An interpretation of the result shown in fig. 4a can be that
a test of the noise reduction system with a test signal based
on discrete-interval binary signals may be more sensitive to
changes in noise sensitivity in the noise reduction system under
test than to changes of its filter slopes. This is consistent
with an observation from [1], where experiments with noise
reduction systems of different modulation sensitivity were reported
to show that discrete-interval binary sequences produced
side effects for systems with high noise sensitivity only,
when used as a basis for test signal design.
The result shown in fig. 4b could in turn indicate that test
signals based on multi-sines may not be suited if the filter
slope of bandsplit filters in the noise reduction system under
test is expected to vary in future products. A possible explanation
for the dependency of quality of multi-sine based tests
on filter slopes is given in [1], where results of experiments
with different multi-sine-based test signals are explained by
non-idealities of band split filters in the system under test.
As a result, one would thus choose signals based on
discrete-interval binary sequences for testing noise reduction
systems with low sensitivity to unmodulated noise and
a high likelihood that the slopes of bandsplit filters rather
than the noise sensitivity would change during the further
improvement of the system. On the contrary, one would
choose signals based on multi-sines for systems with a high
likelihood of changes regarding the noise sensitivity. In many
cases, one would have to look for a third test signal candidate