PAPERS

Complete Response Function and System Parameters for

a Loudspeaker with Passive Radiator*

DOUGLAS H. HURLBURT

SM&A (East) Corp., Arlington, VA 22201, USA

The response function, including all losses, for loudspeaker,systems that use a passive
radiator or a closed box is derived. It is shown that preferred system alignments, such as
maximally fiat ones, can be obtained while including both driver losses and internal enclosure
losses that arise from the use of acoustic filling. A methodology for deriving the appropriate
design curves is presented with an example. The process for obtaining system design param-
eters is explained, and it is shown how the properties of the filling material impact the design
parameters. The additive effect of passive-radiator losses is presented, and a design method
that compensates for these losses is outlined, including curves that provide the necessary
correction factors for a particular case. The impact of leakage losses and a method of
measuring them are also presented. Lastly, a method for accurately determining the param-
eters fp and 8 for an unknown passive radiator is presented in an appendix.

0 INTRODUCTION RAL = acoustic resistance of enclosure losses due to

leakage losses
The use of a passive radiator as part of a loudspeaker Us = acoustical volume velocity of source diaphragm
system was first described by Olson in 1934 [1]. The Up = acoustical volume velocity of passive-
basic acoustic elements of a loudspeaker system with a radiator diaphragm
passive radiator are shown in Fig. 1, where UL = acoustical volume velocity out of enclosure

VB = enclosed volume through leakage.-

M^s = acoustic mass of source (driver) diaphragm l
plus in situ coupled mass I The diaphragmmass includes the voicecoil mass, if pres-
MAP = acoustic mass of passive-radiator diaphragm ent, plus the mass of any moving suspension components.
plus in situ coupled mass
Sl_l = projected surface area of source diaphragm
So2 = projected surface area of passive-radiator MAS .___CAs

CAs = acoustic compliance of source diaphragm SOURCE [] •

ENCLOSURE
• (Area SD1 ) mr
us
CAp = acoustic
diaphragmcompliance
suspension of passive-radiator die- VB I__ ('_ Ira"
=m._
phragm suspension CAB-"-'-

CAn = acoustic compliance of enclosed volume VB RAB

RAS = aCOUStic resistance of source diaphragm Acoust_j

suspension Filling
RAp = acoustic resistance of passive-radiator dia-

RAn = acoustic resistance of enclosure

phragm suspension losses due to RA, _ _ Up

internal energy absorption UL (_Z_"_

I" _ ,,I (AreaSD2)_MAp/I
_ CAPRAp
* Manuscript received 1998 February 25; revised 1998 Au-
gust 22, 1999 September 15, and 1999 December 3. Fig. 1. Loudspeaker system with passive radiator.

J.AudioEng.Soc.,Vol.48, No.3, 2000March 147

HURLBURT PAPERS

In 1974 Small published a pair of papers [2], [3] that Both Rvc and Xvc can be functions of frequency. The net
laid the modern foundation for the analysis and under- acoustic pressure P produced by the driver is given by
standing of passive-radiator loudspeaker systems. 2
However, Small's analysis did not directly address the P = PG -- Us ' ZAG •
presence of internal enclosure and passive-radiator
losses. Since these losses play a significant role in the Appendix 1 presents expressions for RABand CAB-We
performance of the complete system, a formulation of assume that RAB = 0 without enclosure filling and that
the response function that includes these losses plus a the in situ masses, from within the enclosure, which are
methodology that includes them in the system design coupled to each diaphragm, are given by Eqs. (27).
were desired.
In this paper the response function, including losses 1.1.2 Definitions of a and 8
associated with the passive radiator, .enclosure filling, The effective compliance ratios ot and _ are defined
and leakage, is derived. This formulation also includes by [2]
sealed enclosure systems, and it can be extended to
vented loudspeaker systems if desired. In addition, a or- CAS
methodology for passive-radiator loudspeaker system CAB
design that incorporates the losses due to both acoustic
fill and the passive radiator is presented. The impact CAp
of enclosure leakage on the response is also shown in _ - CAB "
conjunction with a means of determining the magnitude
of the leakage losses in a typical passive-radiator, or For a system without a passive radiator, _ = 0. Using
sealed; system. Eq. (28) we can relate CASand CApto effective volumes
VAS and VAp, respectively, given by
1 ANALYSIS

1.1 Introduction and Review VAS= _/PoCAs = a VA. (la)

1.1.1 Analogous Acoustical Circuit VAp = _/PoCAp = _ VAB (lb)

Fig. 2 shows the analogous acoustical circuit to be
used for the loudspeaker system of Fig. 1. On the left where VABis the effective acoustic volume of the enclo-
there is the source of the driver acoustical pressure P, sure. (Manufacturers typically specify both C A and an
consisting of an open-circuit pressure source Po and a equivalent volume VA.)
series acoustic generator impedance ZAG, where
1.2 Response Function
eo " (Bl) 1.2.1 With Passive Radiator
Po = SO1 . (R ° + Zvc) The acoustic response function G(s) is given by [4]

_ 62
ZAG Re + Zv C G(s) - SMAsU°
Pc (2)

f|- Bl _2
/ where U0 -- Us + Up + UL is the total acoustic volume
\Sol / " velocity leaving the system, and s = tr + jto is the
complex frequency. Solving the loop equations indicated
Wealsohave inFig.2 gives
eo = open-circuit output voltage of electrical
source G(s)
B = magnetic flux density, in driver air gap
l = length of driver voice-c0il conductor in mag- SMAs •ZAp
netic field of air gap Z'AsZAB+ Z'AsZAP+ ZABZAP+ (Z'AsZABZAp)/RAL
Re = output resistance of electrical source (3)
Zvc = electrical impedance of driver voice coil,
= Rvc + jXvc The acoustic impedances in Eq. (3) are given by
with
Z_S = ZAS + ZAG

Rvc = electrical resistance of driver voice coil

Xvc = electrical reactance of driver voice coil. S 2 MApCAp + SRApCAp + 1
ZAp =
SCAp
2 It is possible to utilize a driver as a passive radiator by
leaving its electrical terminals unconnected. However, typical
passive radiators are loudspeakers that have neither voice coil ZA B -- SRABCAB + 1
nor magnet. SCAB

148 J.AudioEng.Soc.,Vol.48,No.3, 2000March

PAPERS LOUDSPEAKER
WITH
PASSIVE
RADIATOR

Assuming that [Xvc[ << Rvc over the frequency range xa = RABCAB
where the passive radiator is effective, then
,rL = RALCAB •
62
ZAG--_ -- RAG
RG + Rvc The terms ,rB and "rL have units of seconds and represent
the decay time constants associated with the internal
and enclosurelossesand leakagelosses,respectively.
In the preceding,
R_s = RAS + RAG.
_os = 2,trfs
Thus,
_op = 2"rrfp
sZMAsCAs + SR'AsCAs + 1
Z_s = where
SCAs
fs = resonance frequency of driver
We now make the following substitutions: fp = resonance frequency of passive radiator.

