52 WIRELESS WORLD, MAY 1978

Loudspeaker system design

Three-enclosure system with active delay and crossovers
by Siegfried Linkwitz

This detailed description of a

multiple-driver loudspeaker design is in
two parts and covers driver selection,
enclosure design, the active crossoyer,
equalization and positioning. Sufficient
information is given to duplicate the
system or to improve existing systems by
equalizing the low-frequency response or
adding a separate woofer box.

THE SYSTEM DESCRIBED has evolved over

years and out of experimentation with
many different configurations and
types of drivers and enclosures. Many
people have contributedtheir ideas. It is
not “the ultimate loudspeaker”, but it
reveals enough about microphone
placement and recording practices to
suggest, that the recording studio is the
next weak link in the chain between
original and reproduction. The few
recordings with good spatial definition
are proofthat the full potential of stereo
has not been exploited. Possibly this
potential has gone unnoticed because
hardly any commercial loudspeaker
reproduces the depth perspective
adequately, giving either a diffuse or
thin-walled stereo image. Fig. 1. Directivity pattern for a rigid circular piston in the end of a long tube as
Every driver becomes more direction- function of d/X (d is piston diameter, X is radiated sound wavelength). Wide
al as frequency increases. The radia- dispersion can only be obtained for frequencies where d/X~1.
tion pattern of a rigid piston mounted at
the end of a long tube’ is omnidirection-
al at frequencies where the ratio of
piston diamter d to the wavelength X of lOin
radiated sound is small, Fig. 1. As d/X
increases the on-axis pressure increases
but the pressure at 450 off-axis de-
creases relative to it. Experience shows
that wide dispersion of sound is desir-
able for natural reproduction. Allowing
for a maximum 6dB drop-off at 450 im
off-axis requires that a driver be only
used over a frequency range where its
equivalent piston diameter is less than 343m/sc
one wavelength. This is an idealized
assumption because real drivers do not
behave exactly like rigid pistons but the
general principle still holds that uni- 100mm -
form, wide dispersion can only be ex-
pected for frequencies where d/X ~ 1.
In all loudspeaker designs the
physical dimensions of driver, box and
room have to be compared to the
wavelength of the radiated sound to 10mm -

determine whether a dimension is 1UMZ 100Hz 1kHZ 10kHz

acoustically small, as when d/X<0.5, or FREQUENCY f
large, Fig, 2. A 200mm diameter driver Fig. 2. Dimension (d) of driver, box or room must be compared to wavelength of
for example should only be used up sound to determine whet her a dimension is acoustically small.
to 1.5kHz to maintain wide dispersion. _________________________________________________________________
 WIRELESS WORLD. MAY 1978 53

This is indeed a popular crossover

frequency but it is also well within the
critical frequency range of fundamen-
tals and lower harmonics of many
musical instruments. It is unavoidable
that some change in the radiation
pattern is introduced around the transi-
tion frequency from a larger to a smaller
driver. In the design described a 100mm
diameterdriver is chosen which crosses
over to a 25mm diameter unit at 3kHz.
The radiation pattern change occurs
.4
therefore an octave higher in a
relatively less critical frequency range,
but still care has been taken in the
design of the crossover circuitry to
minimize irregularities in the transition
region.
Some designers try to obtain wide
dispersion or “omni-directionality” by Fig. 3. Use of multiple drivers to give
using multiple drivers, covering the wide dispersion produces interference
same frequency range. The fallacy in effects for d>O (a) single driver, (b) two
this approach can be seen by comparing drivers one wavelength apart, (c) two
the radiation pattern of a single driver drivers four wavelengths apart, all with
to the resulting dispersion when two of same signal.
these units radiate together, Fig. 3. If
the distance d between the drivers is
greater than half a wavelength signal
cancellation can occur. The two outputs
will be 180~ out of phase whenever the
path lengths from each of the drivers to
the listener differ by an odd multiple of
a half wavelength. As frequency in-
creases the two units move relatively
further apart (Fig. 2) and the locations
for which the outputs cancel become
more frequent. Such a system can only
be described as multi-directional. Ad-
ditional drive units further destroy the
phase coherence of the direct sound.
output from the speaker system. This
imparts the illusion of wide dispersion d-X
to all program material but lacks the
accuracy in sound perspective which
can be obtained from a single drive unit.
After establishing from the cone
diameter the highest frequency up to
which a driver can be used with good
dispersion, the lower frequency limit
will be determined from the cone ex-
cursion capability of the drive unit and
the desired sound pressure level.
The radiation from the piston in a
long tube was found to be omnidirec-
tional for low frequencies where d/
X«l, Fig. 1. If the piston moves with a
peak-to-peak excursion ~ at
frequency f and radiates into free space,
then the pressure p at a distance r from
the source is
1T~O a0~fd2 d.4Xc
p=. 4
8V2 r x
Assuming a.,,
1, is 6mm and f 70Hz a increase of approximately 6dB can be
Normalizing the pressure with res- direct pressure level of 83dB at 1m can expected. Therefore from a stereo sys-
pect to the reference pressure be obtained from the 100mm driver and tem with 100mm drivers a direct free-
p=2 x 10 ja bar yields an expression for 85dB from a 200mm unit. field s.p.l. of about 95dB can be ex-
the more familiar sound pressure level These s.p.ls may not seem very high pected. Furthermore the normal lis-
(d and ~ increased in mm) but crossing over to a woofer at 70Hz tening environment is a semi-
will double the maximally obtainable reverberant room where sound is
sound pressure (+ 6dB) and because reinforced by reflections from walls and
2Olog (p/p) = —86 + 4Olog f 2Olog r
— signals from the left and right channels objects.
+ 4Olog d + 2Olog a of a stereo system are predominantly in Practical experience confirms that a
phase at such low frequency a further 100mm unit can handle program
54 WIRELESS WORLD, MAY 1978

