simulation to streamline cellular phone...
TRANSCRIPT
Reprinted with permission from the
Spring 2007 edition of
Simulation to Streamline Cellular Phone Development
Ideas and Strategies in Product Development
Ideas and Strategies in Product Development
An Publication
SP
RI N
G 2 0 0 7
Altair Engineering, Inc.
2
6
15
c2rCover 7 Winter-2007:c2rCover 6 Winter/2007 4/19/07 9:17 AM Page covA
Simulation plays a critical role in product
development, with a broad impact from package-
level simulation to system-level modeling. To put
this into perspective, typical design cycles decreased
from about 18 m
onths in the late 1990s to
under a year in 2005. While cycle tim
e has
compressed, the functionality of the devices has
grown tremendously. In order to design new products
in fewer than 10 months, which is a goal at Motorola,
the traditional “build-and-test” approach has been
widely replaced by up-front simulation.
Finding Part Interferences
Cellular phone models typically consist of
hundreds of components packaged within a small
space. Often, after a mesh has been completed, there
are portions of the mesh (elements and nodes) that
interfere with each other. These interferences occur
either because of underlying geometry or result from
the tessellation of the geometry while being meshed.
Using the traditional process, an analyst needed up
to three days to manually find and correct the
interferences in the virtual model. Moreover, these
interfering nodes and elements can often be nearly
impossible to detect manually.
If these interfering elements are allowed to remain
in the model, it can lead to errors during the execution
of the model. This results in unacceptable delays to
debug the problem – and wasted central processing
unit (CPU) cycles on the compute server.
To find and correct these issues, a custom interference-
checking program was built, based on Motorola’s
internal requirements. The proprietary “interference
matrix” shows modeling errors in a graphical format,
and the “interference solution graphical user
Figure 1. Typical FEM model of a cellular phone.
Motorola, Inc. – founded in 1928 and recognized
around the world for innovation and leadership
in wireless and broadband communications –
pioneered mobile communications in the 1930s with car
radios and public safety networks, coining the company
brand that suggests “sound in motion.” With
approximately 68,000 employees, Motorola directs the
activities of 320 facilities in 73 countries from its global
headquarters in Schaumburg, IL.
For more than 75 years, the company has remained
committed to engineering innovation. Harnessing the
power of technology convergence, Motorola made the
equipment that carried the first words back from the
moon in 1969. Motorola led the communications
revolution with the 1983 introduction of the first
commercial cellular phone. Reportedly, the company
spends as much as 10 percent of its annual income in
research and development efforts.
Today, Motorola offers market-changing icons of
personal technology and is a leader in multinode, multi-
band communications products and technologies. The
Plantation, FL-based Mobile Devices division designs,
manufactures, sells and services wireless subscriber
and server equipment for cellular systems; portable
energy storage products and systems; servers and
software solutions; and related software and accessory
products.At M
otorola, that means designing and delivering
“must-have” products, must-have experiences and
powerful networks to keep people connected.
Motorola
at a Glance
12
Concept To Reality/ S
pring 2007
www.altair.co
m/c2
r
MOTOROLA 6 dave_BB:MOTOROLA3 4/19/07 10:09 AM Page 12
Upstre
amm
odel sim
ulatio
n
reduce
s prod
uct d
evelo
pmen
t tim
e
and e
nables
Mot
orola
to m
eet
critic
al m
arke
t dem
ands
.
to S
trea
mlin
e
Dialing
into
Simula
tion
Cellu
lar Pho
ne
Deve
lopmen
t
IInsp
ired b
y a vi
sion of
“sea
mless m
obili
ty,” g
lobal
commun
icatio
ns lead
er M
otorol
a has
transfo
rmed
the
ubiqu
itous
cell p
hone o
f the l
ast ce
ntury i
nto tod
ay’s
univ
ersa
l rem
ote c
ontrol f
or li
fe. R
evol
utio
nizing
broad
band,
embe
dded
syste
ms and w
ireles
s netw
orks,
the c
ompa
ny that
pionee
red th
e firs
t port
able
two-w
ay
radio
for th
e U.S. A
rmy i
n 1943
now help
s indiv
iduals
reach
the p
eople
, things
and i
nformati
on th
ey nee
d in
their h
omes,
cars,
workpla
ces an
d eve
rywhe
re in
betw
een.
