When you want to filter some data you often start by defining the response you're after, for example for a sub-woofer output you might want a low-pass filter that cuts off frequencies higher than 250 Hz.
Then you use some math to figure out the filter coefficients that will give you that response. For many (all?) filters you can get away with far fewer coefficients, often orders of magnitude like 4 vs 1024, if you go for an IIR filter rather than a FIR filter. This of course makes it far cheaper computationally to implement.
IIR filters have some downsides though compared to FIR, for one an impulse "never goes away" (at least in theory) since there's always some of it begin fed back into the next step. It's also difficult to make them have linear phase response, which can be important in certain settings.
When you want to filter some data you often start by defining the response you're after, for example for a sub-woofer output you might want a low-pass filter that cuts off frequencies higher than 250 Hz.
Then you use some math to figure out the filter coefficients that will give you that response. For many (all?) filters you can get away with far fewer coefficients, often orders of magnitude like 4 vs 1024, if you go for an IIR filter rather than a FIR filter. This of course makes it far cheaper computationally to implement.
IIR filters have some downsides though compared to FIR, for one an impulse "never goes away" (at least in theory) since there's always some of it begin fed back into the next step. It's also difficult to make them have linear phase response, which can be important in certain settings.