Tutorial Eb/N0 evaluation

1) a) Using a diagram, explain what is N0, including what are its units. Use the idea of a
single sided AWGN density. Assume the system is affected by AWGN and has a system
bandwidth of B Hz.

b) Evaluate N0 if B=1kHz, and the standard deviation of the AWGN noise is 2volts.

c) For the communication system described in (b) above, find Eb and then Eb/N0 if we use
simple unipolar ASK with logic 1 represented by 5volts and logic 0 represented by 0volts.
You can assume that logic 1’s and logic 0’s are equally probable. Assume a bitrate of
1000bps.

d) Repeat (c) for bipolar signalling in which logic 0 is represented by -5volts. Assume a
bitrate of 500bps.

2) Evaluate Eb/N0 for a communication system which is affected by AWGN, in which the
noise’s Gaussian distribution has 68% laying between +/- 3volts. The signalling of logic 0 is a
1000Hz sinewave of unit amplitude; the signalling of logic 1 is a 1000Hz negative sinewave
of unit amplitude. The bitrate is 1kbps, and the analogue system bandwidth is 1kHz.

3) Using the Chernoff bound for the Q function:

Q(x) ≤ ½ .exp(-x2 /2)

Find the upper bound on the error probability, Pe, for Q1 part c).

4) A channel uses Unipolar NRZ binary signalling, so M=2 levels, and has logic ‘1’
represented by 5 volts while logic ‘0’ is represented by 0 volts. This channel is affected by
AWGN for which the Gaussian distribution has a standard deviation of 2 volts. Assume a
bitrate of 1000bps and a bandwidth of 1000Hz.

a) i) Assuming 1’s and 0’s are equally likely, evaluate Eb. ii) If 1’s are 9 times more likely than
0’s, evaluate Eb again.
b) Assuming a bandwidth of 1kHz, evaluate N0, and then Eb/N0, assuming 1’s and 0’s are
equally likely.
c) Noting that, for Unipolar NRZ binary signalling:

𝐸𝑏
𝑃𝑒 = 𝑄 (√ )
𝑁0

… evaluate Pe, and then the BER for the scenario in which 1’s and 0’s are equally likely. Do
this in two ways:
i) Using The Chernoff bound for the Q function: Q(x) ≤ ½ .exp(-x2 /2)
ii) Using the Table for Q (ignore the Approximation column in the table below).
5) Assume 2-PSK (BPSK) coding, the time for each symbol = 0.01sec, and the carrier wave is
100Hz, with an amplitude of 10V; calculate Eb. If the bandwidth is 100Hz, and the standard
2𝐸
deviation of the AWGN is 2 volts, evaluate the BER using 𝐵𝐸𝑅 = 𝑃𝐵 = 𝑄 (√ 𝑁 𝑏 ). Again do
0

this in two ways:

i) Using The Chernoff bound for the Q function: Q(x) ≤ ½ .exp(-x2 /2)
ii) Using the Table for Q

6) Data (in the form of logic 1’s and logic 0’s) are transmitted using a triangular pulse shape as shown
below. Assume T = 1ms. Find eb/N0 if the standard deviation of the AWGN is 0.02 volts.
ANSWERS
1) a) Using a diagram, explain what is N0, including what are its units. Use the idea of a
single sided AWGN density. Assume the system is affected by AWGN and has a system
bandwidth of B Hz.

Diagram as above.

N0 is the single value on the Y axis of the single sided (i.e. no negative frequencies shown)
density plot.

Total noise power = σ2

So we can write that: Total noise power = σ2 = N0.B as the total area under the curve will
be N0.B.

N0 will have units Watts/Hz = Watt.seconds = Joules

b) Evaluate N0 if B=1kHz, and the standard deviation of the AWGN noise is 2volts.

Total noise power = σ2 = N0.B = 22 = N0.B = N0.1000

Therefore N0 = 0.004 Joules

c) For the communication system described in (b) above, find Eb and then Eb/N0 if we use
simple unipolar ASK with logic 1 represented by 5volts and logic 0 represented by 0volts. You
can assume that logic 1’s and logic 0’s are equally probable.