1 Throughout this paper, fs, fP, Qs, and QP assume that

T_ = MAsCAs = tO_ the masses MAS and MAp include the total in situ coupled
acoustic masses that are appropriate for the enclosure
being used. However, when related to a particular en-
T2 = MApCAp 1
= --_ closed volume, they will be written fsB, fPB, QsB, and
(op QPB"(This is in accordance with the convention forfs B,
which we are extending to the passive radiator. Using
= , CAsR_s the same notation for the Q values emphasizes that they
Qs tOsCAsRAS-- _ are specified at fs B and fPB, respectively.)
With these substitutions, Eq. (3) becomes
1 CApRAp

Q--PP
= O_pCApRAp-- _ G(s) = s2T 2 sET2 + s(Tp/QP) + 1 (4)
D(s)
where

D(s)
= ,4.rsr_2 2+ s3L QP + -_s + 'rB(aT2P+ 8

[
+s 2 T2(ot+ 1) + Ts2(B + 1) +_+'rB\Qp + Qs/J

ITs(8 + l) Tp(a + 1) + ,ra(a + 8)] + (a + 8 + l) + _(s).

+ sL a_ + Op J

ZAS
MAS CAS RAS ;

RAL
P

PG ......... <_.....

IJo'_r ....
O 0 ;_
Fig. 2. Analogous circuit for loudspeaker system with passive radiator.

J. Audio Eng. Soc., Vol. 48, No. 3, 2000 March 149

 HURLBURT PAPERS

The term _(s) arises from the inclusion of leakage loss sure. If enclosure volume and acoustic fill are identical
and is given by for both the passive-radiatorsystem and the closed sys-
tem, we expect that "r_ = a-B. However, since some

• (s) = ¢" [s4(T2T 2) + s 3 \-Q--P-P

( T2TP + T2Ts_
Qs / enclosure
of the passive
leakage
radiator,
may we
be associated
must assumewith
that,theinmounting
general,
"r,'# _L.
If _[ --* _, Eq. (6b) can be written in a more familiar

+: + + QsQp:+ s + +1
1,2
where _ = [(sx, + 1)/S'rE]. Gc(s) = (Q_c + y2) + y (6c)
When RAp = 0 and % _ _, Eq. (4) reduces to Small's
equations [2, eqs. (22a), (22b)]. Both G(s) and D(s) can where
be used wherever Small has used the lossless response
function G(s) and its equivalent expression for D(s) in
his analysis of loudspeaker systems with passive radia- Y = sTsQsL
tors [2], [3].
F_,q. (4) gives the response to either steady-state or Q_c = (ix + 1)Qs2L

transient excitations. The steady-state response to a sinu-

soidal excitation pc(to), given by G(jto), is found by RAT = RAS + RAB
substituting s = jto in Eq. (4). The magnitude of the
steady-state response is represented by IG(jto)l.The ze- QSL = 1 _ Ts
ros in the numerator of Eq. (4) define the frequencies tosCAsRAT Ts/Qs + ot "rB
fmJLL at which the steady-state system response is zero.
Except for the trivial case of zero frequency, these are The term Qvc is often referred to as the total system Q
givenby fora closedsystem.

fN-ut_ = -+fP [( ,:11 - _] j (5) 2 SYSTEM

2.1 DI:SlGN
Basic Formulation
Assuming "rL = 0% Eq. (4) can be written as a fourth-
As RAp _ 0, fNUEE _ +fP, as expected [2]. order high-pass response function,

1.2.2 Closed-Box System S4_ + bls3T3o + b2 s2T2

Replacing thebypassive
is represented radiator
MAp ---* with 1,
oo in Fig. a stationary"plug"
and thus ZAp G(s) = s4Tg
° + alsaT_ + a2s2T2 + a3sTo + 1 (7)
ooin Fig. 2. Eq. (3) then becomes
(We examine the % _ oo case in Section 2.6.) Eq. (7)

Gc(s) = SMAs . (6a) is equivalent to Eq. (4) if

Z_s + ZAB "4- (ZrAsZAB)/RAL
_ rs
Eq. (6a) can be written To yl/2 _jl/4 (8a)

Gc(S) - s2"
Dc(s)
T2 (6b) al = 0'/--_ " + _s l (y)]
+ F" + y _ (8b)

with

a'2=_ 1i_" ot y+ 1 +y(8+ 1) +_s +F. +

Dc(s)= [s2"T2 + S(_s +'rBct) + (ct + 1)] + dPc(S). (8c)

yl/2 [ _ + 1 et + 1 ]
The term _c(S) is given by a3 = _7_" L Qs + _ + F. (ct + _)a (8d)

q_c(s) = 4'' s2T2s + Qs + 1 _ yl/2

bl Qpt_l/4 (8e)

where d#' = [(s'r_ + 1)/s "r;.].

lossThetime
terms
constants,
"r_and "r_are
respectively,
the internal
for the
lossclosed
and leakage
enclo- b2 - ¢jY/2 (8f)

150 J. Audio Eng. Soc., Vol. 48, No. 3, 2000 March

PAPERS LOUDSPEAKER
WITH
PASSIVE
RADIATOR

where Using Eqs. (8), this is written

Y = Ts = f p Q2. [y4(r282 ) + 2y28 (F2ct _ 1) + F2ct2 - 2(ct + 1)1

TP fs
+Qs'2Fet+ 1 =0. (12b)
F=V_a
Ts In the absence of internal enclosure losses, F = 0, and
Eq. (12b) becomes
_=et+8+l.
y2 = 1/Q2 - 2(or + 1) (13)
In the steady-state case the magnitude squared of Eq. 28
(7) is
Thus in the absence of passive-radiator, leakage, and
to*T_0[(m
2 - _2To2)2+ m 2] internal enclosure losses, Eq. (13) places an upper bound
Ic'(jm)12
= _oq'go(k_
2 -- (02T_)2 + (1 -- -2 2_2,2
K2(O 1_) + b2. 2,1-2"
_3tu 10 on Qs as a function of or, given by

(9) Qs < [2(or + 1)] -1/2 = Qs(max) .

