-4-

Feedback Tutorial week 2 - Question Numbers below are as in textbook!

3.3. The cumulative distribution function for a continuous random variable, X, has the form:
 0, −∞< x ≤−2

P X (x) =  a (1 + sin(bx)), −2< x ≤2
 c, x>2

Find: (a) the values of a, b and c that make this a valid CD; (b) the probability that x is negative; and (c) the
corresponding probability density function.
3.4. A particular random variable has a cumulative distribution function given by:
 0, −∞<x≤0
P X (x) =  −x
1 − e , 0≤x<∞
Find: (a) the probability that x > 0.5; (b) the probability that x ≤ 0.25; and (c) the probability that 0.3 < x ≤ 0.7.
[0.6065, 0.2212, 0.2442]
MSc only question 3.5. The power reflected from an aircraft of complicated shape that is received by a radar
can be described by an exponential random variable, w. The pdf of w is:
(1/w o )e−w/w o , for w > 0
p(w) = 
 0, for negative w
where w o is the average amount of received power. What is the probability that the power received by the radar
will be greater than the average received power? [0.368].
3.6. An integrated circuit manufacturer tests the propagation delays of all chips of one particular batch. He dis-
covers that the pdf of the delays is well approximated by a triangular distribution with mean value 8 ns, maxi-
mum value 12 ns and minimum value 4 ns. Find: (a) the variance of this distribution; (b) the standard deviation
of the distribution; and (c) the percentage of chips which will be rejected if the specification for the device is 10
ns. [2.66, 1.63, 12.5%]
3.11. For a tossed dice:
(a) Use convolution to deduce the probabilities of the sum of two thrown dice being 2, 3 etc.
(b) 24 is the largest sum possible on throwing 4 dice. What is the probability of this event from the joint proba-
bility of independent events? Check your answer by convolution.
(c) What is the most probable sum for 4 dice? Use convolution to find the probability of this event.
(d) A box containing 100 dice is spilled on the floor. Make as many statements as you can about the sum of the
uppermost faces by extending the patterns you see developing in the convolution in parts (a), (b) and (c). [(a)
1/36, 2/36, . . ., 6/36, 5/36, 4/36, . . ., 1/36]; [(b) 7.7 × 10−4 ]; [(c) 14, 1.13 × 10−1 ]; [(d)
Σmin = 100, Σmax = 600, 501 possible Σ′s, most likely Σ = 350, PDF is truncated Gaussian approximation].
7.2. Under what assumptions does the Bayes receiver become a maximum likelihood receiver? Illustrate your
answer with the appropriate equations.
7.3. In a baseband binary transmission system a 1 is represented by +5 V and a 0 by −5 V. The channel is sub-
ject to additive, zero mean, Gaussian noise with a standard deviation of 2 V. If the a priori probabilities of 1 and
0 are 0.5 and the costs C 0 and C 1 are equal, calculate the optimum position for the decision boundary, and the
probability of error for this optimum position.
[ 0 V, 6.23 × 10−3 ]
7.4. What degenerate case of Bayes receiver design is being implemented in Problem 7.3? Assume that the
decision threshold in Problem 7.3 is incorrectly placed by being moved up from its optimum position by 0.5 V.
Calculate the new probability of error associated with this suboptimum threshold. [7. 4125 × 10−3 ]
7.5. Design a Bayes detector for the a priori probabilities P(1) = 2/3, P(0) = 1/3 with costs C 0 = 1 and C 1 = 3.
The conditional pdfs for the received variable v are given by:
1
e− v /2
2
p(v|0) =

√ (2π )
1
e− (ν −1) /2
2
p(v|1) = [v = 0. 905]

√ (2π )
MSc only question 7.6. A radar system operates in, zero mean, Gaussian noise with variance 5 V. If the
 -5-

probability of false alarm is to be 10−2 calculate the detection threshold. If the expected return from a target at
extreme range is 4 V, what is the probability that this target will be detected? [5.21, 0.2945]

Feedback Tutorial week 3

5.0. If a speech signal of bandwidth 3.4 kHz is to be quantised and transmitted with an 8-bit linear PCM what is
the practical channel bit rate? [64 kbit/s]
5.00. In comparison to Problem 5.0 what is the corresponding rate for 10-bit nonlinear companded PCM?
5.1. (a) Sketch the design of TDM and FDM systems each catering for 12 voice channels of 4 kHz bandwidth.
(b) In the case of TDM, indicate how the information would be transmitted using: (i) PAM; (ii) PPM; (iii) PCM.
(c) Calculate the bandwidths required for the above FDM and TDM systems. You may assume the TDM sys-
tem uses PPM with 2% resolution. [FDM 48 kHz; TDM 2.4 MHz]
5.2. Two lowpass signals, each band-limited 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 signal
waveform is filtered by an ideal lowpass filter (LPF) before transmission:
(a) What is the minimum clock frequency (or baud rate) of the system? [20 kHz]
(b) What is the minimum cut-off frequency of the LPF? [10 kHz]
5.7. Twenty-five input signals, each band-limited to 3.3 kHz, are each sampled at an 8 kHz rate then time-mul-
tiplexed. Calculate the minimum bandwidth required to transmit this multiplexed signal in the presence of noise
if the pulse modulation used is: (a) PAM; (b) quantised PPM with a required level resolution of 5%; or (c)
binary PCM with a required level resolution of > 0.5%. (This higher resolution requirement on PCM is normal
for speech-type signals because the quantisation noise is quite objectionable.) [(a) 100 kHz, (b) 2 MHz, (c) 800
kHz]
5.8. A hi-fi music signal has a bandwidth of 20 kHz. Calculate the bit rate required to transmit this as a, lin-
early quantised, PCM signal maintaining a SNq R of 55 dB. What is the minimum (baseband) bandwidth
required for this transmission? (Assume that the signal’s peak to mean ratio is 20 dB.) [480 kbit/s, 240 kHz]
MSc only question 5.10. If the peak signal to quantisation noise ratio, (SNq R) peak , for an n-bit, µ=255-law
companded, communications system is given by:
3 × 22n
(SNq R) peak =
[ln(1 + µ )]2
calculate the degraded output SNR in dB for large signal amplitudes in an n = 8 bit companded PCM system
with typical µ value when compared with a linear PCM system operating at the same channel transmission rate.
[15 dB]
MSc only question 5.11. An 8-bit A-law companded PCM system is to be designed with piecewise linear
approximation as shown in Figure 5.26. 16 segments are employed (with 4 co-linear near the origin) and the
segments join at 1/2, 1/4, etc. of the full scale value, as shown in Figure 5.26. Calculate the approximate SNR
for full scale and small signal values. [Full scale 36.3 dB, small signal 71.9 dB]