Today
’s cell
phon
es are
incre
dibly
comple
x, with
multiple
printe
d-circ
uit boar
ds and la
rge,
multi-
functi
onal
liquid
cryst
al dis
plays
(LCDs).
They
also
have t
he abil
ity to
supp
ort m
ultipl
e data
prot
ocols
such
as co
de div
ision m
ultiple
acce
ss (C
DMA),
globa
l syst
em fo
r mob
ile co
mmunica
tions (
GSM)
and W
orldwid
e Inte
roper
abili
ty fo
r Micr
owave
Access
(WiM
AX) – al
l in a
single
, slee
k han
d-held
devic
e. At t
he sam
e tim
e, de
signs m
ust be
rugg
ed an
d
10
Conce
pt To
Realit
y / Spr
ing
2007
www.altair.co
m/c
2r
by Dr. K
inzy Jones,
Doug Carroll,
Phil Green,
Chris Bates a
nd Dr. Marc Zampino
MOTOROLA 6 dave_BB:MOTOROLA3 4/19/07 10:09 AM Page 10
PR
OC
ES
SA
UT
OM
AT
I ON
(CAE) e
arlier
in th
e desi
gn cy
cle. F
or ex
ample
, the
compa
ny is u
sing A
ltair
Hyper
Wor
ks fo
r up-
front
simula
tion al
ong w
ith pr
ototyp
e test
ing to a
id in th
e
develo
pment p
roce
ss. D
oing s
o has
enab
led th
e
compa
ny to c
aptu
re pr
oduc
t-life
behav
ior, p
redict
areas
needin
g impro
vemen
t and g
enera
te high
-quali
ty
produ
cts w
ithin a
greatl
y red
uced
time f
rame.
Captur
ing prod
uct-l
ife be
havior
is a d
auntin
g task
, so
simula
tion is
combin
ed w
ith pr
ototyp
e test
ing to h
elp
in deve
loping t
he prod
uct. T
he com
monly
used p
hrase
to des
crib
e the e
xtern
al lo
adin
g that
the p
hones
expe
rience
durin
g test
ing is “
large
-moti
on, sm
all-ti
me
scales
.” In u
sing s
imula
tion t
o aid
in pr
oduc
t dev
elopm
ent,
the c
once
rn is
in de
termin
ing w
hat lev
el of
detai
l
shou
ld be
inclu
ded i
n the m
odel
to ca
pture
the r
eleva
nt
physi
cs. M
oreo
ver,
since
it is
extre
mely di
fficu
lt to
predic
t acti
ons t
hat will
result
from
a 5,0
00-g,
0.1-m
s
impa
ct pu
lse, in
tuitio
n is of
little
help.
With
toda
y’s m
odern
algo
rithms a
nd vast
compu
ting
power,
simula
tions n
eed n
ot be
simpli
stic m
odels
of
the a
ctual
problem
. Howev
er, t
his ca
n lead
to a
number of c
hallen
ges w
hen lo
oking a
t even
the
simple
st ce
llular
phon
e mod
els. S
ince
the b
ehav
ior
of man
y of t
he mat
erial
s var
ies as
the s
train
rate
chan
ges, a
dditi
onal
materia
l mod
eling m
ust be
done.
In ad
dition, i
f man
ufactu
ring v
ariab
les m
ust be
consid
ered,
the c
hallen
ges a
re ev
en m
ore co
mplex.
In an
y cas
e, while
it is
an in
timid
atin
g tas
k to
predict
drop per
form
ance
, it i
s not i
mpossible
to
manag
e its
analy
sis. A
nd, as
diffic
ult as
the s
olutio
n
techn
iques
can be
, mod
el set
up is
the la
rgest c
ontri
butor
to th
e ove
rall le
ngth of
the s
olutio
n proc
ess.
Up-Front D
esign
At M
otoro
la, t
he use
and c
ustom
izatio
n of
advan
ced C
AE so
ftwar
e hav
e stre
amlin
ed th
e
mechan
ical p
rototy
pe pr
ocess
while
grea
tly in
creasi
ng
desig
n robu
stness
. Sign
ifican
tly de
creasi
ng sim
ulatio
n
time h
as en
abled
Moto
rola e
nginee
rs to
identif
y and
corre
ct pote
ntial i
ssues
early
in th
e dev
elopmen
t
cycle
. Red
uctio
n in si
mulatio
n cycle
time i
s also
critic
al to
short
en tim
e to m
arket,
which
, in tu
rn, h
as
a posi
tive i
mpact
on th
e com
pany’s
botto
m line.
