Bit Duration 𝜏 (𝑇𝑏 ) 0 1 1 0

𝑇𝑠 1
𝜏 = 𝑇𝑏 = = Assume 𝑛 = 2 bits per sample
𝑛 𝑛 𝑓𝑠 𝑇𝑏 𝑇𝑠
𝑇𝑠 1
𝑇𝑏 = =
2 2 𝑓𝑠
𝑏𝑖𝑡𝑠 𝑠𝑎𝑚𝑝𝑙𝑒𝑠 1
Bit Rate 𝑅𝑏 𝑅𝑏 = × = 𝑛 𝑓𝑠 𝑅𝑏 =
𝑠𝑎𝑚𝑝𝑙𝑒𝑠 𝑠𝑒𝑐 𝑇𝑏

Transmission bandwidth

As the time width of time domain signal increases the frequency domain signal width decreases.
1
The maximum bandwidth 𝑅𝑏 =
𝜏
However practically, minimum two bits are transmitted and the transmission bandwidth is
𝑅𝑏 1
𝐵. 𝑊. = =
2 2𝜏
Example Given the message m t = 10si𝑛 4𝜋 × 103 𝑡 is transmitted over a PCM system, find all the
parameters of the system and maximum quantization error?
Solution If we sample at Nyquist frequency where 𝑓𝑚 = 2000 𝐻𝑧 𝑓𝑠 = 2 𝑓𝑚 = 4 𝑘𝐻𝑧
𝑛 4𝜋𝑛
𝑡 = 𝑛𝑇𝑠 = The samples are m n = 10 𝑠𝑖𝑛 4 = 10 sin 𝜋𝑛 = 0 for 𝑛 = 0, 1, …
4
we therefore notice that for sinusoidal signals we need to sample above Nyquist frequency
i. e., we then sample at 10𝑓𝑚 = 20000 𝐻𝑧 𝑓𝑠 = 20 𝑘𝐻𝑧 𝑣𝑝𝑝 = 20 𝑣
𝑣𝑝𝑝 20 20
If we choose 𝑛 = 4 bits per sample, 𝐿 = 16 ∆= = 4= = 1.25 𝑣
𝐿 2 16
𝑅𝑏
𝑅𝑏 = 𝑛 𝑓𝑠 = 80 𝑘𝐻𝑧 𝐵. 𝑊. = = 40 𝑘𝐻𝑧
2
∆ 1.25
𝑄𝑒 |𝑚𝑎𝑥 = =
2 2
Example Given the message m t = 4si𝑛 4𝜋 × 103 𝑡 is transmitted through 8-level PCM system,
sampling rate is 5 times N.R.
1) find all the parameters of the system and maximum quantization error?
Given sampled values −3.2, −2.8, −0.1, 1.5, 3.9 𝑣
2) Find quantizer and encoder output 3) Plot quantizer characteristics
Solution 1) 𝑓𝑚 = 2000 𝐻𝑧 𝑓𝑠 = 5 𝑓𝑚 = 20 𝑘𝐻𝑧
𝐿 = 8 = 2𝑛 𝑛 = 𝑙𝑛 8 = 3 𝑛 = 3 bits per sample note 𝑙𝑛 = 𝑙𝑜𝑔2
𝑣𝑝𝑝 8 1 𝑇𝑠 1 1
∆= = =1𝑣 𝑄𝑒 |𝑚𝑎𝑥 = 𝑅𝑏 = 𝑛 𝑓𝑠 = 60 𝑘𝐻𝑧 𝑇𝑏 = = = 𝑚𝑠𝑒𝑐
𝐿 8 2 𝑛 𝑛 𝑓𝑠 60
1
𝐵. 𝑊. = = 30 𝑘𝐻𝑧
2𝑇𝑏
 Quantizer levels
Given sampled values −3.2, −2.8, −0.1, 1.5, 3.9 𝑣 S. V. Q. V.
𝑣𝑝𝑝
𝑣𝑝𝑝 = 8 and 𝐿 = 8 levels. ∆= =1𝑣 4.5
𝐿 𝟑. 𝟗
We take levels at ∆ = 1 𝑣 3.5 111 𝟑. 𝟓 111
2 2
2.5 110
S. V. −3.2, −2.8, −0.1, 1.5, 3.9 𝑣
𝟏. 𝟓 1.5 101 𝟏. 𝟓 101
Q. output −3.5, −2.5, −0.5, 1.5, 3.5 𝑣
0.5 100
2) Quantizer output −𝟎. 𝟏
−.5 011 −𝟎. 𝟓 011

