Professional Documents
Culture Documents
A. N. THIELE
Editor's Note: The theory of vented-box or bass-reflex valuable calculations. Those working in the design of
loudspeaker baffles has always seemed to have an air of loudspeakers have used these analysis techniques and
mystery, probably because the total electroacoustic sys- probably asked essentially the same seven questions that
tem has four degrees of freedom and seems four times A.N. Thiele recognized at the turn of the previous
as complicated as the closed-box baffle with its two de- decade.
grees of freedom. Beranek gives a good foundation for The seven questions and their answers were published
theoretical analysis and Novak has performed numerous in the August 1961 issue of the Proceedings of the IRE
Australia, and the elegance of the answers adequately
* Presented at the 1961 I.R.E. Radio and Electronic Engi- justifies republication of Thiele's work in the Journal o!
neering Convention, Sydney, N.S.W., March 1961. Reprinted the Audio Engineering Society. In his classic discourse
from Proceedings of the IRE Australia, vol. 22, pp. 487-508
(Aug. 1961). The author was formerly with E.M.I. (Aust.) Thiele observes that the topology of the equivalent cir-
Ltd., Sydney, N.S.W. cuit (Fig. 1) is simply that of a high-pass filter. If suffi-
ELEC. MECHANICALDIAPHRAGM
BOX VENT VENT The advantage of using this large complete equivalent
PART OF PART OF
SPEAKERLOUDSPEAKER
RADIATION RADIATION circuit in the first place is that the equivalent circuit of
ss _d cos_._uc
_lex-xJ1 the loudspeaker in a totally enclosed box may be shown
(Rg_ ao-__
Ro_ by
sent removing the operated
mesh representing the vent.
baffle, ToCaw
repre-
the speaker in an infinite and
] t_s[ . _ .... .I; Rah are short-circuited. If the speaker is operated in open
[_9_)oSd O O U_ O-- _ air (unbaffled), the circuit is as in an in_nite baffle, but
the values of Rat 1 and M_x are modified [see 4, Fig. 5.2].
Fig. 1. Complete (electromechanical) acoustical circuit of The details of these circuits are very well covered in [3]
loudspeaker in vented box (after Beranek [3]).
from which Fig. 1 and the accompanying symbols are
taken.
To make the circuit more manageable, we simplify it
circuit, but first consider Fig. 1 in which the complete to Fig. 2.
equivalent circuit of the loudspeaker and enclosure is
given in acoustical terms.
We note that there are three possible equivalent cir- _'Z
cults, electrical, mechanical, and acoustical. To convert (%*R_)Sa Mos Co_ aos
l length of wire in air gap. 1) The three acoustic masses Mae, M_l, and Maw are
Again to convert from mechanical to acoustic units, lumped together to make a single mass M,_. However,
we must be careful to remember that this is an artifice.
Z_ = Zm/Sa 2 (2) Mas is not fixed, and some error results by assuming it
where to be so. For example, the reduction of M_b and hence
Za acoustical impedance of Ma8 when the speaker is tested in open air causes a
Sa equivalent piston area of diaphragm (usually rise in resonant frequency, which must be accounted for
taken as area inside first corrugation), in measurements, as in Section XIV.
2) R_ 1 and Rar2 are neglected in the equivalent circuit,
Taking then in Fig. 1 the first impedance after the gen- even though they are responsible for the acoustic output
erator which is the acoustical equivalent of the electrical of the loudspeaker. The whole essence of Novak's theo-
resistance of the amplifier output impedance Rg in series retical model which makes a simple solution possible
with the voice coil resistance Re, we can see that the is that a loudspeaker is a most inefficient device. In mca-
various equivalents for this impedance are surements of fifty loudspeakers using the method of Sec-
tion XIV covering a wide range of sizes and qualities,
Ze = Rg+Re (3) efficiencies ranged between 0.4% and 4%. For this reason,
Z_,, = B_°le/(Rg+Re) (4) the radiation resistances may be safely neglected. Since
Z a = B212/Sae(Rg+Re). (5) radiation resistance varies with frequency squared, this
In Fig. 1, simplifiesanalysisconsiderably.For, as pointed out in [3,
Ea open-circuit voltage of audio amplifier p. 216], the radiation resistance of a loudspeaker in a
MaR (= Mrna/Sa2) acoustic mass of diaphragm and "medium-sized box (less than 8 fta) '' is approximately
voice coil the radiation impedance for a piston in the end of a long
Mrna mechanical mass as usually measured tube. And the radiation resistance of the vent (or port)
Cas acoustic compliance of suspension is the same. Thus
--O
Fig. 3. Simplified mechanical circuit of loudspeaker in Fig. 4. Simplified electrical circuit of loudspeaker in vented
ventedbox. box.