Mor
eove
r, th
e acc
elerat
ion of
the m
odel-
build
ing
proc
ess al
lows s
evera
l sim
ulatio
n passe
s on co
ncept
geom
etry b
efore
costl
y injec
tion m
oldin
g too
ls nee
d
to be
built
– a l
uxur
y that
even
two y
ears
ago w
as
impo
ssible
. For
exam
ple, a
new ce
llular
phon
e mod
el
consis
ting o
f more
than
70 un
ique c
ompo
nents
took
16 hou
rs to
go fr
om ge
ometr
y to a
fully
read
y-to-
run
simula
tion m
odel,
rath
er th
an th
e stan
dard
20-d
ay
model
deve
lopmen
t time. T
his is a
tenfol
d impro
vemen
t
in sp
eed i
n less
than
a ye
ar – an
d has
remov
ed th
e
model dev
elopmen
t cyc
le as
the g
atin
g set
of the
prot
otyp
e sim
ulatio
n proc
ess. S
imula
tion ca
n now
beco
me an up
-fron
t desi
gn to
ol.
On avera
ge, 4
0 syst
em-le
vel d
rop sim
ulatio
ns (Fig
ure
1 on pa
ge 12
) are pe
rform
ed be
fore a
desig
n is rel
eased
for pl
astic
injectio
n tooli
ng. These
simula
tions a
llow th
e
engin
eers
to under
stand th
e per
form
ance
of the
displa
ys, el
ectri
cal c
ompo
nents
and p
lastic
housin
gs.
In 20
00, f
inite
-elem
ent m
esh cr
eatio
n, con
tact
and co
nstrain
t defi
nition, m
odel deb
uggin
g and
result
inter
preta
tion w
ere h
ighly
manua
l pro
cesse
s
and to
ok more
than
two m
onths t
o perfo
rm. A
t
that
tim
e, A
ltair
Hyp
erW
orks w
as ch
osen to
auto
mate t
ime-
consu
ming m
anual
task
s, su
ch as
auto
mated
geomet
ry cl
eanup, m
esh ge
nerat
ion,
inter
feren
ce ch
eckin
g or p
ost-p
roce
ssing.
Using t
he
softw
are r
esulte
d in si
gnifi
cant t
ime s
avin
gs
and r
educ
ed m
odel
setup
time f
rom 20
days
to on
ly
two d
ays.
built
to w
ithsta
nd the d
aily e
quiva
lent o
f a co
mputer
worksta
tion an
d mon
itor t
aking a
five-f
oot f
all.
Like m
any o
f its
compe
titor
s, M
otor
ola is
drive
n
by shorte
ned des
ign cy
cles
, incr
easin
g pro
duct
comple
xity a
nd red
uced
profi
t marg
ins. Desi
gn cy
cles
that
took
18 m
onth
s in 19
98 fr
om dr
awin
g boa
rd to
plant f
loor a
re now
push
ing tow
ard a
six-m
onth
goal
to m
eet b
usin
ess-cr
itica
l pro
duct
launch
es an
d fixe
d
ship
dates
. Abo
ve al
l, to m
aintai
n a hea
lthy p
rofit
margin, M
otorol
a must
keep
down th
e ove
rall c
ost of
producin
g the p
roduct
, esp
ecial
ly to
compet
e in
coun
tries
where w
ireles
s pro
viders
rarel
y sub
sidize
the c
ost of
the p
hone.
Real-World
Reliabilit
y
To unde
rstan
d “rea
l-wor
ld” re
liabil
ity w
ell be
fore
next-g
enera
tion ce
ll phon
es hit t
he shelv
es, M
otorol
a
has in
crease
d its
use o
f com
puter
-aide
d engin
eerin
g
11
www.altair.co
m/c
2r
Conce
pt To
Realit
y / Spr
ing
2007
MOTOROLA 6 dave_BB:MOTOROLA3 4/19/07 10:09 AM Page 11
Upstream model simulationreduces product development timeand enables Motorola to meet
critical market demands.
to Streamline
Dialing into
SimulationCellular PhoneDevelopment
IInspired by a vision of “seamless mobility,” global communications leader Motorola has transformed theubiquitous cell phone of the last century into today’suniversal remote control for life. Revolutionizingbroadband, embedded systems and wireless networks,the company that pioneered the first portable two-wayradio for the U.S. Army in 1943 now helps individualsreach the people, things and information they need intheir homes, cars, workplaces and everywhere in between.
