FIR Parks-McClellan

Hi,

I have a set of FIR Parks-McClellan coefficient :

Numerator: FR_SV = [3.91719822748777E-02, 0.103629842929331, 0.171922134825388, 0.221881476438683, 0.221881476438683, 0.171922134825388, 0.103629842929331, 3.91719822748777E-02];

Denominator: [1.000];

Can i use spectrogram function to plot and see the filter response? I have browse through the help file and could not find Parks-McClellan window. Anyone can help?

Regards, Newbie

Answer 1

[Honglei Chen]

Parks-McClellan is an algorithm for filter design. It does not use window. Spectrogram is normally used to view the signal since it explores both time and frequency domain. You have a set of fixed filter coefficients so there is no time dependence. You can certainly use spectrogram, but it won't give you additional information than freqz does.

[Tan]

Hi Honglei,

Thanks for kind replying. I have actually design a filter using FIR Parks-McClellan in SystemVue2009 (by Agilent). Parks-McClellan is categorized under the windows (Like Hanning, Hamming etc). So i thought it is also one of the window method. When i refer spectrogram helpfile in Matlab, i couldn't find Parks-McClellan as window method. So in my case, what window should i use with the set of coefficient that i have stated in the ori question?

[Tan]

Hi Honglei,

You were saying that freqz is better than spectrogram? If i use freqz to plot out the response, there is not a need to use spectrogram anymore since freqz is good enough?

How to plot the time domain response using the same set of coefficient?

[Honglei Chen]

That depends on what you plan to do. In general it is a trade off between the main beamwidth and sidelobe height. Different window gives different tradeoff, you should be able to find some discussions on windows on any DSP book. That being said, what is your application? There is nothing wrong to add a window to a filter designed by Parks-McClellan, but this may not preserve the the ripple behavior produced by Parks-McClellan. It may be better to include those requirement before using Parks-McClellan

[Honglei Chen]

I'm not saying freqz is better. What I mean is that for a fixed set of filter coefficients, spectrogram will not give you more information than freqz since there is no time variation. If you want to see the impulse response of the filter, use impz

Answer 2

[Wayne King]

You don't want to use spectrogram, use freqz()

freqz(FR_SV,1)

If you know the sampling frequency, you can input that as well. Here I will assume 10 kHz.

   Fs = 1e4;
   freqz(FR_SV,1,[],Fs)

Answer 3

[Tan]

Hi Wayne King,

I have plotted my graph using freqz. but if i still intend to use spectrogram, is it possible? The function does not have window options for Parks-McClellan right?

[Wayne King]

Honglei said it perfectly. That's why I said use freqz().

[Tan]

All right. Thanks :)

Is systemVue2009 reliable enough to design filter? I'm using this software to design a LPF, FIR Parks-McClellan and obtain a set of coefficient that stated in the ori question.

[Tan]

Hi Wayne King,

freqz(FR_SV,1,N,Fs)

N = 512 (default)

Is it good to use as large as possible? For example, use 512 instead of 128?

Answer 4

[Wayne King]

I'm not familiar with systemVue2009 so I cannot answer your question. In MATLAB, you can use FDATool

 >>fdatool

to design a Parks-McClellan FIR equripple filter and export (save) the filter coefficients.

[Tan]

Thanks for your guidance :)