−3.5, −2.5, −0.5, 1.5, 3.5 𝑣 −1.5 010

- Encoder output
−2.5 001 −𝟐. 𝟓 001
−𝟐. 𝟖
000 001 011 101 111 −𝟑. 𝟐
000
−3.5 −𝟑. 𝟓 000
3.5
3) quantizer characteristics
2.5

1.5

0.5
𝑣
−4 −3 −2 −.5 2 3 4

−1.5

−2.5

−3.5
 Example
The quality of video signal using 256 level quantizer PCM and 8 levels quantizer DPCM found to be similar
find the bandwidth for both system.
Solution

The bandwidth for both system is determined as follows:

𝐿 = 2𝑛 = 256 𝑛=8
𝐿 = 2𝑛 = 8 𝑛=3
This means: 8 bits per sample in PCM system are needed, while in DPCM, only 3 bits are required
provided the same quality is achieved.
PCM 𝐵. 𝑊. = 8𝑓𝑠 /2 DPCM 𝐵. 𝑊. = 3𝑓𝑠 /2

Example Lati Ex .2
A band-limited signal 𝑚(𝑡) of 3 kHz bandwidth is sampled at rate of 33% higher than the Nyquist rate. The
maximum allowable error in the sample amplitude (i.e., the maximum quantization error) is 0.5% of the peak
amplitude. Find the minimum bandwidth of the channel to transmit the encoded binary signal.
Solution 𝑓𝑠 = 2 (1 + 0.33) 𝑓𝑚 = 7.98 = 8 𝑘𝐻𝑧
∆ 2𝐴𝑚
maximum quantization error = = = 0.005𝐴𝑚
2 2𝐿
1
𝐿= = 200 = 2𝑛 𝑛 = 𝑙𝑜𝑔2 200 = 7.64 Therefore 𝑛 = 8 We choose 𝐿 = 28 = 256
0.005

8𝑛
The bandwidth is therefore 𝐵. 𝑊. = = 32 𝑘𝐻𝑧
2
Example Problems 6.1.6 Lathi Problems 6.2.1, 2 Lathi
25 signals each band-limited to 3.3 kHz are sampled at an 8 kHz rate and then time multiplexed. Calculate
minimum bandwidth required for the multiplexed signal if the pulse modulation used is: a) PAM b) binary
PCM with a required level resolution of > 0.5%
𝑇𝑠 0.125 1 1
a) PAM 𝑓𝑠 = 2𝑓𝑚 = 8𝑘𝐻𝑧 = = .005 𝑚𝑠𝑒𝑐 𝑅𝑏 = = = 200 𝑘𝐻𝑧
25 25 𝑁𝑓𝑠 .005
1 𝑅𝑏
𝐵. 𝑊. = = = 100 𝑘𝐻𝑧
2𝑇𝑏 2

b) binary PCM 𝑣𝑝𝑝 2𝐴𝑚 1

= 𝑛 ≥ 0.005 × 2𝐴𝑚 2𝑛 ≥ = 200 𝑛 = 𝑙𝑜𝑔2 200 = 7.64 𝑛≥8
𝐿 2 .005
𝑅𝑏
𝑅𝑏 = 25 × 𝑛 𝑓𝑠 = 25 × 8 × 8 = 1600 𝑘𝐻𝑧 𝐵. 𝑊. = = 800 𝑘𝐻𝑧
2
Example
A sinusoidal message signal is transmitted through PCM system where 𝑄𝑒 |𝑚𝑎𝑥 can be almost of 2% of peak
to peak amplitude of message signal, find number of bits per sample required?

