PAPERS

Simulated Free Field Measurements*

CHRISTOPHER J. STRUCK**, AES Member, AND STEVE F. TEMME***, AES Member

Brfiel & Kjcer, DK-2850 Na_rum, Denmark

The development of time-selective techniques has enabled measurements of the free field
response of a loudspeaker to be performed without the need for an anechoic chamber. The
low-frequency resolution of both time-selective techniques and anechoic measurements is,
however, limited by the size of the room. A technique is presented enabling measurements
of the complex free field response of a loudspeaker to be performed, without an anechoic
room, throughout the entire audio frequency range. It is shown that this technique can also
be used for measurements of harmonic distortion.

0 INTRODUCTION HFF(f) = far-field response of loudspeaker

h = length of minor semiaxis of ellipsoid
The purpose of this paper is to present a convenient k = wave number, (= 2-tr/k)
and reliable method for obtaining the complete free field k = wavelength of sound
response of a loudspeaker in an ordinary room, making M = largest linear dimension of source
as few assumptions about the device under test as possi- N = harmonic order
ble. Before attempting to obtain the free field response Po = reference electric powerlevel (e.g., 1 W)
of a loudspeaker, it is useful to know what is meant by Po = reference sound pressure level (e.g., 20
"free field" and "reverberant (or diffuse) field," as well gpa)
as "far field" and "near field." Si> = effective surface area of driver
Se = effective surface area of port
I GLOSSARY OF IMPORTANT SYMBOLS T = time range
At = time resolution
a = radius of driver (or equivalent rigid, flat, ? = time delay
circular piston) % = delay in sound arrival corresponding to
b = length of major semiaxis of ellipsoid reference distance
c = speed of sound, = 344 m/s x0 = dB reference
d = distance from source to measurement mi- Z0 = nominal impedance of loudspeaker
crophone (direct sound path)
do = reference distance (e.g., 1 m) 2 FIELD DEFINITIONS
dR = distance from source to first reflecting sur-
face to microphone (path of firstreflection) A free field describes idealized sound propagation
F = frequency range with no reflections, that is, only the direct sound from
f -- frequency a source can be observed. This occurs naturally in open
fs = transition frequency; near-field/far-field outdoor spaces, high enough above the reflective surface
measurement of the earth. It can also be created artificially by con-
Af = frequency resolution (lower limiting fre- structing an anechoic room. In this case a large room is
quency) built using highly absorptive material on the walls, floor,
H_F(f) = near-field response of loudspeaker and ceiling. The depth of the absorptive material is ap-
proximately equal to l/4k of the lowest frequency that
* Presented at the 93rd Convention of the Audio Engineering
can be effectively absorbed. In practical terms, this
Society, San Francisco, CA, 1992 October 1-4; revised 1993
December 9. translates to a minimum room dimension of
** Currently with Charles M. Salter Associates, San Fran,
cisco, CA 94104, USA.
*** Now an independent consultant in Boston, MA, USA. h = 1.5k (1)

O.AudioEng.Soc.,VoL42,No.6,1994June 467
STRUCK AND TEMME PAPERS

(see Fig. 1). Using the definition of the speed of sound, every doubling of distance from the source, as shown
in Fig. 4. This was verified in Fig. 2, where the level
c = fX (2) of the measured response decreases by 6 dB for each
doubling of the microphone distance, empirically estab-
this becomes lishing that the microphone is in the far field of the
source. For long, narrow line sources, the sound radia-
tion will be cylindrical (see Fig. 5). In this case the
h = 1.5c (3) sound pressure level will change by -3 dB for every
f doubling of the distance from the source· In practice,
these conditions only exist over a limited frequency and
For measurements down to f = 20 Hz, h = 25.8 m distance range for a finite-size source and finite-size
(without absorptive material)· Another restriction is that room.
the size of the device under test be small (approximately When the distance to the source is small compared to
1%) compared to the volume of open space available in the wavelength of sound, one is said to be in the near
the anechoic room [ 1]. The lowest usable frequency field of the source. This typically occurs at low frequen-
is therefore determined by the depth of the absorptive cies, where the wavelength (and radius) of a spherical
material and the smallest dimension of open space avail- wave is so large that the sampled section of the wave is
able, usually floor to ceiling height· essentially a plane.
Often responses measured in smaller chambers (h _ Therefore it can be said that the far field and the near
6 m, f--_ 100 Hz) are plotted down to 20 Hz, perhaps field are determined by the relative size of the source
due to the expense of constructing a large chamber· Fig. compared to the frequency of the sound radiated· The
2 clearly reveals the effect of an anechoic chamber on free field and the diffuse field are functions of the room
the measured response of a loudspeaker, particularly at or environment·
low frequencies, even for an optimization of source and
microphone positioning within the room. This is caused

by the limited
x 6.6 m) and size of the anechoic
insufficient absorptionchamber (7.7 x 6.5
of low-frequency 400 d_ i
/ "_' __1__ '___'".
'"-. . .. , ,_,'

r . ctons. ,:!!
A reverberantfieldoccurswhen soundarrives from _, ',
all directions with equal magnitude and probability. This 60 ...........................
can be approximatedwith a reverberation chamber, 0.5meter'
which is constructed with many irregular, highly refiec- · .. I meter
z meter
tivesurfaces· no[_,

of sound radiated should be large compared to the size 20

of the source· This occurs in practice when the distance
To the
from be inobserver
the far field of source
to the a sound issource, the wavelength
large compared to the 0 [,"
20 200 Hz 2k 20k
size of the source. A so-called point source will radiate
spherically (4_sr) into a free field in all directions if Fig. 2. Frequency response of a loudspeaker measured in an
there are no reflections (see Fig. 3). Note that under anechoic chamber for various microphone distances. Note 6-
dB change in level for each doubling in distance· Ripple in
these conditions the sound field is very well behaved response is due to insufficient absorption of reflections at
and the sound pressure level changes by -6 dB for low frequencies·

¸ ? ?: ii? I'¸
]

Fig. 1. The lower limiting frequency for anechoic measurements is determined by the depth of the absorptive material and the
size of the room.

