3-Way Loudspeaker Design and Construction

Alex Kulyk

PHYS 406 5/10/12

I. Introduction

The objective of this project was to design and build a stereo pair of 3-way “hi-fi”
speakers using off-the-shelf drivers and an original design for the speaker cabinets. A crossover
network would also be designed and implemented to compliment the characteristics of the
speakers. The primary purpose of this design exercise was the associated learning experience,
and good audio fidelity of the final product was not of the utmost importance.

While the quality of the end result was not the most important factor, design decisions
were based around constructing speakers that follow accepted design criteria for good
performance. To this end, the goal was a flat frequency response in order to allow for acceptable
listening for a wide variety of music.

II. Initial Design Considerations

From the begging of the project, the plan was to design a vented 3-way speaker with a
relatively flat frequency response of the range of approximately 50 Hz- 20 kHz. For aesthetic
reasons, it was decided that the speaker cabinet be a traditional rectangular prism shape with a
relatively narrow width compared to most commercially available 3-way, floor standing
speakers. Because of the width constraint, smaller bass drivers would have to be used. In order
not to sacrifice low frequency output, two identical woofers would be used in each cabinet. The
mid and high frequency bands would be handled by a single midrange driver and tweeter,
respectively.

A vented design was chosen in order to compensate for the higher low-frequency roll-off
of the smaller bass drivers. In order to vertically accommodate 4 drivers and a vent per cabinet,
and considering the necessary volume, the speaker would also be relatively tall, which was also
part of the aesthetic goal of the project. The two, smaller woofer arrangement was also chosen
for personal preference, as the designer prefers a tight, fast bass response over the more powerful
and extended, but slower response of larger woofers

The crossover would be designed specifically for the drivers chosen, rather than using an
off-the-shelf, preassembled crossover, in order to allow for flexibility in component choice.
From the outset, the crossover design process seemed daunting due to the inherent non-ideal
behavior of passive electronic components in complex audio circuits, and also due to the
designer’s inexperience with circuit design.

1 
 
III. Driver Selection

As previously stated, each of the speaker cabinets required 2 woofers, 1 midrange and 1
tweeter. A per-driver budget was decided upon, and the following drivers were chosen:

a. Dayton Audio RS180S-8 7” Shielded Woofer

The RS180S-8 was chosen for its combination of small size and low frequency
extension. The external width of the speaker cabinets was set at roughly 8.5”, and
the RS180S-8 is one of the largest woofers available that would fit comfortably in
that dimension. The cone is made of aluminum, resulting in a favorable stiffness
to weight ratio, allowing for larger cone excursion before the onset of distortion
without the sluggish sound of heavier paper cones, which is especially important
in drivers of this size. It also allows the cone to reach higher frequencies before
cone break-up occurs, which happens when the cone ceases to move as a piston.
This higher break-up frequency allows for more flexibility when choosing
crossover frequencies, as will be discussed latter.

b. SEAS Prestige MCA12RC-H1304 Midrange Driver

This midrange driver was chosen for its exceptionally wide and flat frequency
response. The midrange is one the most important range to produce accurately, as
the fundamental frequencies and important harmonics of many instruments fall
directly in this band. The midrange is also the band in which humans are most
sensitive to distortion. The SEAS has an almost flat frequency response from
approximately 200Hz up to its breakup around 10kHz, again giving good
flexibility for choosing both crossover frequencies.

c. SEAS prestige 29TFF/W-H1318 Tweeter

This tweeter was chosen for its flat frequency response and for the fact that it is
one of the few tweeters in its price range that extends beyond 20 kHz. While most
humans can’t hear beyond 20 kHz, and many are limited to significantly lower
frequencies, a higher tweeter breakup frequency generally corresponds to a flatter
response in the audible range. This tweeter also makes use of a waveguide to
cover the magnet mounting screws in order to reduce diffraction. Another
important factor in its selection was its low free air resonance frequency, which
allows for more flexibility in choosing the upper crossover frequency.

IV. Cabinet Design

The cabinet design was to be tailored to the characteristics of the chosen drivers. It was
helpful to have the external aesthetic design requirements before attempting to fully design the
speaker cabinets, as it limited the number of degrees of freedom in the design process.

2 
 
 With the width fixed, the depth was chosen so that the ratio of internal cabinet depth to
internal cabinet width was approximately 1.618:1. This is the golden ratio, and has been shown
to be beneficial in avoiding overlapping resonance frequencies resulting from standing waves
forming between the parallel walls of the cabinet.

At this point, Bass Box Pro 6 loudspeaker enclosure design software was utilized to
determine the internal volume of the cabinet that would optimize the performance of the
woofers. The bass response of a driver is very much dependent on the type and size of enclosure
it is mounted in. It was decided at this point that it would be necessary to separate the midrange
driver from the woofer enclosure because the large pressure forces inside the cabinet caused by
the long excursions of the woofer cones can negatively affect the performance of the midrange
drivers. The tweeter was mounted within the same enclosure as the midrange because its
performance is not significantly affected by the volume of the enclosure in which it is placed.
With the width of the cabinet fixed and the height of the midrange enclosure determined by the
mounting positions of the tweeter and midrange, the depth was determined in order to yield the
optimum volume. From here, the overall height was determined by establishing the necessary
internal volume for the woofers.

