Frequency response estimation uses either sinestream or chirp input signals.
Sinusoidal Signal  When to Use 

Sinestream 
Recommended for most situations. Especially useful when:

Chirp 
Useful when:

A sinestream signal consists of several adjacent sine waves of varying frequencies. Each frequency excites the system for a period of time.
Frequency response estimation using frestimate
performs
the following operations on a sinestream input signal:
Injects the sinestream input signal you design, u_{est}(t), at the linearization input point.
Simulates the output at the linearization output point.
frestimate
adds the
signal you design to existing Simulink^{®} signals at the linearization
input point.
Discards the SettlingPeriods
portion
of the output (and the corresponding input) at each frequency.
The
simulated output at each frequency has a transient portion and steady
state portion. SettlingPeriods
corresponds to the
transient components of the output and input signals. The periods
following SettlingPeriods
are considered to be
at steady state.
Filters the remaining portion of the output and the corresponding input signals at each input frequency using a bandpass filter.
When a model is not at steady state, the response contains lowfrequency transient behavior. Filtering typically improves the accuracy of your model by removing the effects of frequencies other than the input frequencies. These frequencies are problematic when your sampled data has finite length. These effects are called spectral leakage.
frestimate
uses a finite
impulse response (FIR) filter. The software sets the filter order
to match the number of samples in a period such that any transients
associated with filtering appear only in the first period of the filtered
steadystate output. After filtering, frestimate
discards
the first period of the input and output signals.
You can specify to disable filtering
during estimation using the signal ApplyFilteringInFRESTIMATE
property.
Estimates the frequency response of the processed signal by computing the ratio of the fast Fourier transform of the filtered steadystate portion of the output signal y_{est}(t) and the fast Fourier transform of the filtered input signal u_{est}(t):
$$G(s)\approx \frac{\text{fastFouriertransformof}{y}_{est}(t)}{\text{fastFouriertransform}{u}_{est}(t)}$$
To compute the response at each frequency, frestimate
uses
only the simulation output at that frequency.
The sweptfrequency cosine (chirp) input signal excites your system at a range of frequencies, such that the input frequency changes instantaneously.
Alternatively, you can use the sinestream signal, which excites the system at each frequency for several periods.