468 d. Audio Eng. Soc., Vol. 42, No. 6, 1994 June

PAPERS SIMULATED
FREEFIELDMEASUREMENTS

3 TIME'SELECTIVE TECHNIQUES Using time-selective techniques, it is possible to ob-

tain the free field response of a loudspeaker by measuring
All time-selective techniques rely on the constant in an ordinary room. This is done by restricting the time
speed of sound in order to separate the desired direct range to prevent reflections from entering the measure-
sound component from reflections, which arrive at the ment. Fortunately no special treatment of the walls,
measurement microphone delayed due to a longer trav- floor, or ceiling is necessary. The absorption influences
eled path [2]. Therefore differences between the various the reverberation time, however, and consequently the
techniques are found only in terms of signal to noise total measurement time for impulse and gating tech-
ratio, measurement time, and frequency selectivity [1]. niques [2].
In practice the measurement time range T is limited in Fig. 6 shows a loudspeaker and a measurement micro-
order to exclude unwanted reflections and noise. This phone situated in an ordinary room. The distance trav-
can be done as part of the measurement or as postpro- eled by the direct sound is d. From the loudspeaker to the
cessing. The uncertainty principle, in turn, determines nearest reflecting surface to the microphone, the sound
the frequency resolution A f of the measurement. Disre- travels a total distance of
garding the actual shape of the time window for the
moment, this is dR = dRl + dR2 . (5)

1
A f -- _ . (4) The difference in path length between the direct sound
and the first reflection determines the time available for
The relation holds for any analysis or technique.

3r _

Fig. 3. Idealpoint sourceradiating into free field. Soundpres- Fig. 5. Ideal line sourceradiating into free field. Soundpres-
sure level decreases by 6 dB for every doubling of distance sure level decreases by 3 dB for every doubling of distance
from source, from source.

/),
dB

Free · Diffuse

'--I-- I I
I I I I
I I l

_lLI· I l
Field I I _ Field
I I I
I I I
[ I I I I I [ Iii
d 2d Log d

Fig. 4. Definition of sound fields for point source. Near and far fields are functions of the source; free and diffuse fields are
functions of the environment.

J. AudioEng.Soc.,Vol.42,No.6, 1994June 469

STRUCK AND TEMME PAPERS

a free field measurement, usually the floor-to-ceiling height. The eccentricity of

the ellipse _ is defined as the ratio of the distance be-

T - da - d i6) tween the focal points to the length of the ellipse along
c the major semiaxis,

where c is the speed of sound. This, in turn, determines = d

- (11)
the lower frequency limit for the measurement. From b
Eq. (4),
= 0.5.
c
Af - dR -- d (7) The ratio of the length of the major semiaxis to the
length of the minor semiaxis is
It is easily seen that the loudspeaker, measurement mi-
crophone, and nearest reflecting surface define an ellip- b = 1.15 . (12)
sold, with the loudspeaker and microphone at the focal h
points. The ellipsoid is a solid body of rotation, con-
taining no reflective surfaces, about the major semiaxis The problem of optimization then becomes that of plac-
which contains the microphone and loudspeaker focal lng the largest possible ellipsoid with eccentricity E =
points. Any surface touching the outer shell of the ellip- 0.5 within a given room. In an ordinary rectangular room
sold will cause a reflection to appear at the microphone this is usually done by centering the major semiaxis of
that has traveled a distance dR. the ellipsoid b along the longest horizontal room dimen-
In general, the distance traveled by the first reflection sion L. If L > 1.15h, we obtain from Eq. (10)
is equal to the length of the ellipsoid along the major
semiaxis, d = h (13)
x/5
dR = b (8)
0.58h .
= x/g -
Furl < 1.15h,
For the typical case where
L
d = - . (14)
dR = 2d (9) 2

the height of the ellipsoid along the minor semiaxis h is In general it is not worthwhile to carry the optimization
found to be too far, as will be seen in Section 4. It is useful, however,
to keep a mental picture of the ellipsoid when arranging
h = N,/h2 - d2 (10) the test setup in the room in order to avoid the influence
of disturbing objects that may cause reflections [2]. For
= d V_ this application the magnitude of the time response is

-- [------ f -- h

Fig. 6. Loudspeaker, measurement microphone, and nearest reflecting surface define an ellipsoid within which free field measure-
ments can be performed using time-selective techniques.

470 J. Audio Eng. Soc., Vol. 42, No. 6, 1994 June

PAPERS SIMULATED FREE FIELD MEASUREMENTS

useful for viewing the arrival of the direct-sound compo- difficult to perform free field measurements.
nent, accurately fixing the measuring microphone dis- In contrast to gating, or to the time-delay spectrometry
tance, and determining the delay before the arrival of technique, where the time window is determined by the
the first reflection, use of a tracking filter [3], it is advantageousto apply
For the response to be measured in the far field of the the time window as a postprocessing operation [4]. This
source, d should be at least 3 times the largest linear is done in order to have a more well-defined time window
dimension of the source M, and to obtain the optimum frequency resolution. It also
allows the entire measurement, including reflections, to
d > 3M. (15) be viewed in the time domain before applying the win-
dow. This implies an inherently linear, constant band-
Depending on the size of the source, this may be some width data format for the far-field measurement for use
distance other than 1 m (or other reference distance), of the forward and inverse fourier transforms. Reflec-
Because simulated free field conditions have been ob- tions are removed from the far-field measurement by
tained (that is, the sound field is well known), the mea- multiplying the measured response (in the time domain)
sured response can easily be normalized to any desired by a window function. Consequently the frequency re-
reference (see Appendix 1). For measurements in the sponse is convoluted with the Fourier transform of this
far field of the source, the distance d is then determined window. Empirically, through much trial and error, we
by the size of the device under test. In general, assuming have found that the optimum time selectivity and the
Eq. (15) is fulfilled and substituting Eq. (8) into Eq. minimum resulting frequency-domain spreading are ob-
(7), we obtain tained by the use of a variable-widthrectangulartime
window with leading and trailing Hanning (cos 2) tapers
Af = c (16) of 10% of the rectangular width (see Fig. 7). With re-
N/h2 - aa - d spect to worst-case frequency resolution, the effective
time range T is determined by the width of the rectangu-
This expression is deliberately left unsimplified [using lar portion of the time window. It will be seen in Section
the assumptions of Eqs. (9), (10), (12), and (14)] in 4 that the actual shape of the time window becomes less
order to show the dependence of A f on both room size critical if the low-frequency response of the system can
and test object size if Eq. (15) is also fulfilled. Therefore be obtained by another method.
the lowest measurable frequency using any time-selec- As mentioned previously, the choice of excitation sig-
tive technique is a function of the room size, where h nal and test method should be based on speed, its effect
is the smallest room dimension (usually floor to ceiling on the test object, selectivity in both time and frequency,
height), and the test object size, which determines the and the maximum available signal-to-noise ratio. In ad-
distance to the measurement microphone d according to dition, it should be possible to specify the input power
Eq. (15). to the device under test accurately, unless the system is
Whenever attempting to visualize this problem men- known to be perfectly linear. Transducers, and loud-
tally, it is useful to think of a "small box (the loud- speakers in particular, do not normally fall into the cat-
speaker) inside a large box (the room)." When the ratio egory of perfectly linear systems, particularly not over
between room size and loudspeaker size is large, it will a wide dynamic range. The sine wave has a low crest
be relatively easy to perform simulated free field mea- factor and its power can be determined easily and
surements over a wide frequency range F. As the ratio precisely.
decreases (that is, the room size is decreased or the Sinusoidal excitation also enables ideal frequency se-
loudspeaker size increased), it will become increasingly lectivity, offering the possibility for harmonic measure-

Amplitude

I 1
I

, ]

I I I_
I I Time
I I

F
-j ---i
Left Taper _' ',
width "-i-'J
Right Taper J
Start

Fig. 7. Effect of reflections can be removed from far-field measurement by multiplying measured time response by a time
window. This window is rectangular with Hanning tapers to reduce frequency-domain spreading.

