According to IEEE-Std 802.15.4-2006  spec ( Zigbee ) ,p 299-300
The out of band spurious emission for  European 868 Mhz  ISM band  at 300 Khz offset from fc = 868.3 Mhz are

-36 dBm  for  100 Khz detector BW or resolution BW  in active mode . That is , -86 dBm /Hz.

-57 dBm  for  100 Khz detector BW or resolution BW  in standby mode.  . That is , -107 dBm /Hz.

The spur level is measured  with the transmitter  outputting an  unmodulated carrier and  resolution bandwidth ( RBW ) used on the spectrum analyzer is 100 Khz. The output power at fc is at least -3 dBm /Hz and at most 14 dBm/ Hz. So, signal attenuation at 300KHz offset should be at least  83 dB/ Hz,


Here I am confused about the issue of the unmodulated carrier. When baseband signal will modulate the carrier, its spur -energy will spread. Right ?   But, what about the baseband spectrum ?

The baseband signal I have got from other digital guys have an interesting spectrum which is attached herewith. In this spectrum, the  signal level at dc is  -19.68 dB and at 300 Khz is  -68.12 dB.

So,  signal attenuation  is  48.44 dB which is much less than  83 dB !!!

Then , after upconversion  and power amplification , the transmitted signal will violate the Transmit spectrum mask.  Am I right ?

@ RFICDude,

 Thanks  for your response. Well I have used MATLAB finction "psd" which gives  dB/ Hz values , not integrated power over a band. So, frequency resolution is not needed for this graph. Am I right ?  

I have used the following MATLAB code to plot the spectrum.
------------------------------
 % x is the data vector.
 % Fs1 is sampling frequency.

      h2 = spectrum.welch('Hann',512,50);
      hopts = psdopts(h2,x);  % Default options
      set(hopts,'Fs',Fs1,'NFFT',4096,'SpectrumType','twosided','centerdc',true);
     
      hpsd = psd(h2,x,hopts);
      dataplot6 = figure;
      plot(hpsd);
-------------------------------------------------------------------

The following info is taken from MATLAB help.

-------------------
The power spectral density is intended for continuous spectra. Note
    that unlike the mean-squared spectrum (MSS), in this case the peaks in
    the spectra do not reflect the power at a given frequency. Instead,
    the integral of the PSD over a given frequency band computes the
    average power in the signal over such frequency band. See the help on
    AVGPOWER for more information.
 
    The Hpsd object also contains a vector of normalized frequencies W at
    which the PSD is estimated.  W has units of rad/sample.  For real
    signals, W spans the interval [0,Pi] when NFFT is even and [0,Pi) when
    NFFT is odd.  For complex signals, W always spans the interval
    [0,2*Pi).
------------------------------------------