Exam Questions Example

Questions
Q1
The signal below is a seismograph drawn by a pen on a drum
during an earthquake
Which kind of signal is it: analog or digital?
Is it continuous or given at discrete times only?
If you want to classify it according to the time domain
classification, to which family does it belong to and why?
Q1
The signal below is a seismograph drawn by a pen on a drum
during an earthquake
Which kind of signal is it: analog or digital? Analog, because it is directly written
without analog to digital conversion
Is it continuous or given at discrete times only? Continuous, since there is no sampling
If you want to classify it according to the time domain
classification, to which family does it belong to and why?
It is a random signal, in case of no earthquake, it is transient in case of earthquake
Q2
The spectrum of a signal is continous: what can you say about the
signal which has generated it?
Do you think that such a spectrum can be limited to a maximum

frequency and why?
Q2
The spectrum of a signal is continous: what can you say about the
signal which has generated it? It was recorded for an infinite amount of time
Do you think that such a spectrum can be limited to a maximum

frequency and why? If it is sampled, the frequency is limited to the Nyquist
frequency. If it is continuous in time domain, it is not
limited
Q3
Which is/are the parameters you can use to detect a peak in a signal?
Q3
Which is/are the parameters you can use to detect a peak in a signal?
Crest factor
Kurtosis
RMS could be used but it is ambiguous, because the RMS could be influenced by fluctuations of the signal
Q5
Guess for the skewness in both cases?

Which one has the highest kurtosis?
Which one has the highest RMS?
Q5
Positive, for the left plot; very close to zero for the right plot, since the signal is
Guess for the skewness in both cases? symmetric with respect to average
Which one has the highest kurtosis? They have the same kurtosis (a very high value), given by peaks that influence
Which one has the highest RMS? kurtosis in the same way, in case of positive or negative values with respect to
mean
Very low change in the RMS since the mean change slightly between the two
signals but STD is the same (left plot higher RMS)
Q6
What does it happen if I convolve the sine by
the impulse given at time t=0?
What’s the mathematical expression of the
impulse at time t = 0?
Q6
What does it happen if I convolve the sine by
the impulse with unitary height given at time
t=0?
What’s the mathematical expression of the
impulse at time t = 0?
The result is the same function, since if the

impulse with unitary height is shifted it simply
replicate the sine function at the time shifted
Mathematical expression of the impulse:
1
𝛿 𝑡 = ቐ 𝜏 𝑓𝑜𝑟 𝜏 → 0, 𝑓𝑜𝑟 𝑡 = 0
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Q7
Given the signal aside
• Plot the complete (2 sided) spectrum
• Which is the best window to be adopted for
this signal?
Q7
Given the signal aside
• Plot the complete (2 sided) spectrum
• Which is the best window to be adopted for
this signal? Rectangular since there is no leakage
1 A
0.5
f
-3 0 3
df = 0.5 Hz
Q8
T
In applying the DFT I want to replicate the signal given in the figure to
make it periodic. How can I get this?
Q8
T
In applying the DFT I want to replicate the signal given in the figure to
make it periodic. How can I get this?
It can be obtained by convolving the function by a train of pulses, spaced of T
Q9
Can you provide the spectrum of this
signal?
-2T -T 0 T 2T
Q9
Can you provide the spectrum of this
signal?
The spectrum is given by a train of pulses spaced of
1/T
-2T -T 0 T 2T
Q10
I have a signal made of a sine wave at 5 Hz with Gaussian noise
superimposed. I calculate both the square root of the power spectrum
and the spectrum (same measurement units). I keep on averaging the
signal for both the calculation of the power spectrum and the spectrum
(vector averaging). Which is the best choice?
Q10
I have a signal made of a sine wave at 5 Hz with Gaussian noise
superimposed. I calculate both the square root of the power spectrum
and the spectrum (same measurement units). I keep on averaging the
signal for both the calculation of the power spectrum and the spectrum
(vector averaging). Which is the best choice?
The best choice is averaging on power spectrum, since the spectrum contains the phase information and time
intervals may not start with the function on the same phase.
If I use a trigger to divide the signal into different time intervals, they start all with the same phase and it is
better to average on the spectrum, since the noise is canceled
Q11
Which among the following frequency domain defined functions allow
you to get the Inverse Fourier Transform to get back to the original time
record?
PSD
Autopower
Spectrum
Q11
Which among the following frequency domain defined functions allow
you to get the Inverse Fourier Transform to get back to the original time
record?
PSD
Autopower
Spectrum
Only the spectrum since it contains the phase information
Question 20 4pt I have the frequency domain representation of a sine +
noise, after averaging
• I can say that the red plot and the black plot have
different window lengths, though they refer to the
same process YES, because they have different frequency resolution
• They represent square root power spectra as the
noise floor is the same and the sine amplitude is
different NO, because the line at 6 Hz is deterministic and it
should be the same amplitude for power spectra
• They represent PSDs as the noise floor is the same
NO, because they could represent the
same phenomenon in case of PSD
independent from the frequency resolution; the sine
amplitude is not constant in the two records
• They represent PSDs as the noise floor is the same
YES, because for deterministic signals independent from the frequency resolution; the sine
the sine amplitude is reduced by df amplitude is biased by the definition of PSD
Question 26 YES, because otherwise signals would not be in phase but
Regarding time-averaging: probably its frequency would be out of phase
• It can only be applied on periodic signals to reduce noise;

