Apologies in advance if this has been covered before, but I couldn't find anything clearing up this question.
I'm attempting to get Ubuntu up and running with BruteFIR to do active crossovers digitally. In thinking about filter design and I'm trying to understand the number of taps required. I've seen the equation for frequency resolution: Fres = Fs/Ntaps. I'm trying to understand, does this determine the limit on the resolution of the filter frequency response that can be specified only(slope, width of peaks dips that can be equalized out, etc.)? Or does it determine some kind limit on the frequency resolution of the signal that is output from the filter? For example let's say my sample rate is 44100kHz and I use 441 taps to give a resolution of 100 Hz. If I have a signal feeding the filter that is two sine waves of frequency 325 Hz and 375 Hz, are the frequencies blurred together or distorted in some sense?
thanks,
Dan
I'm attempting to get Ubuntu up and running with BruteFIR to do active crossovers digitally. In thinking about filter design and I'm trying to understand the number of taps required. I've seen the equation for frequency resolution: Fres = Fs/Ntaps. I'm trying to understand, does this determine the limit on the resolution of the filter frequency response that can be specified only(slope, width of peaks dips that can be equalized out, etc.)? Or does it determine some kind limit on the frequency resolution of the signal that is output from the filter? For example let's say my sample rate is 44100kHz and I use 441 taps to give a resolution of 100 Hz. If I have a signal feeding the filter that is two sine waves of frequency 325 Hz and 375 Hz, are the frequencies blurred together or distorted in some sense?
thanks,
Dan
Dan,
more taps you'll have, sharper and steeper the frequency response can be. That formula probably derives from the Fourier transform method, where you "draw" the frequency response of the filter, and inverse Fourier transform is then performed to calculate the filter impulse response. Or, the frequency domain specification is directly used for filtering the input data using FFT.
Frequency resolution in this sense affects only frequency response, it does not distort the signal by any way. So feeding those 325 and 375 Hz signals to filter with 100 Hz frequency resolution outputs same frequencies.
Regards,
Janne
more taps you'll have, sharper and steeper the frequency response can be. That formula probably derives from the Fourier transform method, where you "draw" the frequency response of the filter, and inverse Fourier transform is then performed to calculate the filter impulse response. Or, the frequency domain specification is directly used for filtering the input data using FFT.
Frequency resolution in this sense affects only frequency response, it does not distort the signal by any way. So feeding those 325 and 375 Hz signals to filter with 100 Hz frequency resolution outputs same frequencies.
Regards,
Janne
Have a look at http://www.dspguide.com/ and download chapters 14 to 21 (or even the whole book). FIR filter can be found in chapter 17.
It's a great book, written easily, with a bunch of examples. I can really recommend it for beginners, students and everybody else who's involved in this numeric audio stuff.
It's a great book, written easily, with a bunch of examples. I can really recommend it for beginners, students and everybody else who's involved in this numeric audio stuff.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.