material down to 70Hz at adequate sion can only be obtained from a small frequency range covered. The volume
levels and low distortion. This moves drive unit which will also have higher inside the enclosure will exhibit cavity
the crossover to the woofer to a less distortion than a larger unit. It appears resonances which have to be elimin-
critical frequency range and a single though that psychoacoustically the in- ated. Acoustic energy is stored when-
large woofer can be used to cover the creased distortion is outweighed by an ever one of these resonances is excited
remaining frequency range below 70Hz. improved sound perspective which and gradually released after the excita-
The large woofer enclosure can be gives a greater sense of realism. Some tion has been removed. Most of this
placed separately from the relatively further investigation of this subject is acoustic energy exits through the cone:
small midrange and tweeter enclosures needed. the speaker regurgitates its own cha-
and still be acoustically close because racteristic box sound. Fortunately it is
d/X is small. Further consideration is Speaker enclosures not difficult to dampen cavity
given to this aspect of the system design Usually the size of a loudspeaker en- resonances. The technique of filling the
later. closure is dictated by the required low enclosure with long-fibre wool is well
2 ~.
The frequency range below 70Hz frequency response and efficiency. A established and very effective.
Another form of undesired secondary
could be covered by two 200mm units different approach is taken here where radiation comes from the enclosure
which will generate 90dB of direct s.p.l. the enclosure is optimized for minimum walls themselves. The walls can be ex-
at 35Hz and lm or two 250mm diameter secondary radiation over as wide a cited to vibrate by the internal air
drivers with 94dB s.p.l. assuming 6mm frequency range as possible. The low volume pressure changes, but more
peak-to-peak excursion capability. frequency output capability is treated serious is the direct transmission of the
The particular drivers chosen for this as a separate problem. mechanical vibration of the driver’s
design are the 100mm KEF BilO low The purpose of the enclousre is to cone to the enclosure. The walls then
frequency/midrange unit, the 25mm control the radiation from the back of radiate the transmitted mechanical en-
KEF T27 tweeter and the KEF B139 the cone. A closed box design is chosen ergy as sound, particularly when its
woofer. A different unit like the KEF as the simplest form of enclosure. If the frequency coincides With a panel
B200 or some other make with adequate largest box dimension is less than a resonance.
excursion capability and linearity could quarter wavelength at the highest It is not unusual that more energy is
be substituted for the B139. frequency from the driver then the box radiated directly from the enclosure
There are of course considerations is acoustically small and the air volume walls than from the cone at resonance
other than dispersion and cone excur- inside the enclosure will act like a uni- frequencies. If for example the vibrating
sion which must enter into the selection form soring. The box has to be made enclosure surface has ten times the area
of a drive unit, such as smoothness of sufficiently stiff so that the internal air of the cone then its acoustic output will
frequency response, freedom from high pressure changes will not deflect the already equal that of the cone if it has
Q resonances, minimum phase behavi- walls and cause secondary radiation. only one tenth of the cone excursion.
our, and low non-linear distortion. Un- The woofer enclosure can be made The output of most loudspeakers is
fortunately few meaningful data are small relative to the 70Hz maximum coloured by radiation from the enclos-
published by many manufacturers. frequency. It will therefore have no ure walls.
Knowing the magnet weight and flux internal air volume resonances and It has been verified experimentally
density is of little help. With some resonances of the box panels can be that vibration coupling between the
training though the ear can sort out pushed above 70Hz by crossbracing of driver and the walls occurs primarily
those drive units that seem worth the walls. through the rigid mounting of the driver
further investigation and the units The enclosure for the Bi 10 presents to the enclosure. Vibration-mounting
chosen for this design proved to be very greater difficulties because of the wider the driver to the enclosure with some
satisfactory. form of complaint suspension will
The selection ofdrivers was primarily significantly reduce the wall excitation,
guided by the desite for wide, uniform but it poses some difficult mechanical
dispersion and crossover frequencies as design problems. The natural frequency
high and low as possible. Had emphasis Fig. 4. Loudspeaker enclosure of the driver mounting has to be well
been placed on high power output dimensions. As there is little stereo below the acoustical output frequency.
capability or lower non-linear and information below the 70Hz limit of the The mount has to seal the enclosure air
Doppler distortion then larger diameter enclosures (left). a centre woofer covers tight and provide sufficient mechanical
drivers would have to be chosen, or the the remaining range down to 25Hz support for the driver. Another
crossover frequencies shifted to a more (right). approach would be to enclose the box
critical frequency range. Wide disper-

Mahogany 20 thick
& 40x40

3,

180 all dinins. mm

WIRELESS WORLD, MAY 1978 55
and then encounter an abrupt transition
• to which the driver is mounted by a where it ends, Fig. 5. A second wave is
second box avoiding all rigid coupling Enclosure design objectives
generated at the cabinet edge which will
between the two.
• Narrow frontal area for optimum interfere with the original wave. If a
For the design described here a single,
horizontal dispersion; tweeter pulse is radiated from the T27 then a
totally enclosed box was chosen, Fig. 4.
mounted directly above the mid- secondary pulse of lower amplitude is
The BilO driver was attached to it with
range unit. generated at a time t = d/c = 26Oiis later;
soft rubber grommets in the four
• Box edges rounded to reduce the original pulse is smeared out. This
mounting holes of the basket and the
scattering. scattering of sound should be avoided
sealing foam gasket barely compressed.
• Drivers mounted to minimize by eliminating sharp discontinuities
Comparing this to a directly mounted
direct transmission of vibration. through bevelling of the cabinet
driver by tapping on either basket indi-
• Air cavity resonances attenuated edges.5.
cates a significant reduction in coupling
to the box. Some further investigation with filling materials to eliminate Figure 6 shows the on-axis amplitude
of this subject is in progress. delayed re-transmission through response of a small driver mounted in
The relatively small size for the cone. the center of a cube and of a sphere.6
Bi 10/T27 assembly has the advantage Clearly the sphere with its surface
that the internal air volume resonances gradually receding from the source
occur at high frequencies where they produces a much smoother response
can be damped effectively with wool plane therefore the radiation pattern at than the cube with its sharp edges.
filling. The lowest cavity resonance the crossover frequency should follow Therefore the larger box relative to the
occurs at 600Hz, the next ones at 800, the previously discussed behaviour of driver the more closely should it
1000, 1200Hz etc. The resonances are two drivers contributing equally, Fig. approach the shape of a sphere. It fol-
measured easily with a small 3b. lows that the midrange/tweeter enclos-
Ideally the sound from the T27 should ures might be further improved by
omnidirectional electret microphone
protruding into the box and applying a be able to disperse freely in all direc- reducing their size and a more curved
tions, but because of the large width of driver mounting area. Papier mache,
sweep signal to the B1 10. Filling the the front panel relative to the cone cardboard or epoxy fibreglass with
boxes rather tightly with long-fibre
wool attenuates all the resonances to a diameter a wave emanating from the damping materials applied to it rather
cone will initially be blocked by the panel than wood might be more suitable
smooth frequency response inside the
box and at the outside cone surface.
The boxes are constructed out of
20mm mohagony boards. The panels are
quite stiff. The lowest panel resonance
was observed at 430Hz using a magnetic
phono pickup in a makeshift tonearm as ZN
N
vibration transducer. A 430Hz tone was
measured to decay 40dB in l20msec
~7U7~ N
- ~ \
after the electrical signal was removed.
This indicates that the resonance was a / - - ~

~“
\\
Q of 36 according to O.
7fRT~dB.
The Q is quite high and the decaytime
,-~

s long. By applying approximately two

K
~ ~ \d¶~\
litresof roof patching tar to the inside of
the box the resonant frequency was
lowered to300Hz due to the added mass
to the panels. The decay time decreased
to 4Oms, corresponding to a Q of 8.4.
While this treatment proved effective it Fig.~. Sound scattering from the sharp corners of a loudspeaker enclosure
does point out the problem that a small producing a smeared out transient behaviour of the system.
panel can have a high Q which is diffi-
cult to dampen because of its high stiff-
ness and large mass. Better results
might be expected from a thin plywood
construction, with thick layers of damp-
ing material to attenuate resonances
and to reduce the direct transmission of
sound from the inside of the box.4.
Ideally of course the panels should not
be excited in their resonances at all,
neither from the air pressure changes
inside the box nor from the mechanical
coupling to the driver.
A small box presents a small obstacle
to omnidirectional sound propagation.
Ii
+10dB-

This is a clearly audible advantage +5

when properly placed in the room. As
the box is only marginally wider than 0
the Bi 10 driver it can be assumed that
the radiation pattern for a piston at the
end of a long tube is an adequate first-
order approximation to its sound dis-
persion, Fig. 1. The T27 tweeter is RELATIVE SIZE OR FREQUENCY
mounted as closed as possible to the
Bl1O. At the crossover frequency of Fig. 6. On-axis frequency response of a point source mounted in different
3kHz the spacing corresponds to a dis- enclosures shows smoother response of sphere (after ref. 6).
tance of one wavelength. In the vertical
56 WIRELESS WORLD, MAY 1978

materials for the unconventional con-

tours of such an enclosure.
For the given design the frequency
range extends down to 70Hz where the
BllO has its enclosed resonance. As
there is little meaningful stereo infor-
mation below this frequency a single
centre-channel woofer box can cover
the remaining range down to 25Hz, Fig.
4. This is built with internal bracing to
stiffen it and to push panel resonances
to frequencies above 70Hz. In addition
25mm- thick heavy felt is glued to all
panels to reduce direct transmission.
The box is loosely fitted with long-fibre
wool.

