Handbook for Sound Engineers

1406 Chapter 36

has to be reasonably linear and exhibit a well-extended
frequency response. Unfortunately many paging and
voice alarm systems do not fulfill these criteria and so
can give rise to readings of questionable accuracy.
(However, this still takes account of a wider frequency
range than the traditional D/R and %Alcons methods.)
Fig. 36-35 shows a system response simulated by the
author and evaluated via RaSTI. Although the majority
of the speech spectrum is completely missing, the result
was an almost perfect score of 0.99 STI!

Figure 36-33. Principle of STI and RASTI showing octave band spectrum and speech modulation frequencies.

1 s intervals

Intensity

A. An example of the intensity envelope

of a segment of human speech.

* * * * * * * * *

Amplitude

1 s

B. RASTI signal modulation (as applied

in the 2 kHz octave).

60

50

40

60

50

40

125 250 500 1k 2k 4 k 8k

dB RASTI

Octave

level

1.0

0.8

0.6

0.4

0.2

(^0) 0.25 0.5 1 2 4 8 16 Hz
Modulation index
(rms within 1/3
octave bands)
Octave center frequency
A long-term averaged octave spectrum
of normal human speech, at 1 m
distance (Leq.,A = 60 dB). The shaded
portions indicate the carrier signal used
in the RASTI method.
Curve showing the modulation spectrum
of human speech. The discrete modulation
frequencies used in the RASTI method are
marked " * ". Four modulation frequencies
are applied in the 500 Hz octave and five
in the 2 kHz octave as follows:
500 Hz octave: 1 Hz, 2 Hz, 4 Hz, 8 Hz
2 kHz octave: 0.7 Hz, 1.4 Hz, 2.8 Hz,
5.6 Hz, 11.2 Hz.
C. D.
Figure 36-34. STI subjective scale and comparison with
%Alcons.
15 10 5
Equivalent
%Alcons
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Bad Poor Fair Good Excellent
STI/RaSTI scale
Table 36-2. STI Modulation Matrix
Carrier/Modulation
Frequency (Hz)
125 250 500 1K 2K 4K 8K
0.63 XXXXXXX
0.80 XXXXXXX
1.0 XXXXXXX
1.25 XXXXXXX
1.6 XXXXXXX
2.0 XXXXXXX
2.5 XXXXXXX
3.15 XXXXXXX
4.0 XXXXXXX
5.0 XXXXXXX
6.3 XXXXXXX
8.0 XXXXXXX
10.0 XXXXXXX
12.5 XXXXXXX