Handbook for Sound Engineers

Loudspeakers 639

frequencies will be integer multiples of this frequency.

In order to have infinitesimally small frequency resolu-

tion (i.e., perfectly resolved frequency data), a test

would have to be conducted for an infinite amount of

time. It follows that all realizable response tests have a

limit on their frequency resolution.

The effect of frequency resolution on a transfer func-

tion measurement is to smooth the appearance of a plot

of the results, thereby possibly obscuring some of the

details of the transfer function. This smoothing is

present to some degree in all transfer function measure-

ments. In the case of electronic devices, transfer func-

tions are typically well behaved enough that the

frequency resolution of a response test does not cause

meaningful loss of detail. With loudspeakers, the oppo-

site is often true: a loudspeaker’s transfer function often

has so much fine structure that a practical test will

noticeably smooth out the peaks and dips in the

speaker’s response. The degree to which this fine struc-

ture is audibly significant is a matter of some contro-

versy. As a result, there is no widespread agreement in

the industry on the minimum desirable frequency reso-

lution in testing loudspeakers.

17.11.2 Chart Recorders

Prior to the advent of computer-based measurement
systems, the most commonly employed loudspeaker
measurement instrumentation comprised a strip chart
recorder and a signal sweep generator. The two devices
are synchronized such that, for a given frequency in the
sweep, the pen on the recorder is in the appropriate x
(frequency) position on preprinted graph paper. The
pen’s y position would correspond to the amplitude of
the signal received from the test microphone, and there-
fore, hopefully, to the amplitude response of the speaker
at that frequency.

If the y amplifier is logarithmic, then the amplitude
will be expressed in decibels. As common as the
strip-chart measurement technique was prior to the
1980s, it had several prominent disadvantages:

1. There is no means of measuring a loudspeaker’s
   phase response with this technique.


2. The measurement is incapable of discriminating
   between direct sound from the device under test and
   sound that is reflected from surfaces in the test envi-
   ronment. This necessitated the construction of very
   costly anechoic chambers. Even in such a chamber,
   the inclusion of some reflected sound in a strip-chart
   type measurement is unavoidable.
   3. The measurement technique does not isolate the
   linear portion of a loudspeaker’s transfer function.
   Distortion products are simply added to the ampli-
   tude of the loudspeaker’s transfer function at the
   fundamental frequency that excites them.
   4. There is no direct, accurate way to determine or
   control the frequency resolution of a strip chart
   measurement. Reducing pen speed and/or increasing
   chart (and sweep) speed have the effect of reducing
   frequency resolution, or smoothing, the data, but the
   degree to which this has taken place is not always
   apparent.
   5. Data from this form of measurement is only gener-
   ated in hard copy format.
   6. This measurement technique provides no ready
   means to compensate for propagation delay: the
   time required for sound to travel from the loud-
   speaker to the test microphone.
   7. Since there is no means for distinguishing between
   the output signal from the loudspeaker and ambient
   noise, the test environment must be quiet.

17.11.3 Real Time Analyzers

Although initially developed to measure the response of

sound systems in their operating environments, real-

time analysis has also been used to measure loudspeaker

response in controlled environments. With this testing

technique, a pink noise signal is applied to the loud-

speaker. Pink noise is a random signal that contains

equal energy for each unit of logarithmic

frequency—e.g., for each octave or fraction thereof.

The signal from the test microphone is applied to a

series of bandpass filters of constant percent-octave

bandwidth, each of which is tuned to a different band

center, and the averaged output level of each filter is

displayed, either on a CRT, LCD, or LED display. The

display represents, within the limitations due to the

measurement technique and the test environment, the

amplitude response of the loudspeaker. Because the

frequency content of a random signal has small fluctua-

tions over time, the display may be averaged, or inte-

grated, to produce a stable graph. Real time analysis

suffers from the same general disadvantages as the chart

recorder method of measurement.

17.11.4 Time-Windowed Measurements

The development of inexpensive computers has liter-

ally revolutionized the field of acoustic instrumentation.

This is largely the result of the computer’s ability to

process and store large amounts of signal data. With a