References
1. Beranek, L. L., Acoustics, McGraw-Hill
1954
2. Bailey, A. R., Non-resonant loudspeaker
enclousre design, Wireless World, October
1965. p.483
3. Bradbury, L. J. S., Use of fibrous materials
in loudspeaker162 enclosures, LAES, vol. 24,
April 1976,D.
p. A., Sound output of loudspeaker
4. Barlow,
cabinet walls, 50th AES Convention, London,
1975
5. Heyser, R. C., Energy-time test, Audio,
June 1976, p.74
6. Muller, G. G., Black, R. & Davis, T. E., This article will be concluded with active crossover and equalizer
Diffraction produced by cylindrical and designs and a discussion of speaker location.
cubical obstacles and by circular and square
plates, .JASA, vol. 10, July 1938, p.6

Current broadcasting techniques

“a hybrid rat—bag of improvisation”
TV engineers may propose a 1875-line tv pened? Nothing. The public ignor~d them. micro-circuits. “The reduction in costs is
service for Band III when 405 line transmis- Everyone still listens on antiquated staggering. An integrated circuit now costs
sions are no longer needed, but it seems amplitude-modulation receivers — statisti - about the same as an apple.” Complex func-
unlikely that the frequencies will be used for cally, so to speak.” tions could be performed with devices that
television, according to Dr Boris Townsend, Dr Townsend warned of the problems could be cheaply mass-produced, and circuits
head of the IBA’s engineering information technology could bring if it were directed were getting, and would continue to get,
service. Speaking at the first of the three IBA solely at increasing efficiency and reducing even smaller, and so even more complex.
lectures on tomorrow’s broadcasting, he costs: “Perhaps unemployment is a more While mechanisms grew more and more
said, “I expect that we shall lose the use of important matter than television automa- expensive, electronics became cheaper:
these frequencies for television, which will be tion. If it is, then should design engineers be “Compare the £70 monochrome television
a pity, for band III has excellent propagation devoting their thinking to making this public receiver, of dubious performance, which sold
characteristics for our purposes. Neverthe- service of television an under-capitalised, in 1938, with its contemporary 1938 £100
less, I expectthe engineers to put up acase for more labour-intensive operation? It is a naive small motor car; and then take today’s large,
a new television service in the band using the thought. But the basic dilemma is a matter of bright, reliable monochrome television
latest advance in engineering terms: for concern to many engineers.” The problem of receiver, still at the same £70, and its con-
example, better definition, 1875 scanning deciding among the bewildering array of temporary £2,000 Mini. So,” he added,
lines, larger pictures, tone gradation scales technical possibilities “is not a problem “whenever we can replace mechanisms by
tailored to human physiology and, I hope, a which should be left to engineers.” They electronics we shall.”
certain technical flexibility to make tele- were moving so fast in some areas, and not at
vision more helpful in teaching the handi- all in others, “that we may be unable to
capped child with abnormal physiology.” prevent some massive mistakes on their part
Speaking of the current state of broad- unless we speak quickly, and precisely, and Programme labelling for
regularly with them.”
casting technology he said, “We are stuck
with old-fashioned television transmissions In questions after his speech he said of sound broadcasts
which are a hybrid rat-bag of improvisation surround sound, “We can do it. Whether Electronic “labelling” of sound broadcast
and almost incompatible technologies, which people want it is another matter, whether programmes is definitely on the way, accor-
carry a quite unnecesssary amount of people will like it when they have got it. I ding to a BBC spokesman, and the only
repetitive data yet still reproduce inadequate think they will like it, but whether they will problem is deciding the best way to do it. This
detail, inadequate contrast, inadequate col- think it worth the trouble and the cost I’m technique (described by Duncan MacEwan
our gamut, and pictures which lack most of dubious.” There were also problems of pro- in “Radio in the ‘80s” in the May 1977 issue)
the attributes of real life. It is a system which duction, and-engineers had to get used to it to uses data signals associated with the broad-
gives the viewer and listener little choice and make the best use of the new forms of drama cast signal to identify particular programmes
to which any graduate tea-boy engineer DrTownsend predicted that the changes at or channels so that they can be automatically
could now suggest improvement. What are the transmission end of the broadcasting pre-selected at the receiver for, say, an
we going to do about it? chain would be radical: “Engineers are evening’s listening or recording purposes. An
“In the next few years, not much. We did making a two-pronged thrust forward. One is automatic “search tuning” receiver could be
something about our equally outmoded radio a miniaturisation of the analogue techniques arranged to select programmes of a particu-
transmissions, and duplicated our use of which have been used in our studios since lar type. The labelling code signals could be
precious wavelenfths with our stereo f.m. Savoy Place and Alexandra Palace, while the transmitted on a subcarrier in or out of band
transmissions using circular polarisation. other is based.., on digits.” Both analogue or by frequency modulation of the am.
These transmissions are glorious. What hap- and digital methods were dependent on carrier. D
 WIRELESS WORLD. JUNE 1978 67

Loudspeaker system design

Three-enclosure system with active delay and crossovers — part 2

by Siegfried Linkwitz DipI. Ing.

improvement, particularly in light of the

This is not the “ultimate loudspeaker”. program material it now became quite just-mentioned frequency response in-
but in the first part of this article in the easy to pinpont the location of in- terferences from virtual images. It
May issue Mr Linkwitz says that “the dividual instruments both laterally and might be that the ear-brain combination
recording is the next weak link in the in their distance behind the speaker performs a time domain analysis and is
chain.” The equalized system plane. It might be said that the whole able to allocate the wall reflections
incorporates electronic crossovers and sound stage moved into focus. which occur 4 to 6ms later than the
delay compensation. Part one describes Furthermore, tape hiss and record direct sound to the characteristic of the
the enclosure design and this article gives surface noise became spatially listening room and to the program
sufficient information for the experienced separated from the musical material. material.
constructor to duplicate or to adapt the The noise originated definitely at the Mounting of the speakers away from
electronic design to other needs. speaker boxes while the musical instru- the walls was accomplished by hanging
ments assumed their own space bet- them from the ceiling with a nylon
ween and behind the boxes. In this monofilament. Electrical connections
THIS SPEAKER DESIGN project was started sense the noise and ticks from a record run \from the back of the enclosure to
with the idea of mounting the boxes surface are comparable to tne coughing the wall behind it and also serve to keep
flush with the surrounding wall. and shuffling of people at a live concert the speaker aiming forward. The small
where one can concentrate on the per- hanging units might be appropriately
Positioning the box in front of a wall
causes a severe dip in amplitude res- formance and not be side-tracked by called satellites to the woofer box. The
8.
ponse when the distance from the front unrelated acoustical events woofer itself is located halfway between
It is not clear why the placement only
of the box to the wall equals a quarter a relatively short distance away from the satellites, which are 2.5m apart.
wavelength. For a typical 300mm en- the walls should give such a marked The listening room is 5x8x4m
closure depth the dip occurs around (w x 1 x h), with the speakers in front of
250Hz, Fig. 2 (ref. 7). the narrow walls and the typical lis-
If one imagines the walls near the tening positions 5 to 6m away from the
speakers to be made of mirrors then one satellites.
0.7
can easily visualize the images of the
box in these mirrors. At frequencies Crossover design
where the radiation from the box is Cs The simplest crossover network is the
omnidirectional, each of these images VL —6dB per octave slope filter of Fig. 7.
can be thought of as a separate sound 0.7 Assuming idealized components, the
source, whose output will add or sub- current from the generator will split in
tract from the original source, depend- Amplifier Woofer T~eter such a way that the vector sum of the
ing upon how far the image source is (b) voltages across the low and high
removed in terms of wavelengths. For a (a) frequency driver terminals is equal to
half wavelength distance to the image the source voltage at all frequencies
source, corresponding to a quarter Fig. 7. Schematic network with 6dB per VL4~VH=es.
wavelength distance from the speaker octave slopes and voltage phasor diagram at
the crossover frequency. Correspondingly the vector sum of the
to the mirror, the output from the real sound pressures PL and PH generated by
and the imagined source will cancel the drivers will be directly proportional
A
each other. to the generator voltage PL + PH = ke
This description of virtual sound 5
sources is valid provided that the and independent of frequency, provided
speaker is radiating sound towards the that the distance from the listener to
walls and that the walls, floors and each of the drivers is identical. The Bl 10
ceiling act as acoustically reflecting and T27 drivers though are a
surfaces, i.e. mirrors. Mounting the wavelength apart, which means that
boxes flush with the reflecting surface equidistance is obtained only for a plane
eliminates the virtual image behind the in space, Fig. 8. For points outside this
speaker and produces a smooth plane the sum of the two driver outputs
frequency response. The completed will vary with frequency.
system with flush mounted boxes per- Furthermore, because the two drive
formed very satisfactorily. In particular voltages already have a 900 phase dif-
it gave a good sense of depth perspec- ference the summation will be different
tive for some stereo material. for symmetrical points above and below
It was discovered later that by the plane of equidistance. In the
moving the speaker out into the room crossover frequency region where both
and at least 0.5-lm away from walls and Fig. 8. Plane of points are equidistant from
the high frequency and low frequency drivers contribute equally the syste~rn
floor, a significant improvement in will radiate its maximum pressure at a
drivers. The sum of the sound pressures is
sound perspective was obtainable, see proportional to the sum of the electrical drive 14-degree angle below the plane, Fig. 9.
photograph in part 1. On approprate signals only in this plane. This simple dividing network has a wide
68 WIRELESS WORLD, JUNE 1978
range of overlap between the two
drivers and therefore a tilted radiation
pattern over at least two octaves.
A seemingly attractive feature of this
network is its complete lack of phase
distortion for points which are equi- w
distant from H and L, Fig. 8. At these
0
points perfect square-wave reproduction
is achievable under free-field conditions -J