Today’s cell phones are incredibly complex, withmultiple printed-circuit boards and large, multi-functional liquid crystal displays (LCDs). They alsohave the ability to support multiple data protocolssuch as code division multiple access (CDMA),global system for mobile communications (GSM)and Worldwide Interoperability for Microwave Access (WiMAX) – all in a single, sleek hand-helddevice. At the same time, designs must be rugged and
10Concept To Reality / Spring 2007 www.altair.com/c2r
by Dr. Kinzy Jones, Doug Carroll, Phil Green,Chris Bates and Dr. Marc Zampino
MOTOROLA 6 dave_BB:MOTOROLA3 4/19/07 10:09 AM Page 10
Reprinted with permission from the
Spring 2007 edition of
P R O C E S S A U T O M A T I O N
(CAE) earlier in the design cycle. For example, thecompany is using Altair HyperWorks for up-front simulation along with prototype testing to aid in thedevelopment process. Doing so has enabled the company to capture product-life behavior, predict areas needing improvement and generate high-qualityproducts within a greatly reduced time frame.
Capturing product-life behavior is a daunting task, sosimulation is combined with prototype testing to helpin developing the product. The commonly used phraseto describe the external loading that the phonesexperience during testing is “large-motion, small-timescales.” In using simulation to aid in product development,the concern is in determining what level of detailshould be included in the model to capture the relevantphysics. Moreover, since it is extremely difficult topredict actions that will result from a 5,000-g, 0.1-msimpact pulse, intuition is of little help.
With today’s modern algorithms and vast computingpower, simulations need not be simplistic models ofthe actual problem. However, this can lead to a number of challenges when looking at even the simplest cellular phone models. Since the behaviorof many of the materials varies as the strain ratechanges, additional material modeling must be done.In addition, if manufacturing variables must be considered, the challenges are even more complex.
In any case, while it is an intimidating task to predict drop performance, it is not impossible tomanage its analysis. And, as difficult as the solutiontechniques can be, model setup is the largest contributorto the overall length of the solution process.
Up-Front DesignAt Motorola, the use and customization of
advanced CAE software have streamlined the mechanical prototype process while greatly increasingdesign robustness. Significantly decreasing simulationtime has enabled Motorola engineers to identify andcorrect potential issues early in the development cycle. Reduction in simulation cycle time is also critical to shorten time to market, which, in turn, hasa positive impact on the company’s bottom line.
Moreover, the acceleration of the model-buildingprocess allows several simulation passes on conceptgeometry before costly injection molding tools needto be built – a luxury that even two years ago wasimpossible. For example, a new cellular phone modelconsisting of more than 70 unique components took16 hours to go from geometry to a fully ready-to-runsimulation model, rather than the standard 20-daymodel development time. This is a tenfold improvementin speed in less than a year – and has removed themodel development cycle as the gating set of the prototype simulation process. Simulation can nowbecome an up-front design tool.
On average, 40 system-level drop simulations (Figure1 on page 12) are performed before a design is releasedfor plastic injection tooling. These simulations allow theengineers to understand the performance of the displays, electrical components and plastic housings.
In 2000, finite-element mesh creation, contactand constraint definition, model debugging andresult interpretation were highly manual processesand took more than two months to perform. At that time, Altair HyperWorks was chosen to automate time-consuming manual tasks, such as automated geometry cleanup, mesh generation,interference checking or post-processing. Using thesoftware resulted in significant time savingsand reduced model setup time from 20 days to onlytwo days.
built to withstand the daily equivalent of a computerworkstation and monitor taking a five-foot fall.
Like many of its competitors, Motorola is drivenby shortened design cycles, increasing product complexity and reduced profit margins. Design cyclesthat took 18 months in 1998 from drawing board toplant floor are now pushing toward a six-month goalto meet business-critical product launches and fixedship dates. Above all, to maintain a healthy profitmargin, Motorola must keep down the overall cost ofproducing the product, especially to compete incountries where wireless providers rarely subsidizethe cost of the phone.