∆ 𝑣𝑝𝑝 2𝐴𝑚 1
𝑄𝑒 |𝑚𝑎𝑥 = = = = 0.02𝐴𝑚 2𝑛 ≥ = 25 𝑛≥5
2 2𝐿 2𝑛:1 .04
Example
Two low pass signals, each bandlimited to 4 𝑘𝐻𝑧 are to be multiplexed into a single channel using pulse
amplitude modulation. Each signal is impulse sampled at a rate of 10 kHz if the time multiplexed
waveform is filtered by an ideal low pass filter before transmission:
(a) What is the minimum clock frequency (or baud rate) of the system?
(b) What is the minimum cut-off frequency of the LPF?

Solution
(a) The minimum clock frequency (or baud rate) of the system =𝑁𝑓𝑠 = 20 𝑘𝐻𝑧
(b) The minimum cut off frequency of the low pass filter = 4 𝑘𝐻𝑧
Example A signal 𝑚(𝑡) is band-limited to 3 kHz is sampled at a rate 33.3 % higher than the Nyquist rate. The
maximum acceptable error in the sample amplitude (the maximum quantization error) is 0.5 % of the
peak amplitude. The quantized samples are binary coded. Find the minimum bandwidth of a channel
required to transmit the encoded binary signal. If 24 signals are time division multiplexed, determine the
minimum transmission bandwidth required to transmit the multiplexed signal.

Solution
Nyquist frequency where 𝑓𝑚 = 3000 𝐻𝑧 𝑓𝑠 = 2 𝑓𝑚 = 6 𝑘𝐻𝑧
𝑓𝑠
Actual sample frequency where 𝑓𝑠 + = 8000 𝐻𝑧
3
maximum quantization error
∆ 𝑣𝑝𝑝 0.005 × 𝑣𝑝𝑝
𝑄𝑒 |𝑚𝑎𝑥 = = =
2 2𝐿 2

1 0.005 1 𝑛 ≥ 𝑙𝑛 200 = 7.64

= 𝐿≥ = 200 𝑛=8
2𝐿 2 .005

Therefore 𝑛 to be chosen 8, 𝐿 = 28 = 256

𝑅𝑏
𝑅𝑏 = 8000𝑛 = 8 × 8 = 64𝑘 𝑏𝑝𝑠 𝐵. 𝑊. = = 32 𝑘𝐻𝑧 PCM system
2
𝑅𝑏
𝑅𝑏 = 8000𝑛 = 24 × 8 × 8 = 1.536 𝑀𝑏𝑝𝑠 𝐵. 𝑊. = = 0.768 𝑀𝐻𝑧 TDM PCM system
2
Example
24 voice signals are sampled uniformly and then time division multiplexed. The sampling operation involved
flattop samples with 1 μs duration. The multiplexing operation includes provision for synchronization by adding
an extra pulse of sufficient amplitude and also 1 μs duration. The highest frequency component of each voice
signal is 3.4 kHz.
(1)Assuming a sampling rate of 8 kHz, calculate the spacing (guard) between successive pulses of the
multiplexed signal.
(2) Repeat your calculations using Nyquist rate sampling.
Solution 1) At sampling rate of 8 kHz, 𝑓 = 8 𝑘𝐻𝑧 𝑇𝑠 = 125 𝜇𝑠𝑒𝑐
𝑠

Per frame, 24 pulses + 1pulse for synchronization = 25 pulse

125
The time allocated for each channel is therefore = 5 𝜇𝑠𝑒𝑐
25
Since the pulse duration is 1 𝜇𝑠𝑐𝑒, the time between pulses is 5 − 1 = 4 𝜇𝑠𝑐𝑒
1
(2) Using Nyquist rate sampling: 𝑓𝑠 = 2 𝑓𝑚 = 2 × 3.4 = 6.8 𝑘𝐻𝑧 𝑇𝑠 = = 147 𝜇𝑠𝑒𝑐
147
6.8
The time allocated for each channel is therefore = 5.88 𝜇𝑠𝑒𝑐
25
Since the pulse duration is 1 𝜇𝑠𝑒𝑐, the time between pulses is 5.88 − 1 = 4.88 𝜇𝑠𝑒
Example Given the message m t = 10si𝑛 4𝜋 × 103 𝑡 is transmitted over a 4-PCM system, find:
1- 2- The peak signal to peak (maximum) quantization noise power.

