Vented-Box Loudspeaker Systems

Part IV: Appendices

RICHARD H. SMALL

School of Electrical Engineering, The University of Sydney,

Sydney, N.S.W. 2006, Australia

The appendices present a method of calculating the system parameters required to

obtain a desired alignment defined by transfer-function polynomial coefficients in the
presence of enclosure losses together with diaphragm displacement data for that align-
ment, a derivation of the parameter-impedance relationships that permit parameter
evaluation from voice-coil impedance measurements, and a method of evaluating the
amounts of absorption, leakage, and vent losses present in a vented-box loudspeaker
system.

Editor's Note: Part I of Vented-Box Loudspeaker Systems s4To4

. (57)
appeared in the June issue, Part II in July/August, and G_(s) -- S4To 4 "}- axsaToa + ass2To 2 _}_aasTo + 1
Part III in September.
Study of the magnitude-versus-frequency behavior of
filter functions is facilitated by the use of the magnitude-
squared form
APPENDIX 1
FOURTH-ORDER FILTER FUNCTIONS AND {Gu(/_o)[2 =
VENTED-BOX SYSTEM ALIGNMENT _STo8
(58)
General Expressions _oSTo8 + Alo_eToo + Aeto4To4 q- Aa_o2T02q- 1

The general form of a prototype low-pass fourth- where

order filter function G_(s) normalized to unity in the

passband is A_
A s = aa__ + 2 -- 2alaa
-- 2a2
1 Aa = aa2- 2a 2. (59)
GL(S ) = (55)
1 q- axsTo q- asseToS + aasaT0 a + s4To4
Using Eq. (58) it can be shown that the magnitude
where To is the nominal filter time constant and the response of Gu is down 3 dB, i.e., [aH[ 2 = 1/,_, at a
coefficients ax, as, and aa determine the actual filter frequency fa given by
.characteristic.
Tables of filter functions normally give only the de- fa/fo = d_ (60)
tails of a low-pass prototype function; the high-pass and
bandpass equivalents are obtained by suitable transforma- where
tion. For the high-pass filter function Gn(s), the trans- f0 = 1/(2_rT0) (61)
formation (retaining the same nominal time constant) is
and d is the largest positive real root of the equation
G_(sTo ) = GL(1/sTo). (56)

This leads to the general high-pass form of Eq. (20): d 4 -- A1da -- A2ds -- Aad -- 1 = 0. (62)

_OCTOBER
1973,VOLUME21, NUMBER8 635
RICHARD H. SMALl

Coefficients of Some Useful Responses Because the direct relationships between B and the a
coefficients are very involved, the range of responses is
Butterworth Maximally Flat Amplitude computed by taking successive values of a2 and then
Response (B4) computing a l, aa, ,43, and B.
This well-known response is characterized by [10], [18]

as = (4+ 2 2---)'A = 2.6131 Other Possible Responses

a2 = 2 +_1 2 = 3.1412 Other fourth-order responses which can be obtained
a a = a 1 = 2.6131 with the vented-box system include transitional Butter-
A1 = A2 = A a = 0 worth-Thompson [18], transitional Butterworth-Cheby-
shev [30], Thiele interorder [31], and degenerated
fa/fo = 1.0000
Chebyshev [11].
The degenerated Chebyshev responses of the second
Bessel Maximally Flat Delay Response (BL4) kind (DT2) described by Nomura [11] look particularly
The normalized roots are given in [19]. They yield appealing for cases where a smooth bass lift (similar to
an underdamped second-order response, but with a
a1 = 3.20108 As = 1.4638 steeper cutoff slope) is desired. Nomura's design param-
a2 = 4.39155 A 2 = 1.2857 eters are readily convertible into those of this paper.
aa = 3.12394 A s = 0.9759.
/a/fo = 1.5143
Computationof Basic Alignment Data
Chebyshev Equal-Ripple (C4) and The basic alignment data are obtained by using the
"$ub-Chebyshev" (SC4) Responses coefficient-parameter relationships given by Eqs. (21)-
(24). The steps are as follows.
These responses are both described in [14]; the C4
responses are further described in [32]. The pole loca- 1) For a given response and value of QL calculate
tions may be derived from those of the Butterworth
response by multiplying the real part of the Butterworth cs = asQ_
pole by a {actor k which is less than unity for the C4 c2 = agQL. (67)
responses and greater than unity for the SC4 responses. 2) Find the positive real root r of