Feedback Tutorial week 4

6.10. For Gaussian variables use the tables of erf(x) supplied in Appendix A to find:
(a) The probability that x ≤ 5. 5 if x is a Gaussian random variable with mean µ = 3 and standard deviation
σ = 2; (b) the probability that x > 5. 5.
(c) Assuming the height of clouds above the ground at some location is a Gaussian random variable x with
µ = 1830 metres and σ = 460 metres, find the probability that clouds will be higher than 2750 metres.
(d) Find the probability that a Gaussian random variable with standard deviation of σ will exceed: (i) σ; (ii) 2σ;
and (iii) 3σ. [(a) 0.89, (b) 0.11, (c) 0.023, (d) 0.159, 0.023, 0.0014]
9.1. (a) Consider a source having an M = 3 symbol alphabet where P(x 1 ) = 1⁄2; P(x 2 ) = P(x 3 ) = 1⁄4 and symbols
are statistically independent. Calculate the information conveyed by the receipt of the symbol x 1 . Repeat for x 2
and x 3 . [I x 1 = 1 bit, I x 2 = I x 3 = 2 bits]
(b) Consider a source whose, statistically independent, symbols consist of all possible binary sequences of
length k. Assume all symbols are equiprobable. How much information is conveyed on receipt of any symbol.
[k bits]
(c) Determine the information conveyed by the specific message x 1 x 3 x 2 x 1 when it emanates from each of the
following, statistically independent, symbol sources: (i) M = 4; P(x 1 ) = 1⁄2, P(x 2 ) = 1⁄4, P(x 3 ) = P(x 4 ) = 1/8, [7
bits] (ii) M = 4; P(x 1 ) =P(x 2 ) =P(x 3 ) =P(x 4 ) = 1⁄4. [8 bits]
 -6-

9.2. (a) Calculate the entropy of the source in Problem 9.1(a). [11⁄2 bit/symbol]
(b) Calculate the entropy of the sources in Problem 9.1(c). [13⁄4 bit/symbol, 2 bit/symbol)]
(c) What is the maximum entropy of an 8 symbol source and under what conditions is this situation achieved?
What are the entropy and redundancy if P(x 1 ) = 1⁄2, P(x i ) = 1/8 for i = 2, 3, 4 and P(x i ) = 1/32 for i = 5, 6, 7, 8?
[3 bit/symbol, 2.25 bit/symbol]
9.6. Calculate the loss in information due to noise, per transmitted digit, if a random binary signal is transmitted
through a channel, which adds zero mean Gaussian noise, with an average signal-to-noise ratio of: (a) 0 dB; (b)
5 dB; (c) 10 dB. [0.6311; 0.2307; 0.0094 bit/binit]
S.1 A 625-line black and white television picture may be considered as being composed of 550 picture elements
(pixels) per line. Assume that each pixel is equiprobable among 64 distinguishable brightness levels. If this is
to be transmitted by raster scanning at a 25 Hz frame rate, calculate using the Hartley Shannon theorem the min-
imum bandwidth required to transmit the video signal assuming a 35 dB signal to noise ratio on reception. [4.43
MHz.]