or in an anechoic chamber. In a living 4

room the increased radiation towards w
the floor and the reduced radiation 4-
4
upwards will produce a coloration in -J

sound due to the frequency-selective

change of the reverberant field. The ear
is more sensitive to the amplitude res-
ponse than to phase shift. Therefore this
filter and related designs with even
greater than 900 phase difference bet-
ween the drive signals and correspon- FREQUENCY
dingly greater off-axis intensity peaks
9. Fig. 10. Schematic response for crossover
are not used for the satellite system
A 24dB per octave slope filter was points and driver resonances.
chosen which has no off-axis peaks in
the radiation pattern10 Fig. 9. The steep
filter cut-off narrows the overlap region
where the BllO and T27 interact. The
T27 has its fundamental resonance at
1.4 kHz and the highpass provides 27dB
of attenuation at this frequency. At
5kHz where the BllO exhibits a cone
resonance the filter has reduced the
drive voltage by 18dB, Fig. 10. A 6 or Fig. 11. Group delay frequency response of a
even 12dB per octave filter ~yould have speaker system due to a 3kHz crossover
between midrange and tweeter with conven-
insufficient attenuation to minimize tional first and third-order Butterworth net-

1
exciting these resonances. The 18dB per works, and with second and fourth-order
octave filter was not considered be- cascaded Butterworth sections (ref. 10).
cause it tilts the polar pattern.
All these filters, with the exception of 300Hz 1KHZ
the 6dB per octave network, have a FREQUENCY —~
frequency-dependent phase shift and
consequently some phase distortion. phase shift with increasing frequency Therefore it seemed safe to use the
Only a network of linearly increasing will have no phase distortion. The slope fourth-order filter with its sharp cut-off
of the phase curve is constant in this behaviour which minimizes the overlap
case. Any deviation from the constant between drivers.
slope indicates that some amount of
phase distortion is present. The ques-
tion arises how much slope variation Crossover and equalizer circuits
can be tolerated before it becomes aud- The crossover networks and equalizers
ible and not merely visible on an consists of a variety of active filter
oscilloscope. The slope of the phase circuits. The overall block diagram of
curve, usually referred to as envelope Fig. 12 gives an indication of the system
delay or group delay, has been plotted complexity. Design formulas are pre-
for typically used Butterworth sented for each functional block so that
crossover networks and the new net- the experienced constructor should be
work function10, Fig. 11. Merely able to duplicate the circuits of Fig. 13 or
changing the polarity to one of the adapt the design to particular needs.
drivers drastically changes the group
delay for the summed driver outputs as 3kHz crossover networks
in the case of the first and third-order The fourth-order high and low-pass
Butterworth crossovers. Their on-axis filters are made up from cascaded
amplitude response is unchanged, second-order Butterworth sections, Fig.
unless the drivers are separated some 14. The outputs VH and VL are in phase
distance from each other. Then the with each other at all frequencies and
polar pattern will tilt either up or down the voltage sum is equal to VIN. At the
with the change in driver polarity. crossover frequency f
To investigate the audibility of phase 0, therefore, 2the
or
distortion an all-pass network was built 6dB
outputdown,
fromwhich is different
each filter will be from
VIN/ the
which duplicates the group delay of the typical 3dB crossover point for filters
new second and fourth-order crossover where VH and VL are in phase quadra-
Fig. 9. Radiation of a 6dB per octave networks (12 and 24dB per octave ture10.
crossover network at the cross-over .curves in fig. 11). Listening with
frequency (3dB peak occurs below the plane Delay compensation
of equidistance for non-coincident drivers) headphones to stereo and mono pro-
and the symmetrical pattern of a 24dB per gram material, no audible difference The Bl1O and T27 drivers do not radiate
octave crossover network at the crossover could be detected with either one of the from the same acoustical plane even
frequency (ref. 10). all-pass networks switched in or out. though they are mounted on the same
WIRELESS WORLD, JUNE 1978 69

50W
Power amplifiers

Fig. 12.System block diagram. Design for-

mulas for each functional block are given to
allow adaptation of the circuits of Fig. 13.

Fig. 13. Circuit diagram of crossover net-

works and equalizers incoroporating delay
compensation. Broken lines show optional
h.f boost components.

3kHz hp + 6dB 36mm deloy(~ 3kHz

Resistors not marked = 36k3 Cs =741

 70 WIRELESS WORLD. JUNE 1978
baffle. The electrical signals arrive at Mechanically moving the tweeter back down at this frequency due to the s~mall
each voice coil at the same time but is feasible provided care is taken to internal volume of the satellite enclos-
because the T27 voice coil is located in avoid sharp cabinet edges and their ures. The output continues falling off at
front of the BilO voice coil the sound associated scattering of sound. For a 12dB per octave rate below this
pressure wave generated by the T27 will electrical delay an all-pass network has frequency with approximately second-
be advanced relative to the Bi ~0. The been used, Fig. 16. Its delay varies with order Butterworth response shape.
40mm driver off-set may seem in- frequency from ‘r = 2RC at low frequen- Therefore the driver in the closed box
significant unless it is related to the cies, approaching zero delay at very can be used as one half section of the
144mm wavelength of a 3kHz tone high frequencies. To reduce the required high-pass filter. The other half
where it corresponds to a 100~ phase frequency dependency in the crossover is implemented with an active second-
difference between the two driver out- region of around f~ the component order Butterworth filter section — the
puts. values should be chosen such that first stage in the centre channel of Fig.
The effect of driver off-set on the RG =1/20f~. Several delay stages are 13. The low-pass filter for the B139 is
on-axis frequency response can be quite cascaded to obtain the required total again the two amplifier fourth-order
significant, particularly if both drivers delay. This delay has to be determined network design — the second and
contribute almost equally over a wide experimentally, but the spatial driver fourth stages of the lower channel in
frequency range, Fig. 15. The frequency offset gives a reasonable starting point. Fig. 13.
region of overlap is significantly nar-
rower for higher-order filters because of 70Hz crossover network Woofer equalization
their steeper cut-off. The transition between the woofer and The response of the woofer does not
The driver offset can be compensated the satellite uses 24dB per octave slope extend sufficiently far down in
for by adding electrical delay to the filters similar to the 3kHz crossover. A frequency. The fall-off in acoustic out-
tweeter drive signal, or by mounting the transition frequency of 70Hz was put will therefore be compensated with
tweeter in a different plane. chosen because the BilO output is 3dB a properly increasing drive signal. Over
the frequency range where the driver
acts like a rigid piston its frequency
response when mounted in a closed box
(ref. l1)is
Highpass filter
If~
1~fo)

