Noise in Modulation The primary figure of merit for signals in analog systems is signal-to-noise ratio. The signal-tonoise ratio is the ratio of the signal power to the noise power. While the amount of external ambient noise is not always within the control of a communications engineer, the way in which we can distinguish between the signal an the noise is. The demodulation process for any modulation system seeks to recreate the original modulating signal with as little attenuation of the signal and as much attenuation of the noise as possible. The signal-to-noise ratio changes as the modulated signal goes through the demodulation process. We wish the signal-to-noise ratio to increase as a result of the demodulation process. The extent to which the signal-to-noise ratio increases is called the signal-to-noise improvement, or SNRI. This signal-to-noise improvement is equal to the ratio of the demodulated signal-to-noise ratio to the input signal-to-noise ratio. The calculation of the signal-to-noise improvement (SNRI) is shown in the following diagram. si(t) si(t) Demodulator ni(t) no(t) 2 Psi s i 2 Pni n i 2 Pso s o 2 Pno n o The notation < > means time average. SNR i SNRI Psi SNR o . Pni SNR o SNR i Pso Pni Pno Psi Pso . Pno Pso Pni Psi Pno . In determining the ability of a demodulator to discriminate signal from noise, we calculate Psi, Pso, Pni and Pno and plug these values (or expressions) into the expression SNRI Pso Pni Psi Pno . Noise in Amplitude Modulated Systems • The SNRI will be calculated for DSB-SC and DSB-LC. • A DSB-SC demodulator is linear : the signal and the noise components can be considered separately. • A DSB-LC demodulator is non-linear : the signal and the noise components cannot be treated separately. The expression for the modulated carrier for DSBSC is x c ( t ) m ( t ) cos w c t . Demodulation of this signal is performed by multiplying xc(t) by cos wct and low-pass filtering the result. d(t) si(t) = xc(t) X coswct LPF so(t) The result of this demodulation can be seen by analysis: d ( t ) x c ( t ) cos w c t m ( t ) cos w c t 2 m ( t ) [1 cos 2w c t ]. 1 2 Low-pass filtering the result eliminates the cos 2wct component. s o (t ) 1 2 m ( t ). Based upon these relationships, we can find the ratio of the signal input power and the signal output power. Psi m ( t ) cos w c t m ( t ) 2 2 2 1 2 Pso m ( t ) 2 Pso Psi 1 2 2 2 m (t ) 2 m (t ) m (t ) 2 1 4 1 2 1 1 4 . 2 So, it seems, half of our SNRI calculation is done. The other half of the calculation deals with the noise. Let us use the quadrature decomposition to represent the noise: n ( t ) n c ( t ) cos w c t n s ( t ) sin w c t . It is this signal which will represent ni(t). The noise input to the demodulator (in the above form) is demodulated along with the signal. Because the demodulator is linear, we can treat the noise separately from the signal. ni(t) = nccoswct – nssinwct X dn(t) coswct LPF no(t) d n ( t ) n ( t ) cos w c t n c cos w c t n s sin w c t cos w c t n c cos w c t n s sin w c t cos w c t 2 n c 12 [1 cos 2w c t ] n s 1 2 sin 2w c t After dn(t) passes through the low-pass filter, all we have left is n o (t ) 1 2 n c ( t ). Now, we calculate the power in the noise. Pno n c ( t ) 2 1 2 2 2 n c (t ) 1 4 1 4 2 n (t ) . The ratio of the noise powers becomes Pni Pno 2 n (t ) 1 4 2 4. n (t ) Finally, our signal-to-noise improvement becomes SNRI Pso Pni Psi Pno 1 2 4 2. Thus, the DSB-SC demodulation process improves the signal-to-noise ratio by a factor of two. The expression for the modulated carrier for DSBLC is x c ( t ) 1 m ( t ) cos w c t . (This is a simplified version of the more accurate expression which takes into account the modulation index. In this simplified version, the