Re: [Audacity-devel] New effect - Scientific Filter
A free multi-track audio editor and recorder
Brought to you by:
aosiniao
From: Steve t. F. <ste...@gm...> - 2013-10-02 16:11:11
|
I very much like that the filter shows a graph of the filter curve. I've only had time to do a little testing, but the filter response looks about right. For accessibility the effect will need access keys. Personally I'd find it useful if there was also a slider for the frequency control. For completeness it might be nice to include band pass and band stop filters, though I realize that raises complications for the GUI design. Nice work Norm. Steve On 2 October 2013 11:39, James Crook <cr...@in...> wrote: > Thanks for confirming, and the detail. > > I've committed the new effect to SVN and it should appear in 2.0.5. > I'm personally taking responsibility for any 'risk' involved. Really it > seems very safe to me. > > We may at some future date have a special category for contributed > effects like PaulStretch and ScienceFilter, which are not everyday > effects for the vast majority of our users - so in that sense it might > at a future date get 'demoted' from the main effects menu, but for the > time being it is there. > > Norm thanks very much for this. > > --James. > > > On 02/10/2013 03:51, Norm C wrote: >> Yes, I wrote all the algorithmic code myself, derived from the math. Here's a >> description, maybe more than you want, but trim as needed: >> >> The Scientific Filter effect implements 3 different types of filters which >> together emulate the vast majority of analog filters, and provide useful >> tools for analysis and measurement. The three main types are Butterworth, >> Chebyshev Type I and Chebyshev Type II. Each of those has two subtypes: >> Lowpass and Highpass. These filters (and the analog prototypes on which they >> are based) generally pass through a range of frequencies (called the >> passband) with minimal attenuation, and the remainder (called the stopband) >> with some significant degree of attenuation. The boundary between the >> passband and stopband is called the Cutoff Frequency. >> >> Of the 3 types, Butterworth is the simplest. An analog Butterworth filter >> provides a "maximally flat" passband (ie. no ripples), the magniture >> response at the cutoff frequency is -3 dB, and above (for lowpass) or below >> (for highpass) the cutoff frequency, the attenuation increases at >> approximately 6 dB per octave times the filter order (so eg. 60 dB per >> octave for 10th order). The Scientific Filter approximates this behaviour >> very closely, except at very high frequencies (see later). Butterworth >> filters are simple to implement in analog; in fact, a voltage source driving >> a series resistor which feeds a shunt capacitor is just a 1st order >> Butterworth lowpass filter, and if you reverse the resistor and capacitor it >> becomes a 1st order Butterworth highpass filter. >> >> Chebyshev Type I filters are similar, except that a) the magnitude response >> of the passband has "ripples" in it (usually small), b) at the cutoff >> frequency the magnitude response is equal to the ripple value, and c) above >> (below for highpass) the cutoff frequency, the stopband attenuation >> increases more rapidly, for a given filter order, than Butterworth. >> >> Chebyshev Type II filters are similar to Butterworth, including the flat >> passband response, except that a) at the cutoff frequency the magnitude >> response is equal to the ripple value, b) above (below for highpass) the >> cutoff frequency, the stopband attenuation increases more rapidly, for a >> given filter order, than Butterworth, and c) the stopband attenuation varies >> from infinite to the ripple value. (Here it's common to use a ripple value >> of 20, 30 or more dB). >> >> All three filter types (and both subtypes) are commonly used in analog audio >> circuitry. First order Butterworth filters, as mentioned, are extremely >> simple. Second to 4th order Butterworth and Chebyshev filters are normally >> implemented with opamps or, rarely, inductors and capacitors. Higher order >> analog filters are rare, due to the very high component precision that would >> be required. >> >> The high-frequency response of all 3 filter types diverges from that of the >> analog prototypes, due to the bilinear transformation used in converting >> analog to digital filters. Essentially the response at the Nyquist frequency >> (sampling frequency divided by 2) in the digital version is equal to the >> response at infinite frequency in the analog version. So for lowpass >> filters, the attenuation at Nyquist is infinite (for Butterworth and >> Chebyshev Type I filters). >> >> >> Norm |