The coefficients of Eq. (9) are related to the coefficients 2.2.2 Alignment Options
of Eq. (7)by the following equations: There are three typically used alignments derived
a2 from Eqs. (11) [3].
k2 = a2 2 (10a)
• Maximally fiat, requiring that k 2 = k 3 = 0
• Elliptical, requiring that k 3 = 0 (if k 1 = k2 also, then
k_ + k_ = a 2 - 2ala 3 + 2 (10b) the response is symmetric elliptical)
• Quasi maximally fiat, requiring k2 = 0, k 3 > 0.
k 2 - 2k 2 = a2 -2a 2 (10c)
For the purposes of our analysis, we focus on the maxi-
m 2 = b2 b2 mally flat alignment since it is usually the preferred goal
2 (10d) of any loudspeakersystem designer. Results for the other
alignments are obtained in a similar manner.

m_ = b2b_ b4
4 (10e) For maximally
reduceto fiat alignments Eqs. (10b) and (10c)

With certain choices for the coefficients {ki} and {mi} , b_ = a2 - 2ala 3 + 2 (14a)
Eq. (9) can be related to specific loudspeaker system
alignments [3]. Hence Eqs. (10) are used to derive the a] = 2a 2 (14b)
loudspeaker system parameters for any alignment that
can be expressed in the form of Eq. (9). to be solved in conjunction with Eq. (12a). Note that
we have three equations [(10b), (10c), and (12a)] to
2.2 Alignments define the system, and up to five variables {y, F, ix, 8,
The family of alignments that are typically used for Qs}. Thus in general there are no unique solutions to
loudspeaker design are defined by the following general these equations. However, by using the proper tech-
relationships between the coefficients {ki} and {mi} [3], niques, it is possible to obtain parametric solutions.

m_ = 0 (lla) 2.3 Without Internal Enclosure Loss

2.3.1 Design Curves
m2 = k 2 • (lib) In the absence of internal enclosure loss, "rB = 1" =
0. Thus Eqs. (14) with substitutions from Eqs. (8) give
For Eq. (9) to have nontrivial solutions, Eq. (1 la) re- a pair of equations in the terms y2, Q2, ct, and 8. Elimi-
quires that Qp = 0% which limits the use of Eq. (9) nating Q2 by using Eq. (13), Eqs. (14) become
to the design of loudspeaker systems having " useful"
alignments. A design process for Qp # oo is discussed [8 2 + 28(8 + 1)ly 4 + (2(xS)y 2 - (or + 1) 2 = 0 (15a)
in Section 2.5.
[8(8 + 1)21y4 + [ot8(8 + 1)]y2 - d_(ct+ 1) = 0. (15b)
2.2.1 General Constraint
Eqs. (10) and (11) lead to a general constraint for all Eqs. (15) can each be solved directly for y 2, and equating
alignments defined by Eqs. (11), namely, these solutions gives a single equation in tx and 8. (In
both instances only the positive root gives real solutions
2b2 = 2a2 - a2 . (12a) for y). With et as the independent variable, this equation

J. Audio Eng. Soc., Vol. 48, No. 3, 2000 March 151

 HURLBURT PAPERS

can be solved for 8. (Note that there are no useful solu- [2], and the curve for ct = 5 corresponds tOfs = fp. To
tions for a < 1.5.) Then using Eq. (15a) or Eq. (15b) determine the system parameters we start by choosing
and Eq. (12b) we obtain solutions for y and Qs. Once et to specify the desired normalized system response,
the values of a, 8, y, and Qs are determined, the normal- and f3 to establish the half-power frequency. With these
ized half-power frequency fa/fs can be determined from two parameters selected, Fig. 3 gives fsB, feB, and QsB.
Eqs. (9)-(11). (Depending on the alignment chosen, a Then choosing VASbased on driver availability, or as a
numerical root solving routine may be the quickest way manufacturing specification, we determine VB = VAB
to solve forfa/fs.) These calculations lead to the lossless and VApfrom Fig. 3.
design curves, including Qs(max), shown in Fig. 3. As explained earlier, when specifying system param-
These curves can be compared directly with those of eters, we use fSB and feb for the driver and passive-
Small [2, fig. 5]. radiator resonance frequencies, respectively. This is be-
cause a self-consistent set of system parameters requires
2.3.2 Response and System Parameters that fs and fp include the appropriate in situ coupled air
Using Fig. 3, IG(jco)
I, normalized tofs and as a func- mass when the driver and the passive radiator are in-
tion of or, is shown in Fig. 4. The curve for a = 3 stalled in the specified enclosure. Likewise, QsB = Qs
corresponds to the a = 8 example discussed in Small specified atfs B.

15 I 30
14 17B = U.U 2.8 Y,

13 Qp =oo 2.6 f3/fs

12 '_L =o0 7 "'_ 2.4

10 Qs(ma_
2.0 -- 0.5
9 /- 18
t 1.6 -- 0.4
7 Qs

_ 1.2 -- 0.3
5 ,_ _ J'<'_'_ _" 1.4

4 /_ _ "'_ -_.._
'_- _"_ 0.8 -- 0.2
3 _< / _ --_ 1.o
0.6 Qs
2 _ 0.4 --
1 0.2
ol
0 0.0 __ 0.0
1.0 1.5 2 3 4 5 6 7 8 9 10
(X

Fig. 3. Maximally fiat design curves without internal enclosure loss.

i
;-
/ilji :!x i " ,
-35 T_

0.4 0.5 0.6 0.7 0.8 1.0 1.5 (oTs2.0 3.0 4.0 5.0 6.0 7.0 8.0

_= I _2.0 --_-3.0 4-5.0 --*-8.0 I

Fig. 4. Maximally flat response without internal enclosure loss.

152 J. AudioEng. Soc., Vol. 48, No. 3, 2000 March

PAPERS LOUDSPEAKER
WITH
PASSIVE
RADIATOR

we have
2.3.3 Coupled Mass Correction
In Section 2.3.2 we determinedfs B andfp a. However,
manufacturers typically specify the free-air resonance QSA = QsB .fs_._Ba
fSA (19)
frequencies fSA and fro. Thus we must correct for the
difference between the total coupled mass atfs B andfp a where QSA refers to Qs determined atfs A.
and that atfs A and fro. Also, we must determine QSA = Evaluation of Eqs. (17) for an unfilled enclosure
Qs specified atfsA- Since shows that, for most cases, fsB/fSA _--fPB/ft'A >_ 0.95.
(This also validates our earlier assumption that mABs < <
(O2_ 1 MAS(fSB) and mABP(faB).) Hence assumingfsB, fPa, and
MAsCAs QSBto be the free air, or spec sheet, values will typically
have a negligible impact on the final system response.
we have
2.4 With Internal Enclosure Loss

mABS
fs_= 1 MAS(fSB)
+___%]1/2
J (16) 2.4.1 Design
For _'B# Curves
0, Eqs. (14a) and (14b) become

where Q2. {y4182 + 28 - 2FEB(or + 8)]