With the overall dimensions of the cabinet set, 3 shelf braces were added in order to
increase the stiffness of the cabinet and prevent the long side panels from resonating excessively.
Significant portions of the material in the braces were removed in order to make them effectively
acoustically transparent, while enough remained to provide significant bracing effects.

a. Port Considerations
In a vented cabinet design, a port of a specific length and diameter is constructed,
leaving the inside of the cabinet open to the exterior. This port causes the cabinet
to behave as a Helmholtz resonator. It resonates at a characteristic frequency,
which is determined by the size of the cabinet and the acoustic mass of the air in
the port, which, in turn, is determined by the dimensions of the port. The resonant
frequency of the port is chosen so that the output of the port rises as the output of
the woofers falls, thus extending the bass response further. At the port resonant
frequency, all of the output is from the port. For this design, the port resonant
frequency was approximately 40 Hz.

The length and diameter of the port are constrained by the necessary volume of air
it encloses. It is important that the diameter of the port not be too small, as this not
only results in an impractically large port length, but also results in a higher air
velocity within the port. The movement of air can be audible at high speaker
volumes due to the turbulence caused at the port ends, and is known
onomatopoetically as “chuffing”. For this design, a diameter of 3” was sufficient,
as flared port ends were also employed to further reduce the possibility of
chuffing.
3 
 
 The physical placement of thee port on thee cabinet is ggenerally nott especially
imporrtant, so long
g as it is not placed behinnd a driver, aallowing thee out of phasse
sound
d emitted from m the back ofo the driverr to escape thhe cabinet, ddistorting thee
sound
d heard at thee listening poosition.

With
W all of th he cabinet dim mensions annd variables ddetermined, the cabinet was modeleed in
Pro/ENG GINEER Willdfire 5.0. Frrom this mod del, engineerring drawinggs were prodduced that w would
be in valu
uable in the constructionn process. Thhe cabinet w
was modeled acousticallyy in Bass Box Pro
6, which,, although noot especially
y sophisticateed, was suffi
ficient for thee sake of this design. Thhe
BassBox x simulationss for the basss and midran
nge enclosurres can be seeen in the am mplitude vs.
frequencyy and imped dance vs. freqquency plotss below.

V. Cabin
net Constructtion

The
T cabinets were constru ucted of med dium densityy fiberboard, more comm monly known as
MDF. MDF
M has desirrable acousttic propertiess in that it haas a complettely uniform density
througho
out, unlike noormal wood,, and has a higher
h densitty than most other wood like materiaals.
This has the benefit of
o mass dam mping, reduciing the ampllitude of the resonances of the panels
themselv
ves. The engiineering draw wings were used
u as referrences for alll part dimennsions.

The
T same lack k of grain sttructure that gives MDF its superior acoustic prooperties makkes it
quite diffficult when iti comes to assembling
a thhe cabinets. Because of the way MD DF is
manufacttured, mechaanical fasten ners do not hold
h well in iit and can onnly be used tto hold panells
together while glue is drying. Du ue to the larg
ge size of thee cabinets, poocket screws were used to
hold the panels
p in plaace while thee glue dried,, and the pocckets were suubsequently filled with w wood
filler.

The
T cabinet was w construccted so that th he baffle woould be remoovable to alloow for easier
installatio
on of driverss and the cro
ossover comp ponents andd to allow forr varying thee amount of
damping material witthin the cabiinet to see its effects. Thhis turned ouut to be a maj ajor challenge in
constructtion, as threaaded inserts had to be insstalled in ordder to acceptt the bolts thhat hold the
baffle onn. Most of thee threaded innserts ultimaately pulled oout of the M
MDF, as the M MDF could nnot
support the
t repeated tightening and a loosening g involved rreplacing andd removing tthe baffle.

4 
 
 All
A of the joinnts were sealled internallyy. Even withh a vented caabinet designn, small gaps in
the cabin
net can resultt in unwanteed air noise cause
c by the large pressuure differencces between the
inside on
n outside of the
t cabinet. Because
B the baffle was nnot yet perm
manently attaached, gasketting
tape was used on the edges where the baffle met m the cabiinet to providde an airtighht seal.

The
T following
g series of ph
hotos showss the cabinett constructionn process.

5 
 
VI. Crossover Design

In a multi-driver loudspeaker, each driver is chosen because of its favorable operating

characteristics over a specific frequency band. To ensure that the drivers are not subjected to
signals outside of their specified frequency band, it is necessary to insert a set of filters, called a
crossover, in the signal path. The function of the crossover is to distribute the desired frequencies
to each driver, while also preventing unwanted frequencies from reaching the drivers. In a
parallel, 3-way design such as this, it is necessary to employ 3 filters: a low pass filter for the
woofers, a band pass filter for the midrange, and a high pass filter for the tweeter.

