Analysis of phase noise on OFDM signal An OFDM baseband signal generated from a inverse discrete Fourier (IDFT) can be written as: N 1 rn s[n] d r exp( j 2 ) N r 0 The received baseband signal is given by: r[n] s[n] exp( j[n]) where [n ] is a random process representing phase noise. After OFDM demodulated (equivalent to DFT), the signal in the kth subcarrier y[k ] is given by: 1 y[k ] N N 1 N 1 m exp( j[m]) d r exp( j 2 (r k ) ) N m 0 r 0 Wireless Communication Technologies 2.5.3 1 Analysis of phase noise on OFDM signal To simplify analysis, the term exp( j[n]) 1 j[n] is assumed, hence above Eq. can be approximated to be : 1 y[k ] N N 1 N 1 m d r exp( j 2 (r k ) ) N r 0 m 0 j N 1 N 1 m d r [m] exp( j 2 (r k ) ) N r 0 m 0 N d k ek where d k represents the correct information in the kth subcarrier, and ek is the error caused by phase noise. The phase error ( ek ) is considered in two conditions. One is the common phase error, the other introduces ICI. Wireless Communication Technologies 2.5.3 2 Common phase error 1. Common phase error: r = k: The phase error ek j Nc N N 1 r k is expressed as N c , N 1 m j N 1 d r [m] exp( j 2 (r k ) ) r k d k [m] N N m 0 r 0 m 0 Hence one can observe that each sub-carrier constellation has the same rotation called the common phase error, where the rotational angle is equal to N 1 1 Tan ( [m] / N ) . m 0 Since this phase error is constant for every sub-carrier, it can be corrected by the information from the pilot signal. Wireless Communication Technologies 2.5.3 3 ICI error 2. Loss of orthogonality (or inter-carrier interference; ICI): r≠k ek r k is expressed as N i , j N 1 N 1 m N i d r [m] exp( j 2 (r k ) ) r k N r 0 m 0 N N i is the summation of the information of the other N-1 sub-carriers multiplied by some complex function of phase noise. If the number of sub-carriers is large enough and the data point in each sub-carrier is statistically independent, according to the central limit theorem, N i can be treated as additive Gaussian noise that could not be compensated. Intuitively, the value of N i will increases with the number of sub-carriers. Wireless Communication Technologies 2.5.3 4 ICI error The ICI effect can be treated as a white noise with zero mean. The average power of the white noise can be expressed as [2] Ni 2 fs / 2 S ( f ) sin( Nf / f ) fs / 2 2 S Ps S ( f )df ( ) df fs / 2 fs / 2 N 2 sin( f / f S ) where PS is the average signal power, S ( f ) denotes the PSD (power spectral density) of phase noise, N is the number of sub-carrier, and f s is the DFT sampling rate. From above Eq. , the variance of N i increases with N and PS . As a special case, in conventional single-carrier modulation (N=1) the variance of N i approaches zero since there is no other sub-carrier to cause inter-channel interference. If the common phase error can be corrected completely, the phase noise effect on an OFDM signal is equivalent to the increase of the white noise due to ICI. Wireless Communication Technologies 2.5.3 5 Lorentzian spectrum In general, the PSD of an oscillator signal with phase noise can be modeled by a Lorentzian spectrum, whose two-sided spectrum is given by Phase Noise 0 1 f f c / f12 2 where f1 is the 3-dB linewidth of the oscillator signal. -100 dBc/Hz @100kHz -95 dBc/Hz @100kHz -90 dBc/Hz @100kHz -85 dBc/Hz @100kHz -10 -20 -30 -40 PSD [dBc/Hz] Sd ( f ) 1 / f1 -50 -60 -70 -80 -90 -100 0 10 1 10 2 3 10 10 4 10 5 10 Hz Wireless Communication Technologies 2.5.3 6