J. Audio Eng. Soc., Vol. 42, No. 6, 1994 June 471

STRUCK AND TEMME PAPERS

ments. The time required for the far-field measurement microphone should be placed as near to the center of
should be determined entirely by background noise. In the diaphragm as possible [7]. This technique physically
an ordinary room this should be no more than about I s eliminates reflections and noise. The near-field measure-
[5]. A short measurement time will also minimize time ments can be carried out entirely in the frequency do-
variance, such as effects due to heating. It should be main. The theoretical upper frequency limit for per-
kept in mind, however, that an improvement of 3 dB in forming near-field measurements on a driver mounted
the signal-to-noise ratio can be gained for every succes- in an infinite baffle is given by
sive doubling of the effective measurement time [3], [6].
ka = 1 (18)
4 NEAR-FIELD MEASUREMENTS
where k is the wave number. Substituting Eq. (2) into
The low-frequency response of a loudspeaker can be Eq. (18) becomes
obtained using the near-field technique [7]. This is done
by simply moving the microphone very close to the low- c
frequency driver(s) in the loudspeaker system (see Fig. f_-l - 2_ra (19)
.8). This, of course, assumes that there is little or no
passive radiation from the walls of the enclosure, that is, or
the structure is essentially rigid. To avoid measurement
errors, the measurement microphone should be as close 58.749
as possible to the driver under test. fL=] - a (20)
A microphone distance of
for driver radius a in meters. For convenience, this can
d < 0.11 a (17) be rewritten as

results' in measurement errors of less than 1 dB. The

10 949.86
f_=l _< 2a (21)

in terms of the driver diameter 2a in centimeters. Fig.

9 shows the upper limiting frequency for near-field mea-
surements of drivers mounted in an infinite baffle versus
driver diameter in centimeters. For driver diameters
specified in inches, this becomes

4310.97
ka,=l _ 2_ (22)

For closed or ported loudspeaker systems, this limit will

Fig. 8. Low-frequency measurements can be performed by be lower (see Section 5).
placing microphone in near field of loudspeaker. Microphone
should be placed as near as possible to center of diaphragm For ported loudspeakers or enclosures containing mul-
or portopening, tiple drivers, additional near-field measurements of the

2k-

lk

500

400
g
__ 300

20O

100
0 5 10 15 20 25 30 35 40 45 50
Driver Diameter 2a or Major Source Dimension M (cra)

Fig. 9. Highest frequency at which near-field measurements can be performed is determined by size of driver when mounted in
infinite baffle or largest dimension of enclosure. Upper frequency limits are shown versus driver diameter 2a or major source
dimension M.

472 J. Audio Eng. Soc., Vol. 42, No. 6, i994 June

pAPERS SIMULATED
FREEFIELDMEASUREMENTS

individual sources are required. The complete near-field where both the near-field and the far-field measurements
response is found by first scaling the measurement of should be valid (see Fig. 10). The upper frequency limit
the port(s) and then taking the complex sum of all mea- for the overlap range is explained in Section 6. As it is
surements [7] (see Appendix 2). usually not possible to reduce the size of the loudspeaker
Because the microphone is so close to the driver (or once it is created, it is necessary to perform the far-field
port), the measured sensitivity appears considerably measurements in a room of sufficient size so that an
higher than an equivalent far-field measurement. For a overlap frequency range exists. Combining the two
loudspeaker system radiating into a 4_ space, the far- methods of determining the sensitivity offset and solving
field sensitivity can be calculated from Eq. (23) for the radius a, yields

HFF(f) -- HSF(4d
f)a (23 a) a - 2dHFF(f) (26)
H NF(f )
or

for a frequency f in the overlap range, when HNF(f)

4d [dB] (23b) and HEF(f) are measured independently, for a driver
HFF(f = HSF(f) -- 20 Ioglo _- mounted in a baffle. For individual drivers, this interest-
ing result provides an empirical method for determining
where HNF(f) is the measured near-field response and the effective radius a and subsequently the effective radi-
HR:(f) is the far-field response at a microphone distance ating surface area of the driver SD (SD = _ra2). This
d for a driver of a radius a. For ported loudspeakers the parameter is critical in the determination of the
radius a is taken as the radius of the driver. For a driver Theile-Small parameters.
mounted in an infinite baffle, radiating into a 2'_ space, For frequencies above fka=l most drivers no longer
the far-field sensitivity can be calculated as behave like rigid pistons (that is, modal behavior or
"cone breakup") and the relationship between near-field
HFF(f) -- HSF(f)a (24a) and far-field response becomes complex. The frequency
2d range for the near-field measurement should therefore
be chosen to an upper frequency no greater than fka= _.
or The lower frequency should be less than or equal to 20
2d Hz (lower for large-diameter or extended-range woofers)
HFF(f) = HNF(f) -- 20 1ogl0-_ [dB] (24b) in order to include the low-frequency rolloff of the sys-
tem. This is important for examining the time-domain
This means that at low frequencies, where the loud- behavior of the system.
speaker behaves like a rigid piston, the measured near- Near-field measurements are entirely independent of
field response is directly proportional to the far-field the measurement technique and offer the optimum mea-
response and is independent of the environment into surement signal-to-noise ratio due to the proximity of
which the loudspeaker radiates [7], [8]. Alternatively, the microphone to the sound source. This provides a
the level offset can be determined directly by comparison high rejection of both correlated noise (that is, reflec-
to a far-field measurement in the "overlap" frequency tions) and uncorrelatedrandom background noise. Using
range, stepped sine excitation, tests can be performed at dis-
crete frequencies in a logarithmic (ISO preferred) fre-
Af _ f _< fka= 1 (25) quency format, typically 1/12 octave (ISO R40) or 1/24

I
Near Field
I
Measurement
I
I
I
I

Far Field
Measurement

I t Im
1_._ c__ f
T _M

Fig. 10. For a sufficiently large room, a frequency range exists where both near- and far-field measurements are valid.