MAS(fs B) = effective acoustic mass of driver for res- + y212(a + 1)(8 + 1) + 20 - 2F2ot(a + 8)] + (a + 1)2}
onance atfa
mAas = acoustic mass coupled to driver from - Qs " (4y 2Fet) + [y 4F282 - 2y2( 8 + 1)] = 0 (20a)
within enclosure
es = difference between external coupled air and
mass when driver is mounted in enclo-

sure and when in free air. Q2. {y2[F2(o t + 8)2 _ 20(8 + 1)] - 2O(a + 1)}
We assume that es << rnAB s << MAS(fSB). Then using
Eq. (la), we write Eq. (16) as + Qs" (2y 2For) + [Y2(8 + 1)2] = 0 (20b)

respectively, and must be solved simultaneously with

fsB
fSA _ 1 -- mABS" 2"lr2VAsf2a
"YPo (17a) Eq. (12b). Solutions can be obtained by fixing one vari-
able and letting another be independent. However, the
In a similar manner, complexity of the equations suggests a numerical
approach.
27r2VApf2B We choose to fix the value of F and let et be the
fp___a
_ 1 - mABP (17b) indePendent variable, thus ensuring that the resulting
fm _/Po design curves will be of the same form as those of Fig.
3. However, there are two apparent problems with this
From Eqs. (27) (Appendix 1), with Ix = 0, we also have
choice. First F is given by

B1
= " "r-a= 2_rfs •RABCAB (21 a)
mABS (qi.SDl)l/2 P0 (18a) F = Ts

B2
mAap = (Ti.SD2)I/2" DO" (18b) or, using Eq. (la),

Using a value of SD1 based on available drivers, or F = 2¢rfs • RABVAs (21b)

ot_/P0
chosen as a "spec" value for manufacturing purposes,
and a desired baffle area Sa, the value of B 1 is deter- Thus for any fixed value of F, we expect to have a range
mined. (For example, see Leach [5, fig. 2].) Thus we of options for fs, et, VAS, and RAB. (However, we will
can estimate fs A using Eqs. (17a) and (18a). Likewise, see that our choices are more limited than they appear.)
by choosing a value for So2, we estimatefp A using Eqs. The second problem is that we have no aprior knowledge
(17b) and (18b). With these results we can use Eqs. (18) of what represents a reasonable range of F for use in
to confirm our earlier assumption that e s << mga s and picking a value for the calculations. We address the
ep << mAaP [6, eqs. (8.12) and (8.15), p. 222]. Fi- latter problem by choosing a maximally flat response
nally, since from Fig. 4 and evaluating 16(j.,)l for values of I" =
1 [0, F(max)]. This shows how sensitive IG(jto)l is to
Qs - changing values of F, and it also helps to identify a
2wfsCAsR_s reasonable upper bound for F(max). Using the et = 2

J. Audio Eng. Soc., Vol. 48, No. 3, 2000 March 153

HURLBURT PAPERS

curve from Fig. 4 and varying the value of F gives the a range of a and find the results shown in Fig. 7. We see
results shown in Fig. 5. As expected, for fixed a, inter- that for the same a, the value offa/fs and the stopband
hal enclosure loss has virtually no impact on the stop- response are increased relative to the case without inter-
band response orfNuLL- However, it has a definite impact nal enclosure loss. However, Fig. 7 shows that with
on the flatness of the response and f3, as shown by the the proper assumptions, maximally flat designs can be
horizontal dashed line in Fig. 5. obtained while including the effects of internal enclo-
Based on Fig. 5 we choose F = 0.2 for the remaining sure loss.
analyses. Using the numerical method described in Ap-
pendix 2, Eqs. (12b), (20a), and (20b) can be solved 2.4.2 System Parameters
simultaneously to produce the design curves shown in We start by choosing a based on the desired normal-
Fig. 6. (It is necessary to derive new curves for each F ized response, plus f3 and VAS. From Fig. 6 we then lind
to be considered.) Comparing the curves of Fig. 6 with fSB, VAB, VAp,QSB, andfpB. The values of Sol and SD2are
those of Fig. 3 it is seen that the inclusion of internal selected by either choosing both a driver and a passive
enclosure losses has a major impact on the maximally radiator that approximates the preceding requirements,
fiat design parameters. Most notably, for equal values or sp6cifying them for the purposes of manufacture. This
of a, 8 is decreased while y and f3/fs are increased, still leaves us with the need to determine VB.
Using the curves of Fig. 6, we calculate IG(jco)lfor For fixed F we have R_ given by Eq. (21b). However,

0
-5 i i

-10 , ct =2 !
IG(j_)l _ = 6.16 i
(dB) -15 y =0.44
Qs = 0.347 i

-20 _" i ! ' Qp =oo --_

-25 _ xL = oo

-35
-30 _ :'

0.4 0.5 0.6 0.7 0.8 1.0 1.5 (oTs2.0 3.0 4.0 5.0 6.0 7.0 8.0

F= I --0.0 -*-0.1 +0.2 -k-O.4 I

Fig. 5. Response with internal enclosure loss.

15 3.0

F = 0.2 2.6

11
12' '_L
Q, =o0
=oo ,_ // 2.2
24 fs/fs
10 / / / 2.0 -- 0.5
9
/ // 1.8
8
8 Qs _ / / 1.6 -- 0.4
7 __ -- 1.4
6 _ 1.2 -- 0.3

5 / Y 1.o Qs
4 \/ 0.8 -- 0.2

3 / ,_ 0.6
2 _ _- ..._... 0.4 -- 0.1
1 _ _ 0.2
0 0.0 0.0
1.0 1.5 2 3 4 5 6 7 8 9 10

Fig. 6. Maximally fiat design curves with internal enclosure loss.

154 J. Audio Eng. Soc., Vol. 48, No. 3, 2000 March

PAPERS LOUDSPEAKER
WITH
PASSIVE
RADIATOR

from Appendix 1, Eq. (25c), idate the assumptions that led to Eqs. (17).] Hence as-
sumingfsB,fp,, and QsB to be the free air, or spec, values
RAa -- B(IX • no • rF) may result in a system whose response is measurably
('trSD)1/2 different from the desired response.

Thus RABis also related to (1_ • no • re), which is deter- 2.4.3 Importance of Filling Parameters
mined by the particular filling material used and the The importance of knowing (no • rF), Ao, PF, and cF
filling factor IX. Using the selected values of Sm and for various filling materials is readily apparent. Without
SD2, and the desired baffle area SB, the value of B for them we cannot specify all of the necessary design pa-
use in Eq. (25c) is readily determined. (See, for exam- rameters for a passive-radiator--or any other--system
pie, Leach [5, fig. 2].) Then if we know (no • rF), Ao, that employs filling material. The accurate measurement
PF, and cF for the filling material, the value of IX is of these parameters, for different fill types, will enable
determined by equating (25c) and RAB from Eq. (21b). system designers to employ the design curves and meth-
The enclosure volume VB is then found from Eq. (29). odology presented. (Note that Leach [5] provides these
Lastly,fsA andfa Aare given by the equation pairs (17a), parameters for common fiberglass insulation, except for
(27a) and (17b), (27b), respectively, with the appropri- (no • rF). )
ate value of IX.The value of QSAis found from Eq. (19). Also, if we know (no • rF) for a particular filling that
Alternatively, we can choose the filling factor IX.Then we wish to use, we can work the solutions backward by
using Eq. (25c) and RAB from Eq. (21b), the required choosing IX, or,fs, VAS, Sin, and SD2. Then by solving
value of (no " rF) is specified. Once we find a filling for F by means of Eqs. (21b) and (25c) and using the
material that provides this value of (no • rF), VB is deter- technique described in Appendix 2, we determine the
mined from Eq. (29) using the values of A0, PE, and cE remaining parameters for the desired alignment.
for the material. Once the filling parameters are known,
the values OffsA, fPA, and QSA are found as before. 2.5 Designing with Passive-Radiator Loss
From Eqs. (27) the internal coupled mass density, 2.5.1 Impact of Loss
relative to air, is given by Unlike drivers, passive radiators have only mechani-
cal loss, typically 1.0 _< Qp _< 5.0. Using a maximally

