Professional Documents
Culture Documents
الفصل 3 - اتصالات رقمية
الفصل 3 - اتصالات رقمية
الفصل 3 - اتصالات رقمية
Chapter Three
Source Coding Techniques
1- Introduction:
Analog waveform or signals are sampled into pulses. In digital pulses
modulation methods, the analog amplitude pulses are converted to digital
form. Thus each sample of the message signal is represented in binary (1, 0)
format.
2- Pulse Code Modulation:
The PCM technique samples the input signal x(t) at frequency 𝑓𝑠 ≥ 2𝑊. This
sampled pulse is then digitized by the analog to digital converter as shown in
Fig. 2-1.
The x(t) is bandlimited to W by LPF. The sample and hold circuit then samples this
signal at the rate above of Nyquist rate 𝑓𝑠 ≥ 𝑊. The sampled signal 𝑥(𝑛𝑇𝑠 )is discrete
in time and continuous in amplitude. The quantizer is convert it to q discrete level
by rounding each sample to fixed digital level with minimum error (quantization
error). The input to the quantizer 𝑥(𝑛𝑇𝑠 ) (for example) can take any values between
(−4𝛿 𝑡𝑜 + 4𝛿), the output of quantizer (𝑥𝑞 (𝑛𝑇𝑠 ) are available at
Digital Communication د محمود فرحان. ا: مدرس المادة
𝛿 3𝛿 5𝛿 7𝛿
± ,± ,± 𝑎𝑛𝑑 ± as shown in Fig. 2-2. Thus the maximum quantization
2 2 2 2
𝛿
error is ± .
2
The quantized signal (𝑥𝑞 (𝑛𝑇𝑠 ) is converted by encoder to v digits binary word, and
then converted to serial bit stream to generate single baseband signal as shown in
Fig. 2-3.
Since 𝑓𝑠 ≥ 2𝑊
∴ 𝐵𝑟 ≥ 𝑣𝑊
Figure 2-4 shows the block diagram of PCM receiver. The regenerator is to reshapes
the pulse and removes the noise. The signal is then converted into parallel digital
words for each sample.
The digital word is converted to its analog value 𝑥𝑞 (𝑡) along with sample and hold
(S/H), then passed through lowpass reconstruction filter to get 𝑦𝐷 (𝑡). There is
quantization error between reconstructed signal 𝑥(𝑘𝑇𝑠 ) and original signal 𝑥(𝑡) as
shown in figure 2-5. This can reduced by increasing bits ‘v’, but this increases the
bandwidth.
𝜀 = 𝑥𝑞 (𝑛𝑇𝑠 ) + 𝑥(𝑛𝑇𝑠 )
The range of amplitude of input 𝑥(𝑛𝑇𝑠 ) is −𝑥𝑚𝑎𝑥 𝑡𝑜 + 𝑥𝑚𝑎𝑥 and it is mapped into
q levels. So that total amplitude range 2𝑥𝑚𝑎𝑥 is divided into q levels with step size
𝛿.
2𝑥𝑚𝑎𝑥
𝛿=
𝑞
𝛿 𝛿
We have the maximum quantization error is ± , or 𝜀𝑚𝑎𝑥 = | |
2 2
𝛿 𝛿
2 2 𝛿
3 2
1 1 𝜀
𝐸 (𝜀 2 ) = ∫ 𝜀 2 𝑓𝜀 (𝜀)𝑑𝜀 = ∫ 𝜀 2 𝑑𝜀 = [ ]
𝛿 𝛿 3 −𝛿
𝛿 𝛿 2
−2 −2
1 3 1 3
2
1 (𝛿 ) ( )
𝛿 1 𝛿3 𝛿3 𝛿2
𝐸 (𝜀 ) = [ + ]= [ + ]=
𝛿 3 3 3𝛿 8 8 12
2
𝑉𝑛𝑜𝑖𝑠𝑒
The noise power=
𝑅
2
𝑉𝑛𝑜𝑖𝑠𝑒
Assume R=1, then the noise power (normalized)=
1
2)
𝛿 2 /12 𝛿 2
𝐸 (𝜀 = =
1 12
We have 𝑞 = 2𝑣 , so:
2𝑥𝑚𝑎𝑥 2𝑥𝑚𝑎𝑥
𝛿= =
𝑞 2𝑣
𝑆
Substituting in above equation:
𝑁
𝑆 𝑃 3𝑃
= 2 = 2
× 22𝑣
𝑁 2𝑥 (𝑥𝑚𝑎𝑥 )
( 𝑚𝑎𝑥
𝑣 ) /12
2
This equation shows that the signal to noise power ratio of quantizer increases
exponentially with increasing bits per sample. For normalized 𝑥𝑚𝑎𝑥
𝑆
= 3 × 22𝑣 × 𝑃
𝑁
𝑆 𝑆
( ) 𝑑𝐵 = 10𝑙𝑜𝑔10 ( ) 𝑑𝐵 = 10𝑙𝑜𝑔10 (3 × 22𝑣 ) = (4.8 + 6𝑣)𝑑𝐵
𝑁 𝑁
For normalized values of power the destination signal power ‘P’ is less than 1
So that
𝑆
( ) 𝑑𝐵 ≤ (4.8 + 6𝑣 )𝑑𝐵
𝑁
Digital Communication د محمود فرحان. ا: مدرس المادة
Example 1:
A television signal with bandwidth of 4.2 MHz is transmitted using binary PCM.
