IIR vs FIR filters

Thread Starter

Dritech

Joined Sep 21, 2011
907
Hi all,

What are the advantages of using IIR over FIR filters in DSP?

Is the fact that IIR filters have an infinite output a disadvantage?

Any help would be highly appreciated.
 

Papabravo

Joined Feb 24, 2006
22,082
In an FIR filter, the lack of an input implies the lack of an output. That is the filter will not continue to produce an output in the absence of an input.
In an IIR filter, an output will continue to be produced even after the input goes away.
IIR filters involve the use of feedback and can be simpler to implement.
 

Thread Starter

Dritech

Joined Sep 21, 2011
907
"In an IIR filter, an output will continue to be produced even after the input goes away."

Is that considered as a disadvantage when compared to FIR?
 

Papabravo

Joined Feb 24, 2006
22,082
The IIR filters closely approximate the corresponding analog filters. Inductors and capacitors have "memory" that decays exponetially with time and can be ignored past a certain point but never becomes identically zero. The FIR filter is only possible in the digital domain, and like all digital filters the input must be band limited by an appropriate anti-aliasing filter.
 

Thread Starter

Dritech

Joined Sep 21, 2011
907
Thanks for the reply.

I found that IIR filters have much better frequency response than FIR filters. Why exactly does that make it an advantage please? I cannot exactly understand that is meant by better frequency response.

Thanks in advance.
 

Papabravo

Joined Feb 24, 2006
22,082
A better frequency response could mean any one or all of the following:
  1. Sharper rolloff in the transition band
  2. More attenuation in the stopband
  3. Less ripple in the passband or the stopband
 

JWHassler

Joined Sep 25, 2013
308
FIR filters do not feed back the output to the input. An advantage to this is that they will not oscillate ( oscillation is almost always undesirable.)
FIR filters require more filter stages for a given attenuation than IIR filters: they are computationally expensive.

Read this.
 
Top