v = IxA0 (Po
PF_ 1 ) + 1. (22) flat alignment
[G(jto)[ (Qp = of
as a function _)withot = 2, intheFigs.
Qp is shown variation
8 and9in
for F = 0 and F = 0.2, respectively. In both cases it is
Using the typical values for fiberglass fill [5], we have seen thatf3/fs increases with increasing passive-radiator
loss, resulting in significant deviations from the maxi-
v -------
5IX + 1. mally flat response for typical values of Qp. In addition,
as Qe decreases, the stopband response increases and
Thus when using filling, mAas and mABe will be a larger the response null is eliminated. (This is obvious in the
fraction of MAS and MAp, respectively, than in the un- case with internal enclosure loss of Fig. 9. However, it
filled case. (However, they are not large enough to inval- is also true without internal loss.) It is readily shown

0
-5
i ' _"
I
__ i_i__:
!

-lO i ! I
IG(j_ot I
_ i] [:
it
i
(dB) -15 i i
! i =

m20

i _II I !
-25 E _ 1 i ! ' [

: 1 ' i ! I 1

0.4 0.5 0.6 0.7 0.8 1.0 1.5 mTs2'.O 3.0 4.0 5.0 6.0 7'.0 8.0

Ct = I --2.0 -- 3.0 4- 4.0 I

Fig. 7. Maximally flat response with internal enclosure loss.

J. Audio Eng. So¢., Vol. 48, No. 3, 2000 March 155

HURLBURT PAPERS

that as Qp ---->0, the response approaches that of the filling since this is the normal situation for any loud-
equivalent sealed box for both cases. This is intuitively speaker system.
obvious since a passive radiator with infinite damping Starting with a maximally flat (Qp = oo) design that
resistance is the equivalent of a "plug." includes the desired internal enclosure loss, we achieve
compensation by lowering fsB and fPs through the addi-
2.5.2 Design Compensation tion of mass to the respective•diaphragms. This results
While introducing passive-radiator loss causes unde- in new values for fSB, Qss, fPB, and QPB based on the
sirable effects, some relatively simple changes in the initial value of Qp, However, the remaining system pa-
system design parameters can compensate for the worst rameters are unchanged. (It is possible to provide some
of these effects. The resulting "compensated response," degree of compensation by changing i_; however, this
while no longer maximally flat, has a similar value of impacts numerous design parameters and is a less desir-
f3 and provides an acceptable response characteristic able approach.)After compensation, the original driver
over the frequency range of the system. While the corn- and passive radiator can be used with the applied mass
pensation can be applied to a system without internal or new ones selected (or manufactured) to meet the new
enclosure losses, we will focus only on the case with specifications.

i i
't
_,oZ ! ,,:
(dB)
F- T

-20 _ _ i ]

-35 :', 'Tj [ ' ',

0.4 0.5 0.6 0.7 0.8 1.0 1.5 (OTs2.0 3.0 4.0 5.0 6.0 7.0 8.0

Qp = [ -- oo _5.0 _2.0 +1.0 +0.5 ]

Fig. 8. Impact of adding passive-radiator loss, without internal enclosure loss.

-10 F = 0.2 {

(dB) -15 XL =oo

-20 -- -- !

-30

-35
-25

0.4 0.5 0.6 0.7 0.8 1.0 1.5 o_Ts2.0 3.0 4.0 5.0 6.0 7.0 8.0

ap = _ oo _ 5.0 + 2.0 4-1.0 --x- 0.5 ]

Fig. 9. Impact of adding passive-radiator loss, with internal enclosure loss.

156 J. Audio Eng. Soc., Vol. 48, No. 3, 2000 March

PAPERS LOUDSPEAKERWITH PASSIVE RADIATOR

The easiest way to converge on a compensated solu- the passive radiator to reduce fe by _ 12%, with an
tion uses an empirical approach using a digital computer increase in Qp given by Eq. (23b). The Compensa-
with software that rapidly calculates IG(jto)lfor different tion--Final curve shows that f3 is virtually unchanged,
choices and presents the results graphically. (It is con- but there is a very slight rise in the response (20.15 dB)
ceivable that compensated values for fs and fp could at toTs(0) ----3.5, relative to the maximally fiat response.
be determined from Eq. (9) by requiring that f3 remain While not maximally flat, the compensated response is
unchanged. However, without an additional constraint probably acceptable for a high-quality loudspeaker
that is relevant to the desired response, there are a large system.
number of potential solutions, most of which have an Using a = 2 and F(0) = 0.2, this empirical process
undesirable response characteristic.) gives the results shown in Fig. 11. The values required
From Eq. (21b) it can be seen that lowering the value to achieve a compensated response are plotted against
offs, while making no other changes to the system pa- QPB, the final compensated value of Qp. (If a different
rameters or the acoustic filling material, will also reduce value of a or F(0) is employed, a different set of compen-
the value of F while RABremains unchanged. In addition, sation curves will result.)
when the resonance frequency is reduced by mass load-
ing the diaphragm, the effective Q increases proportion- 2.5.3 Determining the Compensated System
ately. From Eq. (19), the "compensated" Q is written Parameters
To determine the compensated system parameters, we
Qs = Qs(0) fs(0) (23a) start with the designcurvesof Fig. 6, assumingthat
fs F(0) = 0.2. We choose to define the system response
by letting a = 2 and f3 = 40 Hz. Then from Fig. 6,