J. Audio Eng. Soc., Vol. 42, No. 6, 1994 June 473

STRUCK AND TEMME PAPERS

octave (ISO R80). This optimizes low-frequency resolu- under test. This enabled direct measurement of the sys-
tion and test time and enables measurements of harmonic tern response at the input and output terminals of the
distortion to be performed [6]. device under test. All postprocessing was carried out
directly in the analyzer. In addition to the graphs shown
5 FULL-RANGE RESPONSE MEASUREMENTS here, the measurement results are available in ASCII
format for further processing (such as polar response,
If a full audio range response (typically 20 Hz to 20 directivity, and statistics).
kHz) is desired, the measurement can be performed in First a measurement of the loudspeaker response in
two passes. The first measurement is performed in the the far field is performed. The result is shown in Fig.
far field using a calibrated microphone. The microphone 11, including the effect of room reflections. The time-
can then be repositioned into the near field to measure response magnitude clearly shows the arrival of the di-
the low-frequency response. Assuming an overlap fre- rect sound as well as the delay before the arrival of
quency range exists, a transition frequencyfs within this reflections. The delay before the arrival of reflections
range can be selected at which to join these measure- determines the time range T available' for a time-selec-
ments. With the exception of the room size, no other tive measurement. If the time range is insufficient, the
assumptions are made. Postprocessing is then performed microphone-loudspeaker setup must be repositioned or
to account for the time delay to the far-field microphone, the measurements may need to be performed in a larger
for the application of a time window to the far-field room. Reflections are convoluted with the response of
measurement to isolate the direct-sound component, and the loudspeaker in the frequency domain, resulting in a
to account for the sensitivity offset between the far- "ragged" frequency response.
field and near-field measurements. All block arithmetic The peak in the time-response magnitude indicates
operations should be comple x so that other functions the delay in sound propagation x for the arrival of the
such as phase, group delay, and the time response are direct sound at the microphone, and consequently the
available for the final, full frequency range response distance d to the microphone position (typically the tweeter
data. arrival). A time shift can be applied to the far-field mca-
To test the application of this method, measurements surement to compensate for this delay, which is not an
were performed on asmall (180by ll0by ll0mm)two- actual part of the loudspeaker response (see Fig. 12).
way closed-box loudspeaker, with an 85-mm-diameter This allows the final phase response to be displayed
woofer using a Briiel & Kja_r Type 2012 audio analyzer without "wrapping." It is followed by the application of
and a calibrated Briiel & Kja_r Type 4133 free field the time window to the far-field measurement (see Fig.
microphone. The inherent phase inversion of the mca- 13). The resulting frequency response appears
surement microphone is corrected during the postpro- "smoothed," due to the removal of time-domain reflec-
cessing. The output of the measurement system was cali- tions which cause frequency-domain ripple.
brated to account for the gain and frequency response The microphone is then repositioned in the near field
of the power amplifier used to drive the loudspeaker of the woofer, and the low-frequency response is mea-
sured. This appears with a higher sensitivity than the
far-field measurement (see Fig. 14). The cursor can be
80 I

Direct _ [ i _ Room Reflections I

6osound
i---i ', l
-- 'i __i / 6o 'I
i I
40 i I ..... _......................................................
i I i I ._1 _ Time Window

I = I
20 60 I

dB 0 1
dB t

9o I i i

70
80 ..........................................................................
_ 20

6O

56 0 I I
40 -5m 0 5m s 10m 15m
20 200 Hz 2k 20k

(b) Fig. 12. A time shift is applied to measured response to account

for delay to measurement microphone. This is followed by
Fig. 11. (a) Magnitude of time response shows arrival of direct application of a time window to isolate the direct-sound cum-
sound at measurement microphone followed by arrival of re- portent. Inner lines indicate rectangular portion of window (T
fiections. Effect of reflections appears as ripple in frequency = 5.67 ms); outer lines indicate leading and trailing Han-
response (b). ning tapers.

474 J. Audio Eng. Sue., Vol. 42, No. 6,.1994 June

 PAPERS SIMULATED FREE FIELD MEASUREMENTS

used to read the offset between the measurements to with respect to frequency, this delay can be found by
match the sensitivity of the high-frequency measure- examining the phase difference A_b (see Fig. 14) be-
ment. The magnitude of the offset can also be calculated tween the measured near-field phase and the compen-
according to Eq. (24). After rescaling the magnitude of _ sated far-field phase atfs. For "unwrapped" phase func-
the near-field measurement, an overlap region should be tions the delay compensation to be applied to the near-
clearly visible. A transition frequency fs can then be field measurement can be calculated as
selected within this region.
In order to avoid a discontinuity in the phase response, ,1 _bNF(fs) -- _bFF(fS)
both near-field and far-field measurements must be re- Xx = - 360 fs (27)
ferred to the same acoustic reference plane. This is ac-
complished by applying a time shift to remove the appro- for phase measured in degrees (see Fig. 16). The applied
priate delay from each measurement. The reference time-shift compensation has the opposite sign of the
plane is established after time shifting the far-field mca- propagation delay to the microphone.
surement as previously mentioned by x. If the intrinsic The near-field measurement is shown along with the
delay in the near-field measurement is _'r_F,then the delay far-field measurement in Fig. 16, after the application
to be removed from this measure in order to align it to of Eq. (24) and after delay compensation. Note the align-
the already compensated far-field measurement is % (see merit of the phase responses and the overlap range in
Fig. 15). Noting that delay is the derivative of phase the magintude responses.

80 120 ' fs _i

60 110 '; I"r ~"' _";:"'.

''''- i
dB dB ,' i

40 100 ........... _! .......................................... _-.....................

2o i

o '~~
-Sm 0 5m s 1Om 15m 20 200 Hz 2k 20k

(a) (a)

0 ", [
100 i "
-9o
dB 90 i -- Oegr. ' .......... _..............
'_';:', i
]

o'"
, - _t i
' J ".
i i
70 ...............................................
._;._.
.......................... !

60 ................ ......, -27o

;.. ........................................................................................................................................................ '_

-_o I i
50 20 200 Hz 2k 20k
20 200 HZ 2k 20k

(b) (b)
Fig. 14. Near-field measurement before magnitude and delay
Fig. 13. (a) Time response (re 1 Pa/V) after application of compensation. Low-frequency measurement shows higher appar-
time window. (b) Resulting frequency response (re 20 IxPa/ em sensitivity compared to far-field measurement due to de-
V), valid to A f = 1/T (176 Hz). creased distance from microphone. (a) Magnitude. (b) Phase.

