Determining Bit Error Rate in Digital Optical

Transmission Network Using the Q-Factor

Kiril K. Koitchev1, Krasen K. Angelov2, Stanimir S. Sadinov3
Abstract – Knowledge of the ratio of the signal power to the
ρ ρ 2
noise power (signal-to-noise ratio – SNR) is important because it ρ ρ E (t ) ρ E (t )
is directly related to the bit error ratio (BER) in digital PO (t )= E (t )× H (t ) = ⋅ E (t ) = . (1)
communication systems, and the BER is a major indicator of the η η
quality of the overall system. The SNR ratio related to the optical
communications has its own specifics. In practice direct measuring of optical power is rarely the
The purpose of this paper is to show the relationship between case. The usual practice involves the use of PIN photodiode
the electrical and optical signal-to-noise ratio introducing the Q- whose current is proportional to the optical power [5].
factor. Although some of the principles can be applied to the A point of interest is the studying of photo-receiver’s
existing coaxial cables, the scope of the paper is limited to binary
performance. By synchronizing with the incoming optical
digital signals over optical fiber. Application of codecs based on
Read-Solomon codes will be considered as an example, where signal the photo-receiver performs periodic processing of the
using this error correction method can reduce SNR in the received signal at the optimum moment. It registers the optical
receiver keeping the same value of BER. signal intensity and finally decides, on the basis of a certain
threshold value, what type of signal is received: 0 or 1. Quali-
Keywords – Bit Error Rate, Q-factor, Optical Signal-to-Noise ty of the digital communication system performance could be
Ratio, Probability Density Function, block codes evaluated by way of the so called eye-diagram (Fig. 1).

I. INTRODUCTION
So far BER estimation of characteristics in digital
transmission systems, that use quadrature modulations, is
important for studying especially in recording SNR. This ratio
if related to optical communications is relatively sparsely
represented and has its specific features [1].
The aim of this paper is to disclose the relationship between
electrical and optical signal-to-noise ratio SNR by means of Fig. 1. Typical appearance of eye-diagram while using a code of the
the relating parameter Q-factor. Although some of the type RZ (return to zero) in optical signal receiving
principles could be applied in the coaxial cables, the particular
case of this study will regard the transfer of binary digital Let us assume that in the optical receiver the converted
signal along an optical fiber because in this way both bit and voltage v(t) is compared with one definite value γ, which is
symbol rates will be equivalent. Application of codecs based referred to as threshold level. When v(t) is greater than γ this
on Read-Solomon codes will be considered as an example, means that a binary “1” has been sent. When v(t) is less than γ
since by this method of error correction it is possible to lower – a binary “0” has been sent [1,2].
SNR on the receiving side with unchanged BER. In presence of white Gaussian noise (AWGN), it is possible
to calculate the statistical error probability. The probable
density of v(t) in presence of AWGN can be mathematically
II. CALCULATION OF ERROR PROBABILITY expressed using Gaussian probability density function (PDF)
[9,10,11]:
As is known the power of the optical signal is presented as
1 ⎛ v (t ) − v ( s ) ⎞
2

the product of electric and magnetic field vectors: − ⎜⎜ ⎟

PROB[v(t ), σ x ]=
1 σx ⎟
e
2⎝ ⎠
, (2)
2πσ x2
1
Kiril Koitchev is with the Faculty of Electrical Engineering and
Electronics, Technical University – Gabrovo, 4 H. Dimitar St., 5300 where v(s) is the mean value of the density function of the
Gabrovo, Bulgaria, E-mail: [email protected] receiver, v(t) is the moment value of the receiver’s voltage,
2
Krasen Angelov is with the Faculty of Electrical Engineering and and σ is the standard noise aberration. The graphic
Electronics, Technical University – Gabrovo, 4 H. Dimitar St., 5300 interpretation of Eq. 2 is shown in Fig. 2.
Gabrovo, Bulgaria, E-mail: [email protected]
3
Stanimir Sadinov is with the Faculty of Electrical Engineering
and Electronics, Technical University – Gabrovo, 4 H. Dimitar St.,
5300 Gabrovo, Bulgaria, E-mail: [email protected]

53
 – on the standard noise aberrations (σL and σH);
– on the difference between voltages vL and vH.
When σL = σH, then γ = (vH − vL ) 2 , i.e. γ is the average
value between the high and low level. In most cases σ L ≠ σ H .
Then the optimum value of BER will be larger or less than the
value of (vH − vL ) 2 .
Fig. 2. Gaussian probability density function
To solve Eq. 4 it is necessary to calculate the result of
If we assume that v(s) could take one of the two voltage integrating Gaussian distribution PDF (PROB[v(t ), σ x ]) which
levels designated as vL and vH then the probability for wrong is defined through Eq. 2. This is possible by employing
decision making in the receiver is: numerical methods. For this purpose Eq. 2 can be solved by
setting it into standard form and applying z = (x − μ ) σ where
P[ε ]= P[v(t ) > γ | vS = vL ] P[vS = vL ]+ P[v(t ) < γ | vS = vH ] ,(3)
x = v(t ) and μ = vS .
where P[ε] is the error probability, P[x,y] is the conditional For the sake of facilitation let us apply this substitution for
probability of x for assigned value of y. Suppose that the error function taking into account Eqs. 2 and 4,
transmission and receiving probability is 50% then P[vS=vL] =
∞ ∞ 1 ⎛ x−μ ⎞
2

