Question about 16 bit signed integer samples

I’m interested in extracting samples from USRP to perform some spectrum
analysis on them. I use an USRP and a XCVR2450 to receive signals in the
ISM 2.4GHz band. My setup in GRC is a USRP (decim 64, gain 0dB, baseband
2.405GHz) connected to a file sink. This file is then processed by
read_complex_binary.m located in the directory …/gnuradio-core/utils/.
Its output, for a sampling rate of 1MS/s, gives (I printed out only a
small part):

 -18 -    19i
 -22 -    12i
  -2 -     8i
 -23 -    31i
  21 +    62i
 -62 -   231i
  76 +   510i
-128 -  1023i
 240 +  1925i
-571 -  3749i
3643 + 10423i

14675 - 14352i
21379 + 3772i
-1807 + 4089i
851 - 2080i
-483 + 1112i
251 - 553i
-135 + 256i
54 - 96i
-24 + 13i
-3 - 12i
-8 - 1i
-8 - 1i

Now I would to know how the 12 bit samples from ADC are mapped in 16 bit
signed integers. Since now I know that the samples (12 bit, 4096 levels)
come into the USB after a mapping to USB’s 16 bit signed integers with 4
bits zeroed. I would understand better the process of sampling in the
USRP from the analog signal, to 12 bit samples (signed? so 2^11 levels =
2048), to 16 bit signed integers (2^15 = 32768) as seems to be appened
from my printout. How could I correlate the signal on antenna with these
16 bit samples?

BR,
Sergio Benco

On Fri, Mar 12, 2010 at 11:17:13AM +0100, Sergio Benco wrote:

 -22 -    12i

21379 + 3772i

Sergio Benco
Please read the USRP FAQ.

Eric