The filter-function coefficients are then given by r4 --Cl )'a --[-c2r- 1 = 0. (68)

k(4+2 2)% 3) Then, using Eqs. 60-62 to obtain fa/fo, the align-
a a
D% ment parameters are

1 +kC(1 q-_/ 2) h = r2
i:12
D_A fa/fa = h% (fa/fo)

as I-' 1-- k2 _ a = a2h- h 2-1- (1/Q_ 2) (aah%QL-1)

as D%
_ L 1 2_-_ J (63) Q_, _ hQL/(aah%Q L -- 1). (69)
For infinite QL the above expressions reduce to Thiele's
where formulas:
k4 + 6k 2 + 1
D -- h = aa/as
8 fa/fa = h_/_(fa/fo)
For the C4 responses, the pasaband ripple is given by a = aah -- ha -- 1

dB ripple = Q_ = 1/(alas) %. (70)

10 logs0 [ 1 + K4/(64 + 28K + 80K 2 + 16K a) ] (64)

where Computation of Displacement Maxima

K =llk 2 -- 1. Eq. (14) may be written in the generalized form

bss2To 2 + b2sT o + 1

Quasi-Third-Order Butterworth Responses (QB3) X(s) = cato 4 -4-assaTo a + a2s2To2 + aasro + 1 (71)
This class of response is described in [10] and [32].
In this paper, the response is varied as a function of the where To, as, a2, and aa are given by Eqs. (21)-(24) or
parameter B given by by the alignment specification and

B = Aa%. (65) bs = 1/h

b2 = 1/(h'AQ_). (72).
The other coefficients are given by
The magnitude-squar ed form of this expression 'is
As = A2 = 0
....... Ix(i_,)l_=
at = (2a2) % (73)
a a = (a22+2)/(2aO. (66) coSToS+AlcoOToO+A2o_4To4+Aaco2To2+ 1

636 JOURNALOF:THE"AUDiO;E_IGjNE_EItING
SOclET'_'
 VENTED-BOX LOUDSPEAKER SYSTEMS PART IV: APPENDICES

Zvc(ko) =
-- rH

'_'_ _--_[,,, _4TB2Ts2+1

> + _0s [(_ + 1) TBs + TsS]
N ca IX: '4- ](coTs/Q]is ) (1 -- wsr_ 2)

This expression has minimum magnitude and zero phase

when the numerator of the second term is zero, i.e.,
_-_[ RE4-RBs ](cTs/Q_s)(1-2TB2) (78)
RE+RBM rM when co = 1/Ts. Thus for this case, the frequency f3x of
RE 1 Fig. 20 is equal to lB. The expression .also has zero phase,
with maximum magnitude, when the real part of the
'" fl fM fH denominator of the second term is zero, i.e., for

Fig. 20. Voice-coil impedance magnitude of vented-box eo2 _

loudspeaker system as a function of frequency.
Ts24- (a+l)TBS-+' v Ts4+ (a+l)erB4+ (2_--2)TB2Ts 2

where the A_ coefficients are given by Eq. (59) and 2TBSTs2

(79)
B1 = bx2
B2 = b22 --2bi. (74) Let the solution using the plus sign be cot_2 and the
solution using the minus sign be e%s. Then
The value of JX(]o0[m._.? for any alignment is found
by differentiating Eq. (73), setting the result equal to eous 4- col2 ----,_Bs 4- (a 4- 1) e0s2 (80)
zero, solving for the value of _ST0S, and then replacing
this solution in Eq. (73) and evaluating the expression, and
There are always at least three frequencies of zero slope (eoH2- c0L2)s = coB4+ (a + 1)scos4 4- (2a- 2)coBeros2.
for Eq. (73): zero, near lB, and above lB. For the ex- (81)
treme C4 alignments, there is a fourth frequency, below
lB. The first of these frequencies gives unity displace- Combining Eqs. (80) and (81), it can be shown that

ment; the second is not of interest because it gives a (co_2- eoL2)2 = (eous 4- toL2)2--4e0Betos 2 (82)
displacement minimum. The third frequency gives the
displacement needed to evaluate the displacement-limited which simplifies to
power capacity for bandwidth-limited drive conditions.
The procedure is as follows. WH2'}L2 = eS'2WB2

