# Unexpected oscillations in the impulse response

Started by July 9, 2008
Hi all! My question is about getting the impulse response from the transfer function of a system.

I have simulation data for the transfer function of an LTI system. I make the data Hermitian symmetric so that when I do an IFFT on it, I get a real impulse response. So far so good. But the problem is the shape of the impulse response I get after the IFFT. It oscillates at the beginning, decays to zero and then starts oscillating again towards the end.

It is this oscillation at the end that I don't like. (Physically, the response of the system should settle down after some time of the application of an impulse).

What can I do to get rid of the oscillation at the end? Can anybody help me with this? Thanks a lot!
Probably the problem is related to the sampling frequency. If the signal has
a limited spectrum, you have to chose the sampling frequency at list twice
the unilateral band of the signal. Greater is this value more the replica of
the signal are separated in frequency. If the replica are far each other,
it easy to use a low pass filter to select only the fundamental replica of
the signal located in the origin. If the fs is not adequate, the replica
of the signal are superimposed (starting from the lateral part an reducing
the fs the overlapped part become bigger). This phenomenon is called
ALIASING.
If I'm right, increasing fs you have to see a more regular spectrum.
Otherwise sorry and good luck!

On Wed, Jul 9, 2008 at 4:42 PM, wrote:

> Hi all! My question is about getting the impulse response from the
> transfer function of a system.
>
> I have simulation data for the transfer function of an LTI system. I make
> the data Hermitian symmetric so that when I do an IFFT on it, I get a real
> impulse response. So far so good. But the problem is the shape of the
> impulse response I get after the IFFT. It oscillates at the beginning,
> decays to zero and then starts oscillating again towards the end.
>
> It is this oscillation at the end that I don't like. (Physically, the
> response of the system should settle down after some time of the application
> of an impulse).
>
> What can I do to get rid of the oscillation at the end? Can anybody help me
> with this? Thanks a lot!
>
>