Real-World ReliabilityTo understand “real-world” reliability well before
next-generation cell phones hit the shelves, Motorolahas increased its use of computer-aided engineering
11www.altair.com/c2r Concept To Reality / Spring 2007
MOTOROLA 6 dave_BB:MOTOROLA3 4/19/07 10:09 AM Page 11
Simulation plays a critical role in product development, with a broad impact from package-level simulation to system-level modeling. To putthis into perspective, typical design cycles decreasedfrom about 18 months in the late 1990s to under a year in 2005. While cycle time has compressed, the functionality of the devices hasgrown tremendously. In order to design new productsin fewer than 10 months, which is a goal at Motorola,the traditional “build-and-test” approach has beenwidely replaced by up-front simulation.
Finding Part InterferencesCellular phone models typically consist of
hundreds of components packaged within a smallspace. Often, after a mesh has been completed, thereare portions of the mesh (elements and nodes) thatinterfere with each other. These interferences occureither because of underlying geometry or result fromthe tessellation of the geometry while being meshed.Using the traditional process, an analyst needed upto three days to manually find and correct the interferences in the virtual model. Moreover, theseinterfering nodes and elements can often be nearlyimpossible to detect manually.
If these interfering elements are allowed to remainin the model, it can lead to errors during the executionof the model. This results in unacceptable delays todebug the problem – and wasted central processingunit (CPU) cycles on the compute server.
To find and correct these issues, a custom interference-checking program was built, based on Motorola’s internal requirements. The proprietary “interferencematrix” shows modeling errors in a graphical format,and the “interference solution graphical user
Figure 1. Typical FEM model of a cellular phone.
Motorola, Inc. – founded in 1928 and recognizedaround the world for innovation and leadershipin wireless and broadband communications –
pioneered mobile communications in the 1930s with carradios and public safety networks, coining the companybrand that suggests “sound in motion.” Withapproximately 68,000 employees, Motorola directs theactivities of 320 facilities in 73 countries from its globalheadquarters in Schaumburg, IL.
For more than 75 years, the company has remainedcommitted to engineering innovation. Harnessing thepower of technology convergence, Motorola made theequipment that carried the first words back from themoon in 1969. Motorola led the communicationsrevolution with the 1983 introduction of the firstcommercial cellular phone. Reportedly, the companyspends as much as 10 percent of its annual income inresearch and development efforts.
Today, Motorola offers market-changing icons ofpersonal technology and is a leader in multinode, multi-band communications products and technologies. ThePlantation, FL-based Mobile Devices division designs,manufactures, sells and services wireless subscriberand server equipment for cellular systems; portableenergy storage products and systems; servers andsoftware solutions; and related software and accessoryproducts.
At Motorola, that means designing and delivering“must-have” products, must-have experiences andpowerful networks to keep people connected.
Motorola at a Glance
12Concept To Reality / Spring 2007 www.altair.com/c2r
MOTOROLA 6 dave_BB:MOTOROLA3 4/19/07 10:09 AM Page 12
Reprinted with permission from the
Spring 2007 edition of
P R O C E S S A U T O M A T I O N
Figure 4: Green color indicates a “clean” model.
interface (GUI)” supports the analyst in resolvingmesh problems.
Using the custom tool, the analyst can select thecomponents that are of interest and generate the interference matrix (Figure 2). Then, each pair ofcomponents can be visualized by selecting the appropriate box on the GUI. This generates the populated interference solution GUI. A table thendisplays the nodes that interfere, as well as theamount of the overclosure. (This is shown in whitein Figure 3).
This GUI also allows analysts to set themaster/slave relationship of the surfaces. Once all theinterference pairs are fixed, the interference matrixwill show all green (Figure 4). That signals the modelis clean and ready to run.
As a result of this automation, model checks cannow be performed in just one hour, compared to thetraditional 15 to 20 hours. This represents an approximately 80-percent to 90-percent time savings.
Automating Post-ProcessingNow that model development has been largely
automated and compute server speeds are faster thanever, post-processing of drop simulations remains themost time-consuming manual process. Interpretationof the results is the most critical step, and as such,there are no plans to automate that process. However, Motorola analysts agreed that HyperWorks’ automatic report generation – consistingof images and 3D interactive animation result filesusing Altair’s compact .h3d format – greatly speedsup the process.
