Re: [Discuss-gnuradio] unknown attenuation after analysis filterbank

[Brian Padalino]

On Fri, May 15, 2009 at 8:18 AM, feldmaus <address@hidden> wrote:
> Hi Gnuradio folks,
>
> i am analysing the spectrum from 1MHz to 2Mhz. I splitted this Spectrum
> into 3 parts:
> 1MHz - 1,333MHz
> 1,333MHz - 1,666MHz
> 1,666MHz - 2MHz
>
> Therefore i used the Analysis Filterbank. Because i am searching for si 
> function
> in the time domain as a response of a peak in the frequency domain, i want
> to use a band pass filter in the Analysis Filterbank.
> Do i have to use the band pass filter or can i sill use the low pass filter.
> The response should be another one by using the Low pass filter ?
>
> But my main Problem is, that my signal goes down in the second and third band.
> I doesn't know why.
>
> Here is my Flow Graph:
> Usrp Source --> Analysis Filterbank+Band pass filter --> Selector -->
> fft sink graph
>
> My band pass looks like this:
> samp_rate=1e6
> Intermediate Frequency of the usrp= 1MHz
> firdes.complex_band_pass(1.0, samp_rate, samp_rate*0, samp_rate*1/3,
> samp_rate*0.01)
>
> Any hints how to correct this ?
> Did i anything wrong ?

So you are looking at 1MHz of spectrum with a sample rate of 1Msps
complex centered at 1MHz coming in from the USRP?  Or do you receive a
complex stream directly from the USRP at baseband?

The CIC filter has roll off which could be your culprit?  Do you have
a spectral plot of your bands as well as the entire incoming spectrum?

Brian