er plus amplifier for a vented box will usually lie between shape of the resulting electrical impedance curve of
0.3 and 0.5. The Q of the vent, on the other hand, can be Fig. 5 above /n' However, this will be of greater impor-
found by combining [3, Eqs. (5.54) and (5.55)] to give tance when we come to testing procedures in Section XIV.
/_ kinematic coefficient of vkscosity; for air at Now we have seen already that the radiation resistances
NTP,/_ = 1.56 × 10 -5 me/s. of speaker and vent must always be the same. And since
Thus if Sv = 4 in 2, the bottom limit specified by No- the radiated sound depends on the sum of the volume
yak, and / = 25 Hz, then Qv = 64. velocities Uc and Up (or rather their difference, since U r
Since these are the smallest values of S_ and / likely to derives from the back pressure of the speaker), then the
be found in practice, it is clear that little error will result acoustic power output is
from this source, and this is confirmed in Section XI. In Wao ---_]Uc--UpI2Rarl (9)
the preceding discussion, the effect of M_2 and Rat 2 has
been neglected, but in no case investigated has the total while the nominal electrical input power is
ly between 3 and 10, so that this does not affect matters _ W,o/Wei
greatly, but since Ras can be lumped with the equivalent _ --
electrical resistance (see Eq. (8)) and because it has some = [IUc--U_IeRarl(Rg+Rc)e]/(E_2RO. (11)
importance in the loudspeaker measurements of Section Analyzing the circuit, we find that
by the conversion factor Sa2 as in Eq. (2). Thus these p4MasMavCasCa,-}-paMavCasCabRat '
impedances
from Fig. 2 represent the mechanical
by multiplying impedances
all the acoustical at the
impedances [{ '{-p2(MasCasq-MavCasq-MavCab
p4MasMaveasCab ) -[-PCasRat-t-1 t]
loudspeaker diaphragm due to the whole acoustical- (12)
mechanical circuit. Since the conversion is obtained by To make the expression easier to manage we write
multiplying by a constant, the form of the circuit remains E(p) for the expression inside the square bracket on the
the same. However, when the conversion is made from
right-hand side which is a fourth-order high-pass filtering
Fig. 3 to Fig. 4, the electrical equivalent circuit, it can function. Also if ]to is written for p, the steady-state re-
be seen from Eq. (1) that an impedance inversion sponse E(jto) is found. We also convert PMas from the
takes place. Thus all series elements become parallel operational form to the steady-state form jtoMas, and then
elements, inductances become capacitances, and vice substitute
versa. Thus Lees is the electrical inductance due to the
compliance of the loudspeaker suspension, Cme_ is the M,,,_ = MasSa 2. (13)
electrical capacitance due to the mass of the loudspeaker This puts the expression for mass into a more intelligi-
cone, C,.e. is the electrical capacitance due to the mass
of the vent, and Leeb is the electrical inductance due to
the compliance of the box. In Fig. 4 an additional pair of
elements contribute very small effects at the low fre- Fig. 5. Typical impedance curve of loudspeaker in vented
quencies we are considering, but show the reason for the box.
bio form, but it is emphasized that the total loudspeaker nominal cutoff frequency) and three coefficients x_, x2, xa
mechanical mass Mms includes not only the mass of the which determine the shape of the response curve. In fact,
cone plus voice coil, but also the mechanical equivalent the general expression is often written with a constant
of the acoustic air load. The latter is only a small part x0 and x4 instead of the two unity coefficients in the
of the total, but varies with the speaker's environment, denominator of Eq. (21); but the expression can always
e.g., box volume [3]. Thus if we substitute Eqs. (6), (12), be reduced to the form of Eq. (21) by division of the
and (13) in Eq. (11), whole expressionby a constant, and suitable adjustment
of To and the x coefficients. Considering Eq. (20) now
v = poBSleSaS[E(]
°J) ]2/4_rcReMms
2 (14) from the viewpoint of what can be done with a given
or speaker, the parameters Cas and Ts are fixed. Thus there
= (po/4rrc)(B212Sa2/ReMms2) [E(jro)[2. (15) are three variables Qt, Th, and Cab, and it is possible to
achieve any desired shape of curve (i.e., any desired cum-
Thus the expression for efficiency contains three parts:
bination of the three x coefficients); but Ln doing so To is
1) a constant part containing physical constants,
2) a constant part containing speaker parameters, determined (see Eq. (27)).
