Target freq vs Actual freq

Hi,

I am sending a stream on packets between two USRP N210, and at the Tx
side i
am setting a center freq. of 2.490 GHz for eg.
when outputting the Tune Result, I am always getting a difference of
some
khz. (Target Freq: 2.490000 GHz

vs

Actual freq: 2.489993GHz)

  • I wonder if that is related to the low accuracy of the frequency
    synthesizers of the RF front end or is it a software setting issue

  • Is the information of the actual freq being centered on returned from
    the
    FPGA, or from the GNU Radio.

  • Then, is that frequency mismatch is it specific to the target freq and
    the
    RF daughter-board or what.

  • If that is not the unique- freq. offset to account for (am using the
    actual freq. of the Tx to set the Center Rx freq.)
    or I will need a PLL implemented in the Rx Side to have center frequency
    as
    exact as possible.

All help will be appreciated.


View this message in context:
http://gnuradio.4.n7.nabble.com/Target-freq-vs-Actual-freq-tp41115.html
Sent from the GnuRadio mailing list archive at Nabble.com.

On 05/02/2013 05:59 PM, NaceurElOuni wrote:

  • I wonder if that is related to the low accuracy of the frequency
    synthesizers of the RF front end or is it a software setting issue
  • Is the information of the actual freq being centered on returned from the
    FPGA, or from the GNU Radio.

The frequency-setting precision of the USRP+daughtercard combinations is
under 1Hz. THe accuracy (remember the difference between the two
from high-school math/physics) is dependant on the master frequency
reference used. The on-board TCXO on the N210 is specced, as I recall,
to +/- 2.5PPM, and if you use an external GPSDO reference, you can
achieve 50PPB accuracy. The +/- 2.5PPM of the on-board TCXO puts
it in the “pretty darned good for most applications” category. The
average hand-held “walkie-talkie” typically is about an order of
magnitude
worse than this, whether for commercial or amateur-radio purposes.

What you’re seeing in the printout is standard anomalies of the way IEEE
floating-point works, and the “error” you show there is roughly
equivalent
to an error of 7PPM, if I"ve done my math correctly.

But having said that, in REAL LIFE, all RX algorithms that are
demodulating “stuff” particularly “stuff” that is narrowband, will
require a mechanism
to provide for frequency offset between the TX and RX. It’s a fact
of life, as sure as the Sun will rise in the morning. No two crystal
oscillators
are exactly the same, and if you want to build RF systems that are
broadly interoperable among devices that may have somewhat-different
frequency offsets, stability, and accuracy, you’re going to have to
deal with it in the receiver.


Marcus L.
Principal Investigator
Shirleys Bay Radio Astronomy Consortium