MODAL SPACE - IN OUR OWN LITTLE WORLD by Pete Avitabile

My coherence is better Am I doing

There are definitely some
some measurements something
issues to discuss
others when impact

Illustration by Mike Avitabile

My coherence is better in some measurements than others when impact testing. Am I doing something wrong?
There are definitely some issues to discuss here.

OK – so this yet another area of measurement quality that needs parameters are selected such that the input force and response
to be discussed. Impact testing is by far the most common and acceleration are totally observed signals within one sample
most popular of the approaches for obtaining frequency record of data. This eliminates the need for any window
response functions for the description of a structural system. functions on the input or output signals measured.
The impact test is a very economical approach for frequency
response testing. In addition, impact testing is very easy to The measured frequency response (lower trace) and coherence
setup and is extremely portable for field testing. Due to the ease (upper trace) are shown in Figure 1. Notice that the frequency
with which measurements can be made, impact testing is widely response function appears to be a very good measurement and
used in many industries and applications. the coherence is very good for this measurement. The
coherence for most of the frequency range is extremely close to
But there are a wide range of issues that need to be recognized one. The coherence has a slight dip in antiresonant regions but
when performing impact testing. Some of these relate to double is not a problem for this measurement. (Note that drops in the
impacts, pre-trigger delay, high peak voltages compared to coherence in antiresonant regions are expected due to the fact
overall RMS level of the signal, nonlinear systems, etc. Some that the structure has no response at these frequencies and
of these are commonly cited “areas of concern” when impact therefore the response of the system is not coherently related to
testing. These often become the stated reasons why impact test the measured input signal.)
results may have coherence values that are not as acceptable as
may be desired. But these may not be the only reasons – one
very important consideration that I would like to discuss in this
article is the effect of impact location on the resulting frequency
response function and its coherence.

When performing impact testing, the input impact location can

have a very significant effect on the resulting frequency
response function. And this can be seen in the coherence
function measured for each set of averaged data. First, let’s
take a set of measurements where care is exercised in the impact
location during the test to show a very good high quality
measurement. Then some “less than perfect” impact
measurements will be made on the same structure to show the
degradation of the coherence.

The structure is a very simple structure with what are expected

to be some very good measurements. A typical impact
measurement is going to be made for the frequency response
function at the drive point on the structure. Sampling Figure 1 – FRF & Coherence for a Well Controlled Impact
_________________________________________________________________________________________________________________________
My coherence is better in some measurements than others when impact testing. Am I doing something wrong? Copyright 2008
SEM Experimental Techniques – June 2008 Page 1 Pete Avitabile
Now in that first measurement, extreme care was exercised to coherence is also degraded. There is also a lack of consistency
assure that each average was the result of an impact at the same in the antiresonant regions for this measurement.
location in the same direction. This is a very important concern
when impact testing.

To illustrate what happens where this care is not exercised, a

measurement is made where each average is intentionally made
within a region that is very close to the desired input location
but there is some slight variation in the actual input location.
With the same number of averages, the frequency response
(lower trace) and coherence (upper trace) are shown in Figure 2
for this measurement where there is some variation for each
impact location. While the frequency response function looks
reasonable, the coherence is seen to have some significant
degradation across the entire frequency range. While the
coherence is acceptable in the immediate region of the peaks of
the frequency response function, overall the coherence is poor.

Figure 3 – FRF & Coherence for a Skewed Impact

For both of the cases shown in Figure 2 and 3, the coherence is

not nearly as good as the measurement shown in Figure 1. This
is due to the inconsistency of the impact location – whether it be
not impacting the same location for each measurement or for
not maintaining a consistent strike angle for each measurement.
Both cases clearly show a degradation of the measurement
coherence. A very well controlled, precise impact excitation
needs to be maintained for each average that makes up the
complete measurement.

These cases are presented here because this is a very common

problem during impact testing. This is especially true when the
test lasts for a long period of time for measuring many
Figure 2 – FRF & Coherence for a Poorly Controlled Impact locations. Generally, as time goes on it is very easy to become
bored and not maintain the consistent impact during the entire
The coherence is most significantly affected in the antiresonant test. This is also very common when the impact locations are at
regions of the frequency response function. This is due to the inconvenient locations around, on top or underneath the test
fact that while resonances are global characteristics of a system, structure. When climbing all around the structure (and often in
the antiresonant regions are absolutely not global in character at very unnatural positions), it is very easy to not maintain a
all. The antiresonant regions are highly dependent on the consistent impact for all averages making up a measurement.
particular input-output measurement location. Because care
was not exercised during the impact test to assure than all So as a word of caution when impact testing… be very sure to
impacts were made at the same location, the antiresonant region impact the same point, in the same direction, for each of the
changes for each input output measurement that makes up the averages that make up the frequency response function to assure
total average for the measurement. Therefore, from one that an overall acceptable coherence is obtained for all
measurement location to the next there is no consistency in the measurements.
measurement and therefore the coherence reflects this.
I hope that this clears up the concerns about possible coherence
One additional set of averages was made where the impact point degradation when impact testing. While there are many more
was kept the same but the angle of the impact excitation was items that could affect coherence, this is one that has an effect.
allowed to vary during each of the averages. The frequency If you have any more questions on modal analysis, just ask me.
response (lower trace) and coherence (upper trace) are shown in
Figure 3 for this measurement. Similar to the previous case, the
_________________________________________________________________________________________________________________________
My coherence is better in some measurements than others when impact testing. Am I doing something wrong? Copyright 2008
SEM Experimental Techniques – June 2008 Page 2 Pete Avitabile