1) For a given alignment and value of QL, calculate or [10, eq. (105)]

C4 ----(l/2Bl) (A1Bl-4' 3B2) /s -- /_fL (83)

C 3 = (l/B1) (A1B 2 + 2) f/_
C2 = (1/2Bi)(3Ax + AsBe--AaBx) where ]s = fsB is the resonance frequency of the driver
C1 = (1/BO(As--BO for the particular air-load mass presented by the en-
Co = (1/2B1) (Aa -- Be). (75) closure.
2) Find the largest positive real root G of With ]s known, a can be found by rearranging Eq.
(80) into
G5 4- C4G 4 4- C3Ga -1t- C2G2 4- CiG 4- C0 = O. (76)

(The normalized frequency of maximum passband dis- a = -- 1. (84)

placement is then /x m_x//0 = G %). /sS

3) Calculate . Alternatively,substitutingEq. (83) into Eq. (80), it is

easily shown that [10, eq. (106)]
Bl.G 2 4- B2G 4- 1
JX(jw) Jmj = G4 + AsGa + AsG2 + A_G + 1 ' (77) a = (fus -- ]Bs) (ia -- fL2) (85)
/nVLs
The same procedure is used to determine the fre-
quency of maximum displacement below /B for the ex- This expression factors into
treme C4 alignments by finding the smallest nonzero (/_ + J.) (fu --/B) ([B + fL) (lB -- fL)
positive real root in 2). The corresponding maximum a = (45)
value of the displacement function magnitude is then /u2fLs
determined as in 3). Approximate Determination of QB
From Fig. 3, Zv_ will be resistive when the portion
APPENDIX 2 of the circuit to the right of Rs s is resistive. The steady-
PARAMETER-IMPEDANCE RELATIONSHIPS state impedance of this portion of the circuit is

Determination of fs_ and oc Z(ioO = RgL (ar_QL) [-- tosrn/Q_ 4-/offl -- ,,_T_s) ]
For infinite Q_, the steady-state form of Eq. (16) be- o_4TB2Tss 4- 1 -- _o2 [ (a 4- 1 ) T_s 4- Ts_ ]
comes 4- ]ro(T_/Q_) (1 -- o_2rs2) (86)

OCTOBER1973,VOLUME21, NUMBER8 637

RICHARD H. SMALL

At a frequency of zero phase, the magnitude of Z(j_o) by a combination of the three fixed resistances RxB,
may be evaluated by taking the ratio of either the real RAL, and RAe of Fig. 1.
or the imaginary parts of the numerator and denominator,
because these ratios must be equal. That is, for zero System Loss Data
phase, From the systemimpedance curve, Fig. 20, find the
[Z(j_,)[ = three frequencies /L, /_f, and Jrt, and the ratio of the
--coeTB/QL corresponding maximum or minimum impedance to Re,
Rrr(aTeQL)
oj4T_2Ts 2 + 1 -- co2[(a + 1) Te 2 + Is 2] designated rL, r_, and rn.
1- _02TB2 Using the methods of Section 7 (Part II) or [32], de-
= Rr_(aTeQL) (87) termine the system compliance ratio a. Measure inde-
(TB/QL)(1--o_2Ts 2) pendently the driver resonance frequency /s _nd the
Setting the real and imaginary ratios equal in the corresponding value of Qes as described in [12] or [32].
normal way leads to a very complex set of solutions for The driver mounting conditions for the latter measure-
the exact frequencies of zero phase. However, it can be ments do not matter, because the product I'sQzs which
seen that the first ratio varies relatively slowly with fre- will be used is independent of the air-load mass present.
quency near coB (as indeed does [Zvc(j6o)]) and hence
can be expected to have about the same magnitude at Driver Loss Data
the frequency of zero phase o_ very near to _oBas it has Let the symbol p be used to define the ratio
at c0_. This gives
p = (RBa-F Rz)/RB. (90)
lz(io,_)]_ Iz(i_B)I= R_. (88)
Because Res is in fact a function of frequency for real
The resistive voice-coil impedance measured at /u, drivers, so too is p. Typically the variation is of the order
defined as RB + RB_ in Fig. 20, is thus made up of R_ of 2 to 4 dB per octave increase with increasing
plus the parallel combination of REs and RtL. Evaluat- frequency.
lng this resistance and using Eqs. (5), (7), (8), (10), At the resonance frequency of the driver, p is the
and (11), it can be shown that ratio of the maximum voice-coil impedance to Rr whick

