[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Discuss-gnuradio] question regarding usrp_fft.py

From: adib_sairi
Subject: Re: [Discuss-gnuradio] question regarding usrp_fft.py
Date: Wed, 26 Aug 2009 19:21:28 -0700 (PDT)

Firas A. wrote:
> Hi,
> --- On Sat, 7/11/09, adib_sairi <address@hidden> wrote:
>> so if the bin of fft is treated as individual, what is the relationship >
>> between fft_size and the db calculation? what is the usage of fft_size 
>> in the calculation?
> Ok let me explain. Let x(n) = cos(2*n*PI*f/fs)  , n=0,1,2,......,N
> Now take the FFT 3 times with N = 256, N = 1024 and N = 2048.
> You will see that the output of the FFT X(n) will be different in
> magnitude for the same input sequence because of different FFT size. The
> spectrum shape is similar in the 3 cases, but the the amplitude of each
> spectrum bin is different.
>> if i choose to monitor the 2.5G center frequency, so the evt.data[0] is
>>  the place where the 2.5G is right? if the fusb is 4Ms/s and our
>> fft bin is 256, so our bandwidth of monitoring is only 2MHz right? and
>> evt.data[1] to evt.data[128] is the positive side of my signal while
>> evt.data[129] to evt.data[256] is my negative side.. am i right?
> No some numbers are wrong. Assuming complex samples, when Fusb is 4MB/s
> the bandwidth is 1 MHz (2 bytes for I and 2 for Q). For FFT 256,
> evt.dat[0] is the center frequency (2.5G in your case) evt.data[1] to
> evt.data[127] is the positive side and evt.data[128] to evt.data[255] is
> the negative side.
> BR
> Firas

Good day every one,
I am trying to calculate the sensing time for my system with using
usrp_fft.py as the core of my sensing method. my fft bin is 256 with Fusb =
4MHz and my bandwidth is 1MHz. 

with this parameters, each sample of 256 bin FFT will take 1 second. am i
right? means, between one evt.data[0] to the next evt.data[0] will take 1
second. this is due to the maximum USB BW that i use for the sensing.

but i can see that in the usrp_fft GUI, the update of the plot is much
faster than 1 second. so i think my calculation is wrong but i am not sure
what i miss. 

i see in some paper, they divide the Fusb to the number of their bin. for
example if the Fusb is  4MHz and the FFT bin used is 128, then they divide
4MHz with 128 which give them 31.25kHz of sensing speed (0.032s). but i
think we cannot divide like that because the FFT is being done in GNU Radio.
USRP dosent know about this is it? what bin number we choose i think the
speed is the same (1 second) because the speed of the data flush from USRP
to PC is only depend on the decimation we choose not the FFT bin..FFT bin is
only take place inside GNU Radio.. am i right?
so what is the actual way to calculate the sensing time if i am using
usrp_fft as my core (to be more specific, i use the fft_nongl as the basic
code and do a modification from there). 

please help and guide me for this.. thank you very much in advance..


View this message in context: 
Sent from the GnuRadio mailing list archive at Nabble.com.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]