𝑣𝑝𝑝 20 2
Solution ∆ = = = 1.25 Average signal power = 100 = 50 Average power of quantization noise = ∆ = 0.13
𝐿 16 2 12

1) The signal to average quantization noise power = 50 𝑆 50

0.13 | 𝑑𝐵 = 10𝐿𝑜𝑔 = 25.8
𝑁 0.13
∆2
2) Signal peak power = 100 Peak power of quantization noise = = 0.39
4
100 𝑆 10
The peak signal to peak quantization noise power = == 256 |𝑑𝐵 = 20𝐿𝑜𝑔 = 24
0.39 𝑁 ∆
2
Example
A zero order circuit is often used to reconstruct a signal 𝑚 𝑡 from its samples
Find the unit impulse response of this circuit. Input Delay - Σ Output
Find the transfer function H 𝑓 and sketch |𝐻 𝑓 |
+
𝑡 𝑇
𝑡−2
Solution 𝑕(𝑡) = 𝛿 𝜏 − 𝛿(𝜏 − 𝑇) 𝑑𝜏 = 𝑢 𝑡 − 𝑢(𝑡 − 𝑇) = 𝑟𝑒𝑐𝑡( )
𝑇
0
𝑕 𝑡

𝑇 𝐻 𝑓 T
𝑡−2
𝑕 𝑡 = 𝑟𝑒𝑐𝑡( ) 𝐻(𝑓) = T sinc 𝜋𝑓𝑇 𝑒 ;𝑗2𝜋𝑓𝑇/2
𝑇 𝑡
𝑇 𝑓
1/T

Example Find a signal 𝑚 𝑡 that is bandlimited to 𝑓𝑚 and whose samples are

𝑕 0 = 1 and 𝑕 𝑡 ∓ 𝑇𝑠 ) = 0 𝑕 𝑡 ∓ 2𝑇𝑠 ) = 𝑕 𝑡 ∓ 3𝑇𝑠 ) = ⋯ = 0

Solution
∞

𝑚(𝑡) = 𝑚 𝑘𝑇𝑠 𝑠𝑖𝑛𝑐 2𝜋𝑘𝑓𝑚 (𝑡 − 𝑘𝑇𝑠 ) 𝑘=0 𝑚 𝑡 = 𝑚 0 𝑠𝑖𝑛𝑐 2𝜋𝑘𝑓𝑚 𝑡 = 𝑠𝑖𝑛𝑐 2𝜋𝑘𝑓𝑚 𝑡
𝑘<;∞
 ∞

𝑚𝑠 𝑡 = 𝑚𝛿 𝑡 ∗ 𝑝 𝑡 = 𝑚𝛿 𝑛𝑇𝑠 𝑝(𝑡 − 𝑛𝑇)

𝑛<;∞
𝑃 𝑓
Reconstruction of the signal is called interpolation 𝑝 𝑡 1

sin 𝜋 𝑓𝑚 𝑡
𝑝 𝑡 is a sinc function given by 𝑝(𝑡) = 2𝑓𝑚 𝑡 𝑓
𝜋𝑓𝑚 𝑡 𝑓𝑚
−𝑓𝑚
∞
1
𝑀 𝑓 ∗ 𝛿(𝑓 − 𝑛𝑓𝑠 ) A sinc function filter
𝑇𝑠
𝑛<;∞

