Forward Backward IIR EQ

[JulianW]

Forward–backward IIR filtering appears to offer zero-phase behavior with much less pre-ringing than linear-phase FIR EQs.


This almost sounds like a “free lunch” what am I missing? Where are the hidden compromises compared to FIR?

 

[pkane]

Forward–backward IIR filtering appears to offer zero-phase behavior with much less pre-ringing than linear-phase FIR EQs.


This almost sounds like a “free lunch” what am I missing? Where are the hidden compromises compared to FIR?

A big one is that FIR can be applied (with some delay) to a continuously playing signal. Forward-backward IIR requires the whole signal to be pre-processed before playing.

 

[JulianW]

A big one is that FIR can be applied (with some delay) to a continuously playing signal. Forward-backward IIR requires the whole signal to be pre-processed before playing.

A big one is that FIR can be applied (with some delay) to a continuously playing signal. Forward-backward IIR requires the whole signal to be pre-processed before playing.

And regarding pre-ringing? Is nature being tricked here?

 

[NTK]

And regarding pre-ringing? Is nature being tricked here?

No.

 

[bmc0]

with much less pre-ringing than linear-phase FIR EQs

According to whom?
No "free lunch" in signal processing (except perhaps in computational efficiency and numerical stability). Same magnitude and same phase means same time domain behavior; filter implementation does not matter.

As has been mentioned, forward-backward IIR filtering is only possible offline because the filter is non-causal and infinite length. For real time use, one can truncate the forward-backward IIR filter to some finite length, yielding (wait for it...) an FIR filter. In principle, this happens any time you apply an IIR filter to a non-periodic signal because the result must be truncated. Hence, the idea that FIR must somehow produce a different result is incorrect by definition.