QL = -- (49) down to /L may be measured by weighting (mass load-

h [ 1 1 ] is defined as ro in [12].. The value of p for frequencies
s Qzs(r_-1) Q_ls ing) the driver diaphragm and measuring the maximum
where re is (RF. + RBn)/Rz as defined in Eq. (48) and voice-coil impedance at resonance for a number of pro-
Fig. 20. In many cases the 1/QMs term can safely be gressively lower frequencies as more and more mass is
neglected, added. A convenient nondestructive method of weighting
Now, if the two ratios in Eq. (87) are equal at _o_, is to .stick modeling clay or plasticene to the diaphragm
the second must give the same value as the first. This near the voice coil.
requires that Unfortunately, there is no comparable simple way to
reduce mass or add stiffness which will raise the driver
1 -- aQr, 2
_u e'= (89) resonance frequency without affecting losses. For sim-
TS2--aTB2Qi? plicity, it is necessary to extrapolate the low-frequency
data upward to /u. This is risky if /_ is more than an
which may be rearranged to give Eq. (50). The approxi-
octave above ]s but gives quite reasonable results for
mation made earlier in Eq. (88) seems justified by Eq. many drivers.
(50) for QB values as low as 5, because the difference
Under laboratory conditions, it is possible to fabricate
between ]_ and /_ is then at most a few percent. For
a low-mass driver which is "normally" operated with a
lower values of Qe (which are unusual), substantial in- fixed Value of added mass. This mass is selected so that
accuracy must be expected. Inaccuracy can also be con-
the unloaded driver resonance occurs at a frequency
tributed by a significant voice-coil inductance (see [32]). equal to or greater than the value of/u for the loaded
driver in a particular enclosure. In this case the value
APPENDIX 3 of p can be accurately determined for the entire re-
MEASUREMENT OF ENCLOSURE LOSSES quired frequency range by adding and removing mass.
Measurement Principle Measure and plot (extrapolating if necessary) the
value of p over the frequency range /_ to /,. Find the
In _his method of measurement the system driver is values at /L, /_, and /H and designate these pL, P_u,
used as a coupling transducer between the enclosure and p,.
impedances and the electrical measuring equipment. The These measurements should be carried out at the same
driver losses are subtracted from the total measured time and under the same conditions as those for the
losses to obtain the enclosure losses. Greatest accuracy system loss data above. The signal level should be the
is therefore obtained where the. driver mechanical losses same and should be within small-signal limits at all times.
are small and stable.
The method assumes that R_ remains constant with Enclosure Loss Calculation
frequency (i.e., voice-coil inductance losses are negli-
gible), that the individual enclosure circuit losses cdr- Define:
respond to Q values of about 5 or more (so that Qa H =/H/f_
> > 1), and that any Variation with frequency of the L = )'3t/fL
actual losses present can still be represented effectively F =/.v/(a/sQ_s). (91)