𝑝 𝑡
𝑚𝛿 𝑡

…
𝑡
𝑇𝑠 2𝑇𝑠 3𝑇𝑠
∞
sin 𝜋𝑓𝑚 (𝑡 − 𝑛𝑇𝑠 )
𝑚𝑠 𝑡 = 𝑚𝛿 𝑛𝑇𝑠 Ideal interpolation first order hold filter
𝜋𝑓𝑚 (𝑡 − 𝑛𝑇𝑠 )
𝑛<;∞

If 𝑝 𝑡 is a rectangular pulse it is an approximate interpolation

𝑝 𝑡
𝑚𝛿 𝑡

…
𝑡
𝑇𝑠 2𝑇𝑠 3𝑇𝑠 𝑡

Zero order hold filter is a poor approximation of the low pass filter Approximation interpolation
 Example A TDM is used to multiplex the four signals 𝑚1 𝑡 = cos(2𝜋 × 103 𝑡) , 𝑚2 𝑡 = 0.5cos(2𝜋 × 103 𝑡) ,
𝑚3 𝑡 = 2cos(4𝜋 × 103 𝑡) and 𝑚4 𝑡 = cos(8𝜋 × 103 𝑡)
(a) If each signal sampled at the same sampling rate, calculate the minimum sampling rate
(b) What is the commutator speed in revolution per second.
(c) Design a commutator which will allow each of the four signals to be sampled at Nyquist
rate for individual signals
(d) Where the guard interval 𝜏𝑔 = 𝜏 the pulse width, find baseband signal bandwidth

(e) If the PAM signal is further quantized and encoded, 3-bit PCM, find the system dynamic
range D.R., resolution and the maximum quantization error

Solution (a) the minimum sampling rate 𝑓𝑠 = 8 𝑘𝐻𝑧

(b) The commutator speed in revolution per second is 8000 revolution per second
(c) 𝑚4 (𝑡)
𝑚3 (𝑡)
𝑚2 (𝑡)
𝑚1 (𝑡) 𝑓𝑠 TDM PAM
TDM PAM frame
4 3 3 4 3

1 1 1 1 1 1
2 2 2 2 2 2

1 1
𝑇𝑠 = = = 125 𝜇𝑠
𝑓𝑠 8000
 (d) Where the guard interval 𝜏𝑔 = 𝜏 the pulse width, find baseband bandwidth
𝑇𝑠 125
The maximum sampling interval 𝑇𝑠 = 22𝜏 𝜏= = = 5.68 𝜇𝑠
1 22 22
The bandwidth = = 88 𝑘𝐻𝑧
2𝜏
(f) If the PAM signal is further quantized and encoded, 3-bit PCM, find the system dynamic range D.R.

The system resolution dynamic range 𝐿 = 2𝑛 = 8

Example TDM PAM TDM PCM

𝑚1 (𝑡)
Quantizer Encoder
𝑚2 (𝑡)
𝑓𝑠 = 16 𝑘𝐻𝑧 𝑳 = 𝟏𝟔 𝑛 Sequence of bits
𝑚3 (𝑡)

1- What is the system resolution?

2- Find the pulse rate of the TDM-PAM signal and bit rate of the TDM-PCM signal.
3- How many bits generated in an hour?

Solution
1-The system resolution dynamic range: 𝐿 = 16
2- The pulse rate of the TDM-PAM signal = 3𝑓𝑠 = 3 × 16 = 48 𝑘𝐻𝑧
3- The bit rate of the TDM-PCM signal = ln(𝐿) 𝑓𝑠 = 4 × 3 × 16 = 192 𝑘𝑏𝑝𝑠
Example
(a) Sketch the design of TDM and FDM systems each catering for 12 voice channels (telephone) of 4 𝑘𝐻𝑧
bandwidth.
(b) Calculate the bandwidths required for the above FDM and TDM systems.