An ideal low pass filter allows all of the frequencies below a specified frequency to pass
through unaffected, while blocking the frequencies above that. An ideal band pass filter allows
all frequencies between two set frequencies to pass through the filter, while blocking all those
outside the band. An ideal high pass filter blocks all frequencies below a certain frequency, and
passes all those above that frequency through the filter.

Real world crossover implementations are far from ideal, especially in passive crossovers
like the one used in this design. With passive components, it is impossible to achieve the infinite
slope cutoff of the ideal filters described above. It is good practice to put 1-2 octaves between the
crossover frequency and the closest driver resonance or break up frequency. First order crossover
designs use the fewest number of components, thereby introducing the least amount of distortion
to the signal, but only roll off at 6db/octave past the crossover frequencies. This is usually
insufficient for 3-way designs, as each driver operates over a relatively narrow frequency band.
For this project, a second order crossover network design was chosen because it requires fewer
components than higher order networks, and has a sufficient 12dB/octave roll-off.

After installing the drivers in the assembled cabinets, the frequency response of each
driver was measured in an attempt to locate crossover frequencies. The collected data showed no
significant roll-off in any of the drivers within the measured bands, suggesting that the
performance of the loudspeakers would not be affected significantly by the small variations in
the choice of the crossover frequencies. From this information, the crossover frequencies were
set at 600 Hz and 4,000 Hz.

Xover Pro 3 software was used to determine the circuit design and component values
necessary for a second order network with the above mentioned crossover frequencies. The
network is seen in the image below.

6 
 
 The
T tweeter high-pass
h filtter contains a resistor neetwork know wn as an L-paad, which
reduces the
t output off the tweeterr by approxim mately 4 dB.. This is neccessary becauuse the tweeeter is
a significcantly more efficient
e driv
ver over the frequency bband it is dessigned to opeerate in. Witthout
this netw
work, the tweeeter would sound
s louderr than the othher drivers.

Crossover
C coomponents were
w purchassed on the baasis of valuee-for-money.. All capacitoors
have a 1%% tolerance, so that the two
t loudspeaakers sound sufficientlyy similar. Thee inductors aare
all 18-gauge air-core inductors too maintain a low DC resiistance and aavoid the disstortion causse by
iron-coree inductors.

VII. Meaasurements and

a Analysiss

One
O of the sp peakers was completed
c to
o the point thhat it was poossible to takke frequencyy
response measurements. Measureements weree taken usingg white noisee ranging froom 0 Hz-25 kHz.
The micrrophone wass placed at th
he midpoint between
b the center of thhe tweeter annd the centerr of
the bottom woofer, approximatel
a ly 1m away from
f the speeaker. Frequuency responnse measurem ments
of the am
mplifier used
d showed thaat it was very
y flat across tthe frequenccy range testted. The resuults
of those measuremen
m nts are shownn below on the
t log V^2 vs. log f andd log V^2 vss. f plots.

7 
 
 8 
 
 The most noticeable characteristic of the plot is the consistent high frequency roll-off
starting at about 1300 Hz, resulting in 20 kHz being approximately 6 dB down. This is likely due
to the L-pad network implemented, suggesting that, because of the use of 2 woofers, the tweeter
attenuation is unnecessary. Through the bass frequencies and over a good deal of the midrange,
the response is acceptably flat, not varying by more than 3 dB in either direction. A small dip in
the plot is visible at 4 kHz, the upper crossover frequency. This is characteristic of a second
order filter network, and wiring the midrange out of phase with the tweeter and woofer, as it is in
this design, minimizes the effect of that dip.

The fully assembled speaker was test with music played through a 50W/channel Rotel
RA-1520 integrated amplifier. The frequency response plot is indicative of the sonic results.
There is a pronounced recess to the high frequencies. The most troubling result, however, was a
very “boxy” sound in the midrange. It is believed that this could be improved by reducing the
amount of damping material in the midrange enclosure. It is also likely that the sonic character
would be significantly improved if the baffle were rigidly attached to the cabinet with glue,
rather than being bolted on, as in the current configuration, the braces do not eliminate any of the
resonances in the baffle. Below is a picture of the speakers at their present state of completion.

9 
 
VIII. Future Plans for this Project

In the next few weeks, both speakers will be completely assembled, with the baffles
rigidly attached. I would like to take more comprehensive frequency, impedance and phase
response measurements. From these plots, I could evaluate the validity of the simulations I
produced in BassBox and Xover Pro. I would also like to compare any noted resonance peaks to
calculated resonances.

It would also be interesting to characterize the frequency and phase response of the
crossover network, and to see its effect on the speaker output.

After I finish construction, if the audio quality is good enough to merit it, I will veneer
the cabinets so they are more aesthetically pleasing.

10