638' JOURNALOF THE AUDIO ENGINEERINGSOCIETY

 VENTED-BOX LOUDSPEAKER SYSTEMS PART IV: APPENDICES

Calculate: [8] L. Keibs, "The Physical Conditions for Optimum

1 1 Bass Reflex Cabinets," J. Audio Eng. Soc., vol. 8, p. 258
k_---- (Oct. 1960).
rL -- 1 pL -- 1 [9] E. de Boer, "Synthesis of Bass-Reflex Loudspeaker
1 1 Enclosures," Acustica, vol. 11, p. 1 (1961).
k_t-------- [10] A. N. Thiele, "Loudspeakers in Vented Boxes,"
ru --1 P_r- 1 Proc. IREE (Australia), vol. 22, p. 487 (Aug. 1961); re-
1 1 published in J. Audio Eng. Soc., vol. 19, p. 382 (May
ku -- -- (92) 1971), and p. 471 (June 1971).
r_ -- 1 pn -- 1 [11] Y. Nomura, "An Analysis of Design Conditions of
a Bass-Reflex Loudspeaker Enclosure for Fiat Response,"
Electron. Commun. Japan, vol. 52-A, no. 10, p. 1 (1969).

CL = FkL(L 2- 1) ( 1- L2
1 ) tem[12]
Analysis," IEEE "Direct-Radiator
R. H. Small, Trans. Audio Electroacoust., vol.
Loudspeaker Sys-
AU-19, p. 269 (Dec. 1971); republished in J. Audio Eng.
C u = (Fk_)-l Soc., vol. 20, p. 383 (June 1972).
[13] D. E. L. Shorter, "Loudspeaker Cabinet Design,"

Cx = FkH(H 2- 1) ( 1-- H2
1 ) (93) 1950).
Wireless World, vol. 56, p. 382 (Nov. 1950), p. 436 (Dec.
[14] A. N. Thiele, "Filters with Variable Cut-off Fre-
quencies,'' Proc. IREE (Australia), vol. 26, p. 284 (Sept.

A= ( HAL2-- H2L2 ,)( -- H2--L '_ -- L2--'_ ' [15] J. R. Ashley and M. D. Swan, "Improved Mea-

NL = Cat H2L2- 1 -- Cn -- CL H_-- at the 40th Convention of the Audio Engineering Society,
( H2L2 ) (21)(____2)
, L --_ surement
Los Angelesof (Apr.
Loudspeaker Driver Parameters,"
1971), Preprint 803. presented

Nx = --C_t ( L2- -_-_) 4-Cn(L 2- 1)+CL (1 ---_-1) World (Letter), Reith,"Bass-ReflexEnclosures,"Wireless

[16]B.C. vol. 73, p. 38 (Jan. 1967).
[17] J. F. Novak, "Designing a Ducted-Port Bass-Re-