~l-

This is a high-pass function with a

corner frequency near the closed box
resonance f
0 and some peaking depend-
ing on Q0, Fig. 17. The two parameters
Lowpa55 filter f0 and Q0 can be conveniently deter-
mined from the frequency response1’ of the
of
the driving
speaker point
system, Fig.impedance
18. If the system is
driven from a generator with an inter-
nal impedance much larger than Rmax,
yin
then the terminal voltage becomes pro-
portional to the system impedance and
Q
0’ f0 can be determined from the volt-
age response as in Fig. 19.
For the two B139 woofers in their
closed box, the resonance occurs at
54Hz with a Q0 of 1.2. The response can
C 2TEV~RC now be compensated with a network
which exactly complements the woofer
FIg. 14. Fourth-order 24dB per octave crossover filter sections are made up from cascaded roll-off and extends it to a lower cut-off
second-order sections in both high and low-pass forms. frequency, Fig. 20. This design approach
can be used to equalize other speaker
systems if careful attention is given to

2RC
Delay ‘U 1 -.‘(2TtfRC)

NORMALIZED
4~tc FREQUENCY NORMALIZED
4’fc FREQUENCY
Fig. 16.
Several all-pass phase shift networks
Fig. 15. On-axis response when the tweeter is positioned acoustically in front of the midrange by are cascaded to obtain the required delay
d/\ with 6dBper octave crossover, and 24dB oer octave crossover. compensation.
 WIRELESS WORLD. JUNE 1978 71

The corrected response of the woofer (a) circuit

can be verified by placing a microphone
about 1cm away from the cone to de- R
termine the near-field sound pressure 2
which for a uniformly moving piston is
In
proportional to the far-field sound
pressure 13

System equalization
As active networks are already used for
the crossover filters it seems attractive
to also use them to equalize the com-
NORMALIZED FREQUENCY plete speaker system for a flat Gain
amplitude response at the preferred Ib) Frequency response
Fig. 17. Fall-off in response of a rigid piston in listening location. A microphone at that Adc~
a closed box (ref. 11). Box resonancef
0 and Q position will pick up the direct sound
are determined as in Figs. 18 and 19. coming from the speakers and a large
number of reflections from various I I

~2 ~i ~O Frequency
objects and the walls of the room. The Am
microphone cannot’ distinguish
between the different sources. The
microphone output voltage which cor- (c) Design formulas
responds to the direct sound from the fo =
speakers will be masked by the voltage 2TtR
w due to the reverberant sound field. The 1 vc1 c2
U
z ear-brain combination seems to be Oa= 2~ = 2R1+R2 yc2
0
w taking its clues for locating the details
0.
I of the stereo image from the direct
sound even when the reverberant sound Adc = R3
R1 o0Vc2
energy is much larger than the direct
f = 1
I I I r~’””~”’ sound. This might explain why attempts 1 TtR1c1 Am—
—
R3
R2 1- <1
to equalize for a flat response at the
listening location gave unsatisfactory Adc R21’R3
= 2Tt(R2.I-R3)c2 ~= 2R1
results.
Fig. 18. Schematic response of the woofer The response at one metre from the
driving point impedance measured as in Fig. speaker measured in the room appears
19 from which f0 and Q0 of Fig. 1 7are derived to be a better starting point for FIg. 20. Network extends the woofer low
(ref. 11). equalization. But even for this location frequency response to f1 by providing exact
a completely flat response does not compensation for Q0 and f0 with schematic
Dri~r in
seem to give the most natural-sounding amplitude response, and design formulas.
encIosur~
,-— —, reproduction. Some form of shelving 4. or
600 >lOmV ¶Vmax I sloping
In thisresponse
design seems
a 3dB necessary’
low-frequency
boost is applied to the BilO signal to
Frequency
calibrated obtain flat acoustic output over its
signal range (last stage in the centre channel
generator
)IV
of Fig. 13). The T27 is allowed to follow
its o.wn gradual roll-off, but if a flat Fig. 21. Measured voltages at the driver
high-end response seems desirable then terminals of the complete system. A flat
to the simple network shown with broken response does not seem to give natural-
RIC of speaker system sounding reproduction.
with Ohm-meter
Fig. 19. Measurement setup for Fig. 18 to
determine R0~ /RMAX from VDC /VMAX and
the ~ and f2 from
MAX X DC’ Flat OCQUStiC

the cone excursion capability and the

power amplifier output voltage swing
limitations. Both increase by a factor of
four when the cut-off frequency is low-
ered by an octave. w
(9
For the playback of records much of 4-

