Hi, How can one linearly change (i.e., with an approx. straight line in a magnitude plot) the spectral tilt (st) of an audio sample by smaller steps than -6dB/Oct? For example, I found that for an audio x the st is -1.6 dB/Oct, and I want it to have a st of -2.3 dB/Oct, how can I calculate the coeff. of a filter such as once applied to x yields the desired st? thanks for any hint.
spectral tilt/slope
Started by ●July 12, 2011
Reply by ●July 12, 20112011-07-12
On Jul 12, 6:41�pm, pulpo <villegas.jul...@gmail.com> wrote:> Hi, > How can one linearly change (i.e., with an approx. straight line in a > magnitude plot) the spectral tilt (st) of an audio sample by smaller > steps than -6dB/Oct? For example, I found that for an audio x the st > is -1.6 dB/Oct, and I want it to have a st of -2.3 dB/Oct, how can I > calculate the coeff. of a filter such as once applied to x yields the > desired st? > thanks for any hint.You approximate this with alternating up breaks )poles) and down breaks (zeros). The response will have ripple The more break points you have, the less the ripple. There 3 dB/octave filters in the literature that will give you a general idea. Search. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●July 12, 20112011-07-12
On Jul 12, 9:06�pm, Jerry Avins <j...@ieee.org> wrote:> On Jul 12, 6:41�pm, pulpo <villegas.jul...@gmail.com> wrote: > > > Hi, > > How can one linearly change (i.e., with an approx. straight line in a > > magnitude plot) the spectral tilt (st) of an audio sample by smaller > > steps than -6dB/Oct? For example, I found that for an audio x the st > > is -1.6 dB/Oct, and I want it to have a st of -2.3 dB/Oct, how can I > > calculate the coeff. of a filter such as once applied to x yields the > > desired st? > > thanks for any hint. > > You approximate this with alternating up breaks (poles) and down > breaks (zeros).ain't it the other way around, Jerry?> The response will have ripple The more break points > you have, the less the ripple. There 3 dB/octave filters in the > literature that will give you a general idea. Search.not sure how i would do a 2 dB/oct tilt. maybe souping up the spacing of the alternate poles and zeros? r b-j
Reply by ●July 13, 20112011-07-13
On Jul 12, 9:39�pm, robert bristow-johnson <r...@audioimagination.com> wrote:> On Jul 12, 9:06�pm, Jerry Avins <j...@ieee.org> wrote: > > > On Jul 12, 6:41�pm, pulpo <villegas.jul...@gmail.com> wrote: > > > > Hi, > > > How can one linearly change (i.e., with an approx. straight line in a > > > magnitude plot) the spectral tilt (st) of an audio sample by smaller > > > steps than -6dB/Oct? For example, I found that for an audio x the st > > > is -1.6 dB/Oct, and I want it to have a st of -2.3 dB/Oct, how can I > > > calculate the coeff. of a filter such as once applied to x yields the > > > desired st? > > > thanks for any hint. > > > You approximate this with alternating up breaks (poles) and down > > breaks (zeros). > > ain't it the other way around, Jerry?Oeps!> > The response will have ripple The more break points > > you have, the less the ripple. There 3 dB/octave filters in the > > literature that will give you a general idea. Search. > > not sure how i would do a 2 dB/oct tilt. �maybe souping up the spacing > of the alternate poles and zeros?The closer you put them, the less tilt you get. A single break is asymptotic to 6 dB/octave, but it only gets to 3 dB at the end of the first octave. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●July 13, 20112011-07-13
