Repository navigation
Low Pass filter: different results in versions 0.5.0 and 0.6.0 #15
Description
I have upgraded our project from Math.Net.Filtering 0.5.0 to 0.6.0 Now I get completely different results with LowPass filter.
Example
var frequency = 1d / 360d;
var samplingFrequency = 0.1; // Nyquist: sampling rate must be at least double the frequency
var amplitude = 1;
var seriesLength = 10;
var input = Generate.Sinusoidal(seriesLength, samplingFrequency, frequency, amplitude);
var filter = new LowPassFilter();
Result with 0.5.0 {0, 0.032, 0.203, 0.4, 0.584, 0.751, 0.895, 1.012, 1.098, 1.151};
Result with 0.6.0 {0, -0.003, -0.004, -0.005, -0.008, -0.007, -0.01, -0.009, -0.009, -0.012}
Which one is correct? Do I need to change something when setting up filter?
Within my wrapper class "LowPassFilter" the filter is created as follows:
The values used are: filter.CutOffFrequency = 0.04; filter.OrderOfFilter = 2;
protected override OnlineFilter CreateFilter() {
return OnlineFilter.CreateLowpass(ImpulseResponse.Finite, SamplingFrequency, CutOffFrequency.Value, OrderOfFilter);
}
Thanks for help
Gianni
Activity
SOCMarcel commentedon Jun 12, 2020
Hello,
as far as I unterstand the source code (https://github.com/mathnet/mathnet-filtering/blob/master/src/Filtering/OnlineFilter.cs, https://github.com/mathnet/mathnet-filtering/blob/master/src/Filtering/FIR/FirCoefficients.cs) there seems to be an issue with CreateLowpass(ImpulseResponse mode, double sampleRate, double cutoffRate, int order):
In case of passed mode ImpulseResponse.Finite the following code is executed:
double[] c = FirCoefficients.LowPass(sampleRate, cutoffRate, order >> 1);
return new OnlineFirFilter(c);
Yet as you can see in the history (probably corresponding to version change from 0.5 to 0.6) the signature of FirCoefficients.LowPass(...) changed from
public static double[] LowPass(double samplingRate, double cutoff, int halforder = 0)
to
public static double[] LowPass(double samplingRate, double cutoff, double dcGain = 1.0, int halforder = 0)
So from my point of view the order parameter from CreateLowpass(...) gets now passed as dcGain to FirCoefficients.LowPass(...) instead of halforder which results in incorrect coefficients.
Kind regards.
Marcel