Digital Communications – Tutorial Questions on Source Coding

Problem 1:

A standard High Definition colour television (HDTV) with 1080p format uses 8 bits for each pixel
representation and a frame rate of 25 frames per second. Find the raw bit (i.e. prior to any
compression). Comment on the feasibility of broadcasting raw HDTV signal over a wireless channel.

Solution Problem 1:

For HDTV, the standard resolution for format 1080p = 1920 × 1080 pixels. Each of these is
represented by 8 bits, thus the raw bit rate is

HDTV raw bit rate: 1920 × 1080 × 3 × 8 × 25 = 1.24 Gbits/s

Note: The factor 3 comes from the coloured nature of the video signal, i.e. a digital colour video
signal consists of a brightness component called luma and two colour difference components called
chroma, each of which are sampled and processed separately. However, since human eye is more
sensitive to brightness than to colours, chroma signals are under-sampled as compared to luma
signal, therefore the actual bit rate expansion factor is less than 3, but 3 can be taken as a first
approximation.

The raw bit rate of HDTV is simply too large to be broadcast over a wireless channel realistically,
however HDTV signals are typically compressed to 10 – 15 Mb/s which makes their broadcast
feasible.

Problem 2:

Figure 1. Actual Speech Signal s(t) and Quantized Speech Signal s[n]

Consider s’[n] be the samples representing an actual speech signal s(t) and s[n] be the quantized
version, as shown in Figure 1. The values of two signals for samples at instants 5, 10, 15, 20, 25, 30,
35, 40, 45, 50, (i.e. N=10) are as follows:

s’[n]= -0.12, -0.72, -0.12, 0.6, 0.5, 0.75, 0.5, 0.3, 0.2, 0.1

s[n]=0, -0.5, 0, 0.5, 0.25, 0.75, 0.25, 0, 0, 0

Compute the arithmetic or numerical MSE for the quantized signal.

Solution Problem 2:

The difference between original and quantized signals is:

𝑢[𝑛] = 𝑠[𝑛] − 𝑠 ′ [𝑛]

𝑢[𝑛] =0.12, 0.22, 0.12, -0.1, -0.25, 0, -0.25, -0.3, -0.2, -0.1

The MSE of the quantized signal is therefore:

𝑁−1 10−1
1 1
𝑀𝑆𝐸 = ∑ 𝑢2 [𝑛] = ∑ 𝑢2 [𝑛] = 0.0352
𝑁 10
𝑛=0 𝑛=0

Problem 3:

A band-limited signal has a bandwidth equal to 3400 Hz. What sampling rate should be used to
guarantee a guard band of 1200 Hz?

Solution Problem 3:

The minimum sampling rate to guarantee a guard band of 𝐵𝐺 is 𝑓𝑠 = 2 ∗ 𝐵 + 𝐵𝐺 = 2 ∗ 3400 +

1200 = 8000 Hz.

Problem 4:

Assume 𝑥(𝑡) has a bandwidth of 40 kHz.

a) What is the minimum sampling rate for this signal?

b) What is the minimum sampling rate if a guard band of 10 kHz is required?

Solution Problem 4:

a) The minimum sampling rate is 𝑓𝑠 = 2 ∗ 𝐵 = 80 kHz

b) The minimum sampling rate with guard band 𝐵𝐺 is 𝑓𝑠 = 2 ∗ 𝐵 + 𝐵𝐺 = 90 kHz

Problem 5:

Consider a signal with the power spectral density shown in Figure 2 (a). Find the alias induced SDR if
the signal is sampled at 90% of its Nyquist rate and the reconstruction filter has an ideal rectangular
amplitude response.
Figure 2. Signal Power Density before and after sampling

Solution Problem 5:

The sampling rate is given by:

𝑓𝑠 = 0.9 ∗ 2𝑓𝐻

𝑓𝑠 = 0.9 ∗ 2 ∗ 100 = 180 kHz

The folding frequency (Figure 2(b)) is thus:

𝑓𝑠
= 90 kHz
2

For an ideal rectangular reconstruction filter, using equation:

𝑓𝑠 ⁄2
∫0 𝐺(𝑓)𝑑𝑓
𝑆𝐷𝑅 = ∞
∫𝑓 ⁄2 𝐺(𝑓)𝑑𝑓
𝑠

For such a simple 𝐺(𝑓), however, we can evaluate the integrals from the area under the triangles in
Figure (2 (b)), i.e.:

1⁄ (100 ∗ 1) − 1⁄ (10 ∗ 0.1)

𝑆𝐷𝑅 = 2 2 = 99 = 20.0 dB
1⁄ (10 ∗ 0.1)
2

Problem 6:

A bandpass signal is centred around a carrier frequency of 9.5 kHz with bandwidth of 1 kHz. Find the
lower and maximum bounds of the sampling frequency for this signal if the sampling induced
distortion is to be avoided. Identify the tolerance in the chosen sampling rate.