Reference
plane

Position of Position of
Near Field Far Field

Microphone

/ / Microphone Measurement

ACOeUn_triCa'/_
of woofer TFF

Fig. 15. Delay compensationapplied to both near-field and far-field measurementsin order to refer both to same acoustic
reference plane.

J. Audio Eng. Soc., Vol. 42, No. 6, 1994 June 475

STRUCK AND TEMME PAPERS

In order to preserve the low-frequency resolution and sponse for the complete system, shown from 20 Hz to
to avoid a discontinuity in frequency resolution between 20 kHz, appears in Fig. 17. The phase and the group
the upper and lower frequency ranges at fs, the linear delay are shown in Fig. 18.
(FFT) far-field data are converted to a logarithmic (con- Because the data are complex, the final response can
stant-percent bandwidth) format equ'ivalent to the ISO be reconverted to a linear format (without the effect of
RS0 format of the near-field measurement. This is a reflections) before applying an inverse Fourier transform
straight-line interpolation algorithm on a decibel versus to recalculate the time-domain response from the full-
logarithmic frequency axis. The linear data are 1600 range frequency response. In this way the measured data
points, from I Hz to 40 kHz, so the response should be can be displayed optimally in both the frequency and
more than adequately sampled for an accurate conver- the time domains. Extending the measurement frequency
sion in the frequency range where the windowed far- range F to 40 kHznormally eliminates the need for any
field data are valid. This has the added advantage of windowing of the frequency response prior to calculation
showing the response in the frequency domain in an of the time response. Most loudspeaker systems de-
optimum format, as it is universally accepted as standard signed for the normal audio range will be sufficiently
industry practice to show the frequency response in deci- rolled off, or "self-windowed," at both the upper and
bels, on a logarithmic frequency scale, preferably with the lower ends of a 1-Hz to 40-kHz frequency spectrum.
the IEC 263 axis ratio (25 or 50 dB per decade). A linear This subsequently eliminates artifacts caused by such
data set in the frequency domain tends to concentrate windowing. The resulting time resolution At given by
resolution at the higher frequencies on a logarithmic
frequency axis. An additional benefit is that the final
response contains the ISO preferred frequencies, so it is _00
a simple matter to obtain the necessary values for the
calculation of sensitivity, directivity, and so on. 90 i
Afterward rectangular frequency windows are applied dB
to both measurements. The near-field low-frequency so
measurement is windowed to eliminate data at frequen-
cies above fs- The far-field measurement is windowed 7o ........
to exclude data at frequencies below fs. Assuming fs >
l/T, this also eliminates invalid data remaining after the
application of the time window. Now the two responses 60
no longer overlap. The final process is to add the two 50
responses together to obtain a single, continuous com- 20 __ 206 .z 2k 26k

plex data set. The last step is to divide out the complex Fig. 17. Resulting on-axis frequency response, 20 Hz to 20
response of the power amplifier used to drive the loud- kHz, ISO RS0 format (1/24 octaves), dB re 20 p.Pa/2.0 V at
speaker. The magnitude of the resulting frequency re- 1 m (1 W into 4 11),fs = 400 Hz (two-way, 4-11, closed-box
loudspeaker, grille off).

90

I !
dB

40
20 200 Hz 2k 20k }
-540 I
(a) 28 200 HZ 2k 28k

(a)
0 "-

-98 . ............. t, 1Om ] I

-27°
Oegr
; s X..._
__'ii
_ ' ll!
-360 ··· ! I
20 200 Hz 2k 20k ] I

(b) -Sm
20 200
! HZ
!
2k 28k

Fig. 16. Near-field response after application of level offset (b)

and delay compensation. Note overlap range in magnitude
response where both measurements yield same result. Cursor Fig. 18. (a) Phase and (b) group delay for same two-way
indicates selected transition frequency. (a) Magnitude. (b) closed-box loudspeaker. Note that resulting data set is com-
Phase. plex.

476 J. Audio Eng. Soc., Vol. 42, No. 6, 1994 June

PAPERS SIMULATED FREE FIELD MEASUREMENTS

the uncertainty principle is (except for moving the microphone for the near-field
measurement), including guidance in performing both
1 the measurements and the postprocessing as well as doc-
At = _ . (28) umentation of the results.
Fig. 21 compares the response of a similar loud-
For an "extended-range" loudspeaker, with a frequency speaker measured in an ordinary room using this tech-
response that has not rolled off sufficiently at 40 kHz, nique to the response of the same loudspeaker measured
a frequency window with a right-Ha_nningtaper from 20 using traditional techniques (that is, not time selective)
kHz to 40 kHz can be applied before Calculation of the in an anechoic chamber. Differences at low frequencies
time response. This has the advlantage of not affecting are due to previously mentioned errors introduced by
the frequency response in the audiobandwidth, and es- the size of the anechoic chamber (7.7 by 6.5 by 6.6 m),
pecially not affecting the response at low frequencies, according to Eq. (3).
and it is an improvement over the "half-Hann" window To test the technique further, a ported loudspeaker
recommended in [9]. The magnitude of the time response (39 by 23 by 22 em) was also measured. Fig. 22 shows
is seen in Fig. 19. The real part of the time response the individual measurements of the driver and the port.
(traditional impulse response) can also be displayed (see The measurement of the port is first scaled by the square
Fig. 20). root of the ratio of the port area to the effective radiating
Software can be used to automate the entire process surface area of the driver (see Appendix 2). This scaled
response is then summed (complex summation) with the
near-field response of the driver. The complete near-
80 fieldresponse oftheportedloudspeaker isshowninFig.
23. The resulting simulated free field response of the

100

6O

90
Simulated Free Field _ _

dB _
80
dB , "-"-"_ '"-'

Anechoic Chamber

o
20
_m
,

Fig. 19. Time response

lated from full-range
i_lAI

(magnitude)
s 4m

of system can be calcu-

frequency response by inverse FFT (right-
8m
i 000 20 200

(a)
Hz 2k 20k

Hanning taper, 20 kHz to 40 kHz, applied to frequency re- o

sponse before calculation).
Anechoic Chamber

6k

Degr i
4k

-360 i

2k s i

i
-540

Pa 0

-2k

-720

20 200 Hz 2k 20k

-_k (b)

Fig. 21. (a) Simulated free field technique shows good correla-
-2k i , a .'- , , tion with measurements performed in anechoic chamber at high
-2m -2m 0 2m- "s 4m 6m 8,. frequencies [dB re 20 i_Pa/2.0 V at 1 m (1 W into 4 f_)]
and provides significant improvement at low frequencies. (b)
Fig. 20. Real part of time response (impulse response). Comparison of phase response.