P[vS=vH] = 0,5. In this case Eq. 3 is reduced to:

∫ PROB[x,σ ]dx = ∫
1 − ⎜ ⎟
2⎝ σ ⎠
e dx ,
2π σ
P[ε ] = P[v(t ) > γ | vS = vL ]× 0.5 + P[v(t ) < γ | vS = vH ]× 0.5 = γ γ

and by applying z = (x − μ ) σ
γ ∞
,(4) (where x = zσ − μ
PROB[v(t ), σ L ] dt + PROB[v(t ), σ H ] dt
1 1 and
= ∫
2 −∞ 2γ∫ dx = σ dz ) we obtain:

where PROB[v(t ),σ x ] is determined by Eq. 2. This 1

∞ ⎛ − z2 ⎞
⎜ ⎟

∫e
⎜ 2 ⎟

interpretation is graphically presented in Fig. 3. . ⎝ ⎠

σ dz .
2π σ x = zσ + μ =γ

This result defines the error function Er(z)

∞ ⎛ − z2 ⎞
⎜ ⎟
1
∫
⎜ 2 ⎟
Ez = e⎝ ⎠
dz . (5)
2π z =γ

In this way Eq. 4 can be represent as

1 ⎡ v − γ ⎤ 1 ⎡ γ − vL ⎤
P[ε ]= Er ⎢ H ⎥ + Er ⎢ ⎥. (6)
2 ⎣ σH ⎦ 2 ⎣ σL ⎦

III. ESTIMATING THE Q- FACTOR BY MEANS OF THE

ERROR FUNCTION

The argument in the error function from Eq. 6 is the square

root of the signal divided by the square root of the signal level
(see Eq. 11). This is equivalent to the optical signal-to-noise
ratio and could be written as follows:

P[ε ]= E r [SNROH ]+ E r [SNROL ] ,

1 1
(7)
2 2

where SNROH and SNROL are optical signal-to-noise ratios

for the high and low level respectively.
Optimum initial level γopt is determined by this initial level
Fig. 3. Error probability in binary signal which defines the least bit error probability. In addition when
determining optimum initial level it is important if high signal
Considering Fig. 3 and Eqs. 3 and 4 it is possible to reach levels are transmitted as low level [3,4]. This means that γopt is
the conclusion that the error probability is the area of γ obtained when SNROH = SNROL which results in the following
beneath the graphs of the probability density functions. This representation of Q-factor:
area determines the magnitude of BER which depends on two
factors:

54
 v H − γ opt γ opt − v L 4B/5B and 8B/10B. The 10 Gigabit Ethernet standard uses
Q≡ = . (8) two types of encoding: 64B/66B for transmission along single
σH σL
mode fiber and 8B/10B for transmission along multimode
If Eq. 8 is substituted in Eq. 6 taking in consideration that fiber.
In optical communication lines which contain linear optical
γ = γ opt
amplifiers, the codecs based on Reed-Solomon block codes of
RS(n,k) type with s-bit symbols and in particular
P[ε ]= E r [Q ]+ E r [Q ]= E r [Q ] .
1 1
(9) RS(255,251), RS(255,239) and RS(255,223) codes with
2 2 single-byte (s = 8) symbols are widely distributed [3,4].
In communication lines, the so called “out-of-band”
By taking the value of γopt from Eq. 8 we obtain:
standard FEC ITU G.975 is the most widely used code which
v H σ L + v Lσ H is based on Reed-Solomon code RS(255,239). FEC increases
γ opt = . (10) the bit rate from 9,95 to 10,66 Gbps and allows decreasing of
σ L +σ H BER from 10-5 to 10-15 and improving the SNR with 6dB.
It is a conventional practice to measure the efficiency of
Considering Eq. 10 from Eq. 8 we obatin:
FEC correction according to SNR change; in other words how
vH − vL much does the error correction method allow to reduce SNR
Q= . (11) on the receiving side keeping the previous value of BER.
σL + σH

If both sides of Eq. 11 are multiplied by resistance or

impedance then the expression will be converted into
equivalent expressions which contain current or optical
power:

iH − iL POH − POL
Q= = . (12)
σL + σH σL + σH

Finally, by substituting Eq. 11 in Eq. 9 the outcome will be:

⎡ v −v ⎤
P[ε ]= E r [Q ]= E r ⎢ H L ⎥. (13)
⎣σ L + σ H ⎦

On the other hand BER could be logically represented as a

function of Q-factor and by using error function.
If the necessary BER ratio is assigned, it is possible to
obtain the corresponding value for the Q-factor by using the
following expression: Fig. 4. BER parameters for BPSK modulation and addition of FEC to
Reed-Solomon code
1 ⎛ Q ⎞ exp(− Q 2 / 2 )
BER(Q) = erfc⎜⎜ ⎟⎟ ≈ . (14)
2 ⎝ 2⎠ Q 2π

For example, for BER = 1,0-11 it is necessary to ensure a

value of Q = 16,53; for BER = 4,15.10-6 is obtained Q =
13,00.

IV. RESULTS
The FEC combined with the use of appropriate code will
strongly raise the quality of performance of the
communication line and, in particular, allows keeping the
same value of BER at lower value of the SNR ratio, i.e longer
regeneration section.
The error correction device is usually a basic element of the
modern fiber optic data transmission systems [7,8,12]. There
are various methods of implementation of FEC-codecs. For
example, the application of network standards Ethernet or
Gigabit Ethernet in optical interface features increased bit rate Fig. 5. BER parameters for QPSK modulation and addition of FEC to
by 25% and the concerned encoding circuits are referred to as Reed-Solomon code

55
 In Figs. 4 and 5 are presented the results from digital data REFERENCES
transfer using Reed-Solomon encoding. It is evident that
information transfer is improved when using FEC combined [1] K. Vitthaladevuni, M.S. Alouini, BER Computation of 4/M-
with the Reed-Solomon code while using QPSK and BPSK QAM Hieararchical Constellations. IEEE Transactions of
modulations. Broadcasting, Vol.47, №3, pp.228-239, 2001.
Obtained results indicate that by selection of appropriate Q- [2] J. Lu, K. Letaief, C. Chuang. M-PSK and M-QAM BER
factor line’s length can be extended, keeping all previous Computation Using Signal – Space Concepts. IEEE Transaction
parameters such as the input power, amplification coefficient of Communication, Vol.47, №2, pp 181-184, 1999.
[3] K. Nagayama, T. Saitoh. Ultra Low Loss (10,151 dB/km) Fiber
and amplifiers’ noise coefficient and the distance between two
and its Impact on Submarine Transmission Systems. OFC USA,
contiguous amplifiers. Post-didline Papers FA10, 2002.
[4] M. Trukitani, M. Matsui. Ultra Low Nonlinearity Pure-Silica-
Core Fiber with an Effective Area of 211mm2 and Transmission
V. CONCLUSION Loss of 0,159 dB/km. ECOC Copenhagen, SS.6.3, 2002.
[5] B. Sklar, Digital Communications: Fundamentals and
The Q-factor estimated by Eqs. 8 and 11 represents optical Applications, Englewood Cliffs, New Jersey: Prentice Hall, pp.
signal-to-noise ratio for binary optical communication 741-743, 2000.
systems. In this way it incorporates individual SNR ratios [6] G. Agrawal, Fiber-Optic Communication Systems, New York,
related to high and low levels into a complete system of SNR John Wiley & Sons, pp. 172, 1997.
ratios. [7] N. Bergano, F. Kerfoot, C. Davidson, Margin Measurements in
The Q-factor which is represented by Eq. 11 facilitates both Optical Amplifier Systems, IEEE Photonics Technology Letters,
measurement of SNR ratio and the theoretical calculation of Vol.5, №3, pp. 304-306, 1993.
[8] G. Frank, Corning Inc., Backbone to the future „Traditionally,
BER in presence of noise in the system. long-haul networks served as the proving ground for
It is evident that, BER can be improved either by increasing technologies that migrated to other parts of the network”,
the difference between high and low levels in the nominator Lightwave, September 2001.
or by reducing the noise in the denominator of the Q-factor [9] R. Ramaswami, K. Sivarajan, Optical Networks: A Practical
equation. Perspective, Academic Press, 1998.
In this way, using the Q-factor, it is possible to perform a [10] Optical Fiber Telecommunications IV-B: Systems and
simplified analysis of the communication system efficiency. Impairments (Optics and Photonics), Academic Press, 2002.
By using appropriate code and error correction it is possible [11] Р. Убайдулаев, Протяженные ВОЛС на основе EDFA,
to increase the digital data transfer efficiency along optical Lightwave Russian edition, №1, стр26-28, 2003.
[12] O. Панагиев, Изследване влиянието на параметрите на HFC
communication channels. This efficiency is determined by the комуникационна мрежа върху BER. НКМУ “ТЕЛЕКОМ
margin of reduced SNR ratio while keeping the previous value 2009”, Варна, стр. 236-241, 2009.
of BER.

56