Solution
(a)
1 2 12
4 12 1 2
3
𝑡 𝑓
0 4 8 48
1 1
𝑇𝑠 = = = 125 𝜇𝑠 FDM system
𝑓𝑠 8000
TDM system

(b) The bandwidth required for the above FDM systems = 12 × 4 = 48 𝑘𝐻𝑧

𝑁 𝑓𝑠 8000
The bandwidths required for the above TDM system = = 12 × = 48 𝑘𝐻𝑧
2 2
Differential PCM (DPCM)
There is still redundancy in the samples when sampling at NR. Removing these redundant samples leads to
a reduction of the number of bits. One way of doing this is by transmitting the difference between
successive samples 𝑑 𝑘 = 𝑚 𝑘 − 𝑚(𝑘 − 1), instead of sending the sample values 𝑚(𝑘). That the
difference between samples values is far less than the samples values, thus reducing the step size and the
number of bits consequently.
Generally Now, a signal 𝑚 𝑡 and its derivatives of all orders at instant 𝑡 can be given
according to Taylor series as follows
𝑚 𝑡 𝑇 2 𝑚2 𝑡 𝑇 2 𝑚3 𝑡
𝑚 𝑡 + 𝑇𝑠 = 𝑚 𝑡 + 𝑇𝑠 𝑑 + 𝑑 + 𝑑 + …
𝑑𝑡 2! 𝑑𝑡 2 3! 𝑑𝑡 3
𝑚 𝑡 + 𝑇𝑠 is a future of the signal value or at 𝑡 + 𝑇𝑠 and it represents the sum of the signal and its
derivatives at instant t. it means the signal can be predicted applying Taylor series. From the first
derivative the signal can be approximately predicted. 𝑚(𝑘 − 1).

𝑚 𝑡
𝑚 𝑡 + 𝑇𝑠 ≈ 𝑚 𝑡 + 𝑇𝑠 𝑑
𝑑𝑡
𝑚(𝑘).
After sampling, where 𝑡 → 𝑘𝑇𝑠 the signal becomes 𝑚 𝑡 → 𝑚 𝑘𝑇𝑠 → 𝑚(𝑘)

𝑚 𝑡 𝑚 𝑘 −𝑚 𝑘−1
and 𝑑 →
𝑑𝑡 𝑇𝑠
𝑚 𝑘 −𝑚 𝑘−1
𝑚 𝑘 + 1 ≈ 𝑚 𝑘 + 𝑇𝑠 [ = 2𝑚 𝑘 − 𝑚(𝑘 − 1)
𝑇𝑠

It shows the signal can be predicted from its previous sample 𝑚(𝑘 − 1)

The right side of the equation is the predicted value of 𝑚 𝑘

 The approximation can be improved by considering more of the signals derivatives and the prediction
formula is generalized

𝑚 𝑘 ≈ 𝑎1 𝑚 𝑘 − 1 + 𝑎2 𝑚 𝑘 − 2 + … + 𝑎𝑁 𝑚(𝑘 − 𝑁)

𝑚 𝑘 = 𝑎1 𝑚 𝑘 − 1 + 𝑎2 𝑚 𝑘 − 2 + … + 𝑎𝑁 𝑚(𝑘 − 𝑁)

Linear weighted sum of previous samples

It is an 𝑁 𝑡𝑕 order predictor, 𝑚 𝑘 is the prediction of 𝑚 𝑘 and 𝑎1 𝑎2 . . . are the coefficients. It is known by a

transversal filter or taped delay line as plotted.

Delay Delay Delay . . . . Delay Delay

𝑚 𝑘 𝑇𝑠 𝑇𝑠 𝑇𝑠 𝑇𝑠 𝑇𝑠

𝑎1 𝑎2 ... . . ... 𝑎𝑁;1 𝑎𝑁

𝑚 𝑘
Transversal filter or taped delay line

1𝑠𝑡 order predictor is given as

𝑚 𝑘 = 𝑎1 𝑚 𝑘 − 1 therefore the first order predictor is a time delay.
In DPCM, not the samples as in PCM, but the differences between samples 𝑑(𝑘) are transmitted. However,
the difference between predicted samples but not the difference between the actual samples is transmitted
Further, the estimation or the prediction is taking place on the quantized samples 𝑚𝑞 (𝑘).