J. Audio Eng. Soc., Vol. 42, No. 6, 1994 June 477

STRUCK
AND
TEMME PAPERS

ported loudspeaker system (20 Hz to 20 kHz) appears = I/T. Choosing this transition frequency reduces the
in Fig. 24. apparent resolution around fs due to the "smoothing"
Fig. 25 shows a comparison of the response of this effect of the time window on the measured far-field re-
ported loudspeaker measured using the simulated free sponse at its lower frequencies, that is, in this frequency
field technique to the response of the same loudspeaker range there is more detail in the near-field measurement.
measured in an anechoic chamber. Here even greater Another method is to see if, after using Eq. (24), the
differences can be observed at low frequencies than for two measurements intersect at some frequency. This fre-
the smaller closed-box system. The effects of inadequate quency can then be used as fs if it conforms to the
low-frequency absorption in the chamber are more pro- aforementioned restrictions. Alternatively, the magni-
nounced due to the use of a greater microphone distance tude of the near-field response could be adjusted for a
for the far-field measurement, necessary because of a visual "best fit."
larger source size. Becausethe drivers are generallymountedin an enclo-
sure somewhat larger than the driver itself, at high fre-
6 CHOICE OF TRANSITION FREQUENCY quencies (smaller wavelengths) the low-frequency driv-
ers will no longer radiate spherically in the near field,
Several possibilities exist for choosing the exact tran- due to the baffle effect of the enclosure [4], [7]. This
sition frequency fs. In general, fs should be chosen as effect is observed as a transition from 4xr (spherical)
high as possible in order to preserve low-frequency reso- to 2_r (hemispherical) radiation, occurring over several
lution, but above 1/T. In any case it should be the near- octaves for increasing frequency, as the wavelength ap-
field response that is manipulated relative to the far-field proaches the size of the source for sound radiation mea-
response. The level of the far-field response should not sured by a microphone placed in the near field. The
be altered, assuming it was calibrated. The first possibil- useful upper frequency limit for near-field measurements
ity would be to simply apply Eq. (24) and choose fs on loudspeaker systems is always observed to be lower
" than the theoretical fka= _ limit for a driver mounted in

I _ Driver

,'
t

i/ ?.............................
i

?
', i Port

-,'.-, ,,,,-
dB

__
I 50 20 200 Hz 2k 20k

80 ..... [ i _ I.... 60
20 50 100 200 soo Hz lk 2k Fig. 24. Complete simulated free field frequency response of
two-way 8-fl ported loudspeaker, dB re 20 IxPa/2.83 V at 1
Fig. 22. Near-field measurements of driver and port performed m (1 W into 8 Il) measured on axis, grille off, fs = 206 Hz.
on vented loudspeaker system (shown before scaling). Port
response is scaled and then these complex responses are
summed to obtain low-frequency response of system.
100

, 90 Anechoi_ Chamber _ _

110 ! ............ ed Free F'eld

_ _ i
dB
l _, ' :_ 80

'°°I i i , 7o

80 ................... 50

20 200 Hz 2k 20k

70' Fig. 25. Frequency response of ported loudspeaker measured

20 50 lOO 200 500 Hz _k 2k using simulated free field technique and measured in anechoic
chamber. Note irregularities in anechoic response caused by
Fig. 23. Resulting near-field response of ported system. (SD low-frequency reflections in anechoic chamber. Frequency re-
= 127.68 cm°, Sp = 21.32 cmO.) sponse; dB re 20 IxPa/2.83 V at 1 m (1 W into 8 fl).

-478 J. AudioEng.Soc.,Vol.42,No.6,1994June
PAPERS . SIMULATED FREE FIELD MEASUREMENTS

a baffle. The near-field measurement will progressively tained using two-tone techniques, but here the one-to-
under estimate the true free field response at higher fre- one relationship between time and frequency is lost and
quencies. Because this limit is governed by the overall we are limited to measurements in an anechoic room or
size of the test object, rather than by the driver radius, quasi-near-field measurements [10].
the entire enclosure must be considered when comparing The measurement of a harmonic response results in a
the wavelength to the size of the source. In practice we frequency multiplication of the desired response by the
have seen that the near-field response can be used with harmonic order for results plotted at the excitation fre-
errors of less than 1 dB (Compared to anechoic measure- quency. Consequently the upper limiting frequency for
merits) at frequencies where the wavelength is greater near-field measurements is corresponding reduced to the
than approximately 3 times the major dimension of the frequency where
source M. An attempt to explain this behavior can be
found by substituting M for 2a in Eq. (18). In terms of 1
an upper limiting frequency, this is ka = _ (31)

fs _< c or
-_ . (29)
c

This is therefore the most critical factor governing the f_= _ - 2'ri'aN (32)
selection Offs.
In practice the upper limit can also be verified by where N is the harmonic order. Fortunately the lower
performing off-axis measurements to determine at what limiting frequency for far-field time-selective measure-
frequencies the source is no longer omnidirectional. For ments is correspondingly lowered as well,
multiway systems with a crossover frequency fx less
thanfs, a lower transition frequencyfs should be chosen 1
such that Af = NT (33)

fs < fx (30) Fig. 26 shows the separation of individual harmonic

components in time and frequency for a linearly swept
to avoid rolloff in the near-field response due to the sinusoidal excitation. Each far-field harmonic must
crossover filter, therefore be measured with an individual sweep in order
to isolate the desired component from reflections and
7 HARMONIC DISTORTION other harmonics. The near-field harmonics can be tested
at discrete frequencies in a single pass. The transition
Using sinusoidal excitation also enables time-selec- frequency for each harmonic, fs(N), should also be a
tive and near-field measurements of any selected hat- function of the harmonic order,
monic components to be performed [5], [6]. Practical
considerations for full audio bandwidth tests such as =
the range of human hearing, the frequency range of the fs(N) rs(l) (34)
measurement microphone, and the low energy level of
upper harmonics outside the passband typically limit wherefs(1) is the transition frequency selected for the
measurements to the second and third harmonics. More fundamental.
useful results for nonlinear measurements can be ob- Harmonic distortion measurements were performed on

& f
dB

T I 2f
/ Reflections /_-- _,
t [\ 3 f

/ / /I _ it'' 4t 5f S=lOHz/ms
/,/, I h_ \ 6f
t/I I I Il /_ \ I I I I .
6OO
'/ 100 '\200 _,\ _00 400 500 f[l_z]
\ '\' '\
_' / I \ ,_ \ Time/Frequency

/ / I \ '_-:¢--'- _nUow
/ / I \ i \1 _ (e.g, 2 Harmonic)

\ \ \ _'
. f L! \l. \
t, + 10ms I I I{ _, I
// 100 200 300 \400:,\ N500 600 f[_z]
/ \ \ \
_' / I ',,. ,_i x

+ 20 ms/I / i I I I'_ I I

[1 _ [sec] 100 200 300 noo 500 600 f [l_lz]

Fig. 26. Use of linearly swept sinusoidal excitation signal enables time-selective measurements of individual harmonic components
to be performed. Diagram shows separation of second harmonic in time and frequency.