Feedback Tutorial week 5

9.5. Estimate the maximum information content of a black and white television picture with 625 lines and an
aspect ratio of 4/3. Assume that 10 brightness values can be distinguished and that the picture resolution is the
same along a horizontal line as along a vertical line. What maximum data rate does a picture rate of 25 picture/s
correspond to and what, approximately, must be the bandwidth of the (uncoded and unmodulated) video signal
if it is transmitted using binary symbols? (If necessary you should consult Chapter 16 to obtain TV scanning
format information.)
[3.322 bit/symbol and 2.08 Mbit/picture; 52.08 Mbit/s; 26.04 MHz]
9.7. An information source contains 100 different, statistically independent, equiprobable symbols. Find the
maximum code efficiency, if, for transmission, all the symbols are represented by binary code words of equal
length. [7 bit words and 94.9%]
9.8 (a) Apply Huffman’s algorithm to deduce an optimal code for transmitting the following source over a
binary channel. Is your code uniquely decodable? m = 4; p(x 1 ) = 1⁄2, p(x 2 ) = 1⁄4, p(x 3 ) = 1/8 ,
p(x 4 ) = 1/8
(b) Define the efficiency of a code and determine the efficiency of the above code.
(c) Construct another code for the source of part (a) and assign equal length binary words ignoring the
probability of occurrence of the symbols. Calculate the efficiency of this source.
[0, 10, 110, 111, YES; 100%; 87.5%]
9.9 Design a Lempel-Ziv coding scheme with a history buffer of 16 8-bit characters and a maximum match
length of 4 characters. Use your coding scheme to encode the character sequence "1415927" where the next
character to be transmitted is to the right. You should assume the history buffer contains "1414213617920408"
with the last character to have been transmitted to the left. Calculate the compression factor obtained by your
code on this character sequence. [0.536]
10.1. Assume a binary channel with independent errors and P e = 0. 05. Assume k digit symbols from the
source alphabet are encoded using an (n, k) block code which can correct all patterns of three or fewer errors.
Assume n = 20. (a) What is the average number of errors in a block? [1] (b) Assuming binary transmission at
20,000 binary digits per second, derive the symbol error rate at the decoder output. [15.8 symbol/s]
MSc only question 10.2. A binary signal is transmitted through a channel which adds zero mean, white, Gaus-
sian noise. The probability of bit error is 0.001. What is the probability of error in a block of 4 data bits? If the
bandwidth is expanded to accommodate a (7,4) block code, (with a single error correcting capability) what
would be the probability of an error in a block of 4 data bits? [0.0040, 0.0019]

Feedback Tutorial week 6

10.3. Assume a systematic (n, k) block code where n = 4, k = 2 and the four codewords are 0000, 0101, 1011,
1110. (a) Construct a maximum likelihood (nearest neighbour) decoding table for this code. (b) How many
errors will the code correct? Are there any errors which are detectable but not correctable? [8 corrected and 8
detected but not corrected] (c) Assume this code is used on a channel with a P e = 0. 01. What is the probability
of having a detectable error sequence? What is the probability of having an undetectable error sequence?
[0.0388, 0.0006]
10.4. For a (6,3) systematic linear block code, the three parity check digits are:
 -7-

P 1 = 1 × I 1 ⊕1 × I 2 ⊕1 × I 3
P 2 = 1 × I 1 ⊕1 × I 2 ⊕0 × I 3
P 3 = 0 × I 1 ⊕1 × I 2 ⊕1 × I 3
(a) Construct the generator matrix G for this code. (b) Construct all the possible codewords generated by this
matrix. (c) Determine the error-correcting capabilities for this code. [single] (d) Prepare a suitable decoding
table. (e) Decode the received words 101100, 000110 and 101010.
[111100, 100110, 101011]
10.5. Given a code with the parity check matrix:
 1110 100 
H =  1101 010 
 
 1011 001 
(a) Write down the generator matrix showing clearly how you derive it from H. (b) Derive the complete weight
structure for the above code and find its minimum Hamming distance. How many errors can this code correct?
How many errors can this code detect? Can it be used in correction and detection modes simultaneously? [3, 1,
2, No] (c) Write down the syndrome table for this code showing how the table may be derived by consideration
of the all-zeros codeword. Also comment on the absence of an all-zeros column from the H matrix. (d) Decode
the received sequence 1001110, indicate the most likely error pattern associated with this sequence and give the
correct codeword. Explain the statement ‘most likely error pattern’. [0000010, 1001100]
10.6. When generating a (7,4) cyclic block code using the polynomial x 3 + x 2 + 1: (a) What would the gener-
ated codewords be for the data sequences 1000 and 1010? [1000110, 1010001] (b) Check that these codewords
would produce a zero syndrome if received without error. (c) Draw a circuit to generate this code and show
how it generates the parity bits 110 and 001 respectively for the two data sequences in part (a). (d) If the code-
word 1000110 is corrupted to 1001110, i.e. an error occurs in the fourth bit, what is the syndrome at the
receiver? Check this is the same syndrome as for the codeword 1010001 being corrupted to 1011001. [101]
10.7. A (7,4) block code has the parity check matrix as:
1 1 0 1 1 0 0
H = 1 1 1 0 0 1 0
 
1 0 1 1 0 0 1
This code can correct a single error. (a) Derive the generator matrix for this code and encode the data 1110. (b)
Derive a syndrome decoding for the code as described above and decode the received data 1101110. (c) Calcu-
late the maximum number of errors a (15,11) block code can correct. [1110010, 1]
10.8. Given the convolutional encoder defined by P 1 (x) = 1 + x + x 2 and P 2 (x) = 1 + x 2 , and assuming data is
fed into the shift register one bit at a time, draw the encoder: (a) tree diagram; (b) trellis diagram; (c) state tran-
sition diagram. (d) State what the rate of the encoder is. (e) Use the Viterbi decoding algorithm to decode the
received block of data, 10001000. Note: there may be errors in this received vector. Assume that the encoder
starts in state a of the decoding trellis in Figure 10.20 and, after the unknown data digits have been input, the
encoder is driven back to state a with two ‘flushing’ zeros. [0000]