For identity between the two Eqs. (20) and (21), the
3) a part IE(]co)[2 which varies with frequency, coefficients of the various powers of p must be identical,
that is,
maximally flat (Butterworth)l characteristic for which have response peaks (at 1.76% in this case) of 6 dB and
12 dB, respectively, if fed from a zero output impedance
[E(]c0)] = 1/[1 +(%/co)s]_ (30) amplifier, 12 dB and 18 dB if fed from an amplifier with
or impedance equal to loudspeaker resistance R e (e.g., pen-
]E(j_o)Ie = 1/[l+(%/to) sI (31) rode with 6-dB negative voltage feedback), and even
and, in the operational form, more with higher amplifier impedances. Hence the ex-
pression "boom box."
E(p) = 1/(1 +2.613/pTo-b3.414/p-'2To 2 An amplifier with negative output impedance half that
+ 2.613/paTe _ + 1/p_To4). (32) of the loudspeaker resistance Re, a quite feasible figure,
Note that in Eq. (31) and others which will follow, the would correct the medium quality speaker, and reduce
ratio of any two frequencies, say _oa/o_b, is identical to the peak on the cheaper one to 6 dB. An amplifier with
/_//,. Note also that all Butterworth responses are 3 dB a negative output impedance three quarters of R e, to
down when co = COo,i.e., cote = 1. correct the cheaper speaker, is possible but would need
A characteristic of Butterworth responses, though not care in respect of stability (see Section XII).
peculiar to them, which simplifies calculations even fur-
ther is that in all cases Fifth-Order Butterworth Response
x 1 = x a. (33) This has the characteristic
Thus in the fourth-order case where which is the characteristic of two filters in cascade: 1) a
first-order filter which can be provided by a CR network
xt = xa = 2.613 (38) with a time constant To, and 2) a fourth-order filter pro-
x2 ---- 3.414 (39) vided by a loudspeaker and box for which
we have
Qt = 0.383 (40) TO= T8 = Tb (46)
Ca_/Cao ---- 1.414. (41) Qt = 0.447 (47)
E
~ 8 I i
(34 ] -- I 0..) .64l t
] .841 I .aa9 .518 -- -- -- 1.36 .3!12
[7C-o
! 1_, i.o --1.00o 1ooo-17_o0o-
.--7_4,k00- ----_ - - -
,z
12 c, .4 0.2:,I .724 .ss,) .273 .810 3.._0 .....
_:, 13 Cs .355 0.5 i .704 .882 .227 .924 2.93 .....
'1_--5 1_7-- 6 -- 1.0 ..... _00-- 1_--[ 2._-3 .z,),) 1--7_0 _.'F-72 _0 1.07 -- --
19 C6 .414 0,1 .554 .659 1.25 .399 1.04 i --1.9.51 -_-13.2 1.74 -- --
amplitude, (see curve 8 of Fig. 8). Beyond cutoff, the which C_8and/s are constant is placed in different boxes
response falls at a rate whose maximum is greater than to provide different cutoff frequencies, the box volume
the asymptotic slope. Typical values are tabulated in will vary with inverse frequency squared. This illustrates
Table I with the type names C4, C5, and C6 representing a fact long known to designers of vented boxes, but
Chebyshev responses of fourth, fifth, and sixth order. It rather blurred by the exponents of "revolutionary new
will be seen from the table that a considerable change in concepts," that the bigger the box, the better the low-
alignment occurs before the ripples become serious in frequency response. It is also interesting to note that
magnitude. For our purpose here, the Chebyshev re- Casfs2 = 1/4_r2Mas = Sae/4_r2Mm_ 1.41CaJ32 (56)
spouses provide a means of carrying the useful response
of the speaker plus box combination well below the that is, for a given cutoff frequency of the combination,
speaker resonant frequency /s (which is also cutoff fre- the box size varies with the square of diaphragm area
quency /o in the Butterworth cases). This is done by Sa2 and inversely with Mms. In other words, if the mass
tuning the box to below rs, but not as low as the cutoff of the loudspeaker M,, s is fixed and the compliance Cas
frequency (defined here as fa, the frequency where the is varied to give a different resonant frequency L, then
response is 3 dB down). The box size Cab is increased, the box volume C_b for a given cutoff frequency fa re-
and to some extent, so is Qt. mains substantially constant. To this extent, and also ia
The increase in useful low-frequency response is con- the expression for efficiency (Eq. (66)) the compliance
of the loudspeaker is unimportant.
