamforming in wind

Aachen A

tunnels correcti

Amiet icy by reto the ar

Unaccoer whicshifts.

F

With thbased o[8] and plementtion betw

To rate the freqtion, typ

In FigurThe mabeamforof the m

The beaand beacapabilidimensi

Acoustics Co

to aero-acoon accordin

ntroduces aeducing the rray and sca

unted in thich lead to c

Figure 1:

he Porsche Aon the CSM

a steering vted basic alween GFaI,

the performquency-depepically calcu

re 2 a typicain lobe reprming algor

main lobe an

amwidth (Bamwidth areity of an arion, the num

lloquium 201

oustic sourcng to Amiet

a model whshear layer

an plane (Fi

is model arecorrelation l

Sketch o

Array SysteM with diago

vector formgorithms w, DLR Götti

mance of anendent arrayulated by a

cal PSF of tpresents the rithms. The nd the highe

BW) definese the basic irray. Both pmber and lay

5

ces, it is ind[12].

ere the diffrr to a flat anigure ).

e additionallosses of th

of wind tunn

em, most oonal remova

mulation conwere validate

ingen, Volk

n array, they response tsimulation.

the Porschereal sourcedynamic ra

est side lobe

the width oindicators fparameters dyout of micr

dispensable

fraction effend infinite th

l scattering he signals w

nel shear lay

of the analyal. An Ami

ncentrating ed in great dkswagen gro

e Point Spreto an omnid The result

e arrays is pe, whereas tange (DR) ie.

of the mainfor the qualidepend strorophones an

to consider

ects are takehin extent s

effects due with a degra

yer and para

ses are carriet shear layon an exactdetail in theoup research

ead Functiodirectional is also calle

plotted for tthe side lobis defined by

n lobe at 3 dity of soundongly on thnd the meas

r this shear

en into accospecified by

to the turbuading effect

ameters of th

ried out in yer correctit source loce context ofh and Porsc

n (PSF) canpoint sourced beam pat

he third-octbes are artey the differe

dB level decd source loce hardware

surement dis

1

r layer effec

ount with goy its angle a

ulence in tht evaluating

he Amiet mo

the frequenion modifiecation is usef an intensivche.

n be used. ce with a dettern.

tave band oefacts generence betwe

crease. Dyncation and t

e parametersstance (focu

137

ct, e.g. by a

ood accura-and distance

e shear lay-g the phase

odel.

ncy domained by Puhleed. The im-ve coopera-

The PSF isefined loca-

of 2500 Hz.rated by theen the level

namic rangethus for thes: the arrayus).

a

-e

-e

n e --

s -

. e l

e e y

13

Figur

Array DiThe requirlower the fture enougor respectiv

To minimisions with frequencieThis effect

Taking intlysed, the o3 in additiogation of hrable result

Figure

Number The numbeta acquisitiproject bud

38

e 2: Si

imension red size of afrequency a

gh informatively phase

ize the beama minimal s correlatiot results in p

to account aoverall dimon, the centhigher frequts, all array

e 3: Ov

and Layoer of micropion system ddget. Otherw

1

1 m

imulated Po

an array firsand thus theion of the acshifts in the

mwidth ovefocus are as

on losses inphase errors

all these effmensions of t

tral part of tuencies in a s (left, right

verall arrayprefer

out of Micphones and defines a mwise, to rea

B = 5 m

1.5 m

int Spread F

st of all depe bigger thecoustic wave frequency

er the wholespired. Und

ncreases togs for distant

fects and adthe Porschethe arrays (3preferred m

t and on top

y dimension:rred measur

crophonethereby the

major part ofach an excel

H

Function (P

pends on the wavelengtve. That mea

domain for

e frequencyder flow congether with

microphon

djusting thee arrays wer3 x 2 m) ha

measuremenp) were cons

: 5 x 3 m, cerement wind

es e number off the hardwallent dynam

= 3 m0.5

PSF), f = 250

he lowest freth, the largeans to measr the observ

y range, basnditions in athe array d

nes.

e arrays to re defined t

as been designt window ostructed wit

ntre array: dow: 1.5 x 1

f measuremare costs of

mic range of

1 m5 m

Aachen Acou

0 Hz, F = 4

equency to er the array sure enoughed wavelen

sically maxia wind tunndimension a

the size of o 5 x 3 m. Agned especiof 1.5 x 1 mth the same

3 x 2 m, ext m.

ent channelf an array syf the source

1.5 m

ustics Colloqu

m, DR = 21

be considedimensions

h runtime dingth.

imum arraynel, howeveand the flow

the cars toAs shown iially for the

m. To obtaindimensions

tension: 3 x1

ls needed byystem limite

maps, as m

uium 2015

dB.

red. The s to cap-ifference

y dimen-r, higher w speed.

be ana-n Figure

e investi-n compa-s.

