6.
Delta Modulation (DM)
Delta modulation is a simplified version of the simple DPCM shown in Sec. 5. In DM, the quantizer is a 1-bit (two level) quantizer with magnitudes . In other words, in DM, - Quantizer: 2 level, i.e., L = 2 * - Prediction filter: a pure delay device, i. e., xn = xn1 a simple DPCM described in Sec. 5 Remark. In DM, 1 bit/sample, the quantization noise , unless the dynamic range of en the correlation of .
x n and x n 1
Justification: By over sampling the input message signal x(t) (i.e., at a rate much higher than the Nyquist rate), the correlation between adjacent samples is increased purposely. In this way, the prediction error, so that the quantization error, can be very small.
A. Structures of DM Encode and Decode
DM Encode
DM Decode
xn
en + +
n 1
2-level Q D: Ts
yn
xn
...
yn
x n
x 1 n
D: Ts
Low-pass filter
xn = x
Output Accumulator Accumulator Fig. 6.1 DM System
where en = xn xn = xn x*1 n
* xn = xn 1
The 2-level quantizer: step - size
= 2
output
+
Special property of DM: y n = sgn( e n ) Accumulative property: if we choose the * initial state x 0 = 0 , then we have
input
x =x
* n
* n 1
+ yn = ( x
* n2
+ y n 1 ) + y n = L =
y
i =1
+ x = sgn( ei )
* 0
i =1
This means that to obtain x , one only to accumulate the values of n simplifies the block diagram Fig. 6.1 to Figure 6.2.
yn . This
xn
+ +
en
2-level Q
yn =
...
yn
Accumulator
x n
Low-pass filter
Accumulator Fig. 6.2. DM with integrators
B. Staircase Approximation
* x n 1 = x n
a staircase approximation to x n
Staircase approximation x*(t)
Ts
x(t) t
C. Quantization Noise
1. Slope-overload distortion
The step size = 2 is too small * for the staircase{xn } to follow a steep segment of x(t).
2. Granular noise
The step size is too large relative to the local slope characteristics of the input waveform x(t), therefore causing the staircase approximation to hunt around a relative flat segment of the input waveform. This type of noise is analogus to quantization noise in PCM. x(t) Staircase approximation x*(t)
Fig. 6.3. Small and slopeoverload distortion Fig. 6.4. Large and granular noise
Notice that Then
* xn = xn + q n
en = xn x*1 n
= ( x n x n 1 ) qn 1
To avoid the slope-overload distortion, we require:
dx(t ) max Ts dt
discrete approximation to the derivative of the message signal
Example 6.1. For sinusoidal signal x(t ) = A cos(2f ct + )
max dx (t ) = A2f c A2f cTs dt
, we have
Techniques to reduce the quantizati on noise : adaptive delta modulation (ADM)
Principle : i) increase during a steep segment of x(t ); ii) decrease during a flat segment of x(t )
Example 6.2. For the DM depicted in Fig 6.1., find the maximum output signal-toquantization noise ratio under the assumption of no slope-overload for sinusoidal signal
x(t ) = A cos( 2f ct + )
2 (2 )2 2 = = = 12 12 3
2 Q
Solution. Then
Assume that quantization error is uniformly distributed over [-, ].
We also assume that the psd of the error is flat in the interval: , . The filter T Ts in the DM decoder is a low-pass filter whose bandwidth is equal tos the message bandwidth W. Then the average noise power at the filter output is given by
2 2 Qout = Q
1 1
2W 2 2 = WTs Q = WTs 2 / Ts 3
For a sinusoidal input,
A 2 f c
Max
dx(t ) = A2f c dt
Since there is no slope-overload, we have
A2 2 P max = = Ts 2 8 2 f c 2 T s2
SNR
o, max
Pmax 3 = 2 Q out 8 2Wf c2 Ts3
Remark.
By doubling
f s , SNR o, max is increased by 9dB
Summary of Chapter 3 ( Chapter 5 in the text)
1. Transmission of PCM signals : block diagram from x(t ) at Tx to x* (t ) at Rx.
2. Binary signal detection in AWGN: the match filter as an optimal detector, block diagram of the Tx and Rx, computation of average probability of bit error for the on-off signaling and antipodal signaling, SNR and ratio E/N0. 3. Time division multiplexing (TDM) of PCM signals
2 2 X 4. Uniform quantization : = , and SNRo = 2 12 Q 2 Q
5. DPCM: transmitter and receiver structures, linear predictor and optimal predictor in the sense of minimizing the mean square error.
6. DM : (a) transmitter and receiver structures (b) staircase approximation dx(t ) (c) to avoid slope overload : max Ts dt