J. Audio Eng. Soc., Vol. 42, No. 6, 1994 June 479

STRUCK AND TEMME PAPERS

the closed-box loudspeaker using this technique (see The sine-based test methods employed to test technique
Fig. 27). Note that these are true harmonic responses optimize speed and the selectivity in both time and fre-
(not THD plus noise). The total harmonic distortion quency, and maximize the available signal-to-noise ra-
(THD) referred to the total is calculated as tio. In addition, sinusoidal excitation provides a low
crest factor signal and enables accurate specification of
THD = N/H:(f)2 + H3(f)2 the input power to the device under test. This technique
N/Hi(f) 2 + H2(f) 2 + H3(f) 2 (35) also presents the resulting data optimally in both the
frequency and the time domains. These results are avail-
where the denominator (total) is the power sum of the able in any desired coordinates and can be exported for
fundamental plus the measured harmonics (in this case, further processing. Furthermore it has been demon-
the second and third). The resulting THD compared to strated that, using sine-based analysis, this technique
the total is shown in Fig. 28. Good correlation was found can be extended to include measurements of harmonic
with equivalent measurements performed in an anechoic distortion. The only limits for this technique are, in fact,
chamber, especially at higher frequencies (see Fig. 29). imposed by the size of the room used for performing the
tests. These limits, however, are much less critical than
8 CONCLUSION with either purely time-selective techniques alone or tra-
ditional measurements in an anechoic chamber. The ef-
A technique has been presented enabling measure- feets of the anechoic chamber on the measured response
merits of the complex free field response of a loudspeaker at low frequencies, caused by its limited size and insuf-
to be performed, without an anechoic room, throughout ficient absorption of low-frequency reflections, are elim-
the entire audio frequency range. One measurement is inated. A comparison to traditional anechoic measure-
performed using a time-selective technique in the far ments clearly reveals the magnitude of these chamber-
field of the source. The low-frequency response is then induced errors. This technique offers reliability and con-
obtained using the near-field technique. Assuming an venience without making unnecessary assumptions
overlap frequency range exists, a transition frequency about the environment or about the device under test.
can be selected at which to join these measurements.
9 ACKNOWLEDGMENT

_00 The authors wish to thank Thomas V. Petersen for his

80
help in preparing the manuscript.

_B 10 REFERENCES
60

[1] C. H. Biering and O. 'Z. Pedersen, "Free Field

40 Techniques-- A Comparison," BrQel & Kj a:r ( 19 81).
[2] R. Heyser, "Acoustical Measurements by Time
Delay Spectrometry," J. Audio Eng. Soc., vol. 15, p.
20 370 (1967 Oct.).
0 [3] C. H. BieringandO. Z. Pedersen,"SystemAnal-
20 200 Hz 2k 20k ysis and Time Delay Spectrometry, Part 1," Briiel &
Kjeer Tech. Rev., no. 1 (1983).
Fig. 27. Fundamental, second, and third harmonic responses [4] L. R. Fincham, "Refinements in the Impulse Test-
of closed-box loudspeaker for 2.83-V input, measured using
simulated free field technique. Responses are in dB SPL re 20
gpa referred to I m.
50

-10
40

dB
-20 10%

-3o ..........................
;..........................................
§-_;
20

dB _.
' ili 30
-40 1%

-50
0
20 200 Hz 2k 20k

-60 __ ...........
._.._ 10
20 200 HZ 2k 20k Fig. 29. Comparison between third-harmonic measurements
performed using simulated free field technique and measure-
Fig. 28. Resulting total harmonic distortion in dB re total and merits performed in anechoic chamber. Frequency response,
in %. dB re 20 ixPa/2.83 V at 1 m (1 W into 8 _).

480 J. Audio Eng. Soc., Vol. 42, No. 6, 1994 June

PAPERS SIMULATED
FREEFIELDMEASUREMENTS

ing of LoudsPeakers," J. Audio Eng. Soc., vol. 33, pp. to the loudspeaker's specified nominal impedance,
133-140 (1985 Mar.).
[5] C. J. Struck and C. H. Biering, "A New Tech- V2
nique for Fast Response Measurements Using Linear Po - Zo [W] (38)
Swept Sine Excitation," presented at the 90th Conven-
tion of the Audio Engineering Society, J. Audio Eng. so
Soc. (Abstracts), vol. 39, p. 384 (1991 May), preprint V = N/PoZo [VI (39)
3038.
[6] C. J. Struck, "An Adaptive Scan Algorithm for
where V is the corresponding input reference voltage.
Fast and Accurate Response Measurements," presented This refers the measured response to the reference
at the 91st Convention of the Audio Engineering Soci-
power, regardless of the actual level of the excitation
ety, J. Audio Eng. Soc. (Abstracts), vol. 39, p. 1004 signal. This of course assumes that the device under test
(1991 Dec.), preprint 3171. behaves linearly (that is, no power compression).
[7] D. B. Keele, Jr., "Low-Frequency Loudspeaker Because the output terminals of a loudspeaker are not
Assessment by Nearfield Sound-Pressure Measure- intrinsically obvious, it is also necessary to specify a
ment," J. Audio Eng. Soc., vol. 22, pp. 154-162 reference distance. Typically this is 1 m. For free field
(1974 Apr.). measurements of a point source (that is, spherical sound
[8] L. L. Beranek, Acoustics (American Institute of radiation), the sound pressure level is inversely propor-
Physics for the Acoustical Society of America, Cam- tional to the distance (a -6-dB change in output level
bridge, MA, 1986). for every doubling of distance). This is easily verified
[9] J. Vanderkooy and S. P. Lipshitz, "Uses and empirically. As explained earlier, it may be necessary
Abuses of the Energy-Time Curve," J. Audio Eng. to perform measurements at some distance other than
Soc., vol. 38, pp. 819-836 (1990 Nov.). the reference distance in order to be in the far field of
[10] S.F. Temme, "Why and How to Measure Distor- the source. Fortunately this can easily be incorporated
tion in Electroacoustic Transducers," Proc. AES llth into the dB reference using either distance or time delay,
Int. Conf. on Audio Test and Measurement, Portland, assuming a constant speed of sound. The complete refer-
OR, 1992 (May 29-31). ence then becomes