𝑚 𝑡
𝑚 𝑘 𝑑 𝑘 L-level 𝑑𝑞 𝑘
sample + Quantizer
Encoder 𝐷𝑃𝐶𝑀
+
−
+
𝑚𝑞 𝑘
Predictor 𝑚𝑞 𝑘

𝐷𝑃𝐶𝑀 transmitter.
𝑑 𝑘 = 𝑚 𝑘 − 𝑚𝑞 (𝑘)
and quantized to get 𝑑𝑞 (𝑘) where 𝑑𝑞 𝑘 = 𝑑 𝑘 + 𝑄𝑒 (𝑘) where 𝑄𝑒 (𝑘) is the quantization error.

The predictor output 𝑚𝑞 𝑘 is fed back to its input so that the predictor input 𝑚𝑞 (𝑘) is

𝑚𝑞 𝑘 = 𝑚𝑞 𝑘 + 𝑑𝑞 𝑘

𝑚𝑞 𝑘 = 𝑚 𝑘 − 𝑑[𝑘] + 𝑑𝑞 𝑘
𝑚𝑞 𝑘 = 𝑚 𝑘 + 𝑄𝑒 (𝑘)
At the receiver 𝑚𝑞 𝑘 is generated locally and added to from the received 𝑚𝑞 (𝑘) is reconstructed.
𝑑𝑞 𝑘 𝑚𝑞 𝑘 Smoothing
𝐷𝑃𝐶𝑀 Decoder +
filter

𝑚𝑞 𝑘 Predictor
Logarithmic units
Usually applied to power
When large dynamic scale you need to use the logarithmic scale whether the quantity is a ratio or not
and is given by
𝑆
𝑥 𝑑𝐵 = 10log( )
𝑁
𝑆
𝑥 𝑑𝐵𝑤 = 10log(𝑆) Expressed in mw 𝑥 𝑑𝐵𝑚𝑤 = 10 log = 10 log 𝑠 − 30
1000
Digital Audio recording
The analog audio recording suffers a dynamic range problem. The dynamic range of i.e., an orchestral music
is in the range of 100-120 dB. This level has to be reduced in order to be able to record the music, since the
dynamic range of analog recording is limited to around 70 dB.
The digital audio recording is better since the dynamic range degradation can be avoided.
The audio signals coming from the left (L) and right (R) microphones in a recording system are sampled,
quantized and encoded into binary PCM digits, (ADC).

The frequency of the audio signal is limited to approximately 20 kHz. Thus the Nyquist frequency is 40
kHz. However the sampling frequency in audio system is chosen to be 44.1 kHz to allow some frequency
guard-band.
The PCM system is 16-bits and the signal to quantization noise ratio in this case is over 90 dB (6n).
The output is recorded then on a compact disc. The compact disc is first introduced in 1982.

Digital Audio transmission in telephone systems

The speech or voice signal represents almost all of audio transmission. The voice is sampled at 8000 samples
per second. The speech frequency content is limited to 3.4 kHz and the rest represents a guard interval.
The samples are quantized using a non-linear quantizer with 𝜇 = 255 compander, which reduces the required
bits to 8 bits per sample instead of 12 bps.
PCM or DPCM are widely used for digital speech transmission. Therefore typical data rate is 64000 bps for
PCM and 32000 bps for DPCM.
Several encoded speech signals (voice channels) are transmitted simultaneously over a single channel.
In other words several subscribers share the channel on time basis. It is known time division multiplexing TDM.
 Delta modulation DM)
Delta modulation in comparison with PCM or in comparison with DPCM is much simpler in hardware
implementation in the transmitter or in the receiver.
𝑚 𝑡 𝑚 𝑘 𝑑 𝑘 L-level 𝑑𝑞 𝑘
+ +
sample Encoder
Quantizer
− 𝐷𝑃𝐶𝑀
+
𝑚𝑞 𝑘
Predictor 𝑚𝑞 𝑘

𝐷𝑃𝐶𝑀 transmitter.