1 m,

y the da-ed by the much mi-

Aachen Acoustics Colloquium 2015 139

crophones as possible should be applied. As a compromise, each Porsche array was equipped with 192 microphones.

Another important point, especially regarding the dynamic range, is the layout of the micro-phones. A lot of effort was provided to find the best arrangement of the microphones. Therefore, a completely new optimisation strategy was developed and applied by Dr. Hartmann from Volkswagen group research.

The optimisation performed by a generic algorithm in combination with a gradient based search of local minima operates directly to all microphone coordinates – not just to some layout pa-rameters. Thus, for 192 microphones with adjustable x- and z-positions 384 degrees of freedom have to be optimised.

The target function for the optimisation was generated by a rating of weighted sums for each cri-terion below:

beamwidth

distance of the main lobe to the highest side lobe

level of the highest side maximum

average of the side lobe levels in the preferred measurement window

Further restrictions were introduced to concentrate microphones in the centre array. This could be important for the high frequency region if the use of a shading procedure to the outer micro-phones could reduce negative correlation loss effects as discussed before. In this case only few microphones would be affected by shading. Future work is scheduled regarding this topic.

The result of the optimization is an irregular microphone layout as depicted in Figure 4, where simulated source maps are compared to the best symmetric standard layout considered.

14

Figure

To comparrithmic whmized layourement w

The only dasymmetri

According equal for a

MeasureFor the intespecial railby several tem rigidity

Due to thebeen reachof approx. ably short investigatea frame wi

40

4: Sim

re on the saheel standarout is clearl

window.

disadvantagc representa

to the arraall arrays ins

ement Disegration of l system to pneumatic y is achieve

e given condhed by a dir

6 m to the measuring

ed in a studyith the array

mulated sour

ame basis, rd layout. Tly visible f

ge observedation of the

ay dimensiostalled to ge

stance or Fthe arrays ibe moveabcylinders b

ed by this m

ditions, a mrect installatscan plan adistance (f

y carried ouy dimension

rce maps, 6

the numberThe improvfor all frequ

d regarding sources in t

on, the numet comparab

Focus in the wind ble in flow dbracing the wmechanism.

measuring dition of the assumed at thfocus) is imut with 5 mins of 5 x 3 m

sources with

r of micropvement of tuencies and

the asymmthe source m

mber and lable results.

tunnel, a pudirection. Twhole fram

istance of 7arrays to ththe side of amportant foricrophones

m (Figure 5)

h same level

phones was the dynami

d particularl

metric dimemaps.

ayout of m

urpose-builtThe fixing toework to th

7 m to the we framewor

a car with a r excellent mounted at

).

Aachen Acou

l, F = 4 m, 19

extended toc range achy clear in t

ension and

icrophones

t frameworko a defined

he wind tunn

wind tunnelrk. This cortypical widresults, shot the edges

ustics Colloqu

92 microph

to 192 for thieved by tthe preferre

layout is a

was define

k was instalposition is

nel walls. H

l centre wourresponds todth. But, as orter distancand in the c

uium 2015

ones.

the loga-the opti-ed meas-

slightly

ed to be

lled on a realized

High sys-

uld have o a focus a prefer-ces were centre of

Aachen A

The evawind tuna furthelayer. Hthe expewas dec

Array The arrconsistswinch sparking

All arraminutes

Fig

To avoiminutessystem to everyvent any

Acoustics Co

Figure

aluation of nnel centre

er reduction,However, foectable impcided.