siderable. In alignment no. 9, a response down to 0.6/s is
obtainable without amplifier assistance, if a ripple of 2) Qtfo/L lies around 0.38. If Eq. (18) is rewritten as
1.8 dB can be tolerated. In alignment .no. 25, where a Qt = %Mas/Rat (57)
maximum lift of 6 dB is required from the amplifier be- then the expression above becomes c%M_s/Rat which can
fore its response falls off, a flat response can be obtained be thought of as the total Q of the speaker at the box
down to nearly 0.4rs. resonant frequency. This remains nearly constant through-
out alignments no. 1-9.
Certain alignments, no. 13, 14, and 27 with no. 12 as
- __rs_----
r a borderline case, which require auxiliary filtering with
, _ _EG*_,vE
'X// _----__ large attenuation at the cutoff frequency of the whole
/ As_o,_ // _---- system, must be considered suspect, since they postulate
ber that the basis of the theory is that the overall efficiency
OUTPUT 12 dB PER x _POSITNE
is low. In the borderline case, no. 12 for example, the
°_/ high
peakacoustic efficiencies
efficiency in the
will be just region
above of cutoff.
cutoff Remem-
frequencyand!
,_._ _,_o_
...... will be approximately 2.52 times the loudspeaker effi-
Fig. 6. Typical curves for second-order auxiliary filter, ciency. If the loudspeaker is 4% efficient, this means a
illustrating terms used in Table I. maximum overall efficiency of 25%. Around this point,
the basic assumptions will become inaccurate, especially
if resistive losses in the box are large.
Ouasi-Butterworth Third-Order Responses Similarly, for reasons of cone excursion (considered
This long name disguises a class of responses charac- in Section X), alignments with smaller values of /a/.f_,
terizedby such as nos. 17-19 shouldbe avoidedif possible.These
particular alignments which do give good low-frequency
[E(]oJ)12 = 1/[l+y3(%/co)°+y4(%/co)s ] (55) responses in small box volumes would probably be un-
that is, in the expression for the modulus of the fourth- popular anyway since they make such great demands or_
order filter, there are zero coefficients for the second and amplifier output in the region of cutoff.
fourth powers of frequency, with nonzero coefficients Alignment no. 28 is interesting in that it represents the
result of "pure" bass lift. In the other alignments which
for both the eighth and sixth powers. This type of re-
use "amplifier aiding," the response often rises near cut-
spouse yields a series of alignments, nos. 1-4 of Table I, off, but always falls off ultimately at lower frequencieg
in which the cutoff frequency (again defined here as the
frequency f3 where the response is 3 dB down) is above at a rate of 6 or 12 dB per octave. In this way, althougk
the speaker resonant frequency. So also is the box reso- increased amplifier output may be required over a cum-
nant frequency, but again, not to the same extent. As paratively narrow range of frequencies, a greatly de-
the cutoff frequency is made higher, these alignments creased output, and with it, a greatly decreased cone ex-
require smaller box volumes, and lower values of Qt. cursion, is required at the lower frequencies. But in align-
ment no. 28, a simple low-frequency lift of 6 dB, such ag
VI. GENERAL DISCUSSION OF TABLE I results from a network with two resistors and a capacitor,
is required. The mean frequency of lift (at which the lift
It will be seen that alignments no. 1-9 provide a means is 3 dB) is 1.08f3. However, since the maximum lift con-
of varying the cutoff frequency of a loudspeaker-box tinues to the lowest frequencies, the amplifier would be'
combination over a wide range. The last two columns more likely to cause intermodulation distortion witl_
for these alignments illustrate two interesting properties "rumble" components. However it does give some de-
which remain substantially constant (-----5%) over this crease of box volume compared with alignment no. 5.