For unipolar signalling:

𝑇 52 . 0.001
𝐸𝑏 = 𝑉 2 . = = 12.5𝑚𝐽
2 2

Therefore Eb/N0 = 12.5e-3/0.004 = 3.125

d) Repeat (c) for bipolar signalling in which logic 0 is represented by -5volts. Assume a bitrate
of 500bps.

1
𝐸𝑏 = 𝑉 2 𝑇 = 52 . ( ) = 50𝑚𝐽
500

Therefore Eb/N0 = 50e-3/0.004 = 12.5

2) Evaluate Eb/N0 for a communication system which is affected by AWGN, in which the
noise’s Gaussian distribution has 68% laying between +/- 3volts. The signalling of logic 0 is a
1000Hz sinewave of unit amplitude; the signalling of logic 1 is a 1000Hz negative sinewave
of unit amplitude. The bitrate is 1kbps, and the analogue system bandwidth is 1kHz.

…has 68% laying between +/- 3volts means that σ = 3, so σ2 = 9 Watts

σ2 = N0.B = 9 = N0.B = N0.1000

therefore: N0 = 0.009 Joules

Similarly for logic 1, we will also get 0.5mJ.

Therefore:

Eb/N0 = 0.0005 / 0.009 = 0.055

3) Using the Chernoff bound for the Q function:

Q(x) ≤ ½ .exp(-x2 /2)

Find the upper bound on the error probability, Pe, for Q1 part c).

From Q1, part c) we have that:

Eb/N0 = 12.5e-3/0.004 = 3.125

𝑉 𝐸𝑏
𝑃𝑒 = 𝑄 ( ) = 𝑃𝑒 = 𝑄 (√ ) = 𝑄(√3.125 ) = 𝑄(1.767 )
2𝜎 𝑁0

1 𝑥2 1 1.7672
𝑄(𝑥) ≤ . exp (− ) = . exp (− ) = 0.1048
2 2 2 2

4) A channel uses Unipolar NRZ binary signalling, so M=2 levels, and has logic ‘1’
represented by 5 volts while logic ‘0’ is represented by 0 volts. This channel is affected by
AWGN for which the Gaussian distribution has a standard deviation of 2 volts. Assume a
bitrate of 1000bps and a bandwidth of 1000Hz.

a) i) Assuming 1’s and 0’s are equally likely, evaluate Eb.

𝑇
𝑇
𝐸1 = ∫ 𝑉 2 . 𝑑𝑡 = |𝑉 2 𝑡| = 52 (0.001) = 0.025 𝐽
0
0

𝑇
𝑇
𝐸0 = ∫ 𝑉 2 . 𝑑𝑡 = |02 𝑡| =0 𝐽
0
0

Eb = (E0 + E1) / 2 = 0.0125 J if both 1’s and 0’s are equally likely.

ii) If 1’s are 9 times more likely than 0’s, evaluate Eb again.

Eb = E0(0.1) + E1(0.9) = 0.0225 J if 90% are 1’s.

b) Assuming a bandwidth of 1kHz, evaluate N0, and then Eb/N0, assuming 1’s and 0’s are
equally likely.

σ2 = N0.B, so N0 = σ2 /B = 22/1000 = 0.004 J so Eb/N0 = 0.0125 / 0.004 = 3.125

c) Noting that, for Unipolar NRZ binary signalling:

 𝐸𝑏
𝑃𝑒 = 𝑄 (√ )
𝑁0

… evaluate Pe, and then the BER for the scenario in which 1’s and 0’s are equally likely. Do
this in two ways:
i) Using The Chernoff bound for the Q function: Q(x) ≤ ½ .exp(-x2 /2)

2
𝐸𝑏 1 (√3.125)
𝑃𝑒 = 𝑄 (√ ) ≤ exp (− ) = 0.104
𝑁0 2 2
Therefore as this is BPSK and therefore BER = Pe, BER < 0. 104