the linear excursion range of the woofer -J

is used to reproduce the large

amplitudes of record rumble. This
wastes driver linearity. Fortunately the w
left and right-channel vertical rumble
outputs from the pickup are out of -J
w
phase and therefore cancel when the -30
left and right channels are summed for a
center channel woofer, as in this design.
Separate left and right channel woofers
-40
can easily be tied together electrically to
eliminate the unnecessary movement of 10HZ 100Hz 1kHz 10kHZ
air at subsonic frequencies
2. from one
speaker box to the other’ FREQUENCY
72 WIRELESS WORLD. JUNE 1978
lines at the input stage will give the
necessary high-frequency boost.
for reduced spurious resonances but
even more so the enclosure design,
System X speed-up
An analogy might help to describe the materials and shapes will need further An Integrated Systems Development (ISDD)
subjective impression of a properly de- 5. Nevertheless has been set up by the Post Office to
study and development’
it is possible to design a highly satisfac=
signed and equalized system by com- accelerate the development of System X, the
tory system even with today’s standard electronic exchange equipment “that will
paring it to the colour photograph of a revolutionise the nation’s phone system.”
familiar scene. A fair sound system components.
(Wireless World, passim). The department is
might correspond to an out-of-focus to be headed by Mr John Martin, who has
picture, possibly with the wrong ‘reds worked on electronic exchange equipment
and blue or an overall colour tint. Com- for 15 years, and on System X for four.
paring two such systems to each other In a statement the BPO said they intended
is like looking at’two blurry pictures of to place contracts for the first production
reality, where one might prefer one over exchanges before the end of the year. The
the other because of its colour balance first exchanges would come into service by
but there is no question of either being a the early 80s. The new department is an
offshoot of the TSSD (Technical Systems
realistic reproduction. A high accuracy Strategy Department), set up in 1974 which,
sound system corresponds to a photo- References
7. Allison R. F., Influence of room boun- under Roy Harris, has been responsible for
graph which is focused and without daries on loudspeaker power output, the overall design of the system. ISDD will
unnatural emphasis on any colour. JAES, vol. 22, June 1974, p. 314. liaise with Plessey, GEC and STC, who are
When a high standard of reproduction is 8. Heyser R.C., Geometry of sound percep- developing System X with the P0, and other
being approached it becomes possible to tion, 51st AES Convention, Los Angeles, Post Office departments.
hear clearly areas of slight imperfection May 1975. The Carter Committee on the running of
like in a picture which is not exactly Heyser R.C. Some new audio measure- the Post Office criticised the slow develop-
ments, 51st AES Convention, Los An- ment of System X, which it feared would
focussed or has just a slight colour tint. arrive too late to be competitive in world
IFor the high-frequency equalization of geles, May 1975.
9. Pramanik S.K., Crossover networks and markets.
the speakers this means that too much phase response, Wireless World, vol. 81, The Post Office revealed some of the ideas
output shift’s the sound image out of November1975, p. 529. underlyingSystem X atthe Communications
focus. The image depth becomes blur- Ashley JR. and Henne L.M., Operation- 78 exhibition and conference at Birmingham.
red because the high frequency over- al amplifier implementation of ideal System X will use integrated digital trans-
tones seem to be less distant than the electronic crossover networks, JAES, mission and switching, stored programme
virtual sources which generated them. voL 19, January 1971, p.7. control and common channel signalling. The
The chosen speaker equalization 10. Linkwitz S.H., Active crossover net- devices used would be based on low power
appears to match the greatest variety of works for non-coincident drivers, JAES, Schottky t.t.l., c.m.o.s. and n.m.o.s. w
program material. A properly func- vol. 24, January 1976, p. 2.
tioning treble control in the pre- Ii. Small RH., Closed-box loudspeaker sys-
amplifier is needed though to correct for tems, part 1: analysis, JAES, vol. 20,
differences in recordings. The final res- December 1972, p. 798.
Hunt F.V., Electroacoustics, Harvard
IERE hits out at
ponse of the drive voltages for the three U.P., Wiley, 1954. “degree obsession”
speaker units, Fig. 21, could have been 12. Oldfield M.L.G., Stereo rumble filter,
generated or approximated with passive Wireless World, vol. 81, October 1975, p. “The national obsession with education to
networks. The practical implementa- 474. ‘degree standard’ has deprived the traditional
tion might prove to be difficult though 13. Keele D.B., Low-frequency loudspeaker technician engineer training pattern
and no attempt has been made to design assessment by nearfield sound-pressure (ONC/HNC) of both its status and much of
a passive crossover/equalizer. The de- measurement, JAES, vol. 22, April 1974, its best raw material; this, coupled with the
p. 154. inflexibility of CEI’s qualification rules and
sign flexibility of active networks far procedures which has robbed the technician
Jacobsen 0., Some aspects of the self and
outweighs the possible cost saving of mutual radiation impedance concept engineer of the facility to proceed by practi-
passive networks when only a single with respect to loudspeakers, JAES, vol. cal achievement to chartered engineer
system is being built. 24, March 1976, p. 82. status, has divided the electronic engineering
14. Harwood H.D., Some factors in loud- profession into a rigid two-class structure to
Conclusion speaker quality, Wireless World, vol. the detriment of the profession and of the
It is hoped that some of the design 82, May 1976, p. 45. national interest.”
15. Berman M.J., Cooke RE. and Fincham This broadside was delivered in the IERE’s
techniques and ideas expressed here evidence to the Finnistoii Committee on the
will stimulate a more rational design of L.R., Loudspeaker testing using digital
techniques, KEF Electronics Ltd, March future of the engineering profession. In other
loudspeaker systems. Certainly the 1975 D respects, however, the IERE seems to agree
drivers will be continuously improved with the views of the Engineering Professor’s
Conference that degree courses should be
longer and that there should be more co-
operation with industry. The IERE recom-
Siegfried Linkwitz became interested in mends that there should be an extensive
publicity campaign aimed at raising degree
sound reproduction as a hobby, recognizing the
many dualities between the propagation of course entry standards to restose the status
microwave and acoustic fields. “A 1 GHz elec- of ONC, HNC, HND and TEC courses and the
trical wave and a 1 kHz acoustical wave have associated work levels, to stimulate recruit-
ment and to boost the morale of “this vital
about the same wavelength.” After modifying
and equalizing several commercial louds- element of the engineering workforce.”
peakers, he set about designing his own system ~ This, as we said in our September, 1977
— out of frustration with available units.
leading article, is more like it. w
He is program manager for a high perfor-
mance microwave spectrum analyzer with
Hewlett-Packard, and he’s been involved with
the design of signal generators, a vector-
voltmeter (8450A) and was project leader for a
spectrum analyzer (8554L). Joined Hewlett- Nabbed by satellite
Packard Company in California as a develop-.
American GEC has demonstrated a system for
ment engineer for r.f. test equipment following preventing drugs being smuggled over remote
graduation in 1961 from Darmstadt University.
parts of the Mexican-uS border which uses
mobile radio and a geostationary satellite.
WIRELESS WORLD. DECEMBER 1978 79

Loudspeaker system design

Changes and refinements to the system described in the May and June issues
by Siegfried Linkwitz, DipI. Ing., Hewlett-Packard Co., Santa Rosa, California

It is unfortunate that there are still so few resonant frequency f

0 and flat sound
loudspeakers commercially available pressure output above it. The height of
which achieve a high standard of the peak near f0 is governed by Q0. Both
accuracy, according to Mr Linkwitz. parameters f0 and Q0 are easily deter-
“After all,” he says, “the design mined from an impedance measure-
concepts are rather straightforward and ment of the driver, Fig. 18. This general
rational.” The design of a loudspeaker transfer function between terminal vol-
system has to include a large number of tage and sound pressure output applies
electrical, mechanical and acoustical to woofers, mid-range units and
parameters if optimum results are tweeters as long as their cone dime~-
desired. There is not one single sions are small acoustically, Fig. 2, and
parameter which by itself will turn a poor must be taken into account when de-
loudspeaker into a superior one; attention NORMALIZED FREQUENCY signing a crossover network.
has to be given to all parameters, f/fo As an example, consider the high-
including the driver, enclosure and pass section of a crossover to a 25mm
crossover. Fig. 22. Frequency response of a dome tweeter which has a resonance of
moving-coil driver with dimensions
800Hz with Q0 of 0.9, Fig. 23(a). The
THESE NOTES are intended to encou- small compared to a wavelength, which
desired acoustic output should follow
rage further development of loud- must be taken into account when de-
the fourth-order high-pass characteris-
speakers, and bring increased enjoy- signing crossover networks.
tic of the 24dB/octave crossover with
ment to those who want to undertake 1.5kHz as the —6dB crossover
the task of building their own systems. frequency (b). At first glance it seems
The changes and refinements made to sufficient to shape the driver terminal
the original loudspeaker system, de- voltage to follow the 24dB/octave
scribed in the May and June issues, are high-pass function of (b) because the
presented to show the completeness of Fig. 23. To achieve an acoustic or overall
the analytical design approach, and high-pass filter response with 24dB/
should not be taken as an indication octave slope (b), requires the terminal
that the previous system is obsolete. Fig. 24. Required drive voltage (c) has to
voltage to follow a 12dB/octave slope be constant below the driver resonance
The audible effects of the changes are below resonance to compensate for the
subtle and the added complexity of the frequency f0 to give the desired acoustic
effects of the driver, whose sound h.p. response (b) (cone excursion shown
circuits would be worthwhile only to pressure and phase response are shown
someone trying to achieve greatest dashed), as a result of driver response
at (a). (a).
accuracy of reproduction. But the tech-
niques described should be of general
interest to any loudspeaker design.
I feel the weakest link in recreating
the illusion of a life source with loud-
speakers lies at the microphone pick-up
end of the signal chain. It seems likely
that more than two loudspeakers are
needed, but first a much better under-
standing for recording and reproducing
the appropriate sound field has to be
developed and demonstrated. Then it
may be possible to transport oneself to
Symphony Hall without moving out of
the living room chair. Meanwhile the
loudspeaker as the necessary electro-
acoustic transducer can approach a
high state of development.
Any moving coil driver has the
general frequency response of Fig. 22
(Fig. 17 ref 16*) when driven from a
constant voltage source. This is a
second-order filter with an asymp-
totic slope of 12dB/octave below the

