
Echo Speech Source Type 4270 – User Manual10
9) The estimated reverberation time (RT) is 2 s, so the product, SNR*RT = 40 dB. Because
this is below 120 dB, Intermittent Noise (5) is the proper stimulus type to be selected
according to the rule of thumb described in section 4.3.
10) In DIRAC start the measurement. DIRAC automatically detects the Echo Speech Source
settings.
NOTE: Do not make loud noises until DIRAC finishes recording and starts processing the
recording.
11) Save the resulting impulse response file, containing the auditorium impulse response in
channel 1 and the background noise in channel 2.
From this file, speech intelligibility related parameters are obtained in DIRAC through the
Parameters menu or the Graph/Statistics button.
12) Repeat steps 7 and 8 for the other receiver and source positions of interest.
Positions close to the source may require another source level (1) or (2) than positions at a
greater distance. DIRAC supports statistical analysis over multiple measurements.
4.2 High Background Noise
If the SNR is substantially lower than 15 dB at the reference speaker level, the resulting
impulse response in channel 1 may become very noisy, and the derived parameters less
reliable. This can be improved by pressing the Raised Level Speaker button (1) on Echo
Speech Source Type 4720 or by increasing the Pre-Average value in DIRAC.
On the other hand, if the SNR exceeds 15 dB, raising the level may not in all cases lead to a
further reduction of the noise in the impulse response because the inherently higher distortion
at the raised level will also add some noise that might dominate a low background noise level.
The obtained relative noise level in a measured impulse response can be judged through the
Impulse response-to-Noise Ratio (INR), which can be viewed in DIRAC as a Level parameter
in the Parameter menu or the Graph window. The INR should preferably exceed 20 dB in each
octave frequency band from 125 Hz through 8 kHz.
4.3 Intermittent versus Continuous Noise
The use of an intermittent stimulus enables DIRAC to determine the noise level separately
from the impulse response measurement, thereby minimising mutual influences and
maximising derived parameter accuracy. This is particularly true if, in the silent part of the
stimulus, the level of the reverberation tail is negligible with respect to the background noise
level, which translates into the following rule of thumb:
• SNR*RT < 120 dB – select Intermittent Noise
• SNR*RT > 120 dB – select Continuous Noise
In DIRAC an Intermittent Noise measurement produces a dual-channel impulse response file
containing the impulse response (plus noise) in channel 1 and the background noise in
channel 2.