Positioniay position

s of severalsystem oper position re

ays can be s. Various se

gure 6: mea

id any tapins, the entire is moveabley interestingy negative e

lloquium 201

5: Tes

the study leand thus a

, the micropor a focus reprovements

ing Systemning system l control arrated by a celiably (Figu

moved auensors are i

Frameworasuring posi

ng of loosewiring of t

e on the railg part of theeffect to aer

5

st set-up to d

eads to the focus of ap

phones neareduction, anin result qu

m has been s

rms connectontrol unit

ure 6 and Fi

utomaticallynstalled to p

rk and arrayition.

e cables anhe array syl system, the test objecrodynamic t

determine th

result that pprox. 4 m tr to the collen additiona

uality, the im

specially deting the arrand a pneu

igure 7).

y to the meprevent ope

ys in

nd to be reastem is rout

he preferablect and a partesting (Figu

he optimal m

a minimal o the assumector would

al mechanicmplementati

esigned forrays rotatab

umatic locki

easuring or erating error

Figure 7:position

ady performted throughe measuremrking positioure 8).

measuremen

measuring med scan plad be stronglyal construction of an arr

the new Poble to the fing system t

parking pors.

Framen Arrays in p

ming measuh a cable trament windowon can be re

1

nt distance.

distance of ane can be ry affected btion is need

rray position

orsche winframework, to fix the ar

osition in l

ework in meparking pos

urements wack. Becausew can be easeached quic

141

f 5 m to therealised. Byby the shearded. Due toning system

d tunnel. Itan electric

rrays in the

less than 3

easuring sition.

ithin a fewe the wholesily alignedckly to pre-

e y r o

m

t c e

3

w e d -

14

A nearly prigidity of array positmaximum

By the usethe deviatiments duri

Side Top

Table 1:

PerformSetup For the detwith a 1:1 tem, the stWhen selec

Protyp

LSPcur

No

Pro

Two typesmembrane between 1 nearly the

42

erfect fixingthe position

tioning and speed. This

of a FAROions of the ng wind tun

Arrays

p Array

Arra

ance Test

terminationmodel of atudy was ccting the lou

oviding enoupically analy

P membranracy.

generation

oviding a pr

of LSP, shdiameter oand 10 kH

same chara

Figure

g of the framning systemlow displac

s is essentia

O Laser Tracarray posit

nnel operati

Max. 0.0.

ay positionin

ts

n of the basia test vehicleconducted wudspeakers

ugh power ysed at a spe

nes as small

of addition

referably lin

hown in Figof 36 mm pHz. LSP of cteristics w

e 8: Frparki

mework to m and the fracements andl, especially

cker with a ioning, region were me

Deviation06 mm 33 mm

ng accuracy

ic array proe (Figure 9)with generic, a lot of att

to generate eed range b

as possible

nal sound du

near frequen

gure 10, werprovides enotype 2 exte

with a memb

ramework aing position.

the wind tuamework, led vibrationsy for optima

measuremearded at mueasured.

y. Table 2

operties, ext). To be as c sound sotention was

source leveetween 100

e for a prec

ue to the sha

ncy response

re found fuough poweends the low

brane diamet

and arrays in.

unnel wall steads to a ves during the al source loc

ent accuracyultiple oper

Flow Spee140 km/h300 km/h

: Arr

ensive perfoflexible as

ources, genepaid to fulf

els represen0 and 200 km

cise analysis

ape of the L

e over a wid

lfilling thesr and a neawer frequenter of 50 mm

Aachen Acou

n

tructure, coery good rep

wind tunnecalisation o

y of 0.03 mmrations and

ed Mh h

ay displacem

formance tespossible for

erated by lofil the follow

nting real aem/h.

s of the sou

LSP themsel

de frequenc

se requiremarly linear fncy limit tom.

ustics Colloqu

ombined witproducibilityel operationon a 3D mod

mm at 18 m dthe array d

Max. Displac0.22 mm1.32 mm

ment during

sts were carr analysing oudspeakerwing require

ero-acoustic

urce localisa

lves.

cy range.

ments. Type frequency so 500 Hz p

uium 2015

th a high y for the

n, even at del.

distance, displace-

cement m m

g operation.

rried out the sys-

s (LSP). ements:

c sources

ation ac-

1 with a spectrum roviding

Aachen A

By the uscript, iThe signeach LS

In total,sive dismountedel wheretially recle. Thesources the vehi

For comtent are distancetunnel tgations

Acoustics Co

Figure

use of a suitt is possiblenal levels a

SP separatel

, 16 LSP pocussions bed absolutelye the whole

easons, all ree LSPs are tthe rear vie

icle model.