Solution Problem 6:

The highest and lowest frequency components in this signal are:

𝑓𝐿 = 9.0 kHz, 𝑓𝐻 = 10.0 kHz

10.0 kHz
The Quotient 𝑄 is this: 𝑄 = 𝑓𝐻 ⁄𝐵 = 1.0 kHz
= 10

Thus, the bandpass signal criterion becomes

 𝑄 𝑄−1
2𝐵 {𝑛 } ≤ 𝑓𝑠 ≤ 2𝐵 {𝑛−1}, where 𝑛 is a positive integer and 𝑛 ≤ 𝑄

10 10 − 1
2 × 103 { } ≤ 𝑓𝑠 ≤ 2 × 103 { }
𝑛 𝑛−1

Since 𝑄 is an integer in this problem, therefore 𝑛 can be set equal to 𝑄, i.e. 𝑛 = 10. Then

2 × 103 ≤ 𝑓𝑠 ≤ 2 × 103 (Hz)

There is zero tolerance in the sampling rate if distortion is to be completely avoided. Note: We can
achieve some tolerance if 𝑛 is chosen to be next lower integer, i.e. 9 but at the cost of an increase in
the sampling rate. If 𝑛 = 9, then

2222 ≤ 𝑓𝑠 ≤ 2250 (Hz) or 2.222 ≤ 𝑓𝑠 ≤ 2.250 (kHz)

In this case, sampling rate could be chosen to be 2.236 ± 0.014 kHz. The accuracy of sampling clock
will be ±0.63%.

Problem 7:

For a sinusoidal signal of amplitude 𝐴𝑚 that is uniformly quantized, how many bits per sample are
required to obtain SQNR of 37.76 dB.

Solution Problem 7:

For a sinusoid,

𝐴2𝑚
𝑃𝑋 =
2

𝑥𝑚𝑎𝑥 = 𝐴𝑚

𝑃𝑋
𝑆𝑄𝑁𝑅 = 3 ( 2 ) 22𝑣
𝑥𝑚𝑎𝑥

𝐴2 3
= 3 (2𝐴𝑚𝑎𝑥
2 ) 22𝑣 = 2 22𝑣
𝑚𝑎𝑥

𝑆𝑄𝑁𝑅|𝑑𝐵 = 10𝑙𝑜𝑔(3⁄2) + 6𝑣

37.76 = 1.76 + 6𝑣

𝑣 = 6 𝑏𝑖𝑡𝑠

Problem 8:

A compact disc (CD) records audio signals digitally. Assume that the audio signal bandwidth equals
15 kHz:
 (a) If the signals are sampled at Nyquist rate and the resulting samples are uniformly quantized
into L=65,356 levels and then binary-coded, determine the number of binary digits required
to encode each sample.
(b) If the audio signal has average power of 0.1 W and peak voltage of 1 V. Find the resulting
signal to quantization noise ratio (SQNR) of the uniform quantizer output in part (a). Also,
determine the number of binary digits per second (bits/s) required to encode the quantized
audio signal.
(c) For practical reasons, signals are sampled at a rate well above the Nyquist rate. In the case of
CDs, a sampling rate of 44,100 samples per second is used in practice. If N =65,356,
determine the number of bits per second required to encode the signal and the minimum
bandwidth required to transmit the encoded signal.
Solution Problem 8:

(a) Since 65536=216 , 16 binary digits are needed to encode each sample.
(b) Given that

𝑥𝑚𝑎𝑥 = 1, and 𝑃𝑋 = 0.1 W, 𝑁 = 2v = 65536. Thus

PX
SQNR = 3 × 22v × 2 = 91.1 dB
𝑥max

Also, the bandwidth is 15 kHz. The Nyquist rate is 30 kHz. We have 30000 samples/sec × 16
bit / sample = 480 kbits/s.
(c) The new bit rate with practical sampling is 44100 samples/sec × 16 bit / sample = 705.6
kb/s. The minimal bandwidth required is 705600 / 2 = 352.8 kHz.

Problem 9:

What is the resulting SQNR for a signal uniformly distributed on [-1, 1], when uniform quantization
with 256 levels is employed? The probability density function of uniform distributions is shown
below.

