Differential PCM (DPCM) When a voice or video signal is sampled at a rate slightly higher than the Nyquist rate, the resulting sampled signal is found to exhibit a high correlation between adjacent samples. This means that in an average, the signal does not change rapidly from one sample to the next, with the result that the difference between adjacent samples has variance that is smaller than the variance of the signal itself. When these highly correlated samples are encoded as in PCM, the resulting signal contains redundant information. By removing this redundancy before encoding a more efficient coded signal is obtained. Knowing the post behavior of a signal up to a certain point in time, it is possible to make some inference about its future values; such a process is commonly called prediction.The fact that il is possible to predict future values of s(t) provide motivation for the differential quantization scheme shown in figure 3.3-1. The input to the quantizer, which is the difference between not quantized signal s(a7s) and a prediction signal, $(n7s) 1s called the prediction error e(n7s} since itis the amount by which the prediction filter fails to predict the input exactly enTs)=s(nTs)-s(nT)..Sampled input _} e(nTs @q (nT (ts) ouantizer | £218) ‘ Encoder JOPCM, Transmitter Sq (nTs) je —_______] —_| Prediction Filter €q (nTs) + —» Decoder |} —___—_>| Sq (nTs) >| LPF o> + Prediction Filter S(aTs) ¥ Receiver Figure 3.3-1 Differential Pulse Code Modulation systemA simple and effective approach to implement the prediction filter is to use a tapped-delay-line filter, with the basic delay set equal T,. The block diagram of this filler is shown in figure 3.3-2. The predicted sample value S(nTs) is modeled as a linear combination of p- past sample values of the quantized inputs, where, p is the prediction order. The quantizer output may be expressed as Cy (AT ANT H(A). srsnnnnneininersnnnnnnneeend 33-2) where g(nTs) - is the quantization error. According to DPCM transmitter Sq (1173) = S(NTS) +g (TS) oscsnrntnnaseninnnnenninnsnnnsennns (3.3-3)sa(nTs) Delay Delay Delay + |___z s(nTs) Figure 3.3-2 Tapped-Delay-Line Prediction Filter Substituting (3.3-2) in (3.3-3) to have: $4 (nTs ) = S(nTs ) + e(Ts) + q(Ts )But from (3.3-1), s(nTs) = e(nTs)+ S(nTs) therefore 5q(nTs) = s(nTs) + q(Ts) + (3.3-4) Equation (3.3-4) represents a quantized version of the input signal s(nTs). If the prediction is good, a quantization with less number of levels can be used to produce error the same as that of PCM with more levels, so in cach code word bits can be saved. The receiver for reconstructing the quantized version of the input is shown in figure 3.3-3. It was found that SVRy, for DPCM is the same as that for PCM and follows the 6-dB rule [Jayant and Noll] 984] and it is: SNRyg = 6.020 +a Where value of a for DPCM and speech signal ranges from (-3 the present sample.SqT, ~T5)* the latest approximation to it eq (Ts): the quantized version of nT.) . According to equations (3.4-1), (3, plotted as shown in figure 3.4-2 Sampled input e, (nTs) 4-2) and (3.4-3) the DM system can be >) Encoder > L____ | ‘Transmitter €q (nTs) »| Decoder LPF > Ts jt Accumulator Figure 3.4-2 Delta modulation system ReceiverThe quantizer output is then applied to an accumulator, > producing the result. Sq(nTs ) = ay sen[e(iTs )]= Ss eg (iT). isl If the input signal i is analog then a simple DM system can be plotted as shown in figure 3.4-3. -(3.4 -4) Pals) ~ eagle vee (t) ane e ye zi d(t} ae Hf x \—_ | Encoder | toe OR ) Sampler — J ‘Transmitter —+[ Decoder fm} LPF) }—» Figure 3.4-3 Delta modulation system Receiver‘The input to the comparator is e() = s()-5(0)... where S(t) the message signal. S(9 the reference signal. fA, e(t)> 9 d(t) = Asgal e(0)] | —A, e(t) <9 The output of the delta modulator is Ypy (4) = A sent e(t)] rot — nT.) ‘ee You =a) senfe(nT,)}6 ((—nT,) n=-00Thus the output of the delta modulator is a series of impulses, each having positive or negative polarity depending on the sign of e(t) at sampling instants. Integration of y p,, (t) gives: Fe SAsenlelIT, eoererennararnenmnmnnnnnnennennenenen 34-8) n=“ This is a staircase approximation of s(t) as shown in figure 3.4-1. Thus the receiver consists of integrator (accumulator) and a LPF with a band equal to original input bandwidth. The LPF rejects the quantization noise in the high frequency staircase waveform5(?).A binary sequence 11 10100000101000000101 is transmitted by DM system. Evaluate and sketch the resulting analog waveform that appears at the receiver output. Solution The analog waveform that appears at the receiver output is shown in figure 3.4-4. Data Sequence 441010000001 01 Stair-case 1 “approximation \fter smoothing filter WT Figure 3.4-4 Solution Of Example 3.4-1 w/ ma 7Example: Consider the input samples Xm) = (22, 2.2, 2.3, 2.6, 2.7, 2.8). Explain fitter (NTs) =xa(n-2) e(nTs)= x¢nt) XS (nts) xa(nTs) = X (nTs)+ea(nTs)prem gf oecoaer of ter Dice Siena! X (nts) Receiver eq(nTS) X (nTs)=Xq(n-1) Xa(n)= X (nTs)+e4(n) Initially ook OON ° 2 2 2 3 3 Received signal: {2, 2, 2, 3, 3, 3} Transmitter signal: {2.1, 2.2, 2.3, 2.6, 2.7, 2.8}