Hello Fred, Alberto and all,

I'm investing an interesting time in the SdR theory. One particular message of 
this group makes me problem.
 
The problematic sentence is the following:
"1° Mix the time domain Q & I signal with a NCO to make a near to zero 
signal"(this before the FFT)
It seems that the "NCO" must mean something as "numerical oscillator". I don't 
suppose that it is an analytic oscillator (cos wot / sin wot) but a simple 
oscillator (cos wot). This simple oscillator applies to I and Q.
I suppose that "I" is a sum of cos (wt) from 0 to the sampling frequency/2
I suppose that "Q" is a sum of sin (wt) from 0 to the sampling frequency/2 

Now you want to slightly shift the spectrum by w0 (1 KHz for example), you will 
have at the output:
- terms in (w-w0) which are the basebands components 
- terms in (w+w0) which are non-desired components.

As w0 is weak, you can't discriminate (by filtering) w-w0 from w+w0 as the 
bands will be in the band from 0 to 3 KHz (now, with a big w0 it would be 
possible to discriminate w-w0 from w+w0).

Other thing: the very high components (the last 1 KHz in the example, close to 
the sampling frequency/2) are going to fold in the base band with the w+w0 
operation. It can't be avoided either.

Note 1: with an analytical mixer, it would possible to discriminate w-w0 from 
w+w0 without filter but you will lose the quadrature configuration.
Note 2: the following FFT will not help to discriminate the two components.

TKS for ideas.

73
Patrick



  ----- Original Message ----- 
  From: Fred Krom 
  To: [email protected] 
  Sent: Tuesday, September 26, 2006 3:32 PM
  Subject: [soft_radio] Change sample rate FFT


  I got a question that maybe can be answered in this group. 
  Most of the SDR radio's I know are using the following steps:
  1e Mix the time domain Q & I signal with a NCO to make a near to zero signal
  2e Take the FFT to convert it to frequence domain
  3e Filter the signal by setting most of the bins to zero (multiply by filter)
  4e Take the reverse FTT to convert to time domain
  5e Down sample the signal to something like 8Khz sample rate.
  6e Somekind of demod.

  What I like to do is take step 4 and 5 together, but can not get the code 
working.
  If for example the sample rate S=48KHz, FFT size N=1024 then BIN=S/N=47Hz. 
  Then taking the reverse FFT with N=256 will result in a signal with sample 
rate S=12KHz (using the same BIN size). 
  That will result in a smaller FFT and no down sample, so less cpu is used.

  What is wrong, the code or the idea?
  Fred
  PE0FKO

  [Non-text portions of this message have been removed]



   

Reply via email to