modulation index is equal to one.) The demodulation process extracts m(t) from xc(t). The input signal power is Psi 1 m ( t ) cos w c t 2 2 1 m ( t ) 2 1 2 1 2 m ( t ) m 1 m (t ) 2 1 2 2 (t ) 1 2 1 m (t ) 2 1 2 . The modulating signal m(t) is assumed here to have zero mean. If m(t) = coswmt, then Psi 1 1 2 1 2 3 4 . The output power is simply Pso m ( t ) 2 which is equal to ½ if m(t) = coswmt. Thus, when m(t) is a sinewave, our signal power ratio is Pso Psi 1 2 3 4 2 3 . Now, we deal with the noise. As mentioned previously, since the demodulation process is non-linear, we cannot treat the signal and the noise separately. To deal with the noise, we retain the carrier, but set the modulating signal to zero. Thus, the signal that we demodulate is 1 cos w c t n ( t ). The resultant output from demodulating this signal will be the output noise power. To demodulate this signal (analytically), we use the quadrature decomposition for n(t). The input to our demodulator becomes cos w c t n c ( t ) cos w c t n s ( t ) sin w c t . We then work with this expression: cos w c t n c ( t ) cos w c t n s ( t ) sin w c t 1 n c ( t ) cos w c t n s ( t ) sin w c t . At this point we make an assumption (which turns out to be quite reasonable in many cases): n s ( t ) 1 n c ( t ) . In words, the noise (either the in-phase or quadrature component) is much smaller in magnitude that that of the signal coswct. With this assumption the input to the demodulator becomes 1 n c ( t ) cos w c t . We recall that when the input 1 m ( t ) cos w c t is applied, the output is m (t ). Similarly, when the input is 1 n c ( t ) cos w c t is applied, the output is n c (t ). Thus, the noise output is n o ( t ) n c ( t ). The output noise power is 2 Pno n c ( t ) n ( t ) Pni . 2 Thus, Pni 1, Pno and SNRI Pso Pni Psi Pno 2 3 1 2 3 . We see that the signal-to-noise improvement for DSB-LC is not as good as that of DSB-SC. The reason for using DSB-LC is that it is relatively easy to demodulate. (Standard broadcast AM [530-1640 kHz], uses DSB-LC.) Noise in Frequency Modulated Systems • The frequency modulation and demodulation process is non-linear. • As with DSB-LC, we cannot treat the signal and the noise separately. The expression for the modulated carrier for FM x c ( t ) cos[ w c t k f m ( t ) ]. As with DSB-LC AM, he demodulation process extracts m(t) from xc(t). This extraction process consists of three steps: 1. Extract argument from cos(). 2. Subtract wct. 3. Differentiate what is left to get kfm(t). Based upon these three steps we can quickly get the input signal power and the output signal power. s i ( t ) x c ( t ) cos( something Psi cos ( something 2 ). ) 1 2 . s o ( t ) k f m ( t ). 2 Pso k f m ( t ) k f 2 2 2 m (t ) . Without much loss of generality, we can assume that m(t) = coswmt. Pso 1 2 2 kf . We now have the signal power ratio: Pso Psi 1 2 kf 1 2 2 2 kf . As with the DSB-LC AM, the effective noise input comes along with an unmodulated carrier: cos[ w c t 0 m ( t ) ] n ( t ). If use the quadrature decomposition of the noise, we have, as the effective noise input cos w c t n c ( t ) cos w c t n s ( t ) sin w c t , or, 1 n c ( t ) cos w c t n s ( t ) sin w ct. This expression can be combined into a single sinewave R cos( w c t ), where 2 R [1 n c ( t )] n s ( t ). 2 tan 2 1 n s (t ) 1 n c (t ) . At this point we make a simliar assumption to what we made with DSB-LC AM: n s ( t ) 1 n c ( t ) 1 . (The new part is the 1.) Using this assumption we have tan 1 n s ( t ) n s ( t ). (This last is true because tan-1 x x for small values of x.) With the assumptions given, our effective noise input becomes R cos( w c t n s ( t ) ) We may now apply our three demodulation steps: 1. Extract argument from cos(). 