very basic question on sampling frequency..need help

[mdzafar]

hi

Refer to following code, please tell me why we need sampling frequency 'fs', for our signal.


fs=250 % why need sampling frequency
N=512
y=wavread('m. wav')
Y= fft(y,N)
p= abs(Y)/(fs/2) % Why we divide here with ( fs/2), what this tells

please explain these questions , i need these basics informatio for my concepts.

best regards

 

[A.Anand Srinivasan]

actually the sampling frequency is not a constant and it depends on the bandwidth of the signal you are sampling....

 

[mdzafar]

let i have suppose this is my sampling frequency of the my signal .
how it depands ??

please try to clear it to me...........
please ans the related Questions on which i have marked ' %'



regards

 

[dozy_walia]

Sampling frquency must be double than the higest frequecny!

 

[A.Anand Srinivasan]

actually the sampling rate must a little more than double the signal bandwidth in order to reproduce the signal from its samples.... this is called nyquist criteria....

 

[dozy_walia]

If it is less than souble it will produce intersymbol interference!

If it is much more than double .. we are not using proper resources provided to us\!

 

[mdzafar]

dear anand ;

please help me that what a difference b/w these two ststements

p= abs(Y)/(fs/2) % Why we divide here with ( fs/2), what this tells ??

and

p=abs(Y)/min(N ,fs)

plz help me...........................

best regards

 

[A.Anand Srinivasan]

if would post the full code then i would know what is being done... with just part of the code and without explanation of what p is used for it is hard to tell.....

 

[vjfaisal]

dear anand

i only wants to know why we use fs in the above code, wht it tells us ??
and what this statement in the code tells us.........

Pyy=abs(Y)/(fs/2)



y r u not geeing QUestion??? i think now you will understand it...

best regards

 

[A.Anand Srinivasan]

if i dont know what p is used for then how would i know why they have divided by 2....

 

[sohiltri]

p is the power...tryn to plot power spectral density i suppose...

p= abs(Y)/(fs/2) % Why we divide here with ( fs/2), what this tells ??

This statement means u r findin normalised power ...why (fs/2) because in fft we get the replica of positive freq on negative side....so neglectin either one u r tryin to find total normalised power at each discrete frequency given by FFT....hEre major criteria u consider is u r readin wave file which could have a max samplin freq of 250 Hz.....

and

p=abs(Y)/min(N ,fs)

now in your above code u have taken N and fs independent of each other....usually we take N=fs/2....so it is obvious again u r findin same norm,alised power.....

and abs(Y) takes only real part into consider ...neglectin imaginary part for power calculation....N defines resolution....


If u find it useful please dont forget to press 'helped me' button...

 

[lordsathish]

Hi sohiltri.. could you tell me how we are normalising it... i still dont get you... what you meant by normalising... becoz... the maximum frequency in fft is (N-1)fs/N...does normalising the frequency means the value of freq will be from 0 to 1...

 

[sohiltri]

I am not normalizing freq...only power ....so freq wil run from 0 to fs/2.....power need to be normalized....in FFT we write 1/N(...)....but in matlab when fft taken it gives it real integer values ....

 

[vjfaisal]

dear sohiltri
ur statements realy helped me.... i appretiate ur concepts....

now please if u could help in these two QUESTIONS:

1) i want backpropagation algorithm implemented in matlab. For the patter recognition.

2) in kaiser or in Equiripple filter , what is the diffreence between small 'n',

and capital" N".

3) please also show me in diagram (is possible), how we implement this formula
L=N-1 or L=M+N-1 in filter frequency response ( what and where is N ,L, andm) or formula derived

these are imp questions for me , so please i hope u could help me in these ...

best regards