Qp = Qp(0),fp(0) (23b) fSB(0) = 25 Hz and B = 1.95. Thus "rs = 1.273 X

fp 10 -3 from Eq. (21a). We then assume that QPB = 2,
such thatfs s = 23.1 Hz, QSB = 0.423, andfp B = 19.5
for the driver and passive radiator, respectively, where Hz, from Fig. 11.
(0) indicates the precompensation values. (Note that Using these values, Fig. 12 shows the results for the
fs(0), fp(0), and Qs(0) are the values for the maximally maximally flat (Qp = _) and compensated (QPB = 2.0)
flat response that is chosen as the starting point.) cases. We see that f3 has been preserved as desired, and
Fig. 10 shows the compensation process assuming that the variation between the maximally flat response
Qp(0) = 3.0. (The vertical and horizontal scales have and the compensated response forf I>f3 is < 1 dB. Also,
been changed to ease viewing of the different curves.) there is no minimum in the compensated response due to
Note that the frequency axis is normalized tofs(0). Also the high loss of the passive radiator. For comparison, the
included are the critical portions of the curves from Fig. closed-box response using the same values of a, fSB, QSB,
9. The first step in the process, shown by the curve and "rs is shown. As expected, the compensated response
labeled Compensation--Step 1, required mass loading lies between that for Qp = _ and the closed box.
the driver diaphragm to reducefs by - 6.6%. From Eqs. Determination of the remaining system parameters be-
(21b) and (23a), this also results in equivalent changes gins by specifying VAS, which defines VABand VAp. The
in F and Qs- The process is completed by mass loading value of RAB is found from Eq. (21b), and the filling

-1

-2

-3

IGOod-4
(dB)
-5

-6: : / -
I i i
-8 1
i
-9 I i
-lo ! I
1.5 2.0 (oTs(0) 3.0 4.0 5.0 6.0 7.0 8.0

QP= I -- oo --_-3.0 (Uncompensated) I

Fig. 10. Resultsof passive-radiatorcompensation

process.

J. Audio Eng. Soc., VoI. 48, No. 3, 2000 March 157

HURLBURT PAPERS

factor ix is determined in either of the two ways de- flat (Qp = oo) ct = 2.0 examples studied earlier, both
scribed in Section 2.4.2 using our assumed knowledge with and without internal enclosure loss. It is seen that
of (no • rF). Then VB is found using Eq. (29) and the the impact on It;(j_)l increases rapidly with decreasing
other parameters for the chosen fill material. The values "rL, but that for "rL > 0.1 the impact is slight. Thus we
of fs A and fPA are found as before. The value of QSA is consider rL = 0.1 = "rLTto be the "threshold" for enclosure
found from Eq. (19), and QPAis given by leakage loss for these examples, and "rL> 0.1 is equivalent
to xL = oofrom a response standpoint. (Note that changing
QPA= QPa "fPB (24) some of the system parameters in order to achieve a partic-
fl,A " ular response function may alter a'LT.Thus for each new
system design, IG(j_o)l should be reevaluated to determine
2.6 Impact of Leakage Losses the appropriate "rLT.Then knowing "rLT,the enclosure can
In the preceding analyses we assumed that "rL _ oe. be built to ensure that "rL > "rLT.)
However, we do not know if this assumption, which led A method for determining TL is described in Appendix
to Eq. (9), is valid for our cases of interest. This issue is 3. When performing the measurements using a well-con-
addressed by reevaluating IG(j_o) I for specific examples, structed loudspeaker enclosure, we find that typically x L
both with and without internal enclosure loss. Figs. 13 1> 0.1. Hence the decision to ignore the leakage losses in
and 14 show 16(j_o)l as a function of'r Lfor the maximally the derivation of Eq. (9) is valid for a well-built system.

0.500- 1.00 , , 1.75

a = 2.0 8 = 1.95

F(0) = 0.2 "_L ---- O0

0.475-- 0.95 ft_,\ 1.72

,3,,sB \

0.425-- 0.85 _, _, ,_ 1.66

0450-
QSB
0Y,0
0.400-- 0.80 /° _

_
_
159
f31 fsa

1.63

0.375-- 0.75 ' ' 1.60

1.0 2.0 3.0 QPB 4.0 5.0 6.0

Fig. 1]. Passive-radiator compensation curves for a = 2, F(0) = 0.2.

0 i i i i i i i i i _

-5

(dB) -15
,G(f)1-10- C,osed S_ --

_1 a =2.0
_- _ ! XB = 1.273x10 "3 __

2
-30

-35
10 15 20 30 40 50 60 70 80 100 150 200
Frequency (Hz)

QpB= I _ oo -=-2.0 I

Fig. 12. Response for compensated passive-radiator loudspeaker system.

158 J. AudioEng. Soc., Vol. 48, No. 3, 2000 March

PAPERS LOUDSPEAKER
WITH
PASSIVE
RADIATOR

3 SUMMARY 4 presents a method, using the measurement of four

resonance frequencies, which can accurately determine
This paper derives the complete acoustic response the values of feb and 8 for a passive radiator whose
function, including all system losses, for a loudspeaker parameters are unknown.
with a passive radiator or using a closed box. It also
derives the system parameters for a passive-radiator sys- 4 ACKNOWLEDGMENT
tern with and without internal enclosure losses. The anal-
ysis shows that knowledge of the acoustic fill parameters The author would like to thank the reviewers of the
is required to fully specify the system parameters when previous versions of this paper who have challenged him
such fill is used. The additive impact of passive-radiator in numerous ways. Their comments were both beneficial
losses is presented, and we show a method that can and illuminating and have helped him to prepare a
be used to compensate for their effects. The effects of greatly improved manuscript, in addition to helping him
enclosure leakage losses on the response function are increase his overall knowledge of the subject at hand.
shown along with a method for estimating their magni- The author would also like to thank his wife for allowing
tude in a finished loudspeaker system. Finally Appendix him the free time to devote to this effort, knowing that

5 I i

=
-5, i :i i
-10
_ i ;
JG(jco)l I '_ i ;:
(dB) -15 i i I

-20 !

-2s i I_ "! .

-30 . ,

-35

0.4 0.5 0.6 0.7 0.8 1.0 1.5 roTs2.0 3.0 4.0 5.0 6.0 7.0 8.0

"_L= I --oo -,--0.1 ---0.01 --0.001 I

Fig. 13. Impact of adding leakage loss, with internal enclosure loss.

-10
[G(jo,_
(dB) -15

% = 0.0
-20
c_ = 2.0

Qp -- oo
_25
-T---

-30

-35

0.4 0.5 0.6 0.7 0.8 1.0 1.5 roTs2.0 3.0 4.0 5.0 6.0 7.0 8.0

¢L = I -- oo -,--o.1 --o.ol --_-o.ooI I

Fig. 14. Impact of adding leakage loss, without internal enclosure loss.

J. Audio Eng. Soc., Vol. 48, No. 3, 2000 March 159

HURLBURT PAPERS

it was only a hobby, but that it created a good excuse CAB, and the mechanical parameters associated with the
for working with his computer, filling material are represented by the components mAa2,
RAB2, RAB3, and CAB3- The components mAsl and mAB2
5 REFERENCES represent the acoustic mass of air and filling, respec-
tively, which are coupled to the total piston area that is
[ 1] H. F. Olson, "Loudspeaker and Method of Propa- exciting the enclosed volume.
gating Sound," U.S. patent 1,988,250, applied 1934 In the absence of filling [5],
Feb. 17, granted 1935 Jan. 15.
[2] R. H. Small,"Passive-RadiatorLoudspeakerSys- RAB3 = RAB2 = mAB2 = 0 ,
terns Part I: Analysis," J. Audio Eng. Soc., vol. 22, pp.
592-601 (1974 Oct.). CAB3---* oo , CAB---) CAB(0) .