1st output bit

i/p Z
-1
Z
-1 o/p

2nd output bit

10.9 For a convolutional encoder defined by P 1 (x) = 1 + x + x 2 , P 2 (x) = x + x 2 and, P 3 (x) = 1 + x: (a) State the
constraint length of the encoder and the coding rate. (b) The coder is used to encode two data bits followed by
two flushing zeros. Encode the data sequences: (i) 10 (ii) 11. Assume that the encoder initially contains all
zeros in the shift register and the left hand bit is the first bit to enter the encoder. (c) Take the two encoded bit
sequences from parts (b)(i) and (b)(ii) above and invert the second and fifth bit to create received codewords
with two errors in each. Decode the altered sequences using a trellis diagram and show whether or not the code
can correct the errors you have introduced. [3, 1/3, b(i) 101111110000, b(ii) 101010001110, both decode cor-
rectly]
 -8-

Feedback Tutorial week 7

11.1. A rectangular pulse OOK signal has an average carrier power, at the input to an ideal correlation receiver,
of 8.0 nW. The (one sided) noise power spectral density, measured at the same point, is 2. 0 × 10−14 W/Hz.
What maximum bit rate can this system support whilst maintaining a P e of 10−6 ? [17.7 kbit/s]
11.3. Define on-off keying (OOK), frequency shift keying (FSK), and phase shift keying (PSK) as used in
binary signalling. Compare their respective advantages and disadvantages. In a Datel 600 modem FSK tone
frequencies of 1300 Hz and 1700 Hz are used for a signalling rate of 600 bit/s. Comment on the consistency
between these tone frequencies and the signalling rate. What value would the upper tone frequency need to
have, to handle a 900 bit/s signalling rate? [1900 Hz]
11.8. A receiver has a mean input power of 25 pW and is used to receive binary FSK data. The carrier frequen-
cies used are 5 MHz and 5.015 MHz. The noise spectral density at the receiver has a value of 2. 0 × 10−16
W/Hz. If the error probability is fixed at 2 × 10−4 find the maximum data rate possible, justifying all assump-
tions. [10 kbits/s] If the transmitter is switched to PRK what new bit rate can be accommodated for the same
error rate? [20 kbit/s]
11.11. Find the maximum spectral efficiency of ISI free 16-PSK. What is the probability of symbol error of
this scheme for a received CNR of 24 dB if the maximum spectral efficiency requirement is retained? What is
the Gray coded probability of bit error? [4 bit/s/Hz, 1. 227 × 10−5 , 3. 068 × 10−6 ]

Feedback Tutorial week 8

12.X What is the free space path loss in dB for a 6 GHz radio system with a distance between terminals of 40
km? [140 dB]
12.10. Digital MPSK transmissions carried on a 6 GHz terrestial link require a bandwidth of 24 MHz. The
transmitter carrier power level is 10 W and the hop distance is 40 km. The antennas used each have 40 dB gain
and filter, isolator and feeder losses of 4 dB. The receiver has a noise figure of 10 dB in the specified frequency
band. Calculate the carrier-to-thermal-noise-power ratio at the receiver output. [62 dB] Comment on any
effects which might be expected to seriously degrade this carrier-to-noise ratio in a practical link. If a minimum
carrier-to-noise ratio of 30 dB is required for the MPSK modulation what is the fade margin of the above link?
[32 dB]
12.11. A QAM 4 GHz link requires a bandwidth of 15.8 MHz. The radiated carrier power is 10 W and the hop
distance 80 km. It uses antennas each having 40 dB gain, has overall (transmit and receive) filter, isolator and
antenna feeder losses totalling 7 dB and a receiver noise figure of 10 dB. If the Rayleigh fade margin is 30 dB,
calculate the carrier-to-thermal-noise-power ratio (CNR) at the receiver. Use Figure 11.21 to estimate what
complexity of QAM modulation will this CNR support for an error rate of 10−6 ? [32.5 dB, 128 state]
See later additional questions

Feedback Tutorial week 9

15.1 For the GSM cellular system of Table 15.4, with allocations as in Figure 15.8, what is the number of simul-
taneous calls which can be handled in a 7-cell cluster as in Figures 15.4? Please assume an ideal system and
ignore the need for any spectral allocation for access and control channels. [1400 users].
15.2 Compare the answer in Problem 15.1 above with an analogue TACS FDMA scheme with 25 kHz channel
spacing in a 12-cell cluster. [40 users per MHz in both systems]
15.3 If the analogue and digital (GSM) systems have the same capacity what is the advantage of the GSM?
15.4 What is the capacity per base station and total system capacity for a DECT standard system? [144 users].
15.5 What are the fundamental differences between the two CDMA receiver implementations in Figure 15.12?
MSc only question 15.6 Compare the capacity of GSM TDMA and cdmaOne systems in terms of active users
per MHz of available bandwidth for a 10-cell cluster. [both systems 40 users per MHz]