Probability Density Function of Uniform Distribution

Solution Problem 9:
1
𝑏 1 1 1 𝑥3 1 1 1 1 1
We have 𝑃𝑋 = 𝐸[𝑋 2 ] = ∫𝑎 𝑥 2 𝑓(𝑥)𝑑𝑥 = ∫−1 2 𝑥 2 𝑑𝑥 = . |
2 3 −1
= 6 . 1 − 6 . (−1) = 6 + 6 = 3
.
Therefore, using 𝑥𝑚𝑎𝑥 = 1 and 𝑣 = 𝑙𝑜𝑔2 256 = 8, we have

1 1
𝑆𝑄𝑁𝑅 = 3 ( . ) 22(8) = 216 = 65536 ≈ 48.16 𝑑𝐵
3 1

Problem 10:

A digital communications system is to carry a single voice signal using linear quantization. What bit
rate will be required if an ideal anti-aliasing filter with a cut-off frequency of 3.4 kHz is used at the
transmitter with sampling at 10% higher rate than Nyquist sampling rate, and the SQNR is to be kept
above 50 dB. Assume the ratio of peak-to-mean signal power for voice is 10 dB?

Solution Problem 10:

𝑃𝑋
𝑆𝑄𝑁𝑅 = 3 ( 2 ) 22𝑣
𝑥𝑚𝑎𝑥

or in dB

𝑃𝑋
𝑆𝑄𝑁𝑅|𝑑𝐵 = [ 2 ] + 6𝑣 + 4.8
𝑥𝑚𝑎𝑥 𝑑𝐵

50 = −10 + 6𝑣 + 4.8

60 − 4.8 = 6𝑣
𝑦𝑖𝑒𝑙𝑑𝑠
→ 𝑣 = 9.2

Rounding up, 𝑣 = 10 bits.

Rate: 𝑅 = 𝑓𝑠 𝑣 = (2 ∗ 3.4 𝑘𝐻𝑧) ∗ 1.1 ∗ 10 = 74.8 kb/s.

Problem 11:

A hi-fi music signal has a bandwidth of 20 kHz. Calculate the bit rate required to transmit this as a
linearly quantised signal maintaining a SQNR of 55 dB. What is the minimum baseband bandwidth
required for this transmission? (Assume that the signal’s peak to mean power ratio is 20 dB)

Solution Problem 11:

𝑃𝑋
𝑆𝑄𝑁𝑅|𝑑𝐵 = [ 2 ] + 6𝑣 + 4.8
𝑥𝑚𝑎𝑥 𝑑𝐵

55 = −20 + 6𝑣 + 4.8
 75 − 4.8 = 6𝑣
𝑦𝑖𝑒𝑙𝑑𝑠
→ 𝑣 = 11.7

∴ 𝑣 = 12 bits

𝑅 = 𝑓𝑠 𝑣 = (2 ∗ 20 𝑘𝐻𝑧) ∗ 12 = 480 kb/s

480
∴ 𝐵𝑊𝑚𝑖𝑛 = 2
= 240 kHz

Problem 12:

Speech signal is band limited to 3 kHz and sampled at the rate of 8 kHz. To achieve the same quality
of distortion of PCM requires 8bits/sample and DPCM requires 4 bits/sample. Determine the bit
rates required to transmit the PCM and DPCM encoded signals.

Solution Problem 12:

𝑅 = 𝑓𝑠 𝑣 b/s

𝑓𝑠 = 8 kHz

𝑣 = 8 bits/sample

For PCM, 𝑅 = 8 ∗ 8 k = 64,000 b/s

For DPCM, 𝑅 = 4 ∗ 8 k = 32,000 b/s

Problem 13:

It is desired to set up a central station for simultaneous monitoring of the electrocardiograms (ECGs)
of 10 hospital patients. The data from the rooms of the 10 patients are brought to a processing
centre over wires and are sampled, quantized, binary coded, and time-division multiplexed. The
multiplexed data are now transmitted to the monitoring station. The ECG signal bandwidth is 100 Hz.
The maximum acceptable quantization error in sample amplitudes is 0.25% of the peak signal
amplitude. The sampling rate must be at least twice the Nyquist rate. Determine the minimum cable
bandwidth needed to transmit these data.