NO, the time averaging would
• It can be applied on random signals to reduce noise; go to zero
• It can be applied only in case of synchronous sampling; NO, it can be applied also with
a trigger
• The autocorrelation can be applied on a signal to determine its period
and to perform time averaging on those periods; YES, it can be applied to get the main
frequency component of the signal to
average within periods of that
frequency component
• It can be applied only if for each period I have the same number of
samples. NO, it can be applied also for periods
with different number of sample, in
case I resample signals to perform
averaging
Question 25
Consider the transfer functions (H1 and H2) and
coherence plot in the figures on the left computed
for different frequencies resolutions:
• In the point A (7.9 Hz), the coherence drop is
caused by noise on the output;
• In the point A (7.9 Hz), the coherence drop is
caused by leakage;
YES, because for resonance peaks leakage causes
coherence drops, much higher than for other
frequencies
• In point B (11.1 Hz), the coherence drop is caused
by leakage;
• In point B (11.1 Hz), the coherence drop is caused
by noise on the output;
• In point B (11.1 Hz), the reason of the coherence
drop cannot be determined by looking at the
B graphs, but it is not caused by leakage; YES
YES, because if I reduce df, there is a reduction of
A coherence and this cannot be caused by leakage
Question 19 I have a random process whose averaged spectrum amplitude is
given by the black line. The red line represents the same
averaged process, with the application of a low pass filter (cutoff
frequency 100 Hz). Which of the following are OK:
• The RMS is the same for the filtered and the unfiltered signal
• The RMS is lower in the case of the filtered signal YES
• The RMS cannot be derived from the frequency domain
representation
• The representation is given in terms of autopower linear (i.e.
square root of): if I averaged the vectors as complex
RMS can be computed the root sum quantities I would have had almost zero in any case
square of all the spectral lines: YES
𝑁 • In this case the PSD would have been a better representation
𝐴2𝑖
𝑅𝑀𝑆 = ෍ of the situation YES
2
𝑖=0 In case I average on complex quantities (spectra), I would have a
In case the amplitude of spectral spectrum going to zero, so the representation given in figure is not
lines is reduced by a filter, the RMS is a spectrum
reduced
PSD is a better representation in case of random signals since it is
independent from frequency resolution

Exam Questions Example

Uploaded by

Copyright:

Available Formats

Exam Questions Example

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Exam Questions Example

Uploaded by

Copyright:

Available Formats

Questions

Do you think that such a spectrum can be limited to a maximum

Do you think that such a spectrum can be limited to a maximum

Guess for the skewness in both cases?

The result is the same function, since if the

Mathematical expression of the impulse:

• It can only be applied on periodic signals to reduce noise;

You might also like