APPENDIX I Po . do [Pa/V] (40)

dB REFERENCE Xo = Pv'-P-_oo d- '

The magnitude of the frequency response of a loud- Typically Po is 20 IxPa, Po is 1 W, and do is I m. For
speaker is usually plotted in decibels versus frequency, example, for an 8-1) loudspeaker measured at 1 m (Z0
In order to be able to interpret this curve correctly, it is = 8 11 and d = lm), x0 = 7.07 ixPa/V.
necessary to know the decibel reference. Recall that For a constant speed of sound c time delay can be
decibels express a logarithmic power ratio, equated to distance as
P
dB = 101ogl0p_ (36) x =- d [s] . (41)
c

where Po is the reference power.

Normally power is not measured directly. Therefore In terms of time delay, the dB reference can then be
written as
this expression can be rewritten as

x x0 = Po . ?o [Pa/V] . (42)
dB = 20 logl0 _ (37) V_0Z ° x
For c = 344 m/s and do = 1 m, x0 = 2.91 ms. Normally
where x0 is the corresponding signal reference. This is
the time delay can be obtained directly from the magni-
derived from the fact that the ratio of the squares of two tude of the time response by locating the largest peak,
signals
ence forisaequal
transferto the ratio of
function hastheir powers. and
dimensions The can
refer-
be corresponding to the arrival of the direct sound.
For line sources, such as tall, narrow ribbon loud-
found by dividing the reference for the output by the
speakers and tweeter line arrays, the sound pressure level
reference for the input. For a loudspeaker x0 is in pascals changes by -3 dB for every doubling of distance of
per volt. Therefore if the output is desired in dB SPL, the measurement microphone under free field conditions
the output reference is 20 txPa. (recall Fig. 4). Applying this relationship, the dB refer-
Typically the reference for the input is given as a ence becomes
power (such as 1 W). The signal applied, however, is
in volts, so the power will depend on the impedance of /-7,

the loudspeaker.
dissipate In general
the reference the across
power voltage a isresistance
found thatequal
will Xo = PV_o
Po o _/_ [Pa/V] (43)

J.AudioEng.Soc.,Vol.42,No.6,1994June 481
STRUCKAND TEMME PAPERS

or, in terms of time delay, be measured using the near-field techniquedescribed by

Keele [7]. The complex response of each source (driver

Po [Pa/V] (44) responses can then be summed. The overall near-field

_/_ or port) should be measured individually. These complex
x° = PX/-P_°° response is given by

This is also easily verified empirically. Of course, if d /--_

is increased sufficiently, even a line source will appear HNF(f) =HD(f) + S/Se He(f)
to behave like a point source. Provided the measure- X_/SD (45)
merits are calibrated, the loudspeaker sensitivity and di-
rectivity can easily be calculated from these dB values, where HD(f) is the near-field measurement of the driver,
Se is the total radiating surface area of the port(s), SD is
APPENDIX 2 the total radiating surface area of the driver(s), and
NEAR-FIELD MEASUREMENTS OF PORTED Hp(f) is the near-field measurement of the port. When
ENCLOSURES AND MULTIPLE-DRIVER measuringthe port response, the microphone should be
SOURCES positioned in the same plane as the port opening (see
Fig. 8). Passive-radiator systems can also be measured
The low-frequency response of ported loudspeakers with this method. In this case the passive elements are
and loudspeakers containing multiple drivers can also simply treated as ports.

THE AUTHORS

C. J. Struck S.F. Temme

Christopher J. Struck was born in Milwaukee, WI, Acoustical Society of America. He is on the IEEE Sub-
in 1962. He is a graduate of the Electrical Engineering committee on Telephone Instrument Testing and the
School of the University of Wisconsin-Madison. From AES SC-4-3 Working Group on Loudspeaker Modeling
1983 to 1985, he worked in the UW Electroacoustics and Measurement. He is a guitarist and composer and
Lab, developing circuitry for transducer measuremenis, enjoys both live and recorded music. Mr. Struck lives
sound intensity research, and active noise control appli- in San Francisco.
cations. During this time, he also studied electronic mu-
sic composition and musique concrete, producing multi- ·
media concerts and works for modern dance. In 1985, Steve Temme was born in Boston, MA, in 1962. From
Mr. Struck attended Stanford University, where he stud- 1980 to 1985, he attended Tufts University, majoring
ied digital signal processing and participated in the Stan- in mechanical engineering. During this time, he also ran
ford Jazz Workshop. .a successful business selling stereo equipment, which
In 1986, he joined Briiel & Kja:r. From 1988 through helped finance his education. He received the B.S.M.E.
1993, he lived in Denmark, working as an application · from Tufts in 1985.
specialist in electroacoustics. There, he helped to de- He was first employed by Bruel & Kj_er Instruments in
velop new measurement techniques and new instrumen- Marlborough, MA, in October of 1985 as a sales engineer
tation for testing electroacoustic transducers, most nota- for the northeastern region of the United States. In 1988,
bly the Type 4128 head and torso simulator and the Type Mr. Temme joined Apogee Acoustics in Randolph, MA,
20!2 audio analyzer· During the course of his work, he as a loudspeaker design engineer. There, he worked on
has traveled extensively throughout the world providing the development of the Stage full-range, dipole, ribbon
training and lecturing on the topics of electroacoustics, loudspeaker and on the DAX active crossover system. In
telephonometry, linear and nonlinear system analysis, 1990 he moved to Denmark where he was employed by
and measurement techniques· Currently, he is a princi- Briiel & Kj_er as an audio application specialist. There,
pal consultant with Charles M. Salter Associates in San he worked with the Type 2012 audio analyzer, training
Francisco, California. engineers, giving seminars, and developing loudspeaker
Mr. Struck is the author of many technical papers, and microphone measurement programs.
application notes and articles, several of these coau- Mr. Temme currently works as an independent con-
thored with his friend and colleague Steve Temme. He sultant in audio and electroacoustics in Boston. He is a
is a member of the Audio Engineering Society, the Insti- member of the AES and an avid concert-goer and hi-
tute of Electrical and Electronics Engineers, and the fi enthusiast.

482 J. AudioEng.Soc.,Vol. 42, No.6, 1994June