2. Subtract wct. 3. Differentiate what is left to get kfm(t). 1. w c t n s (t ) 2. n s (t ) 3. n s (t ) Thus, the output noise is n o ( t ) n s ( t ). All that remains to be done is to find the noise power from the noise signals [ni(t) and no(t)]. The noise power will be found from the power spectral densities of the input noise and the output noise. The input noise is additive white Gaussian noise. The power spectral density of the input noise is S ni ( f ) N0 . 2 The output noise is the derivative of the quadrature component of the (input) noise [ns(t)]. We know, from a previous exercise, that S ns ( f ) N 0 . We need to find the power spectral density of the derivative of the quadrature component. To find this we multiply Sns(f) by the square of the transfer function for the differentiation operation. The effect of the differentiation operation is shown in three ways: . ns(t) d dt ns(t) Ns(f) j2pf No(f) Sns(f) |2pf|2 Sno(f) Thus, the power spectral density of the output noise is S n 0 ( f ) 2p f 2 N 0. Now that we have the power spectral densities of the input and the output spectra, all that remains to find the power is to integrate the respective power spectral densities over the appropriate frequencies. The input noise is a bandpass process. Let BT be the bandwidth. Sni(f) N0/2 f BT Integrating the power spectral density, we have Pni 2 N0 2 ( BT ) N 0 BT . The input noise is a lowpass process. Let W be the bandwidth. Sno(f)=N04p2f2 f -W W Integrating the power spectral density, we have W Pno N 0 4p 2 2 f df W N 0 4p 2 f 3 3 W W 2 N 0 4p 8 p N 0W 2 2 3 W 3 3 . 3 Thus, the ratio of the noise powers is Pni Pno N 0 BT 8 N 0p W 2 3 3 3 BT 8p W 2 3 . Finally, the signal-to-noise improvement is SNRI Pso Psi Pni Pno 2 kf kf — peak frequency deviation BT — modulated carrier bandwidth W — demodulated bandwidth 3 BT 8p W 2 3 . Exercise: Suppose that the modulated carrier bandwidth is given by Carson’s rule: BT 2 f m ( 1). Further suppose that the demodulated bandwidth is the bandwidth of the modulating signal: W fm Show that the resultant signal-to-noise improvement is SNRI 3 ( 1). 2 Use kf 2p f m . Exercise: Suppose that the modulated carrier bandwidth is simply twice the modulating frequency: BT 2 f m . Further suppose that the demodulated bandwidth is the bandwidth of the modulating signal: W fm Show that the resultant signal-to-noise improvement is SNRI 3 . 2 Improvement of FM SNRI Using De-Emphasis • The signal-to-noise improvement for FM, while good, can be improved by minimizing the noise. • The “weakest link” where it comes to the noise amplification is the differentiator: the noise increases as the cube of the bandwidth. • If the “high-pass” effect of the differentiator can be offset by a low-pass filter, the noise will be attenuated, and the SNRI will be improved. Sno(f)=N04p2f2 The “weakest link”: f -W W W Pno W 8p N 0W 2 N 0 4p f df 2 2 3 3 . Now, let us insert a low-pass filter whose transfer function is H LP ( f ) 1 1 Our output noise power becomes jf f1 . W ' Pno W N 0 4p f 2 1 2 f 2 f1 We can integrate this function by letting f f1 tan . df . tan ' Pno f 1 3 1 W f1 tan N 0 4p 1 tan tan f 1 N 0 4p 1 W f1 2 tan tan f 1 N 0 4p 3 tan 2 2 1 W f1 3 2 2 1 W f1 tan sec d 2 2 sec 2 sec d 2 1 W f1 [sec 1 ] d 2 tan 1 W f1 f 1 N 0 4p ( 2 )[ 3 2 W f1 tan 1 W f1 ]. We now define a quantity called the signal-to-noise improvement improvement (no mistake). This is the improvement in the SNRI as a result of deemphasis: Pno Pno ' . The factor becomes 8p N 0W 2 Pno Pno ' 3 3 3 2 f 1 N 0 4p ( 2 )[ Wf tan 1 W 3 W 1 f1 3f [ 1 W f1 ] 3 tan 1 W f1 . ] A plot of this factor is plotted on the following slide. Increase in SNRI Using De-Emphasis 16 14 12 (dB) 10 8 6 4 2 0 -1 10 10 0 W / f1 10 1