Re: [Discuss-gnuradio] USRP2+WBX - How to use precisely an 8 MHz band?

[Vladutzzz]

Matt,
I was trying to explain that using 8 as a decimation rate would complicate
my rational resampler coefficients, not that it would give me less band.
Anyway resampling seems too heavy for my processor and I decided to use
filtering as you suggested.
The problem is that I don't understand what is the difference between using
an 8, 10 or 12 rate decimation (aside from the obvious band size
difference), if and how they affect the measured power values.
I used decimation 8 and the appropriate 8 MHz filter and measured a signal
power of -44.42 dBm.
I used decimation 10 and the appropriate 8 MHz filter and measured a signal
power of -42.1 dBm.
I used decimation 12 (which gives me a 8.333 MHz band) and the appropriate 8
MHz filter and measured a signal power of -47.07 dBm.
As you can see there is a great difference between the 3 results,
considering they are measuring the same band.
I've read that using even decimation values that are not multiples of 4
result in an output of only one halfband (the low rate one), but I don't
know how to interpret this information nor if it's even relevant to my case.
What are your thoughts on this?
Thanks.

Vlad.

P.S.: I don't think my power measuring method is the problem: I unbuffer the
filter's output, input it in a magnitude FFT block (fft squared) after which
I buffer the resulting coefficients, sum them, divide the result by the
buffer size and the sample time and transform the output value to dBm.



--- On Tue, 11/16/10, Matt Ettus <address@hidden> wrote:


    From: Matt Ettus <address@hidden>
    Subject: Re: [Discuss-gnuradio] USRP2+WBX - How to use precisely an 8
MHz band?
    To: "Vladutzzz" <address@hidden>
    Date: Tuesday, November 16, 2010, 8:09 PM

    On 11/16/2010 09:36 AM, Vladutzzz wrote:
    >
    > Thank you all, for your replies!
    > I am a bit confused about the 10 rate decimation, which just gets one
    > halfband (the low rate one) - as explained in USRP2 FAQ. Does this
affect
    > the power level that I will later measure?
    > I thought about using 8 which is also a multiple of 4, but it is a
very
    > unfortunate fraction and I would like to keep as much as possible from
the
    > original band,

    I don't understand this.  Decimation of 8 will give you more bandwidth,
    not less.

    > plus the 10 rate will also decrease the load on my processor.
    > So, is the 10 rate decimation in fact a good candidate for what I'm
trying
    > to do or I'm better of using 8 with a 1.5625 rational resampler?


    Why do you need to resample?  Just use a filter of 8 MHz bandwidth.

    Matt

-- 
View this message in context: 
http://old.nabble.com/USRP2%2BWBX---How-to-use-precisely-an-8-MHz-band--tp30191243p30238749.html
Sent from the GnuRadio mailing list archive at Nabble.com.