DSP basics,fundamentals,software

walters said:

aren't the filters IIR and FIR already designed from the GUI all u do it adjust the GUI's parameters from the GUI interface

The IIR and FIR filters are already done for us and designed already right? all the hardwork is done for us the code has been writen?

The GUI interface can changed or modify the filter coeffecients without programming in code just using the mouse and GUI interface?

What does the vector variables do to the filters coeffecient?

Click to expand...

Yes you only specify your frequency response requirements by adjusting the options in the GUI to generate the proper coefficients. That's called filter design. But after the coefficients are generated you may need to use them in the workspace. For example let's say you designed (generated coefficients) of a low pass filter and you want to filter an audio file. So first you open the wav file using a MATLAB wavread command and you store the sound samples of the wav in a vector named X , and after you designed the LP filter you named the vector that contains the filter coefficients H , now you can obtain the new filtered signal using convolution by using this command y = conv(X,H)

Thanks alot for the help

What generates coefficients?
or what is a coefficient generator?

What are the basics of DSP filter design?
1.) Coefficients
2.) Frequency response
3.) What else please for DSP filter design?

walters said:

Thanks alot for the help

What generates coefficients?
or what is a coefficient generator?

What are the basics of DSP filter design?
1.) Coefficients
2.) Frequency response
3.) What else please for DSP filter design?

Click to expand...

Coefficients are the time domain values of the Filter (It's impulse response). For example. Let's say you have an input signal x and a filter h , you can get the filtered output y using convolution.

But since we can't get many information by looking at a fuction in time domain, we use Fourier Transform FT to obtain the frequency domain representation of the signal or system (filter).
The Fourier Transform of the impulse response of the filter is the frequency response of this filter. Impulse response is a function of time, Frequency response is a function of frequency. FT of a signal tells you the magnitude of each frequency component in that signal. FT of the system (AKA frequency response) will show you the response of this system to various frequencies. Convolution of two signals becomes multiplication in frequency domain. So , if you get the FT of x which is X(ω) and and the FT of h which is H(ω) , you can get Y(ω) by multiplying
X(ω) and H(ω) : Y(ω)=X(ω)H(ω) . In English, Y(ω) is your filtered X(ω).

Example:

Look at this plot of a low pass filter :
**broken link removed**

You'll notice that there's unity attenuation at low frequencies because these are the frequencies which the designer of that filter wanted to keep. And the attenuation of high frequency is very high because it's a low pass filter and you want to get rid of the filter components after some point.

Note that the magnitude in this plot is in DB. That's why I use the word attenuation. If it was plotted without conversion to DB then I'd say (It has unity magnitude at low frequencies) and almost zero at frequencies you want to remove.

Assume the filter in the plot is our H(ω) and we have a signal X(ω) that contains frequencies from 0 to 1 MHz. Imagine how the Y(ω) plot will look like. Y(ω) will look like (equal to) X(ω) at frequencies in the zero attenuation region of the filter (the passband of the filter which in this example ends at 500 KHz, the point where passband ends is also known as cutoff frequency) and will start to attenuate as you go to the right.

Now let's say that you want to design a low pass filter. The main specification that you will think of is the point at which the heavy attenuation will start (cutoff frequency). So let's say you chose 1k Hz as your cutoff freuqncy. So the filter that you imagine looks like this

**broken link removed**

So now you think you're done because "you know your requirements". So next you say let's do the inverse fourier transform so that I get the impulse response (AKA filter coefficients,AKA time domain representation) of this filter that I've just desgined. Then you find out that this ideal filter that you've just "designed" has an infinite number of coefficients and is uncausal in time domain! Which means that this filter can't be realized in real life. Then you get a book on DSP and read the chapter of filter design , and you find out that there are methods that allow you to specify less strict specifications for your filter and will give you the coefficients that you need (A realizable filter).

After you get your realizable filter coefficients , now you happily write your convolution code or design the hardware that implements the convolution.

The FDAtool that you're using implements algorithms of filter design. IF you get any basic DSP book and see the chapter on filter design, you'll see that there are various methods for FIR filter designs and IIR filter designs. FDAtool does all the calculations for you and you just click buttons and enter your specifications to generate your coefficients.