*Figure numbers prior to 22 refer to the author’s

previous articles, ref. 16.
FREQUENCY (Hz) FREQUENCY (Hz)
80 WIRELESS WORLD, DECEMBER 1978

filter has 22dB of attenuation at the (a) Circuit (d) Design formula~
driver resonance. Indeed, this was the
procedure in the original crossover de- (1) Specify to. 00. f p. 0 p
sign for the T27 tweeter, Fig. 10. Such
terminal voltage, however, causes a
36dB/octave roll-off in acoustic output
(2) to
~1,
_ 00
OP k > a required
from the driver for frequencies below
resonance f0. To achieve the exact Op ~o

acoustic frequency response of (b) the

13) Choose c2 (4) R1= 21Tf0 C2120o(1+k)
terminal voltage must follow a l2dB/
octave slope below the 800Hz driver
resonance (c). This then compensates (5) R2= 2kR1 (6) C1 =c2120011+kl 2
2
exactly for the phase shift and group (7) C f
Gain (b) rrequency response
delay which the driver would otherwise 3 =C1~.f-) (B) R3’R1(T2)
Op
add to the acoustic high-pass function. Ad
The additional phase shift would cause [d BI
(9) Adc = 40 log
a tilting of the radiation pattern as the fp
sound pressures from the tweeter and fip
mid-range unit would add 0. The
to aamount
maxi- te) Circuit analysis
mum
of theat phase off-axis’
a pointshift introduced by a 00
second-order high-pass filter can be
calculated for Q
0~ 0.5 from Gain
tpl= irC
1 R1 1 f~1 ~i required
=
1TC.~ R~
I
Op
00= 2R+R~V~T
fo to = 2ITRp’C1 C2
fp
= 1 2R
For the above example, the driver con- 2ff
tributes 400 of phase shift at 1.5kHz. 3
A
Sound pressures form the mid-range dc’
unit and tweeter are therefore not in 00 FIg. 25. Useful network for compen-
phase unless the measures described are sating driver resonance at f0 and exten-
taken. ding frequency response to f1, for woofer
Ic) Pole-zero location equalization or providing cut-off at 4
Driver terminal voltage for mid-range or tweeter high-pass, res-
The acoustic high-pass function of the jW/2 it
ponses. Calculated values should be
previous example requires an exactly- checked with the circuit analysis equa-
shaped terminal voltage to compensate tions.
for the driver’s own frequency response.
A fourth-order high-pass response is
equivalent to the cascade of two 0.
second-order
The first step Butterworth
then is to equalize
sections’
the
driver output to follow a second-order fpl , ~zl point source. While the mid-ranged unit
Butterworth function by shaping the cr/2ir has to cover one octave less in
terminal voltage applied to it, Fig. 24. frequency, the tweeter must now have
Design formulas were developed for a four times the excursion capability to
very useful network, Fig. 25. It is a maintain the same acoustic output. The
modification of Fig. 20 and will later be Son-Audax unit works well in this
used also to extend the woofer response. application and there is no sacrifice in
A note to those familiar with the de- overall smoothness of response
scription of transfer functions by poles compared to the T27. The new unit does
and zeroes in the complex frequency not roll off towards the high end. For
plane: This network will generate a pair Crossover frequencies and drivers most commercial recordings a slight
of complex zeros (f The techniquie described could be used droop of about 3dB between 2k and
0, Q0) which are to modify the original T27 high-pass 15kHz seems subjectively preferable
positioned to cancel the complex poles filter (f0 1.2kHz, Q0 1.1). Instead, I used a and such response can be easily ad-
of the driver (f0,Q0). In addition, a pair of Son-Audax HD12x9 D25 soft-dome justed with properly designed treble
complex poles (f1,,Q1,) is available which tweeter with a 1.5kHz crossover controls.
are placed at the crossover frequency in frequency to the BilO. At 3kHz, the The crossover point between woofer
the case of the tweeter highpass or at previous crossover point, the BilO cone and mid-range units has been raised
the lower cut-off point of the woofer in diameter is about one wavelength, so from 70 to 100Hz, thus reducing the
the case of woofer equalization. The that a certain amount of directionality maximum cone excursions for the B 110
factor K in the design formulas is can be expected, Fig. 1. Further, the by a factor of two for constant sound
necessary for cancelling a pole-zero pair mid-range and tweeter units are output. Experience has shown that only
(f1,1,f~1) which would otherwise be intro- separated by one wavelength at 3kHz 50 the mid-range power amplifier is
duced by the network. that the combined radiation patter be- occasionally driven into clipping. If
The second step in designing the gins to narrow in the crossover carefully fused a 100W amplifier might
acoustic high-pass filter is to follow this frequency range, Fig. 3(b). The lower be consideredfor driving each Bi 10. The
network with a standard second-order crossover reduces the acoustical di- three-way system is very forgiving to
Butterworth section to achieve the mensions by a factor of two so that a clipping of the mid-range amplifier. It is
overall drive voltage of Fig. 23(d). The wider and more uniform dispersion is not audible on short transients because
complete circuit of Fig. 26 is only obtained over all frequencies in both the the woofer and tweeter channels still
slightly more elaborate than Fig. 14 but vertical and horizontal planes of radia- reproduce their undistorted portion of
it achieves the exact fourth-order tion, Fig. 2. The loudspeaker then the total signal. The reduced frequency
acoustic output, Fig. 23(b). approaches more closely the acoustical coverage of the BilO at both low and
WIRELESS WORLD. DECEMBER 1978 81

Vin

0.9 O~, 0.7 1c 15kHz

~o 800Hz t~ 1 5kHz

Fig. 26. Network for a 1.5kHz 24dB/

octave acoustic high pass filter for a it would mean that its cone excursion
Son-Audax HD12 X 9D25 dome tweeter. has to continue to increase at 6dB!
The first op-amp stage compensates octave even below the 3dB corner. The
exactly for the driver resonance at only practical way is to use a l2dB/
800Hz and gives a 12dB/octave 1.5kHz octave rate. If the Q of this high-pass
acoustic high-pass response. The second network is kept low at 0~5 then a mini-
op-amp stage is a conventional Butter- mum of overshoot is combined with a
worth section. Design formulas for this minimumof cone excursion.
network are from Fig. 25 and Fig. 14. The original network Fig. 13 is a good
approximation. The revised crossover
uses the circuit of Fig. 25 with f~19.3Hz
FIg. 27. Shaped toneburst used to and Q
‘evaluate the audibility of phase distor-
high frequencies improves the amplifier 1,0.5 which gives a 30Hz, 3dB
tion. corner frequency.
power distribution between the drivers.
The crossover frequency between The high-pass nature of the woofer
woofer and mid-range units was not channel introduces phase shift at the
raised further because the centre 100Hz crossover to the mid-range unit
woofer is positioned 0.84m behind the according to the previous formula forQo
mid-range unit and the phase shift due 0.5:
0—2arctan __ 220
—
to this path length would become ex- 4= 180 fP
cessive. Further, the stereo effect might
suffer from the blending of left and This amount of the phase shift by itself
right-channel information for too high a is insignificant, but combined with the
crossover frequency. phase shift due to the woofer location of
In the future it could become 0.84m behind the mid-range it becomes
necessary to have truly full range, necessary to add delay to the mid-range
channel. It is implemented with the
separate speakers for reproducing an
appropriately recorded sound field. network of Fig. 16 which has a phase
Previously the mid-range resonance at shift of
(a)
70Hz was used as one section of the = —2arctan(2ITfRC)
24dB/octave acoustic high-pass func-
tion. The second section was provided Both the absolute value of the phase
by an active network. Now, both sec- shift and the slope of the phase curve, or
tions are implemented electronically the group delay, can be made to
using the circuit of Fig. 25 to compen- coincide between woofer and mid-range
sate for the BilO’s resonance in its channel. The specific network com-
enclosure, with f0 and Q0 determined ponent values R and C depend upon the
set-up of the loudspeakersystem and no
from Fig. 18 (f0 73Hz, Q0 0.6). The com- compensation is needed when mid-
plete network has therefore a configu- range and woofer radiate from the same
ration similar to that of the tweeter, Fig. plane.
26.
Audibility of crossover networks
Woofer equalization Lowering of the tweeter crossover to
The centre channel woofer covers a 1.5kHz raised some concern over the
relatively narrow frequency range. Of audibility of phase distortion. The com-
particular interest is the lower cut-off bined mid-range of tweeter sound
point and cut-off rate. There is some ‘pressure has all-pass characteristic.
(b)
indication that the low-end phase be- Sound pressures from the two drivers
haviour of a system can have audible FIg. 28. Passive network for 24dB/ are in phase at all frequencies relative to
effects. A 5Hz square wave for example, octave acoustic slopes and 1.6kHz each other but the overall sound press-
which sounds like a sequence of clicks, crossover frequency. If constant ter- ure has a frequency-dependent phase
will change its tonal character when minal impedance is assumed for the shift relative to the electrical signal at
transmitted
0. Fromthrough
network
an all-pass
theory it net-
is drivers then the network and its ele- the input to the crossover network. The
work’ that any high-pass filter with a
known ment values can be determined ana- group delay is not constant with
slope of more than 6dB/octave will lytically~’ (a). Computer optimized net- frequency’0. -
produce some amount of ringing to a work for actual drivers shown at (b). A new form of test signal was used
step input’7. It is impractical to roll off which consists of a five-cycle tone burst
the woofer at a 6dB/octave rate because of variable frequency. The tone burst is
82 WIRELESS WORLD, DECEMBER 1978

‘B

100Hz 1.5kHz

unmarked Cs: 741 24dB/oct

B.