Feedback Tutorial week 10

See separate sheet
 -9-

Feedback Tutorial week 10 - ALL QUESTIONS FOR MSc class ONLY

19.2 The number of packets arriving at a particular node in a packet switched computer network may be
assumed to be Poisson distributed. Given that the average arrival rate is 5 packets per minute, calculate the fol-
lowing:
(a) Probability of receiving no packets in an interval of 2 minutes. [4. 5 × 10−5 ]
(b) Further to (a) probability of receiving just 1 packet in the next 30 s. [0.2]
(c) Probability of receiving 10 packets in any 30 s period. [2. 16 × 10−4 ]
19.3 In Problem 19.2: (a) What is the probability of having a gap between two successive packet arrivals of
greater than 20 s? (b) What is the probability of having gaps between packet arrivals in the range 20 to 30 s
inclusive? [0.189, 0.106]
19.4 For a single server queueing system with Poisson distributed arrivals of average rate 1 packet/s and Poisson
distributed service of capacity 3 packets/s calculate the probability of receiving no packets in a 5 s period. Also
find the probabilities of queue lengths of 0, 1, 2, 3. If the queue length is limited to 4 what percentage of pack-
ets will be lost? [0.00674, 0.6694214, 0.2231405, 0.0743801, 0.0247933, 0.826446%]
19.5 Assuming an M/M/1 queue, determine the probabilities of having: (a) an empty queue, (b) a queue of 4 or
more ‘customers’. You should assume that the ratio of arrival rate to service rate (called the utilisation factor) is
0.6. [0.4, 0.13]
19.6 If the queue length in Problem 19.5 is limited to 10, what percentage of customers are lost to the service?
[0.00629171%]
19.8 An X.25 packet switch has a single outgoing link at 2 Mbit/s. The average length of each data packet is
960 bytes. If the average packet delay through the switch, assuming an M/M/1 queue, is to be less than 15 ms,
determine: (i) the maximum gross input packet rate to the switch; (ii) the average length of the queue; (iii) the
utilisation factor through the switch if each packet in the input is converted into ATM cells having 48 bytes of
data and a 5 byte overhead (see section 20.10 for a full explanation of ATM). [193.7, 2.91, 0.821]
18.7 A data-link has a path delay of 10 µs and a transmission rate of 500 kbyte/s. The data frame has 72 bytes of
information, a 2 byte start field, a 16 byte header and an 8 byte check field. The acknowledgement information
is ‘piggy-backed’ on or added to returning data frames in this balanced transmission system. When using a four
frame buffer the observed data rate is 800 kbit/s. Determine the average bit error rate on the link and the bounds
on the processing time at the stations assuming go-back-N ARQ. [2. 55 × 10−4 , 0.174 - 0.372 ms]
18.9 A satellite link operates at a data rate of 100 Mbit/s, a path delay of 2.5 ms and a bit error ratio of 5×10−4 .
The data frames have a 128 byte data field, a 1 byte start field, a 5 byte header and a 6 byte check field. The
acknowledgement frame has a 2 byte data field. Calculate the average time for correct transmission and the
effective data rate and efficiency under the selective repeat ARQ. [21.68 µs, 47.24 Mbit/s, 47.24%]
 - 10 -

insert 1 .....................
 - 11 -

insert 2 .........................
 - 12 -

DIGITAL COMMUNICATIONS FORMULAE (as supplied in both exams)

Probabilities
P (J errors) = P eJ (1 − P e ) N −J N
CJ (3.8/10.1)
N!
N
CJ = (3.9)
J! N − J!
∞
P (X > x) = ∫ p X (x) dx = 1 − P X (x) (3.11)
x
dP X (x)
p X (x) = (3.13)
dx
or we often simplify this to
dP(x)
p(x) =
dx
∞
∫ p X (x) dx = 1 (3.14(a))
−∞
x2
P(x 1 < X < x 2 ) = ∫ p X (x) dx = P X (x 2 ) − P X (x 1 ) (13.14(b))
x1

mean
∞
X = ∫ x p(x) dx (3.16)
−∞

2nd central moment, variance, σ 2 or s2

∞
X 2 = ∫ (x − X)2 p(x) dx (3.17)
−∞

Uniform distribution
 1
 , (x 1 < x ≤ x 2 )
p X (x) =  x 2 − x 1
 0, (x ≤ x 1 , x > x 2 )

Gaussian distribution
−(x)2
1
p X (x) = e 2σ 2 (4.59)
σ √

2π
Exponential distribution
 1 −(x − a)
 e b , x≥a
p X (x) =  b
0 , x<a

Rayleigh distribution
 −r 
2
r
p R (r) = e 2σ 
2
(4.64)
σ2
Speech signal SNR
SNRQ = 4. 8 + 6n − α dB (5.23)
≈ 6(n − 1)
 - 13 -

Symbol and bit error rates for baseband transmission

1  1  S  ⁄2 
1

P e(NRZ unipolar) =  1 − erf   (6.10)

2  2 N  

1  1  S  ⁄2 
1

P e(NRZ =  1 − erf  (6.12)

2  N  
polar)
2  √
SER = P e R s (6.13)
BER = P e R s H (6.14)
Decision theory:
p(v RX | 1TX ) P (1TX )
P (1TX | v RX ) = (3.4/7.9)
p (v RX )
p(v RX | 1TX ) P(1TX ) C
choose 1 if > 1 (7.20)
p(v RX | 0TX ) P(0TX ) C0

∞
PD = ∫v p(v|s + n) dv (7.21)
th
∞
P FA = ∫ p(v|n) dv
v th
(7.22)

Information:
I = − log2 P(m) (9.1)
Entropy:
H = Σ − P(m) log2 P(m) (9.3)
1
H = ΣΣ P( j, i) log2 (bit/symbol) (9.6)
i j P( j|i)
1
H = Σ P(i) Σ P( j|i) log2 (bit/symbol) (9.7)
i j P( j|i)
P(iTX |i RX )
I RX (i RX ) = log2 (bits) (9.12)
P(iTX )
Equivocation:
1
E = ΣP( j RX )ΣP(iTX | j RX ) log2 (bit/symbol) (9.16)
j i P(iTX | j RX )
H
Efficiency = (9.21)
L
L = ΣP(m)l m (9.20)
Error correction:
P (J errors) = P eJ (1 − P e ) N −J N
CJ (10.1/3.8)
R
P( > R errors) = 1 − Σ P(J errors) (10.2)
J=0