mparable anessential. T

e. In additioto analyse ware perform

5

1

lloquium 201

e 9: Set

Figure 10:

table signale to run 8 Land frequenly. The phas

ositions weretween the py flat to thee glazing is esults reporthereby pictew mirrors,

nd reliable rTherefore, aon, the freqwith a similmed based o

6 78

43 2

5

tup of perfo

Loudsp

l conditioneLSP time sycy charactese relation b

re defined oproject partn surroundinmade of foa

rted here aretured at the one of the

results, sourall LSP werequency specar signal ton uncorrela

7

4

ormance test

peaker used

er and ampliynchronousleristics are tbetween the

on the side ners. For a

ng surface, ram materiale mapped toexact samedominant a

rces emittine calibratedctra were ado backgrounated sources

t, conducted

for the perf

ifier combinly with signthereby adjue sources ca

window anpreferably realized usil instead of o an image

e positions. aero-acousti

ng the samed to radiate tdjusted to thnd noise ratis and a mod

3

4

2

1

d with 8 loud

formance te

ned with a nnals indepenustable by an also be va

nd the windsgood soundng a special

f glass. Howof a currenTo focus thic sources h

sound levethe same sohe backgrouio for all freified white

8 7 6

1

dspeakers.

ests.

new develondent from a set of pararied.

screen based radiation, al 1:1 test vewever, due tnt productionhe attention has been rem

els and freqound level inund noise o

requencies. noise signa

5 8

143

ped Matlabeach other.ameters for

ed on inten-all LSP are

ehicle mod-o confiden-n line vehi-on the LSP

moved from

quency con-n one meterof the windAll investi-

al.

b . r

-e ---P m

-r d -

14

Source LoThe sourceportance wsent, while

Figure 11 switched o

Figure

However, separation frequency be discusse

In Figure 1array, the sas well wit

On the sidThese devitical calibrIndeed, the

A frequencindicates d1/3-octavebrane, regatic source l

An optimisound sourlocation qu

44

ocalisation e localisatio

was thereby e applying th

shows the on, respectiv

e 11: Lo

for LSP wiis possibledependent ed below.

12 the localisource locath a flow of

de window, iations are mration of thee results also

cy dependedifferent dis-band. Thisarding the frlocalisation

ized matchirce with a vuality. An ac

on accuracyattached tohe shear lay

quality of vely emittin

ocalisation o1/3

ith the short. Two sourceffect can b

isation qualalisation on f 140 km/h.

slight devimost likely e array beco confirm a

nt evaluatiostances of ts effect resufrequency ran to a limited

ing proceduvery local sccuracy of a

y was regardthe compar

yer correctio

source locng sound sig

of 8 uncorre3-octave, f =

test distanceces appear abe controlle

lity is depicthe windscr

iations for tcaused by sause they a

a perfect fun

on of the sothe located ults probablange of inted extend.

ure for the sound radiatat least +/- 5

ded first conrison of theon.

calisation fognals with th

elated source= 4000 Hz, D

es, the locaas a single oed by the be

cted for singreen is exce

the localisasmall inaccuare nearly idnction of the

ource localiSPL maxim

ly from a vaerest. Theref

acoustical tion is alrea5 mm is exp

nducting the results wit

or the left ahe same lev

es with the sDR = 10 dB.

alisation is bone with a eamwidth an

gle LSP souellent for th

ation with thuracies in mdentical wite implemen

isation, as sma to the Lariable sounfore, LSPs a

and opticaady schedulpected.

Aachen Acou

e performanhout flow a

and the topel and frequ

same level a

biased becasource centnd the dyna

rces (sum lehe case with

he left arramatching thethout and w

nted shear la

shown in FiLSP membrand radiationare only sui

al calibratioled. It will

ustics Colloqu

nce tests. Gand with a f

p array withuency conte

at u = 140 km

ause no cleatre in betweamic range

evels). Withhout a flow

ay can be oe acousticalwith a flow ayer correcti

igure 13 forane centre n by the LSitable for th

on, using aimprove th

uium 2015

Great im-flow pre-

h 8 LSP ent.

m/h,

ar source een. This and will

h the top and just

observed. l and op-present.

ion.

r LSP 1, for each

SP mem-he acous-

a special he source

Aachen A

Fig

BeamwThe beabeamwitimationculated

In Figurthe beam

Acoustics Co

gure 12:

Figure 13

width amwidth is idth the shan of the beaaccording t

re 14, a typmwidth dete

lloquium 201

Localisatio

3: Freq

the basis frper a sourc

amwidth basto Sijtsma [

pical beamwermination

5

on of a single

quency depen(Scale 1

for the source appears ised on the b11] at a leve

width is plotin the beam

e source wit

ndent sourc:1, printed o

rce separatiin the map abasic array el decrease

tted for a fremforming m

thout and w

ce localisatioon DIN A4).

ion in a beaand the bettparametersof 3 dB.

equency of map at a 3 dB

ith a flow pr

on of LSP 1 w.

amforming ter a separat, dimension

800 Hz to iB level decr

1

resent, DR =

without flow

map. The tion is possin and focus

indicate therease. Due t

145

= 1 dB.

w

smaller theible. An es-can be cal-

e manner ofto the irreg-

e --

f -

14

ular designtherefore s

A comparigood agreebeamwidth

Figure 14: 1/3

So far, no mid frequeflow speedbig array. T

Source SepFor an aerois importanpreferably a 3 dB criemitting thbest localis

46

n of the Poources appe

ison of the ement and h of the Pors

Determ3-octave, f =

Fig

clear explaency regionds are probaThese effec

paration o-acoustic ant. Therefosmall beam

iterion or ahe same levsation accur

orsche arrayear asymme

estimated aa strong fr

sche arrays

mination of 800 Hz, DR

gure 16:

anation couln. The obserably causedts will be in

analysis of are a good s

mwidth. Figa visual evavel are causracy.

ys, the horietric in the m

and measurrequency demeasured a

the beamwiR = 3 dB.

Measured

ld be found rved higherd by the cornvestigated

a car, an indseparation oure 17 depi

aluation. Thsed by the u

zontal beammaps.

red beamwidependency at various fl

idth, Figurth

d beamwidth

for the sligr values for rrelation losin detail by

dividual andof the sourcicts the wayhe slight levused steerin

mwidth diff

dth, pictureof the bea

low speeds.

re 15: Ehe horizonta

h of the Pors

ghtly better the beamw

ss effects foy an already

d detailed evces is requiy of determivel differenng vector fo

Aachen Acou

fers from th

ed in Figuremwidth. Fi

stimation anal and vertic

sche arrays.

beamwidthwidth at highor distant mscheduled

valuation ofired, basicaining the sonces betweeormulation,

ustics Colloqu

he vertical

e 15, revealigure 16 sh

nd measuremcal beamwid

.

h at 200 kmher frequen

microphonesstudy.

f each singlally dependiource separaen the two

chosen to

uium 2015

one and

ls a very hows the

ment of dth.

m/h in the ncies and s using a

le source ing on a ation, by sources, give the

Aachen A

Fig

Figure 1analysisby step and the realised

Withoutual soursent, theready m

The useprove thand thefraction

DynamThe dynmicrophmicroph

Acoustics Co

gure 17:

18 shows ths without a reduction ovisual imp

d discrete di

t a flow, therce separatie source sep

mentioned co

e of advanche source seir side lobe

ns from the d

mic Range namic rangehones. For thones, irreg

lloquium 201

Determina

he capabilityflow. By si

of the sourcpression. Thstances betw

Figure 1sources o

e agreemenion is therebparation at orrelation lo

ed post-proeparation sies. By a sudirty map, t

e (DR) of anthe Porsche

gularly arran

5

ation of the s1/3-

y of the Porimulations, ce distance hese results ween the lou

18: Sepan the side w

nt between sby nearly idhigher spee

oss effects.

ocessing tecignificantly

uccessive suthe originall

n array is marrays, a re

nged in an o

source separ-octave, DR

rsche arraythe limitatifor each 1/are comparudspeakers

aration of hwindow with

simulations dentical to eds and freq

chniques, e.y. Weaker soubtraction oly masked s

mainly definemarkably hoptimized la

ration for L= 3 dB.

s to separations of sepa/3-octave-bared with the.

orizontal arout flow, 1/3

and measuthe horizonquencies is

g. algorithmources may

of the domisources can

ed by the nuhigh DR couayout as des

LSP 5 and LS

te sources waration wereand, evaluate measurem

rranged 3-octave.

rements is vntal beamwi

also slightl

ms like CLEy be maskednant sourcebe revealed

umber and tuld be achie

scribed befo

1

SP 6 withou

with the same determineting the 3 d

ments, repre

very good aidth. With ly disturbed

EAN SC [1d by domines and theird.

the arrangeeved by the

ore.

