MATLAB: How to use the “filtfilt” function with a “dffir” filter object created using Filter Designer (fdatool)

datadesignerdffirfdatoolfilterfilteringfiltfiltSignal Processing Toolbox

How can I use the "filtfilt" function with a "dffir" filter object created using Filter Designer (fdatool)?

I have designed a FIR (finite impulse response) filter using the Filter Designer and I have generated MATLAB code for it. The output of my code, "Hd", is a filter object of type "dffir". How could I apply this filter on some data using the "filtfilt" function?

Best Answer

The "filtfilt" function can be executed by inputting the numerator and denominator of a filter, followed by the signal to be filtered. So if you could extract the numerator and denominator of "Hd", you could simply pass these as input arguments to "filtfilt".

The numerator coefficients of "Hd" can be accessed as properties of the filter object with the syntax "Hb.Numerator". Notice that "Hd" is a finite impulse response filter, and hence its denominator coefficient is simply 1.

Based on this information, you could use "filtfilt" with your filter object "Hd" as follows:

y = filtfilt(Hd.Numerator,1,x);

Where "x" is the signal that you are filtering and "y" is the filtered output data.

Related Solutions

MATLAB: Error using designfilt – parseAndDesignFilter: Coefficients must be finite

Consider the requirement to design a Low-Pass FIR filter with a Pass band frequency of 0.2 Hz, Stop band frequency of 0.3 Hz with a Sample rate of 1000 Hz. A filter can be designed using the 'designfilt' MATLAB function for the given specifications as shown in the below command:

>> lpFilt = designfilt('lowpassfir', 'PassbandFrequency', .2, 'StopbandFrequency', .3, 'SampleRate', 1000);

Upon executing the above command, 'designfilt' function throws an error: 'Error using designfilt>parseAndDesignFilter: Coefficients must be finite'. The input frequencies to the function by default are assumed to be 'Hz', and the filter design tool is unable to compute a FIR filter solution with a 0.20 Hz passband and 0.30 Hz stopband frequencies with sample rate set at 1000 Hz. The passband and stopband frequencies are very low, and the difference between them is very small to design a filter as the function uses a default Passband attenuation of 1 dB and a Stopband attenuation of 60 dB. To design a FIR filter with these constraints, consider parsing the 'FilterOrder' as one of the inputs. A very high order is necessary to design a FIR filter with these constraints as shown below:

>> lpFilt = designfilt('lowpassfir','FilterOrder', 100,'PassbandFrequency',.2, 'StopbandFrequency',.3, 'SampleRate', 1000);
>> fvtool(lpFilt)

If the provided passband and stopband frequencies are normalized, then 0.2 and 0.3 for a sample rate of 1000 Hz gets scaled to 200 Hz and 300 Hz respectively, and below is the suggested implementation. Specifying filter order is not necessary as the constraints with respect to sample rate are not tight, and MATLAB computes a filter of minimum order which meets the requirements:

>> lpFilt = designfilt('lowpassfir','PassbandFrequency',200, 'StopbandFrequency',300, 'SampleRate', 1000);
>> fvtool(lpFilt)

If the* objective is indeed to design a filter for 0.25 Hz and 0.35 Hz with a Sample Rate of 1000 Hz*, then a Low-Pass *IIR filter *can be designed instead due to the robustness properties of IIR filters. Below is the command to compute a low-pass IIR Filter, and to view the response:

>> lpFilt = designfilt('lowpassiir','PassbandFrequency',.2, 'StopbandFrequency',.3, 'SampleRate', 1000);
>> fvtool(lpFilt)

Please check the below provided documentation link for more information regarding the usage of the 'designfilt' function and input argument constraints for certain filter type implementations:

https://www.mathworks.com/help/signal/ref/designfilt.html#inputarg_resp

Alternatively, filters can be best designed and analyzed by visualizing them in a filter visualization tool. The Filter Designer GUI tool can also be accessed by typing the below command in the MATLAB Command Window:

>> filterDesigner

Please check the below provided documentation link for more information regarding the usage of the 'Filter design and analysis' tool:

https://www.mathworks.com/help/dsp/filter-design-and-analysis.html

MATLAB: How to obtain the quantized filter coefficients as fixed-point objects in Filter Design Toolbox 4.0 (R2006b)

The ability to obtain the quantized filter coefficients as fixed-point objects is not available in the Filter Design Toolbox 4.0 (R2006b).

To work around this you can manually convert the coefficients to fixed-point objects as shown below:

d=fdesign.lowpass(0.40,0.54,0.05,50);
hd=design(d,'kaiserwin','filterstructure','dffir');
hd.Arithmetic='fixed';
set(hd,'filterInternals','specifyprecision');
hd.productWordLength = 16;
hd.accumwordLength = 40;
myfi = fimath('maxproductwordlength',16,'maxsumwordlength',40);
numerator = fi(hd.numerator,hd.signed,hd.inputwordlength,...
          hd.inputfracLength, 'fimath', myfi)

Best Answer

Related Solutions

MATLAB: Error using designfilt – parseAndDesignFilter: Coefficients must be finite

MATLAB: How to obtain the quantized filter coefficients as fixed-point objects in Filter Design Toolbox 4.0 (R2006b)

Related Question