widerange. It shouldbe emphasizedthat thesealignmentsare by
1) The expression Casfs2/Cabf3 2 is substantially con- no means the only ones possible. They have been chosen_
stunt around 1.41. This means that if a given speaker for as the ones most likely to be useful and as showing the:
MAY 1971,VOLUME19,NUMBER
5 389'
A_ N. THIELE
,_ '_'_, , the vent (say 10%) must be subtracted from the gross
J._ - _v_ e volume.
the ratio From thisis value
found, and the
thence eithervalue
known
1.2 -'_' 7 Ca_/C_b and from ofFig.V_s,
7
/ or interpolation from Table I, the values of fa//s, fa//b,
trend of results. If more sophisticated filtering in the output from small boxes at the expense of higher electrical
power output from the amplifier, while alignments no.
amplifier is possible, the choice widens greatly, e.g., 20-25 are the most suitable if a fair sized box is available
there are six alignments for the eighth-order Butterworth
and only moderate lift is required from the amplifier,
response, each with its fourth-order amplifier filter and although in all the fifth- and sixth-order cases, the power
the ratios Cas/Cob of 0.518, 0.681, 1.000, 1.316, 1.932, required from the amplifier and the excursion demanded
and 2.543. of the speaker decrease rapidly below cutoff.
Another possibility would be the use, instead of the
Having .found/b and Vb, the vent dimensions may be
"quasi-Butterworth" responses, of "sub-Chebyshev" re- found using the methods of the standard texts [8]. How-
sponses, i.e., response functions derived by multiplying
ever, the following adaptation of the method has proven
the real coordinates of the Butterworth poles by a con- useful for calculation. The standard form is
stant k which is greater than 1.
In answer to the question proposed in 1) of Section Vb = 1.84 X lOSSJcob2L, (60)
I--What is a large box?--it would appear that a medium
sized box would be one for which Vb is about the same where Sv is the cross-sectional area of the vent, in square
value as V_s, say C,o/Cab lies between 1 and 1.414. For inches, and L_ is the effective length of the vent, in inches,
large boxes, Cas/Caw is less than 1, for small boxes Cas/ which includes its actual length together with an end
Cow is greater than 1.414. Table I shows that smaller correction.
boxes demand a smaller value of Qt. Thus if Qt is not
properly controlled, the smaller boxes will tend to cause '+2
a greater peak at J_,, while larger boxes will cause the _ _
peak to diminish. Fig. 7 is plotted from the points of ° /,'-"__'-'--- /
TableI.
First, the
Typicalresponse
following threecurves
loudspeaker
for alignments
parametersno. must
3, 5, -,se/////
_e
be known: 1) the resonant frequency rs, 2) the Q values o.5 0.7, ,.o _i_, 2.0 2.e3 ,o
Qa and Q,, the latter being-usually the controlling fac-
Fig. 8. Typical response curves for identical loudspeakers,
tot. This is discussed in more detail in Section IX, Eqs. but different box sizes. C,,/C_b = 0.56, 1.41, and 4.46, cor-
(71) and (72), and 3) the acoustic compliance Cas. This responding to alignments no. 8, 5, and 3 (types C,, B,, and
is expressed most conveniently as Vas, the volume of air QBe) of Table I.
Thisis writtenmoreconveniently
as
L_/S, = 1.84 X lOS/cofVb. (61)
THE AUTHOR
Neville Thiele was educated at the University of ment of E.M.I. Australia's television receiver. He was
Queensland and the University of Sydney, graduating appointed advance development engineer in 1957.
as Bachelor of Engineering in 1952. He joined the staff Mr. Thiele joined the Australian Broadcasting Corn-
of E.M.I. Australia Ltd. in 1952 as a development en- mission in 1962, and is now senior engineer, design and
gineer in the Special Products Division. During 1955 development, responsible for the Federal Engineering
he spent six months in England, Europe and the United Laboratory, and with design investigation of equip-
States and on return was responsible for the develop- ment and systems for sound and vision broadcasting.