[3] R. H. Small, "Passive-Radiator Loudspeaker Sys- Thus a lossless unfilled enclosure is represented by a
tems Part II: Synthesis," J. Audio Eng. Soc., vol. 22,
pp. 683-689 (1974 Nov.). coupled acoustic mass mABl, in series with a compliance
CAB(0). If we wished to allow for a small amount of
[4] R. H. Small, "Direct-Radiator Loudspeaker Sys- internal enclosure loss in the absence of filling, we
tern Analysis," J. Audio Eng. Soc., vol. 20, pp. 383-
395(1972
¢
June). wouldwrite
[5] W. M. Leach, Jr., "Electroacoustic-Analogous R_B3 = RAB3 + A
Circuit Models for Filled Enclosures," J. Audio Eng.
Soc., vol. 37, pp. 586-592 (1989 July/Aug.). where A is the internal enclosure loss without filling.
[6] L. L. Beranek, Acoustics (Acoustical Society of In the limiting case, where the compliance of a single
America, New York, 1996). fiber --->oo, and the aerodynamic drag of a single fiber
[7] V. Dickason, The Loudspeaker Design Cookbook, is much greater than its damping resistance, the electro-
5th ed. (Audio Amateur Pres, 1995), chap. 8. acoustic circuit of Fig. 15(a) reduces to that shown in
Fig. 15(b) [5], where
APPENDIX 1
B
In 1989 Leach [5] showed that for frequencies << 1 mABI = (,nSD)1/2" P0 (1 -- IX"A0) (25a)
kHz, the complete thermodynamic and mechanical
model for a closed volume VB, which is leak free and B
uniformly filled with a fibrous acoustic absorbing mate- mAB2= (,rrSD)l/2" IX" A0" OF (25b)
rial can be represented by the analogous electroacoustic
circuit given in Fig. 15(a). In this circuit the thermody- _ B(IX. no" rF)
namic effects of the filling material are represented by RAB = RAB2 (,nSD)t/2 (25C)

CAB = "/Po
VB " (1 - Ix" Ao)" [ 1 + {1 + _[(IX • Ao)--i -- 11 (pocJpFcF)} ] (25d)

To Driver To Passive Radiator

mABt --, -_,

mAB1S mABIP
r ..... ;
i
'1 mABS n_ABP

i mAB QAS2S mAg2p

mAB2 _
--- -4.,
i mAB2
mABI
CAB3

CAB -- CAB

RAB2 RAB2
R'_z I
L.
CAB

(a) (b) (c)

Fig. 15. Analogous circuit for filled enclosure.

160 d. Audio Eng. Soc., Vol. 48, No. 3,2000 March

PAPERS LOUDSPEAKER
WITH
PASSIVE
RADIATOR

In these equations, It is evident from Eq. (29) that, in the absenceof filling
SD = total piston area, = SD1 + SD2 (_JI, = 0), VAB "-> V B.
B = Beranek B factor for piston area SD
Po = density of air, = 1.18 kg/m 3 APPENDIX 2
_/ = ratio of specific heat at constant pressure Ce to
specific heat at constant volume c v for gas This appendix presents a numerical methodology for
within enclosed volume, = 1.4 for air simultaneously solving Eqs. (12b), (20a), and (20b). To
cv = 717 J/(kg • °C) for air achieve solutions, we fix I"and solve for (y, 8, Qs) while
Po = pressure within enclosed volume, = 1.013 × letting ct be independent. We begin by choosing a set
of discrete values {Yi} which span the range 0 _<
105 N/m 2 for standard atmosphere
Pv = density of fiber material used in filling Yi _ Ymax"We choose Ymax_ any expected y and, for
cF = specific heat of fiber material used in filling our example, Ym, = 2. Then for each value of Yi and
our chosen et, we use a numerical minimizing process
rF = damping resistance of single fiber
to determine the values 8j that cause both Eqs. (20) to
?to = total fractional
for filling fiber volume, per unit volume,
material vanish. This requires that for each value of Yi and 8j we
solve Eq. (12b) for the appropriate value of Qs and
no = number of fibers in filling material per unit
volume substitute it into Eqs. (20). (The positive root gives the
p, = filling factor, equals fraction Ofenclosure filled only real solution.)
The result of this process is a set of values {Yi, 8i}a
with acoustic material, and {yj, 8j}b, which are solutions ofEqs. (20a) and (20b),
Use of the filling factor Ix in the preceding equations respectively. Since the {Yi} are the same for both sets
implies that a partially filled enclosure is equivalent to of solutions, taking the difference [(8i)a - (Sj)b[for every
a fully filled on that uses a filling of correspondingly {Yi} and finding the minimum implies that (Si)a --- (Si)b.
lower fiber density. Combining Eqs. (25a) and (25b) Alternatively, the set of solutions {Yi, _i}_ and {Yi, 8i}b
gives can be displayed graphically to find the intersection in
(y, 5) space.
B Either way, the result is the unique pair (Ym, _m) that
mAS = (_rSo)1/2 " [p0(1 - Ix " Ao) + _" Ao" PF] (26) simultaneously solves equations Eqs. (12b) and (20) for
a given value of ot and the specified value of I'. The
representing the in situ internal acoustic mass coupled value of Qs is then readily found using y = Ym, 8 = 8m in
to a diaphragm of area S o. Eq. (12b). The only errors associated with this approach
come from the granularity in the fixed step size (Yi+l
Assuming that the particle motion within the enclosure
is the sum of independent components from each dia- - Yi) and how close to zero the solutions of Eqs. (20)
phragm, we separate mat I into two parts, are driven by the chosen minimizing routine. If a graphic
comparison is used, we must consider how well the re-
suits can be interpolated.
_ BI
mASS 0rSD01/2 [O0(1-- Ix" Ao) + _" Ao" I_] (27a)
APPENDIX 3

82
mABp-- (WSD2)I/2 [po(1 - _" Ao) + tL"/to" OF] (27b) temWebycan estimate -r
connecting aLlow-voltage
for a finisheddcpassive-radiator sys-
source to the driver
terminals, thus creating an instantaneous displacement
as shown in Fig. 15(c). Terms B 1and B 2are the B factors of the driver diaphragm. This causes the passive-radiator
for the driver and passive radiator, respectively. Eqs. diaphragm to undergo a step displacement x oin the oppo-
(27a) and (27b) represent the in situ internal acoustic site direction. Immediately after this displacement, and
mass coupled to the driver and passive radiator, respec- while keeping the dc voltage applied to the driver, the
tively. When added to the respective diaphragm acoustic passive-radiator diaphragm will slowly decay to its equi-
mass, plus the appropriate external coupled air mass, librium position due to enclosure leakage. (A small leak
we have the acoustic masses MAS and MApused in Fig. 2. is always needed to prevent changes in atmospheric pres-
We can write Eq. (25d) as [6, eq. (5.38), p. 129] sure causing displacements of the driver and passive
radiator diaphragms.)
We assume that this return to the equilibrium position
CAll - VAB
_/P0 (28) is given by

where VABis the effective acoustic volume of the enclo- x(t) = x oexp( - t/Xo)
sure. Thus