Using HyperWorks’ automatic reporting capabilities,analysts were able to create a standard report of drop-simulation results. The tool creates graphical, top-leveloverviews about potential failures from high-impactforces for every part of the cell phone.
The standard report contains an isometric view ofthe phone and cross-sections of the three main axes.Designers can interactively animate the drop in 3D tovisualize the motion of the parts with a Web browseror PowerPoint using Altair HyperView Playerfreeware, a 3D web browser plug-in with ActiveXcontrols.
Figure 3. Interfering nodes.
Figure 2. A custom interference-checking program, built toMotorola’s internal requirements, speeds the model-checkingprocess. The screen above shows the interference matrix for theoverclosure process.
13www.altair.com/c2r Concept To Reality / Spring 2007
MOTOROLA 6 dave_BB:MOTOROLA3 4/19/07 10:09 AM Page 13
P R O C E S S A U T O M A T I O N
As a result, analysts and designers can quickly review the phone’s structural integrity and eliminatethe risk of “missed results” due to the simulationmodel’s complexity. The parts that have failed theirrespective material criteria are displayed on the firstpage, as are the number of failed elements. All thecomponents can be viewed by selecting ALL on thebottom of the main page. Clicking on any of thecomponents launches another page with detailed information on the component, including the valuesof interest for the entire event.
In the future, further automation of the model development process will focus on the use of AltairData Manager. This new system will allow continualupdates of the part meshes through the integration of
Using advanced model development tools, simulation hasmoved from a design validation tool to a design tool,capable of generating system-level drop simulations so
quickly that designs can be modified daily. Traditionally, cycle timeran up to eight weeks.
The model (top right) consists of 77 unique components, includingtwo LCDs and two populated printed-circuit boards. It consists ofshell, hexagonal and tetrahedral meshes.
For the first phone concept geometry, it tooktwo working days from the creation of the InitialGraphics Exchange Specification (IGES)geometry to full drop-simulation results (9.5hours running 12 CPUs on an SGI Altix computeserver). Using the automated geometry-cleanupand mesh-generation tools, it took eight hoursto clean and mesh the parts.
An additional eight to 10 hours were neededfor collision detection and correction, as well asto create surface constraints and boundaryconditions. The total time for the first conceptmodel setup was approximately 20 hours.Several simulations were run with this geometryto test early design concepts.
Improvements to the initial design, labeledConcept Two geometry, were usually available aweek after the first set of results. The models(below right) show the level of part complexity inConcepts One and Two.
Only the new or modified geometry must bemeshed, in this case, only 54 components. Ittook eight hours to go from the computer-aideddesign package exported IGES files to acompleted finite-element model that had passed a
Prototypes Now Drive Design
To receive a copy of the MotorolaHyperWorks success story, visit www.altair.com/c2r or
check 03 on the reply card.
data check. Fifteen hours were required to run one of thestandard drop orientations on 12 CPUs on the SGI Altixcompute server.
In this example, then, the model went from IGES geometry topost-processing of the .h3d file – generated using AltairHyperView Player freeware – in less than 24 hours. Advancedmodel simulation tools reduced the overall CAE cycle time fromeight weeks to 24 hours – a significant savings in time and costs.
the computer-aided design life management tool. Theavailable customization of the tool will allow for manynew additional avenues for process automation.
Dr. Kinzy Jones is a Distinguished Member of theTechnical Staff; Doug Carroll is a Distinguished Memberof the Technical Staff; Phil Green is Senior Staff Engineer;Chris Bates is Senior Staff Engineer; and Dr. MarcZampino is Principal Staff Engineer. All are part ofMotorola – iMDC’s Advanced Mechanics Group inPlantation, FL.
Concept One Back Housing Concept Two Back Housing
Models generated from automated mesh generation and collision detection.
14Concept To Reality / Spring 2007 www.altair.com/c2r
MOTOROLA 6 dave_BB:MOTOROLA3 4/24/07 2:50 PM Page 14
Reprinted with permission from the
Spring 2007 edition of
United States (HQ) Australia
BrazilCanadaChinaFrance
GermanyIndiaItaly
JapanKoreaMexicoRussiaSpain
SwedenUnited Kingdom
www.altairproductdesign.com