𝑑𝑞 𝑘 𝑚𝑞 𝑘 Smoothing
𝐷𝑃𝐶𝑀 Decoder +
filter
𝑚𝑞 𝑘 Predictor

𝑑𝑞 𝑘 𝐷𝑃𝐶𝑀 receiver

𝑚 𝑘 𝑑 𝑘

𝑚𝑞 𝑘
𝑚𝑞 𝑘 𝑚𝑞 𝑘 The integrator acts as a predictor.

𝑚𝑞 𝑘 The integrator acts as a predictor.

Discrete time model of a DM system

 Receiver

Transmitter Hardware implementation of a DM system

The output of the quantizer is only of two values ∆ or −∆ . The output of the quantizer is then encoded
using binary digit per sample.
At the receiver, the decoded value of the difference signal is added to preceding value of the receiver output
The delay unit and the adder represent and integrator as shown the block diagram. The input to the integrator is
an impulse sequence of period 𝑇𝑠 and strength ±∆.
The message signal 𝑚 𝑡 is compared with a stepwise predicted signal and the difference is quantized
into two levels according to the sign of the difference.
The demodulator consists of an integrator and a low pass filter.
Practically, the low pass filter can act as an integrator at the same time, thus the integrator is eliminated
and hence reducing the receiver complexity.
A simple low pass filter acts as an integrator.
There are problems associated with DM as follow: the output of the integrator is step and this step keeps rising as
long as the difference is positive until it catches up and hence this interval is called start up, as shown the figure.
Slope overloading
𝑑𝑚 𝑡 𝑑𝑚𝑞 𝑡
Occurs when the slope of the signal greater than the slope of the stepwise signal
𝑑𝑡 𝑑𝑡
𝑑𝑚 𝑡 𝑑𝑚𝑞 𝑡 ∆
To avoid slop overloading: ≤ = = ∆𝑓𝑠
𝑑𝑡 𝑑𝑡 𝑇𝑠
Example Given m t = 𝐴𝑚 sin(2𝜋𝑓𝑚 𝑡) Find 𝑓𝑚 to avoid overloading case.
𝑑𝑚 𝑡 ∆𝑓𝑠
Solution |𝑚𝑎𝑥 = 𝑤𝑚 𝐴𝑚 ≤ ∆𝑓𝑠 Therefore 𝑓𝑚 ≤
𝑑𝑡 2𝜋𝐴𝑚
𝐴𝑚
Example Find 𝑓𝑠 to avoid slope overloading given 𝑓𝑚 = 600 𝐻𝑧 and = 15
∆

2𝜋𝑓𝑚
Solution 𝑓𝑠 ≥ 𝐴𝑚 = 2𝜋 × 15 × 600
∆
Adaptive DM

Continuously variable slope delta modulation CVSD

Sigma Delta modulation 𝚺 DM)

As demonstrated via block diagram

LPF
𝑚 𝑘 Σ
+
−

The integrator at the receiver

transferred to the transmitter.

LPF
+ Σ
−

Since the integration is a linear process,

the integrator moved behind the sigma.

The two integrators combined

Σ
 As the name suggested, sigma comes in front of a Delta modulator.
The integration takes place at the transmitter instead of at the receiver.
The benefit is the noise is shaped and low frequency degradation in DM is avoided.
The system encodes not the signal but the integral of it. The system performance is
therefore less affected by the rate of change of the signal. Receiver
It is known as a smoothed version of a delta modulator.
Transmitter
Systems comparisons
PCM DM DPCM ADM
Number of bits Can be 4, 8, One bit Less bits One bit
16, 32, …
Step size Step size Step size Step size Step size
fixed fixed adaptive adaptive
It depends
Quantization error It exists It exists It exists
on #levels
Other noise Not exist Over Not exist Not exist
loading
Complexity Complex simple simple simple
Bandwidth highest lowest lowest lowest

Applications Audio and Speech and Speech and Speech and

video images images video
The
Good luck