Hamming bound:
2n
2k ≤ (10.4)
1 + n + nC2 + nC3 + . . . + nCt
 - 14 -

Detection power = Dmin − 1

D −1
Correction power = min
2
C/N = C/kTB = C/N 0 B (11.6)

Symbol and bit error rates for IF/RF transmission:

1  1  < E >  ⁄2 
1

P e( ASK ) =  1 − erf  (11.5)

2  2  N0  
√
E b C/ f b C B
= = × (11.8)
N0 N /B N fb
1  (T o B) ⁄2  C  ⁄2 
1 1

P e( ASK ) =  1 − erf  (11.9)

2  √2  N  
1  1  E  ⁄2 
1

P e(FSK ) =  1 − erf  (11.25)

2  2  N0  
√
1   E  
⁄2 1

P e(PRK ) =  1 − erf   (11.11)

2  N0  

1  
⁄2 1
1⁄2  C 
P e(PRK ) =  −
 N  
1 erf (T o B) (11.12)
2  
 S 
Rmax = B log2 1 + (11.38(a))
 N 
For M-symbol alphabets:
R s = R b / log2 M
  π  E  
⁄2 1

= 1 − erf  sin
 M   N0  
P e(MPSK ) (11.39(a))
 
  π  C  
⁄2 1

P e(MPSK ) = 1 − erf  (T o B) ⁄2 sin

1

 M  N  
(11.39(b))
 
Pe
Pb = (11.40(a))
log2 M
   Eb   
⁄2 1
1  π 
M √
=  1 − erf  sin   N0   
P b(MPSK ) log2 M (11.41(a))
log2 M   
   π  C  
⁄2 1
1
=  1 − erf  (T o B) sin 
1⁄2
P b(MPSK )  (11.41(b))
log2 M   M  N  

√

 M 1⁄2 − 1    〈E〉  
⁄2 1
3
P e(MQAM) = 2    1 − erf  (11.43)
 M 1⁄2

  2(M − 1)  N 0  


√
 M 1⁄2 − 1   3 T o B  C  ⁄2 
1

P e(MQAM) = 2   1 − erf  (11.45)

 M
1⁄2

  2(M − 1)  N  


√
 M 1⁄2 − 1   3 log2 M  E b  ⁄2 
1
2
P b(MQAM) =    1 − erf  (11.46(a))
log2 M  M 1⁄2   2(M − 1)  N 0  
 - 15 -


√
 M 1⁄2 − 1   3 T o B  C  ⁄2 
1
2
P b(MQAM) =    1 − erf  (11.46(b))
log2 M  M 1⁄2   2(M − 1)  N  
Noise and link budgets
1W = 0 dBW = 30 dBm
N = kTB (12.8)
 T ph (1 − G l ) 
Te = (12.37)
 Gl 
 T 
f = 1+ e (12.45)
 To 
f2 − 1 f −1
f cascade = f 1 + + 3 + etc (12.49)
G1 G1 G2
4π a e
GR = (12.62)
λ2
2
1  4π R 
L = (12.70)
GT G R  λ 
Networks:
P cfs = (1 − P b )n s (18.1)
P cf = (1 − P b )n (18.2)
Stop and wait:
t out ≥ 2t p + t proc + t s (18.5)
t T = t I + t out (18.6)
t t
a = T = 1 + out (18.7)
tI tI
P retrans = P f + P ACK ∼
− Pf (18.8)
tT
tV = (18.9)
1 − P retrans
1 − P retrans
D = ni (18.11)
tT
D  ni  1 − P retrans
= (18.12)
C n a
Go back N :
t out = 2t p + 2t I (18.14)
tT t 2t p
a= = 1 + out = 3 + (18.15)
tI tI tI
P retrans = 2P f ≈ 2nP b (18.16)
 1 + (N − 1)P retrans 
tV = t I   (18.17)
 1 − P retrans 
D  ni  1 − P retrans
= (18.18)
C  n  1 + (N − 1)P retrans
Selective repeat:
 - 16 -

tI
tV = (18.19)
1 − P retrans
D  ni 
= (1 − P retrans ) (18.20)
C n
Queue (infinite)
( λ T )k −λ T
P (k arrivals in T ) = P(k) = e , k = 0, 1, 2, . . . , ∞ (19.1/3)
k!
λ
ρ=
µ
P(k) = ρ k P(0), k≥0 (19.4)
∞
Σ P(k) = 1 (19.5)
k=0
∞
P(0) Σ ρ k = 1 (19.6)
k=0

P(0) = 1 − ρ (19.7)
P(k) = ρ (1 − ρ ),
k
k≥0 (19.8)
S=λ (19.9)
L = λW (19.10)
∞ ρ
L = Σ kP(k) = (19.11)
k=0 1−ρ
1 ρ 1
W = = (19.12)
λ 1−ρ µ−λ
1
Q = W − =ρW (19.13)
µ
Finite length queues:
ρ N (1 − ρ )
P L = P(N ) = , ρ≠1 (19.18)
1 − ρN + 1
L = SW (19.19)
L 1 N
W = = Σ kP(k) (19.20)
S S k=0
S = λ (1 − P L ) (19.X)