𝐸
ii) Using the Table for Q: 𝑃𝑒 = 𝑄 (√𝑁𝑏 ) = Q(√3.125) = 0.044 = BER
0

5) Assume 2-PSK (BPSK) coding, the time for each symbol = 0.01sec, and the carrier wave is
100Hz; calculate Eb. If the bandwidth is 100Hz, and the standard deviation of the AWGN is 2
2𝐸
volts, evaluate the BER using 𝐵𝐸𝑅 = 𝑃𝐵 = 𝑄 (√ 𝑁 𝑏 ). Again do this in two ways:
0

i) Using The Chernoff bound for the Q function: Q(x) ≤ ½ .exp(-x2 /2)
ii) Using the Table for Q

𝑇
2
𝐸𝑠0 = ∫(𝑥(𝑡)) 𝑑𝑡
0
𝑇

𝐸𝑠0 = ∫ 102 . sin2 (𝜔. 𝑡) 𝑑𝑡

0

𝑇
1 1
𝐸𝑠0 = 100 ∫ ( − 𝑐𝑜𝑠(4. 𝜋. 100. 𝑡)) 𝑑𝑡
2 2
0

𝑇
1 1
𝐸𝑠0 = 100 ∫ ( − 𝑐𝑜𝑠(4. 𝜋. 100. 𝑡)) 𝑑𝑡
2 2
0

𝑡 1 0.01
𝐸𝑠0 = 100 | − 𝑠𝑖𝑛(4. 𝜋. 100. 𝑡)|
2 2.4. 𝜋. 100 0
0.01 1 0 1
= 100 ( − 𝑠𝑖𝑛(4. 𝜋. 100. (0.01)) − + 𝑠𝑖𝑛(0))
2 2.4. 𝜋. 100 2 2.4. 𝜋. 100
 0.01
𝐸𝑠0 = 100 ( ) = 0.5
2

Similarly, Es1 is also ½, so Es is ½ also. Eb = Es = ½ as this is BPSK (M=2).

For AWGN N0 . B = total_noise_power = σ2 = N0.B so N0 = 22/100 = 0.04

Therefore Eb / N0 = 0.5 / 0.04 = 12.5

So:
2. 𝐸𝑏
√ = √25 = 5
𝑁0

2. 𝐸𝑏 1 (5)2
𝑃𝑒 = 𝑄 (√ ) ≤ exp (− ) = 1.85𝑒 − 6
𝑁0 2 2

Therefore, as this is BPSK, BER = Pe, BER < 1.85𝑒 − 6 using the Chernoff bound.

𝐸
ii) Using the Table for Q: 𝑃𝑒 = 𝑄 (√2 𝑁𝑏 ) = Q(√25) = 𝑄(5) = 2.87𝑒 − 7 = BER (as this is
0

BPSK, M=2.)

6) Data (logic 1’s and logic 0’s) are transmitted using a triangular pulse shape as shown below. Assume
T = 1ms. Find eb/N0 as a ratio and in dB if the standard deviation of the AWGN is 0.02 volts.

σ2 = N0.B = N0.r/2 , r =
1/T = 1000bps

σ2 = 0.0004 = N0.r/2 =
N0 . 500, therefore N0 =
8e-7 W/Hz

To find eb we need to
integrate. Equation for
this line is: y=mx+c,
where c=0,
so:

y=(1/0.0005)x, or x(t) = 2000t

 𝑇/2 𝑇/2
𝑇 𝑇 3
𝑡 3 ( )
𝐸0→𝑇/2 = ∫ 𝑥(𝑡)2 . 𝑑𝑡 = ∫ (2000𝑡)2 . 𝑑𝑡 = |4𝑒6. | 2 = 4𝑒6. 2 − 0 = 1.6𝑒 − 4
3 0 3
0 0

𝐸𝑇→𝑇 = ∫ 𝑥(𝑡)2 . 𝑑𝑡 = 𝐸0→𝑇/2 = 1.6𝑒 − 4

2
𝑇/2

Therefore E1 = 2x1.6𝑒 − 4 = 0.00033333333333

E0 will be the same as E1, so Eb = 0.00033333333333

Therefore eb/N0 = 0.00033333333333 / 8e-7 = 416.6

In dB:

eb/N0|dB = 10.log10(416.6) = 26.2dB