IBilo

Fig. 29. Cbmplete schematic of crossover network and equalization of the author’s system.
WIRELESS WORLD, DECEMBER 1978 83

not turned on and off in the usual own to control the motion of the panels tweeter and a 100mm woofer/midrange.
at resonance. A better match between Even the computer-optimized network
abrupt fashion but instead it builds up
the two stiffnesses is required20. of Fig. 28(b) has the desired acoustic
and decays gradually, Fig. 27. The en-
velope of the burst follows a raised Building the enclosures out of 6mm amplitude and phase characteristic only
18. The spectral content plywood with a 15mm damping layer for about two octaves either side of the
cosine function
of The shaped tone burst is concentrated consisting of a 3:1 mixture of water- crossover frequency.
in a narrow frequency range. The ear based roof patching tar and sand gave The active network in contrast to this
appears to be very sensitive to phase optimum results. can be exact because the voltage source
distortion of this signal, while a A simple and quite revealing test is to at the driver terminals is able to impose
square wave or rectangular envelope knock on any box to hear how dead any desired acoustic frequency res-
burst are almost useless at higher acoustically it is. ponse to the driver, without interaction
frequencies for such tests. A system between the source’s frequency res-
with 24dB/octave crossover filters has ponse and the driver impedance.
the phase shift of a second-order all- Passive crossovers
pass network with complex poles and Not everyone is at hom.e with the Note. In addition’ to the points noted on
zeroes of Q = 0.7. No audible change electronics and the rather elaborate page 91 of the October issue, Mr Link-
could be noticed on insertion of this op-amp circuits for this loudspeaker witz points out that the horizontal scale
network into the test signal path. The Q system. A passive crossover seems at- for Fig. 6 is d/X.
had to be increased to 2.4 before any tractive as it would consist only of in-
effect was noticed with the test signal at ductors, capacitors and resistors in a References
‘1.5kHz. Observation with an relatively simple interconnection. Un- 10. Linkwitz S.H., Active crossover
oscilloscope indicated ringing of the fortunately it is considerably more dif- networks for non-coincident drivers,
trailing edge of the shaped burst which ficult for the home constructor to arrive JAES, Vol. 24, 1976. p.2.
became increasingly more audible as Q at the correct element values for a 16. Linkwitz S.J., Loudpeaker system
was raised above 2.4. It can be con- passive network than to design active design, Wireless World, vol. 84, May
cluded safely from these tests and networks with their great flexibility to 1978, p.52 and June 1978, p.67.
others with program material that the change transfer functions and gain’9. 17. Blinchikoff H.J. .and Zverev A.I.,
phase distortion of a 24dB/octave To design a passive network for a Filtering in the time and frequency
crossover is insignificant. 24dB/octave acoustic crossover domains. Wiley, 1976.
Often, claims are made for the super- requires a computer optimization 18. Linkwitz S.H., Narrow band impulse
iority of low-order crossover networks routine unless one is satisfied with the testing of acoustical systems, 60th AES
with 6dB/octave slopes. It should be trial and error procedure on which most Convention, Los Angeles, May 1978,
obvious from Fig. 24 that a 6dB/octave loudspeaker design has been based on preprint no. 1342(F-4), to be published in
acoustic response cannot be realized up to today. If a driver could be JAES.
with a passive network because the represented by a resistor then exact 19. KEFTOPICS, International Edition,
driver itself introduces a 12dB/octave network values are easily cah~ulated2’, vol. 1, no.2A, 1976, and vol. 3, no 1,1978,
slope and the aforementioned Fig. 28(a). Real drivers have complex KEF Electronics Ltd, Tovil, Maidstone
associated phase shift. Merely applying terminal impedances, Fig. 18. This not ME156QP, Kent.
a terminal voltage which changes with only affects the component values of 20. Harris G.M. and Crede G.E., Shock
6dB/octave would guarantee an l8dB/ the theoretical network but also the and Vibration Handbook, 2nd edition,
octave slope below the driver resonance topology as can be seen by comparing ‘McGraw-Hill 1976.
and 6dB/octave above it, but with ex- the two networks of Fig. 28. Here a 21. Linkwitz S.H., Passive crossover
cessive phase shift which defeats the prototype design is shown for a 1.6kHz networks for non-coincident drivers,
whole phase argument for this type of crossover between a 25mm dome JAES, vol. 26, 1978, p.149.
network.
Even a 12dB/octave acoustic high-
pass filter would be extremely difficult
to achieve passively as can be seen from
the required terminal voltage of Fig.
24(c).
The lowest-order acoustic high-pass Association of audio consultants
filter which can be realized with a pass-
ive network has 18dB/octave slope, IN AN A’ITEMPT to improve the standard of lar aspects and members of the Association will be
sometimes called an acoustic Butter- audio equipment reviews, an Association of bound by their code of Ethics to restrict their
Professional Audio and Radio Consultants activities to areas where their professional exper-
worth’9. This filter still suffers from the has been formed. Acting secretary is James tise is relevant. The range of expertise available is
phase quadrature between low and Moir, 16 Wayside, Chipperfield, Herts WD4 fully comprehensive. For example, the prospective
high-frequency driver outputs and the 9JJ. The aims are as follows. client will be able to choose consultants from
resulting frequency-dependent irregu- within the membership of the Association to. give
assistance with:
larity in the radiation pattern’0. Sur- Aims of the association 1. The assessment of the performance of audio and
prisingly then, the 24dB/octave The Association of Professional Audio and Radio radio equipment and the associated software.
crossover is the lowest-order function consultants was formed in July. 1978 to improve 2. All aspects of the acoustics of concert halls.
for which the all-important radiation the standard of services offered by consultants. studios and theatres, including speech reinforce-
work towards protecting the interests of their ment, sound recording and associated techniques.
pattern has a stable axis. So-called “lin- clients and advance the reputation of the pro- 3. Fundamental redesign for development of
ear phase” loudspeakers are based on fession. It recognised that the work of unqualified improved products.
wishful thinking and not on physical consultants sometimes fell below desirable MembershIp: (a) Member: An organisation or an
realities. standards and the membership requirements of the individual offering consulting services with no
association ensure that a high level of professional significant financial interest in the product of the
and technical competence is maintained. service. (b) Conditions of membership: Applicants
The objects of the Association are: for membership shall furnish evidence of pro-
Enclosures (a) To maintain and where possible improve the fessional and technical competence.
Further investigation into the con- standards of professional conduct and competence The ethic Which has been adopted by the
of consultants concerned with audio and radio Association is basically similar to that of the
structon of a well-damped enclosure for engineering. existing professional associations. It ensures that
the mid-range and tweeter led to the (b) To represent and make known the views of its the advice proferred by its members is not in-
following conclusion. members upon matters relating to, or affecting the fluenced by financial interests in the products and
A small box with 20mm thick walls is profession. that any commercial involvement is, in any event.
(c) To promote further education and knowledge in limited to a small proportion of the member’s
too stiff for tar-based damping layers. audio, radio and acoustic engineering. activity. Members will always declare the nature of
The tar has not enough stiffness of its Some consultancies tend to specialise in particu-. this involvement when appropriate.