Designing for Speech Intelligibility 1407The first commercially available instrument that could
measure STI was the RaSTI meter introduced by Brüel
and Kjaer in 1985. Modulated pink noise in the 500 Hz
and 2 kHz octaves is generated and transmitted either
acoustically or electronically into the system under test.
A useful feature of the method is that the resultant signal
is very speechlike in nature having a crest factor of
around 12 dB, which compares well with normal speech
(at around 15–20 dB). The relative levels of the 500 Hz
and 2 kHz signals are also automatically transmitted in
their correct ratios as compared to natural speech.
This makes setting up the test signal levels compara-
tively straightforward and enables measurements to be
carried out by trained but nonexpert personnel. The
introduction and adoption of the RaSTI test method has
literally revolutionized the performance of many PA
systems ranging from aircraft cabins and flight record-
ing systems to trains, malls, and cathedrals as, for the
first time, the intelligibility of such systems could be set
and readily verified. As the limitations of RaSTI as a
measure of sound system performance became more
widely known and understood (e.g., Mapp 2002 and
2004) it became clear that a replacement method would
be required. In 2001, STIPA (STI for PA systems) was
introduced and became part of IEC 268-16 in 2003.
Unlike RaSTI STIPA measures over virtually the com-
plete speech bandwidth range from 125 Hz to 8 kHz.
However, a sparse matrix is used to cut down the com-
plexity of the stimulus and associated measurement pro-
cessing time. Table 36-3 shows the modulation matrix
for STIPA.
The current version of IEC 268-16 (Edition 3, 2003)
employs the above modulations. It can be seen that for
some reason the 125 Hz and 250 Hz carriers employ the
same modulation frequencies—although there is a spare
set available (there are no modulations at 1.6 Hz and
8 Hz). However, it is quite likely that a future version of
the standard (2010) will correct this apparent anomaly
and use the missing modulations (indeed some meters
already do this).
Since its introduction, the STIPA technique has rap-
idly taken off with at least four manufacturers offering
handheld portable measurement devices (though some
are more accurate than others—see Mapp 2005). In a
similar manner to RaSTI, the STIPA signal is speech
shaped and so automatically presents the correct signal
for assessing a sound system. (STIPA is a protected
name and refers to a measurement made with a modu-
lated signal. Although STIPA can be derived from an
impulse response, any such measurement must be
clearly indicated as being an equivalent STIPA). At the
time of writing, the STIPA signal has been relatively
loosely defined, but it is understood that the fourth edi-
tion of IEC 268-16 will clarify the issue. This should
also help insure that the various STIPA meters are fully
compatible, so that any 2 meters using the same
IEC268-16 Ed 4 test signal should give the same result.
The typical time required to carry out a single STIPA
measurement is around 12–15 s. However, as the test sig-
nal is based on a pseudorandom signal, there can be
some natural variation between readings. For this reason
it is recommended that at least three readings be taken to
insure that a reliable measurement result is produced.
STIPA correlates very closely with STI and overcomesFigure 36-35. Simulated system frequency response curve
favoring 500 and 2000 Hz, giving an excellent RASTI value
but very poor sound quality.Frequency–Hz, Log100 1k 10k9080706050TDS magnitude–dBTable 36-3. STIPA Modulation Matrix
Carrier/Modulation
Frequency (Hz)125 250 500 1K 2K 4K 8K0.63 X
0.80 X
1.0 X X
1.25 X
1.6?
2.0 X
2.5 X
3.15 X
4.0 X
5.0 X X
6.25 X
8.0?
10.0 X
12.5 X