The number of quantizatin levels is 512. Calculate:
Solution:
i- 𝑞 = 2𝑣 → 512 = 2𝑣
Example 2:
The bandwidth of signal input to the PCM is restricted to 4 kHz. The input varies
from -3.8V to +3.8 V and has the average power of 30 mW. The required signal
to noise ratio is 20 dB.
Solution:
𝑆 𝑆
( ) 𝑑𝐵 = 10𝑙𝑜𝑔10 ( ) 𝑑𝐵 = 20𝑑𝐵
𝑁 𝑁
𝑆
∴ = 100
𝑁
𝑆 3𝑃 3×30×10−3 ×22𝑣
i- = (𝑥 2 )
× 22𝑣 → 100 = (3.8)2
𝑁 𝑚𝑎𝑥
1444
22𝑣 = = 24066.67
0.06
2𝑣𝑙𝑜𝑔2 = log(24066.67)
𝑣 ≅ 7 𝑏𝑖𝑡𝑠
ii- 𝐵𝑇 ≥ 𝑣𝑊 and for 20 multiplexed signals
Example 3:
Solution:
𝜀𝑚𝑎𝑥 = ∓0.001
𝛿
But 𝜀𝑚𝑎𝑥 = | |
2
𝛿
| | = 0.001 → 𝑡ℎ𝑒 𝑠𝑡𝑒𝑝 𝑠𝑖𝑧𝑒 𝛿 = 0.002
2
2𝑥𝑚𝑎𝑥
We have 𝛿= , 𝑎𝑛𝑑 |𝑥𝑚𝑎𝑥 | = 10 𝑣𝑜𝑙𝑡𝑠
𝑞
2×10
Then 0.002 = → 𝑞 = 10000
𝑞
H.W:
Any signal does not change fast, so that the value from present sample to next
sample does not differ by large amount. The adjacent samples of the signal carry
the same information with little difference as shown if figure 2-11.
DPCM works on the principle of prediction. The value of the present sample is
predicted from the past samples as shown in figure 2-12.
The comparator fined the difference between the actual sample value 𝑥 (𝑛𝑇𝑠 ) and
predicted signal 𝑥̂(𝑛𝑇𝑠 ) this is called error 𝑒(𝑛𝑇𝑠 ):
This error will be quantized and encoded by small number of bits. Thus number of
bits per sample are reduced in DPCM. The quantization error can be written as:
𝑒𝑞 (𝑛𝑇𝑠 ) = 𝑒(𝑛𝑇𝑠 ) + 𝑞(𝑛𝑇𝑠 )
The prediction filter input is:
𝑥𝑞 (𝑛𝑇𝑠 ) = 𝑥̂ (𝑛𝑇𝑠 ) + 𝑒𝑞 (𝑛𝑇𝑠 )
Substituting by 𝑒𝑞 (𝑛𝑇𝑠 ) yields
𝑥𝑞 (𝑛𝑇𝑠 )=𝑥̂ (𝑛𝑇𝑠 ) + 𝑒 (𝑛𝑇𝑠 ) + 𝑞(𝑛𝑇𝑠 )
We have
𝑥 (𝑛𝑇𝑠 ) = 𝑒(𝑛𝑇𝑠 ) + 𝑥̂(𝑛𝑇𝑠 )
The quantized error signals are summed up with prediction filter output to give the
quantized version of the original signal. The signal at the receiver differs from actual
signal by quantization error 𝑞(𝑛𝑇𝑠 ).
Digital Communication د محمود فرحان. ا: مدرس المادة
DM transmit only one bit per sample. That is the present sample value is compared
with the previous sample of approximated signal which confined to two levels
(−𝛿 𝑎𝑛𝑑 + 𝛿). If the difference is negative ‘0’ bit is transmitted and ‘1’ bit is
transmitted for positive difference, as shown in 2-6.
- DM transmitter:
The block diagram of DM transmitter is shown in figure 2-7
Figure 10 DM transmitter
The error between sampled value of 𝑥(𝑡) and last approximated sample is
given by:
Digital Communication د محمود فرحان. ا: مدرس المادة
Figure 11 DM receiver
- Advantageous of DM:
i- DM transmit only one bit for one sample. Thus the signaling rate and
transmission channel bandwidth is quite small for DM.
ii- The DM transceiver system is very much simple.
- Disadvantageous of DM:
Figure 12 distortion of DM
i- Slop overload distortion: This distortion arises because of large dynamic
range of input signal. In this case the step size 𝛿 is too small for staircase
signal 𝑢(𝑡) to follow the steep segment of 𝑥(𝑡). Thus there is large error
between those signals. This error called slop overload distortion. To reduce
this error the step size should be increased when slop of signal 𝑥(𝑡) is high.
But since the step size is fixed it is called Linear Delta Modulation (LDM).
ii- Granular Noise (Hunting): It is occur when the step size is too large
compared to small variation in the input signal 𝑥(𝑡) which can be
Digital Communication د محمود فرحان. ا: مدرس المادة
considered flat, while the staircase signal is oscillated by ±𝛿 around it. The
error in this case is called granular noise, so that step size should be small
to reduce this error.
The step size increases with steep segment of input signal and reduces
with small variation. This called Adaptive Delta Modulation (ADM).
At the receiver the logic for step size control is added is added as shown
in figure 2-9.
If one bit quantizer output is high ‘1’ the step size may be doubled for next
sample and vice versa as shown in figure 2-10.
Figure 15
Digital Communication د محمود فرحان. ا: مدرس المادة
The binary data (bk) is applied to the data encoder which generate the pulse
waveform x(t) represented mathematically by:
∞
Here µ is scaling factor and P(t) is the different from that of g(t)
We have 𝑡 = 𝑖𝑇𝑏 then:
∞