Nr = --C_ ( H2 -- _'_
1 ) + Cx ( 1 )_L__
1 + C L (H 2 -- 1). (94) (Jan.1966).
flex Enclosure," Electron. World, vol. 75, p. 25
[18] L. Weinberg, Network Analysis and Synthesis
(McGraw-Hill, New York, 1962), ch. 11.
Then the values of QL, QA, and Qp which apply at the [19] R. M. Golden and J. F. Kaiser, "Root and Delay
frequency /2u are found from Parameters for Normalized Bessel and Butterworth Low-
Pass Transfer Functions," IEEE Trans. Audio Electro-
QL = A/NL acoust., vol. AU-19, p. 64 (Mar. 1971).
Q_ = A/NA [20] A. N. Thiele, "Equalisers for Loudspeakers," pre-
Qp = A/N_,. (95) sented at the 12th National Convention of the IREE
(Australia), (May 1969).
Using the same data, the total enclosure loss QB at the [21] P. W. Klipsch, "Modulation Distortion in Loud-
speakers," J. Audio Eng. Soc., vol. 17, p. 194 (Apr.
frequency f._r is 1969), and vol. 18, p. 29 (Feb. 1970).
[22] R. H. Small, "Closed-Box Loudspeaker Systems,"
QB(/_) = 1/C_ = Fk_t. (96) J. Audio Eng. Soc., vol. 20, p. 798 (Dec. 1972), and vol.
The approximate formula for Qu = QL given in Eq. 21, p. 11 (Jan./Feb. 1973).
[23] H. F. Olson, J. Preston, and E. G. May, "Recent
(49) differs from Eq. (96) only in that Rrs is assumed Developments in Direct-Radiator High-Fidelity Loud-
constant, i.e., that p_ = r0. However, because pu is sel- speakers," J. Audio Eng. Soc., vol. 2, p. 219 (Oct. 1954).
dom very different from r0, and particularly because [24] E. de Boer, "Theory of Motional Feedback," IRE
r:e -- 1 is usually much less than pe -- 1, Eq. (49) pro- Trans. Audio, vol. AU-9, p. 15 (Jan./Feb. 1961).
rides an adequately accurate measurement of total losses [25] H. W. Holdaway, "Design of Velocity-Feedback
Transdu'cer Systems for Stable Low-Frequency Behavior,"
for normal evaluation purposes. IEEE Trans. Audio, vol. AU-il, p. 155 (Sept./Oct.
1963).
REFERENCES [26] R. H. Small, "Simplified Loudspeaker Measure-
ments at Low Frequencies," Proc. IREE (Australia), vol.
[1] A. L. Thuras, "Sound Translating Device," U.S. 32, p. 299 (Aug. 1971); republished in J. Audio Eng.
Patent 1,869,178, application Aug. 15, 1930; patented Soc., vol. 20, p. 28 (Jan./Feb. 1972).
July 26, 1932. [27] J. L. Grauer, "Acoustic Resistance Damping for
[2] B. N. Locanthi, "Application of Electric Circuit Loudspeakers," Audio, vol. 49, p. 22 (Mar. 1965).
Analogies to Loudspeaker Design Problems," IRE Trans. [28] W. Steiger, "Transistor Power Amplifiers with
Audio, vol. PGA-6, p. 15 (Mar. 1952); republished in Negative Output Impedance," IRE Trans. Audio, vol. AU-
J. Audio Eng. Soc., vol. 19, p. 778 (Oct. 1971). 8, p. 195 (Nov./Dec. 1960).
[3] L. L. Beranek, Acoustics (McGraw-Hill, New York, [29] R. F. Allison and R. Berkovitz, "The Sound Field
1954). in Home Listening Rooms," J. Audio Eng. Soc., vol. 20,
[4[ F. J. van Leeuwen, "De Basreflexstraler in de p. 459 (July/Aug. 1972).
Akoestiek," Tijdschrift Nederlands Radiogenootschap, [30] A. Budak and P. Aronhime, "Transitional Butter-
vol. 21, p. 195 (Sept. 1956). worth-Chebyshev Filters," IEEE Trans. Circuit Theory
[5] E. de Boer, "Acoustic Interaction in Vented Loud- (Correspondence), vol. CT-18, p. 413 (May 1971).
speaker Enclosures," J. Acoust. Soc. Amer. (Letter), vol. [31] A. N. Thiele, "Response Shapes for Simplified
31, p. 246 (Feb. 1959). Active Filters," Proc. IREE (Australia), to be published.
[6] R. H. Lyon, "On the Low-Frequency Radiation [32] I. E. Benson, "Theory and Design of Loudspeaker
Load of a Bass-Reflex Speaker," J. Acoust. Soc. Amer. Enclosures, Part 3--Introduction to Synthesis of Vented
(Letter), vol. 29, p. 654 (May 1957). Systems," A.W.A. Tech. Rev., vol. 14, p. 369 (Nov.
[7] J. F. Novak, "Performance of Enclosures for Low- 1972).
Resonance High-Compliance Loudspeakers," IRE Trans. [33] D. B. Keele, "Sensitivity of Thiele's Vented Loud-
Audio, vol. AU-7, p. 5 (Jan./Feb. 1959); also J. Audio speaker Enclosure Alignments to Parameter Variations,"
Eng. Soc., vol. 7, p. 29 (Jan. 1959). J. Audio Eng. Soc., vol. 21, p. 246 (May 1973).

OCTOBER1973, VOLUME21, NUMBER8 639'