Constants

k = 1. 38 × 10−23 J/K
c = 3 × 108 m/s
 - 17 -

Tabulated values of the error

x function
2
erf(x) =
π 0
√ ∫ −y 2
e dy

x erf x x erf x x erf x x erf x

0.00 0.000000 0.40 0.428392 0.80 0.742101 1.20 0.910314
0.01 0.011283 0.41 0.437969 0.81 0.748003 1.21 0.912956
0.02 0.022565 0.42 0.447468 0.82 0.753811 1.22 0.915534
0.03 0.033841 0.43 0.456887 0.83 0.759524 1.23 0.918050
0.04 0.045111 0.44 0.466225 0.84 0.765143 1.24 0.920505
0.05 0.056372 0.45 0.475482 0.85 0.770668 1.25 0.922900
0.06 0.067622 0.46 0.484655 0.86 0.776100 1.26 0.925236
0.07 0.078858 0.47 0.493745 0.87 0.781440 1.27 0.927514
0.08 0.090078 0.48 0.502750 0.88 0.786687 1.28 0.929734
0.09 0.101282 0.49 0.511668 0.89 0.791843 1.29 0.931899
0.10 0.112463 0.50 0.520500 0.90 0.796908 1.30 0.934008
0.11 0.123623 0.51 0.529244 0.91 0.801883 1.31 0.936063
0.12 0.134758 0.52 0.537899 0.92 0.806768 1.32 0.938065
0.13 0.145867 0.53 0.546464 0.93 0.811564 1.33 0.940015
0.14 0.156947 0.54 0.554939 0.94 0.816271 1.34 0.941914
0.15 0.167996 0.55 0.563323 0.95 0.820891 1.35 0.943762
0.16 0.179012 0.56 0.571616 0.96 0.825424 1.36 0.945561
0.17 0.189992 0.57 0.579816 0.97 0.829870 1.37 0.947312
0.18 0.200936 0.58 0.587923 0.98 0.834232 1.38 0.949016
0.19 0.211840 0.59 0.595936 0.99 0.838508 1.39 0.950673
0.20 0.222703 0.60 0.603856 1.00 0.842701 1.40 0.952285
0.21 0.233522 0.61 0.611681 1.01 0.846810 1.41 0.953852
0.22 0.244296 0.62 0.619411 1.02 0.850838 1.42 0.955376
0.23 0.255023 0.63 0.627046 1.03 0.854784 1.43 0.956857
0.24 0.265700 0.64 0.634586 1.04 0.858650 1.44 0.958297
0.25 0.276326 0.65 0.642029 1.05 0.862436 1.45 0.959695
0.26 0.286900 0.66 0.649377 1.06 0.866144 1.46 0.961054
0.27 0.297418 0.67 0.656628 1.07 0.869773 1.47 0.962373
0.28 0.307880 0.68 0.663782 1.08 0.873326 1.48 0.963654
0.29 0.318283 0.69 0.670840 1.09 0.876803 1.49 0.964898
0.30 0.328627 0.70 0.677801 1.10 0.880205 1.50 0.966105
0.31 0.338908 0.71 0.684666 1.11 0.883533 1.51 0.967277
0.32 0.349126 0.72 0.691433 1.12 0.886788 1.52 0.968413
0.33 0.359279 0.73 0.698104 1.13 0.889971 1.53 0.969516
0.34 0.369365 0.74 0.704678 1.14 0.893082 1.54 0.970586
0.35 0.379382 0.75 0.711156 1.15 0.896124 1.55 0.971623
0.36 0.389330 0.76 0.717537 1.16 0.899096 1.56 0.972628
0.37 0.399206 0.77 0.723822 1.17 0.902000 1.57 0.973603
0.38 0.409009 0.78 0.730010 1.18 0.904837 1.58 0.974547
0.39 0.418739 0.79 0.736103 1.19 0.907608 1.59 0.975462
 - 18 -