On Jul 13, 6:29�am, Jerry Avins <j...@ieee.org> wrote:> On Jul 12, 9:39�pm, robert bristow-johnson <r...@audioimagination.com> > wrote: > > > > > > > On Jul 12, 9:06�pm, Jerry Avins <j...@ieee.org> wrote: > > > > On Jul 12, 6:41�pm, pulpo <villegas.jul...@gmail.com> wrote: > > > > > Hi, > > > > How can one linearly change (i.e., with an approx. straight line in a > > > > magnitude plot) the spectral tilt (st) of an audio sample by smaller > > > > steps than -6dB/Oct? For example, I found that for an audio x the st > > > > is -1.6 dB/Oct, and I want it to have a st of -2.3 dB/Oct, how can I > > > > calculate the coeff. of a filter such as once applied to x yields the > > > > desired st? > > > > thanks for any hint. > > > > You approximate this with alternating up breaks (poles) and down > > > breaks (zeros). > > > ain't it the other way around, Jerry? > > Oeps! > > > > The response will have ripple The more break points > > > you have, the less the ripple. There 3 dB/octave filters in the > > > literature that will give you a general idea. Search. > > > not sure how i would do a 2 dB/oct tilt. �maybe souping up the spacing > > of the alternate poles and zeros? > > The closer you put them, the less tilt you get. A single break is > asymptotic to 6 dB/octave, but it only gets to 3 dB at the end of the > first octave. > > Jerry > -- > Engineering is the art of making what you want from things you can > get.- Hide quoted text - > > - Show quoted text -the 6 dB/ octave fundamental building block clearly applies to analog filters... does it also apply to digital filters, a FIR filter for example? or another way of asking the question, are digital filters fundamentally built of poles and zeros or are poles and zeros simply a handy way to analyze them? A FIR filter is all zeroes (i think) so how does one build a 3 dB/ octave slope with a FIR filter? thanks Mark
Reply by ●July 13, 20112011-07-13
On Jul 13, 9:08�am, Mark <makol...@yahoo.com> wrote:> On Jul 13, 6:29�am, Jerry Avins <j...@ieee.org> wrote: > > > > > > > > > > > On Jul 12, 9:39�pm, robert bristow-johnson <r...@audioimagination.com> > > wrote: > > > > On Jul 12, 9:06�pm, Jerry Avins <j...@ieee.org> wrote: > > > > > On Jul 12, 6:41�pm, pulpo <villegas.jul...@gmail.com> wrote: > > > > > > Hi, > > > > > How can one linearly change (i.e., with an approx. straight line in a > > > > > magnitude plot) the spectral tilt (st) of an audio sample by smaller > > > > > steps than -6dB/Oct? For example, I found that for an audio x the st > > > > > is -1.6 dB/Oct, and I want it to have a st of -2.3 dB/Oct, how can I > > > > > calculate the coeff. of a filter such as once applied to x yields the > > > > > desired st? > > > > > thanks for any hint. > > > > > You approximate this with alternating up breaks (poles) and down > > > > breaks (zeros). > > > > ain't it the other way around, Jerry? > > > Oeps! > > > > > The response will have ripple The more break points > > > > you have, the less the ripple. There 3 dB/octave filters in the > > > > literature that will give you a general idea. Search. > > > > not sure how i would do a 2 dB/oct tilt. �maybe souping up the spacing > > > of the alternate poles and zeros? > > > The closer you put them, the less tilt you get. A single break is > > asymptotic to 6 dB/octave, but it only gets to 3 dB at the end of the > > first octave. > > > Jerry > > -- > > Engineering is the art of making what you want from things you can > > get.- Hide quoted text - > > > - Show quoted text - > > the 6 dB/ octave fundamental building block clearly applies to analog > filters... > > does it also apply to digital filters, a FIR filter for example? > > or another way of asking the question, are digital filters > fundamentally built of poles �and zeros or are poles and zeros simply > a handy way to analyze them?IIR filters are.> A FIR filter is all zeroes (i think) �so how does one build a 3 dB/ > octave slope with a FIR filter?See Tim Wescott's reply in the thread "Custom Filter Design". I'm afraid you will have a very long impulse response, which is what makes the task a better fit to an IIR filter. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●July 13, 20112011-07-13
Thank you all, I was wondering why you didn't mention methods such as 'freqsamp' in Matlab. My knowledge on filters is limited, but I understand that this will basically create the shape of the filter in frequency domain that I'm looking for, pretty much like a 'graphic equalizer' if you may. Any thoughts about this?
