| Application
note # 4 |
Sample
Champion
|
How
detecting distortion
in MLS measurements
The
MLS based method has the advantage to spread the measured distortion
and non-linearities throughout the computed impulse response.
Windowing a small portion of the calculated data to perform
a FFT, only a small part of the measured distortion will appear
in the frequency domain data.
Every MLS sequence has its own characteristic
way to spread distortions. Here below some examples of distortion
effects are reported. These measurements have been performed
using the "loop back" configuration (audio device
input directly connected
to the output), in order
to emphasize this effect, since the "anechoic" impulse
is similar to a Dirac impulse. In this test the Sample Champion
latency time value has been manually "uncalibrated"
to show the impulse starting at a time t>0.
The
distortion effects in these examples are obtained by setting
the input level too high and causing an A/D overload (in Sample
Champion the peak meters on the control bar become red).
Figure
1 shows the impulse, measured using a MLS sequence of 16K points,
obtained with the shift-register taps 14, 13, 12, 2, with correct
input level. The response is perfectly clean and clear.
Figure
1
To
see the situation in depth, it is useful to zoom the amplitude
axis by a factor of 128 (see the yellow text on figure 2).
Figure
2
Figure
3 shows the first example of distortion, obtained by increasing
the input level in the mixer of the audio device. Note that
the peak value is now 10.3 Volts. The A/D converter was severely
overloaded. A lot of small peaks now appear.
Figure
3
Figure
4 shows a measurement obtained with the same input level as
before, but using a different MLS sequence (with the same length,
taps 14, 12, 11, 1). Note that the small peaks are located at
different time positions.
Figure
4
Figure
5 plots another measurement obtained with the same input level
as before, using another different MLS sequence (with the same
length, taps 14, 12, 10, 9, 7, 5, 3, 1). The small peaks are
located at different time positions, compared to previous two
measurements. The good news is that these small peaks are not
randomly distribuited, but are always at the same locations
with the same MLS sequence.
Figure
5
It
is useful zooming the Y axis to see these distortions; in fact
following figure 6 shows the same impulse of figure 5, but without
any Y zoom: apparently there is no distortion at all.
Figure
6
Following
figures 7 and 8 show a detail (1:1 time scale) of the time zone
around the impulse peak. Figure 7 is a detail of figure 5 (distortion).
Figure 8 is a detail of figure 2 (clean).
Figure
7
Figure
8
 Download
PDF file (440K) Shift-click
to download
Back to Application Notes
|