V_.._ = VB" (1 - _" Ao)" 1

J. Audio Eng. Sot., Vol. 48, No. 3, 2000 March 161

HURLBURT PAPERS

where % is the in situ decay time constant for the passive- (In Eq. (34) the acoustic mass of the driver and passive-
radiator diaphragm due to leakage. Using Fig. 2, with radiator diaphragms must include the appropriate in situ
the driver "clamped" by the applied dc voltage (that is, coupled air masses. Thus we use 10SBand 10PB,in accor-
Us = 0) and assuming that the decay process is slow dance with the convention we adopted.)
enough that MAp, RAp, and RAB can be ignored, we have Using a sealing plug in Place of the passive radiator,
MAP --. oo. Then 10PB--* 0 and 1032--* 0 and Eq. (34)

x° (30) collapses to a single solution given by

102
= 10&
+102, (35)
Since it is difficult to observe precisely when t = To,
we measure the time tR after which additional motion is where fE = 10E/2_r is the resonance frequency of the
negligible and assume that tR _ 3 • Xo. driver mounted in the sealed enclosure. Eq. (35) can
The same process is used to find T£for a closed system, be written
However, after a time > > "r0, the dc voltage is removed

fromtTo di
nvor
.ra,mt our,oo esamo
type of decay and ( f2 = f2 B 1 + _sa] = fs2a- 1 + CAB//

% (31) (36)
"rL_ Ot+ 1 "
In the same manner, if CAS--->0, then ms ---*0 and o31
APPENDIX 4 --* 0, and Eq. (34) collapses to a single solution given by

This appendix describes a sequence of resonance fre- ( 102B_

quency measurements by which the parameters a, _, 102 = (102B + 1022)= 1022\1 + _-_322
]
and fp for a particular loudspeaker system can be deter-
mined with reasonable accuracy. It is also assumed that
we know the volume VB of the enclosure to be used for _ 1022(_ + 1) _ 102B(8 + 1) (37)
these measurements. We start with Fig. 2 and assume _ " "
that RAs = 0 and RAL = oo. IfpG = 0, nontrivial solu-
tions for Us and Up require that Eq. (37) is identical to Small's equation [2, eq. (11)]
and defines the box resonance frequency with the pas-
(102_ X12)(102_ h2) _ ix2 = 0 (32) sive radiator.
Substituting Eqs. (35) and (37) into Eq. (34) gives
where
2102 = 102 + 102
kl 2 = I/CAs + I/CAB

+ (002- (o2)2 + 4(e,_ 2 z

- - 10SB)
10B
h 2 = 1/CAp + 1/CAB
MAp (38)

1 Eq. (38) leads us to expecttwo distinctresonancefre-

_2 = C2B(MAsMAp) " quencies for a passive-radiator system identified as fu
and fL for the upper and lower resonance frequencies,
The solutions to Eq. (32) are respectively. Using Eq. (38) and letting fn and fL be
given by the + and - signs, respectively, gives
2o 2 = (h_ + h22)--- [(h_ - h2)z + 41_z]_/2. (33)
20o2 = 102
Letting-

1 = (1121 "t'- 032 "t- (102 __ 02)2 q_ 4(102 _ 102B)02 _--"i- .

C ABM As (39a)
1
-- 1022
CABMAp

Eq. (33) becomes

A. 2 , 2 "1.1/2
2102 = ((02 B -1- 10.021) -I.- (0,)2 B -t-- 1022 ) __+{[((1)2 B -_- 1021) - (102pB -{- 10322)] 2 nt- -vw31w32J (34)

162 J. Audio Eng. Soc., Vol. 48, No. 3, 2000 March

PAPERS LOUDSPEAKER
WITHPASSIVE
RADIATOR

o}_ = (o2 + (o2 where

((a2 (02)2
+ 4((02 2 B HE _ _
-- -- - (OSB)(O_ _ f2
(39b)
with
Adding and subtracting Eqs. (39a) and (39b) gives
X2 = f2.f2
f2 + f2 = f2 + f2 (40a)
/,.,
= + -
- = -f_B)f_ • Having solved for B, Eq. (37) gives
(40b)
fPB -- fB (43)
Solving Eq. (40a) for f2 and substituting into Eq. (B Jr 1) 1/2"
(40b) gives
If required, the value of ot is found from Eq. (36). The
(f2 _ f2)2 = [2f2 _ (f2 + f2)]2 compliances of the driver and the passive radiator V^s
and VAp, respectively, are given by Eqs. (1), where
+ 4(f2 2 2 _ VAB : VB-
-- fsa)(frl + f2 _ f2) _ + 1 " (41) As stated in Section 2.3.3, in the absence of internal
enclosure losses it is reasonable to assume thatfs B -------fsA
After some algebraic manipulation Eq. (41) can be and fPB -------
fPA" However, if desired, fsB can be found
written fromfs A using Eqs. (17a) and (18a). Likewise, f_ can
be found fromfp a using Eqs. (17b) and (18b). The values
8 fe-H2
-He - f_B (42) of fSA, rE, described
technique rE, and f. incan
[7]. be found by employing the

THE AUTHOR

Douglas H. Hurlburt obtained his undergraduate de- band spread-spectrum communications systems and air-
gree frornJohns Hopkins University, Baltimore, MD, borne radar systems. From 1988 to 1990 he served as
and his graduate degrees from McGill University, Mon- a Program Manager at the Defense Advanced Projects
treal, P.Q., which awarded him a Ph.D. degree in elec- Agency (DARPA). After leaving DARPA he joined De-
trical engineeringin 1972. cision-ScienceApplications, Inc. (now SM&A, Inc.),
His professional career began at RCA Ltd. (Mon- where he continues to provide consulting services, radar
treal), where he worked on the development of micro- system analyses, and program management assistance
wave integrated circuits and later on the design and for the Department of Defense and other agencies.
development of surface acoustic wave (SAW) devices Audio has been a life-long hobby and Dr. Hurlburt
and systems. In 1976 he moved to MIT Lincoln Labora- enjoys designing and building many of his system com-
tory, where he continued to work on advanced SAW ponents. He performs occasional audio system con-
devices until 1978. At that time he became involved in suiting work, specializing in the design of custom elec-
the analysis, design, and development of ultra-wide- tronics, from his home.

J.AudioEng.Soc.,
Vol.48,No.3, 2000March 163