147

ut flow,

me level fored by a stepdB criterionesenting the

and the vis-a flow pre-d by the al-

2], can im-nant sourcesr correlated

ment of thee use of 192

r p n e

---

-s d

e 2

14

Figure 19 large scan 1m. These Function) lbased on th

The differethe loudspmeasureme

Figure 19:out flow

Typical RIn Figure 2array and tSC algorith

Small devione scan pthis plane aresentationmodel of tconsideredtest object

SummarFor the Nestalled. Byhigh systemacquisitiondemonstrat

To improvferential- aa frequency

48

and Figureplane of apresults are

located at thhe CSM wit

ences obsereaker mements. Advan

Beamfw, 1/3-octave

Results 21 and Figuthe top arrahm are clear

iations of thlane definedare therefor

n in the beathe test objed for every

surface.

ry and Ouew Porsche

y a complexm rigidity, an is achieveted and the

ve the data aand correlaty dependen

e 20 show tprox. the sicompared

he position th diagonal

rved are prombrane and nced algorith

forming mape, f = 6300 H

ure 22 the vy. All experly visible f

he source lod in a certaire calculatedmforming mect will be single grid

utlook e wind-tunnx design andan outstanded. Based olimitations

analysis signtion-beamfont shading pr

he measureze of the arto the DR oof LSP 6. Aremoval inc

obably causadditionallyhms like CL

p for LSP 6 Hz, DR = 25 d

valuation of cted sound for all frequ

cation in thin distance td with an inmap. Henceestablishedpoint, and

nel, a uniqud a perfect ding basic aon a study were determ

nificantly, forming [3], rocedure fo

ed DR regarrays (5 x 3 of a simulatAll analysecluding a po

sed by the iy by reflectLEAN SC [

with-dB.

Figu

f a productiosources and

uencies.

his overviewto the partic

ncorrect foce, to prevend as a standthus for all

ue microphofixing to th

acoustic perwith LSP smined.

further adva[10], have b

or distant mi

arding the sm) and for ted monopo

es are carrieositive effec

inhomogenetion and da[12] can be u

ure 20:arrays w

on line vehid the impro

w plots resulcular array. cus leading tnt this effectdard method sources wi

one array she wind tunformance asources, the

anced evalubeen implemicrophones

Aachen Acou

ound radiatthe optimiz

ole sound soed out in thect to the DR

eous local smping effeused to incr

Dynamic rawithout flow,

icle is pictuovements ap

lt from a focSources thato a slightlyt, the sourc

d. Thereby tith an assum

system was nel walls, cnd a comfoe capability

ation technimented andwas started

ustics Colloqu

tion of LSPzed windowource (Poine frequencyR.

sound generects performrease the DR

ange of the P, 1/3-octave.

ured for the pplying the

cus adjustedat are not loy biased soue mapping the correct med locatio

developedcombined wortable and y of the arr

iques, like 3d an investigd.

uium 2015

P 6 for a w of 1.5 x nt Spread y domain

ration of ming real R.

Porsche .

left side CLEAN

d to only ocated in urce rep-on a 3D focus is

on on the

d and in-with very fast data rays was

3D-, dif-gation of

Aachen AAcoustics Co

Figure 2

lloquium 201

21: Eva

5

aluation of a

a productionn line vehiclee by the left

1

t side array.

149

1550

Figure 222: Evalluation of a

production line vehicle

Aachen Acou

by the top a

ustics Colloqu

array.

uium 2015

Aachen A

AcknoNumero

Perfect for the ways fo

I am pamance ovice at a

My specin the coeffort. Fduring tsupport

Many thand to tsystem p

Refere Am[1]

N Eh[2]

ac D[3]

m D[4]

fo Le[5]

AuBe

O[6]ra

O[7]sis

Pu[8] Sa[9]

V Sc[10]

in Si[11]

A Si[12] Si[13]

13 St[14]

W15

Acoustics Co

owledgemous people h

appreciatioimplementa

ocused on th

articular graof the company time.

