MODAL SPACE - IN OUR OWN LITTLE WORLD by Pete Avitabile

When impact testing, cause any problems?

can the use of the Let's discuss this
exponential window

Illustration by Mike Avitabile

When impact testing, can the use of the exponential window cause any problems?
Let's discuss this

The exponential window can cause some problems if not used response would decay to a reasonably small value by the end of
properly. If an excessive amount of damping is needed to the sample interval thereby minimizing the effects of leakage.
minimize the effects of leakage, then you run the risk of missing The impact excitation, windowed exponential response and the
closely spaced modes. There are a few examples to show the FRF are shown in Figure 1. On the surface, this measurement
relative to the use of the window and what can happen if care is looks acceptable. [Note that the input spectrum (not shown)
not exercised in using the exponential window. was reasonably flat over the entire frequency range thereby
allowing sufficient excitation of the structure. Also note that
First of all, let me clearly state that in many impact testing the coherence (not shown) was also considered very
situations the use of an exponential window is necessary. acceptable.]
However, before any window is applied, it is advisable to try
alternate approaches to minimize the leakage in the
measurement. Increasing the number of spectral lines or 3.5
V
TIME PULSE
halving the bandwidth are two things that should always be
investigated prior to using a damping window. Both of these Real

items will essentially increase the total time for the collected
data. This can often help by allowing the response of the
system to naturally decay before the end of the sample period. -1.5
V
If this can be accomplished, then the use of the exponential -25.391ms 973.633ms

window may not be necessary. 2.5

V
WINDOWED RESPONSE

However, if the response still does not decay by the end of the Real

sample period, then an exponential window may be necessary.

The use of the window should not be employed until these first
two items (mentioned above) are checked as possible ways to -2.5
V
minimize the leakage problem. The arbitrary use of the -25.391ms 973.633ms

exponential window without first looking at the time response is 50

not recommended as the first step in the measurement process.
Let's look at this through the use of a simple example. dB Mag

A very simple, lightly damped structure was subjected to an

impact test. The signal processing parameters were selected for FREQUENCY RESPONSE FUNCTION
-50
a 400 Hz bandwidth which resulted in a 1.0 second time
0Hz 400Hz
window. Since the structure was expected to have a response
that would not decay by the end of the sample interval, an Figure 1 - FRF with slightly too much damping
exponential window was applied such that the windowed
_________________________________________________________________________________________________________________________
When impact testing, can the use of an exponential window cause any problems ? Copyright 2001
SEM Experimental Techniques - April 2001 Page 1 Pete Avitabile
From all aspects, this measurement appears very acceptable.
But we need to look at this measurement in more depth. First, 2.5
V RAW TIME RESPONSE
let's consider the same measurement but add significantly more
damping to the response signal. Figure 2 shows the same data Real

but with a significantly larger value of damping used for the

exponential window. The FRF that results from the impact
measurement of this signal clearly has significantly more -2.5
V
damping than that shown in the FRF of Figure 1. The peaks of -25.3906ms 1.9736s

the FRF show this effect; notice that the peaks are much wider 2.5

due to the excessive use of the damping window. V

WINDOWED RESPONSE
Real

2.5
V RAW TIME RESPONSE
-2.5
Real V
-25.3906ms 1.9736s

50

-2.5
V
-25.391ms 973.633ms dB Mag

1.2
V
WINDOWED RESPONSE
-50 FREQUENCY RESPONSE FUNCTION
Real
0Hz 400Hz

TWO CLOSELY SPACED MODES

-800
mV
-25.391ms 973.633ms

25
Figure 3 - FRF with increased time/spectral resolution

dB Mag

The most important item to notice in Figure 3 is that what

appeared to be one mode at the first peak in the FRF actually
FREQUENCY RESPONSE FUNCTION
-25 turns out to be two very closely spaced modes of the structure.
0Hz 400Hz The use of the damping window in Figure 1 and 2 resulted in an
FRF that appeared to have only one mode at the first peak in the
Figure 2 - FRF with too much damping FRF. The use of the damping window caused these two distinct
modes to appear as only one peak in the FRF.

Now let's look yet a little deeper into this measurement and try While the damping window was necessary to minimize the
some alternate signal processing parameters. In order to leakage, the window distorted the actual FRF in Figure 1 and 2
minimize the use of the damping window, either the bandwidth such that it was difficult to observe that two peaks existed at this
can be shortened or the number of spectral/time lines of frequency. The use of the exponential window, while necessary
resolution can be increased. Both of these changes result in an for digital signal processing considerations, can cause some
increase in the total time necessary to collect the sample of data. significant difficulties when evaluating structures with light
If the total time is increased, then there is less need for a damping and closely spaced modes as seem in this example.
significant amount of damping window to be applied to the
collected time data. Now, I hope you can see some of the effects of the exponential
window in this example. While an exponential window may be
Figure 3 shows a doubling of the number of spectral/time lines necessary to minimize the effects of leakage, the use of the
of resolution. The time sample was increased from 1.0 second window may also hide or distort the modes in the measurement.
to 2.0 seconds. While an exponential window was still It is extremely important to be very careful when using the
necessary to minimize leakage, the overall damping effect that exponential window when performing an impact test. If you
was added to the measurement is far less than that used for the have any other questions about modal analysis, just ask me.
measurements shown in Figure 1 and Figure 2 above.

_________________________________________________________________________________________________________________________
When impact testing, can the use of an exponential window cause any problems ? Copyright 2001
SEM Experimental Techniques - April 2001 Page 2 Pete Avitabile