Solution Problem 13:

𝑓𝑁𝑌𝑄 = 2 ∗ 100 = 200 Hz

𝑓𝑠 = 2 ∗ 𝑓𝑁𝑌𝑄 = 2 ∗ 200 = 400 Hz or samples/sec/signal

For 10 signals,

𝑓𝑇𝑜𝑡 = 10 ∗ 𝑓𝑠 = 10 ∗ 400 = 4000 samples/sec

 ∆ 0.25
Now, 2 = 𝑥
100 𝑚𝑎𝑥

2∗𝑥𝑚𝑎𝑥
Since, ∆= 𝑁

∆ 𝑥𝑚𝑎𝑥 0.25
∴ = = 𝑥
2 𝑁 100 𝑚𝑎𝑥
𝑦𝑖𝑒𝑙𝑑𝑠
→ 𝑁 = 400

Since N needs to be a power of 2, next of power 2 is 512.

∴ 𝑁 = 512

∴ 𝑣 = 𝑙𝑜𝑔2 𝑁 = 𝑙𝑜𝑔2 (512) = 9 bits

∴ bit rate 𝑅 = 𝑓𝑇𝑜𝑡 𝑣 = 4000 ∗ 9 = 36 kb/s

𝑅 36 𝑘
∴ 𝐵𝑊𝑚𝑖𝑛 = = = 18 kHz
2 2

Problem 14:

A signal band-limited to 1 MHz is sampled at a rate 50% higher than the Nyquist rate and quantized
into 256 levels using a 𝜇-law quantizer with 𝜇 = 255.

(a) Determine the signal-to-quantization ratio.

(b) The SNR (the received signal quality) found in part (a) was unsatisfactory. It must be
increased at least by 10 dB. Would you be able to obtain the desired SNR without increasing
the transmission bandwidth if it was found that a sampling rate 20% above the Nyquist rate
is adequate? If so, explain how. What is the maximum SNR that can be realized in this way?

Solution Problem 14:

𝑓𝑁𝑌𝑄 = 2 ∗ 1 MHz = 2 MHz

𝑓𝑠 = 2 ∗ 1.5 = 3 MHz
𝑦𝑖𝑒𝑙𝑑𝑠
𝑁 = 256 → 𝑣 = 𝑙𝑜𝑔2 𝑁 = 8 bits

𝜇 = 255 ∴ data rate 𝑅 = 𝑓𝑠 𝑣 = 3 ∗ 8 = 24 Mb/s

1
𝑆𝑄𝑁𝑅 = 3 [ ] 22𝑣
[ln(𝜇 + 1)]2

1
= 3[ ] 216 = 38.05 dB
[ln(256)]2

If we reduce the sampling rate and increase the value of 𝑁 so that the same data (or bandwidth)
rate is maintained, we can improve the SNR (because of increase in number of bits) with the same
bandwidth.
It is given that a sampling rate 20% above the Nyquist rate is adequate, therefore,

𝑓𝑠′ = 2 ∗ 1.2 = 2.4 MHz

𝑅 = 𝑓𝑠′ 𝑣 ′

R should still be 24 Mb/s if bandwidth must not increase.

∴ 24 ∗ 106 = 2.4 ∗ 106 ∗ 𝑣 ′

𝑦𝑖𝑒𝑙𝑑𝑠 24
→ 𝑣′ = = 10 bits
2.4

Thus 𝑣 has been increased from 8 to 10 bits/sample.

1
∴ 𝑆𝑄𝑁𝑅 = 3 [ ] 22∗10 = 50.09 dB
[ln(256)]2

Problem 15:

The output SQNR of a 13-bit PCM was found to be insufficient at 30 dB. To achieve the desired SQNR
of 42 dB, it was decided to increase the number of quantization levels 𝑁. Find the fractional increase
in transmission bandwidth required for this increase in 𝑁.

Solution Problem 15:

At 13bits/sample, SQNR = 30 dB.

Each bit increases SQNR by 6 dB.

Desired SQNR = 42 dB implies ∆𝑆𝑄𝑁𝑅 = 12 dB. Therefore, 2 more bits are required. Hence, total
number of bits = 13+2= 15 bits.

Fractional increase in B.W = (15-13)/13 = 15.38 %.

Problem 16:
1
A signal 𝑥(𝑡) is band-limited to 3kHz 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 𝑥𝑚𝑎𝑥 . The quantized samples are binary coded. Find the minimum bandwidth of a
channel required to transmit the encoded binary signal. If 24 such signals are time-division-
multiplexed, determine the minimum transmission bandwidth required to transmit the multiplexed
signal.