cial thanks ontext of hiFurthermorethe whole pselecting th

hanks also the project tpossible.

ences miet, R. K.: o. 2, pp. 467hrenfried, K.coustic sourcöbler, D., Pu

ment of a car iougherty, R.

orming”, AIAechner, W.: “uflösung underlin, 2015. erlemans, S.

ays in open anerlemans, S.s, Universityuhle, C.: “Übarradj, E.: “Tector Formuchmidt, S., D

ng”, inter.noisijtsma; P.: “Edvances in Aijtsma, P.: “Cijtsma, P.: “A3, 2012. tumpf, H., Rö

Wallmann, S.:5. Internation

lloquium 201

ment have genero

on to all colation of thishe best techn

ateful to Miplex and uni

to Wim Leis bachelor te, I want toproject and he most suit

to my colleteam of the

“Refraction 7-482, 1978. ., Koop, L.: “

ces”, AIAA puhle, Ch., Hein a wind tun P. “Extensio

AA Paper 200“Evaluierungd Lokalisieru

, Broersma, nd closed wi: “Detection

y of Twente, ber die ScherThree-Dimenulations”, AdvDöbler, D.: “Vse conferenc

ExperimentalAeroacoustic CLEAN baseAcoustic beam

öser, P., Wie: “The new anales Stuttga

5

ously given

lleagues of s great arraynical solutio

ichael Hartmique optimi

chner for hthesis and tothank CarsGünther N

table loudsp

eagues of Pe New Porsc

of Sound by

“A comparispaper 2006-2eilmann, G.: “nnel using CLon of DAMA05-2961, 200g der Leistunungsgenauigk

L., Sijtsma, Pind tunnels”,of aero-acouEnschede, 2

rschichtkorrensional Acouvances in AcVisualizationce, San Francl techniques and Applica

ed on spatial mforming fo

egand, T., Pfaerodynamic rter Symposi

time, energ

gfai tech Gy system. Mon.

tmann fromzation work

his help condo Dr. Siller sten Spehr (

Nubert (Nubpeakers for

Porsche’s acche wind-tu

y a Shear Lay

son of iterativ2711, 2006. “CorrelationLEAN-SC”, AS and Bene05. ngsfähigkeit keit im neuen

P.: “Quantifi, NLR Reporustic sound s

2009. ektur nach A

ustic Source Mcoustics and n of small decisco, 2015.for identifica

ations, VKI Lsource coher

or the routing

fäfflin, B., Ocand aero-acoium, 2012.

gy and encou

GmbH and GMany thank

m Volkswagk finding th

ducting and(TU Berlin

(DLR Göttibert electronthe perform

coustic depaunnel for m

yer”, Journal

ve deconvolu

n of high chaninter.noise c

efits and Lim

einer akustisn Porsche W

fication of airrt NLR-TP-2sources on ai

Amiet”; TechnMapping witVibration, V

esign modific

ation and chaLecture Serierence”, AIAAg of aircraft n

cker, J., Müloustic wind t

uragement t

GFaI e.V. is for the ve

en group ree best array

d evaluationn) supportiningen) for thnic GmbH) mance tests.

artment for making the r

of Sound an

ution algorith

nnel count bconference, S

mitations of D

schen KamerWindkanal”, B

rframe noise 2007-799, 20ircrafts and w

nical Documh Different B

Volume 2012cations using

aracterisationes, 5:15-19, 2A Paper 200noise”; NLR-

ller, R., Ecketunnel of the

1

to this proje

involved in ery good tea

esearch for y layout and

n the performng this thesishe valuablefor his unc

their absolrealization o

nd Vibration,

hms for the m

beamformingSan FranciscDeconvolutio

ra hinsichtlicBachelor The

using micro007. wind turbines

ment GFaI, 20Beamformin2. g differential

n of noise so2004.

07-3436, 200R-Report, NL

ert, W., Kroße Porsche AG

151

ect.

the projectamwork al-

the perfor-d for his ad-

mance testss with greate consultingcomplicated

ute supportof the array

, Vol. 58,

mapping of

measure-o, 2015

on in Beam-

ch Dynamik, esis, TU

ophone ar-

s”, PhD the-

014. g Steering

beamform-

ources “,

7. R-TP-2012-

ß, H.-U., G”,

t -

--

s t g d

t y