The influence of several sources of error on room acoustical measureme
nts have been investigated in a collaboration between the Physikalisch
-Technische Bundesanstalt (PTB) in Braunschweig, Germany and the Norge
s Tekniske Hogskole (NTH) in Trondheim, Norway. The measurement method
is the technique using maximum-length sequences (MLS). At first, basi
c impulse response measurements of a one-third octave band filter and
of a loudspeaker using quite different MLS-systems are compared and th
e differences in the results proved to be negligible. Then algorithms
for the determination of room acoustical parameters are introduced and
the possible influences on the results are discussed. Finally, interc
omparison measurements of room acoustical parameters are described, wh
ich were performed in the PTB auditorium by participants from seven in
stitutes all using their own equipment. The algorithms for the determi
nation of room acoustical parameters used by different analyzers intro
duce systematic differences caused by differences in time-windowing an
d filtering, in reverse-time integration and in noise compensation. Th
e overall standard deviations of the room acoustical parameters measur
ed in the PTB auditorium by the different teams are about 5 to 10% for
T-30, EDT (Early Decay Time), D (Definition) and TS (Centre time), an
d approximately 0.5 dB for C (Clarity) and G (Strength), all measured
in the 1 and 4 kHz octave bands. Larger uncertainties were found in th
e 125 Hz octave band, and for the parameter LF (Lateral Fraction) in a
ll frequency bands. The results of this investigation show that if mea
surements of room acoustical parameters are performed according to ISO
/DIS 3382, the overall uncertainty is of the same magnitude or a littl
e higher than subjectively perceivable changes in these parameters. Ho
wever, the draft standard allows various procedures to be applied in t
he processing of impulse responses. To reduce the overall uncertainty
the standard should, in future revisions, be more specific in the resp
ect. This also applies to the specifications of the omnidirectional so
urce.