Solution Problem 16:

The Nyquist sampling rate is 𝑅𝑁𝑌𝑄 = 2 ∗ 3000 = 6000 Hz (samples per second). The actual
1
sampling rate is 𝑅𝐴 = 𝑓𝑠 = 6000 ∗ (1 3) = 8000 Hz.
The quantization step is ∆, and the maximum quantization error is ± ∆⁄2.

Therefore,

∆ 2𝑥𝑚𝑎𝑥 𝑥𝑚𝑎𝑥 0.5 𝑦𝑖𝑒𝑙𝑑𝑠

= = = 𝑥𝑚𝑎𝑥 → 𝑁 = 200
2 2𝑁 𝑁 100

For binary coding, N must be a power of 2. Hence the next higher value of N that is a power of 2 is
N=256.

Therefore, we need 𝑣 = 𝑙𝑜𝑔2 256 = 8 bits per sample. We require to transmit a total of 𝑅 = 𝑣𝑓𝑠 =
8 ∗ 8000 = 64,000 bits/s. Because we can transmit up to 2 bit/s per hertz of bandwidth, we require
𝑅
a minimum transmission bandwidth 𝐵𝑇 = = 32 kHz.
2

The multiplexed signal has a total of 𝑅𝑇𝑜𝑡 = 24 ∗ 64,000 = 1.536 Mbit/s, which requires a
minimum of 1.536/2=0.768 MHz of transmission bandwidth.

Problem 17:

A message source generates one of four messages randomly every one microsecond. The
probabilities of these messages are 0.4, 0.3, 0.2, and 0.1. Each emitted message is independent of
the other messages in the sequence.

(a) What is the source entropy?

(b) What is the rate of information generated by this source (in bits per second)?

Solution Problem 17:

(a)
𝑃1 = 0.4, 𝑃2 = 0.3, 𝑃3 = 0.2, and 𝑃4 = 0.1.
𝐻(𝑚) = −(𝑃1 𝑙𝑜𝑔𝑃1 + 𝑃2 𝑙𝑜𝑔𝑃2 + 𝑃3 𝑙𝑜𝑔𝑃3 + 𝑃4 𝑙𝑜𝑔𝑃4 ) = 1.846 bits (source entropy)
(b)
Since, it is given that the source generates one message every one microsecond, therefor
the sampling period is one microsecond, or

1
𝑓𝑠 = = 106 𝐻𝑧
10−6

Thus, the information rate is

𝑅 = 𝑓𝑠 𝐻(𝑚) = 106 × 1.864 bits/s

Problem 18:

A television picture is composed of approximately 300,000 basic picture elements (about 600 picture
elements in a horizontal line and 500 horizontal lines per frame). Each of these elements can assume
10 distinguishable brightness levels (such as black and shades of gray) with equal probability. Find
the information content of a television picture frame.
Solution Problem 18:

For a pixel, the probability of acquiring any one of 10 distinguishable brightness levels is same, thus
𝑃 = 1/10
1 1
Information per element = 𝑙𝑜𝑔2 𝑃 = 𝑙𝑜𝑔2 1/10 = 𝑙𝑜𝑔2 10 = 3.32 𝑏𝑖𝑡𝑠

Information per picture frame = 3.32 x 300,000 = 9.96 x 105 bits.

Problem 19:

A radio announcer describes a television picture orally in 1000 words out of his vocabulary of 10,000
words. Assume that each of the 10,000 words in his vocabulary is equally likely to occur in the
description of this picture (a crude approximation, but good enough to give an idea). Determine the
amount of information broadcast by the announcer in describing the picture. Would you say the
announcer can do justice to the picture in 1000 words? Use data in Prob. 18 to estimate the
information of a picture.

Solution Problem 19:

1 1
Information/word = 𝑙𝑜𝑔2 𝑃 = 𝑙𝑜𝑔2 1/10000 = 𝑙𝑜𝑔2 10000 = 13.3 𝑏𝑖𝑡. Information content of 1000
words = 13.3x1000=13,300 bits. The information per picture frame was found in Prob. 17 to be 9.96
x 105 bits. Obviously, it is not possible to describe a picture completely by 1000 words, in general.

Problem 20:

From the Old North Church Tower in Boston, Paul Revere’s friend was to show him one lantern if the
British Army began advancing overland and two lanterns if they chose to cross the bay in boats.

(a) Assume Revere had no way of guessing ahead of time what route the British might choose.
How much information did he receive when he saw two lanterns?
(b) What if Revere were 90% sure the British would march overland? Then, how much
information would the two lanterns have conveyed?