x erf x x erf x x erf x x erf x

1.60 0.976348 2.10 0.997021 2.60 0.999764 3.10 0.99998835
1.61 0.977207 2.11 0.997155 2.61 0.999777 3.11 0.99998908
1.62 0.978038 2.12 0.997284 2.62 0.999789 3.12 0.99998977
1.63 0.978843 2.13 0.997407 2.63 0.999800 3.13 0.99999042
1.64 0.979622 2.14 0.997525 2.64 0.999811 3.14 0.99999103
1.65 0.980376 2.15 0.997639 2.65 0.999822 3.15 0.99999160
1.66 0.981105 2.16 0.997747 2.66 0.999831 3.16 0.99999214
1.67 0.981810 2.17 0.997851 2.67 0.999841 3.17 0.99999264
1.68 0.982493 2.18 0.997951 2.68 0.999849 3.18 0.99999311
1.69 0.983153 2.19 0.998046 2.69 0.999858 3.19 0.99999356
1.70 0.983790 2.20 0.998137 2.70 0.999866 3.20 0.99999397
1.71 0.984407 2.21 0.998224 2.71 0.999873 3.21 0.99999436
1.72 0.985003 2.22 0.998308 2.72 0.999880 3.22 0.99999473
1.73 0.985578 2.23 0.998388 2.73 0.999887 3.23 0.99999507
1.74 0.986135 2.24 0.998464 2.74 0.999893 3.24 0.99999540
1.75 0.986672 2.25 0.998537 2.75 0.999899 3.25 0.99999570
1.76 0.987190 2.26 0.998607 2.76 0.999905 3.26 0.99999598
1.77 0.987691 2.27 0.998674 2.77 0.999910 3.27 0.99999624
1.78 0.988174 2.28 0.998738 2.78 0.999916 3.28 0.99999649
1.79 0.988641 2.29 0.998799 2.79 0.999920 3.29 0.99999672
1.80 0.989091 2.30 0.998857 2.80 0.999925 3.30 0.99999694
1.81 0.989525 2.31 0.998912 2.81 0.999929 3.31 0.99999715
1.82 0.989943 2.32 0.998966 2.82 0.999933 3.32 0.99999734
1.83 0.990347 2.33 0.999016 2.83 0.999937 3.33 0.99999751
1.84 0.990736 2.34 0.999065 2.84 0.999941 3.34 0.99999768
1.85 0.991111 2.35 0.999111 2.85 0.999944 3.35 0.999997838
1.86 0.991472 2.36 0.999155 2.86 0.999948 3.36 0.999997983
1.87 0.991821 2.37 0.999197 2.87 0.999951 3.37 0.999998120
1.88 0.992156 2.38 0.999237 2.88 0.999954 3.38 0.999998247
1.89 0.992479 2.39 0.999275 2.89 0.999956 3.39 0.999998367
1.90 0.992790 2.40 0.999311 2.90 0.999959 3.40 0.999998478
1.91 0.993090 2.41 0.999346 2.91 0.999961 3.41 0.999998582
1.92 0.993378 2.42 0.999379 2.92 0.999964 3.42 0.999998679
1.93 0.993656 2.43 0.999411 2.93 0.999966 3.43 0.999998770
1.94 0.993923 2.44 0.999441 2.94 0.999968 3.44 0.999998855
1.95 0.994179 2.45 0.999469 2.95 0.999970 3.45 0.999998934
1.96 0.994426 2.46 0.999497 2.96 0.999972 3.46 0.999999008
1.97 0.994664 2.47 0.999523 2.97 0.999973 3.47 0.999999077
1.98 0.994892 2.48 0.999547 2.98 0.999975 3.48 0.999999141
1.99 0.995111 2.49 0.999571 2.99 0.999977 3.49 0.999999201
2.00 0.995322 2.50 0.999593 3.00 0.99997791 3.50 0.999999257
2.01 0.995525 2.51 0.999614 3.01 0.99997926 3.51 0.999999309
2.02 0.995719 2.52 0.999635 3.02 0.99998053 3.52 0.999999358
2.03 0.995906 2.53 0.999654 3.03 0.99998173 3.53 0.999999403
2.04 0.996086 2.54 0.999672 3.04 0.99998286 3.54 0.999999445
2.05 0.996258 2.55 0.999689 3.05 0.99998392 3.55 0.999999485
2.06 0.996423 2.56 0.999706 3.06 0.99998492 3.56 0.999999521
2.07 0.996582 2.57 0.999722 3.07 0.99998586 3.57 0.999999555
2.08 0.996734 2.58 0.999736 3.08 0.99998674 3.58 0.999999587
2.09 0.996880 2.59 0.999751 3.09 0.99998757 3.59 0.999999617
 - 19 -

x erf x x erf x x erf x x erf x

3.60 0.999999644 3.70 0.999999833 3.80 0.999999923 3.90 0.999999965
3.61 0.999999670 3.71 0.999999845 3.81 0.999999929 3.91 0.999999968
3.62 0.999999694 3.72 0.999999857 3.82 0.999999934 3.92 0.999999970
3.63 0.999999716 3.73 0.999999867 3.83 0.999999939 3.93 0.999999973
3.64 0.999999736 3.74 0.999999877 3.84 0.999999944 3.94 0.999999975
3.65 0.999999756 3.75 0.999999886 3.85 0.999999948 3.95 0.999999977
3.66 0.999999773 3.76 0.999999895 3.86 0.999999952 3.96 0.999999979
3.67 0.999999790 3.77 0.999999903 3.87 0.999999956 3.97 0.999999980
3.68 0.999999805 3.78 0.999999910 3.88 0.999999959 3.98 0.999999982
3.69 0.999999820 3.79 0.999999917 3.89 0.999999962 3.99 0.999999983

For x equal to, or greater than, 4 the following approximation may normally be used:
e−x
2

erf(x) ∼
− 1−
π x
√
Some complementary error function values for large x are:
x erfc x
4.0 1. 59 × 10−8
4.1 6. 89 × 10−9
4.2 2. 93 × 10−9
4.3 1. 22 × 10−9
4.4 5. 01 × 10−10
4.5 2. 01 × 10−10
4.6 7. 92 × 10−11
4.7 3. 06 × 10−11
4.8 1. 16 × 10−11
4.9 4. 30 × 10−12

Please recall that before calculating erf (z) you may need to use this conversion formulae relating the voltage x and mean value m to the
standard deviation of the noise σ :
x−m
z=
 2σ
√
And the CD or error probability is then given by:
1
P(x) = P e = [ 1 + / − erf (z) ]
2
The +/ − sign is used in front of the erf function depending on whether the CD is to be calculated for values less/greater than x.

pmg/seccy/PER/TEACHING/COMNOTES/handout2008