Solution Problem 20:

(a) Both options are equally likely

1
𝐼 = log 2 ( ) = 1 bit
0.5
(b) 𝑃(2 lanterns) = 0.1

1
𝐼 (2 lanterns) = 𝑙𝑜𝑔2 = 𝑙𝑜𝑔2 10 = 3.322 bits
0.1

Problem 21:
A source emits one of four messages randomly every one microsecond. The probabilities of these
messages are 0.5, 0.3, 0.1, and 0.1. Each emitted message is independent of the other messages in
the sequence.

(a) What is the source entropy?

(b) Obtain the compact binary Huffman code and find the average length of the code word.
Determine the efficiency and the redundancy of the code

Solution Problem 21:

(a)
4

𝐻(𝑚) = − ∑ 𝑃𝑖 𝑙𝑜𝑔2 𝑃𝑖 = 1.69 𝑏𝑖𝑡𝑠

𝑖=1

(b) See Table below for the compact binary code.

Message Probability Code Prob. Code Prob. Code

m1 0.5 0 0.5 0 0.5 0
m2 0.3 10 0.3 10 0.5 1
m3 0.1 110 0.2 11
m4 0.1 111

So the compact binary code is 010110111.

𝐿 = ∑ 𝑃𝑖 𝐿𝑖 = 0.5 × (1) + 0.3 × (2) + 0.1 × (3) + 0.1 × (3) = 1.7 𝑏𝑖𝑡𝑠
𝑖

𝐻(𝑚)
Efficiency = 𝜂 = × 100 = 99.4%
𝐿
The redundancy is 𝛾 = 100 − 𝜂 = 0.6%.

Problem 22:

A zero memory source emits six messages with probabilities 0.3, 0.25, 0.15, 0.12, 0.1, and 0.08. Find
the 4-ary (quaternary) Huffman code. Determine its average length, efficiency, and the redundancy.

Solution Problem 22:

In this case, we need to add one more dummy message to satisfy the required condition of 𝑟 +
𝑘(𝑟 − 1) messages and proceed as usual. The Huffman code is given in the table below:

Message Probability Code Prob. Code

m1 0.30 0 0.30 0
m2 0.25 2 0.30 1
m3 0.15 3 0.25 2
m4 0.12 10 0.15 3
m5 0.10 11
m6 0.08 12
m7 0.00 13
So the compact quaternary code is 023101112. Note that we have omitted the code 13 for the
dummy message as its probability of occurrence is zero. The length 𝐿 of the code is:

𝐿 = ∑𝑖 𝑃𝑖 𝐿𝑖 = 0.3 × (1) + 0.25 × (1) + 0.15 × (1) + 0.12 × (2) + 0.1 × (2) + 0.08 ×
(2) + 0 × (2) = 1.3 4-ary digits

Also,

𝐻(𝑚) = − ∑6𝑖=1 𝑃𝑖 𝑙𝑜𝑔4 𝑃𝑖 = 1.209 4-ary units

𝐻(𝑚) 1.209
Also, Efficiency = 𝜂 = 𝐿
× 100 = 1.3
× 100 = 93%, and the redundancy is 𝛾 = 100 − 𝜂 = 7%.

Problem 23:

In a QCIF (176x144) formatted video coder the residual signal is quantised by a 2-bit scalar quantizer.
If the frame rate is 10 frames per second, calculate the residual bit rate in the channel for:

(a) Binary encoding of the quantizer levels

(b) Optimum variable length encoding of the quantizer levels if quantizer levels, L1, L2, L3 and
L4 have unequal output probabilities of 0.5, 0.2, 0.15 and 0.15 respectively.

Comment on your results.

Solution Problem 23:

QCIF format at 10 frames per second and 2 bit/pixel coding

(a) The bit rate for binary is 176x144x2x10 = 506880 bits/sec

(b) For variable length

L1 = 0.5

L2 = 0.2

L3 = 0.15

L4 = 0.15

Encoding

For optimum variable length coding, we can proceed like the design of uniquely decodable prefix
codes as shown below:

L1 0
----------------
L2 1 0
------------
L3 1 1 0
-------------
L4 1 1 1
So we will have 1, 01, 001 and 000 encoding for the L1 – L4 quantization level.

The bit rate per pixel will be 0.5x1bit + 0.2x2bits + 0.15x3bitsx2 = 1.8 bits/pixel

Therefore total bit rate = 176x144x10x1.8 = 456192 bits/sec

Comment: We have a bit rate saving of 0.2/2 which is 10% by way of variable length coding to
match the statistical characteristics of the signal