OPTF_C12.

qxd 11/6/08 11:00 Page 742

742 Optical fiber systems 1: intensity modulation/direct detection Chapter 12

12.7.1 Direct intensity modulation (D–IM)

A block schematic for an analog optical fiber system which uses direct modulation of the
optical source intensity with the baseband signal is shown in Figure 12.43(a). Obviously,
no electrical modulation or demodulation is required with this technique, making it both
inexpensive and easy to implement.
The transmitted optical power waveform as a function of time Popt(t), an example of
which is illustrated in Figure 12.40(b) may be written as:

Popt(t) = Pi(1 + m(t)) (12.59)

where Pi is the average transmitted optical power (i.e. the unmodulated carrier power) and
m(t) is the intensity modulating signal which is proportional to the source message a(t).
For a cosinusoidal modulating signal:

m(t) = ma cos ωmt (12.60)

where ma is the modulation index or the ratio of the peak excursion from the average to the
average power as shown in Figure 12.43(b) and ωm is the angular frequency of the modu-
lating signal. Combining Eqs (12.59) and (12.60) we get:

Popt(t) = Pi(1 + ma cos ωmt) (12.61)

Furthermore, assuming the transmission medium has zero dispersion, the received opti-
cal power will be of the same form as Eq. (12.61), but with an average received optical

Figure 12.43 (a) Analog optical fiber system employing direct intensity modulation.
(b) Time domain representation showing direct intensity modulation of the optical
carrier with a baseband analog signal
OPTF_C12.qxd 11/6/08 11:00 Page 743

Analog systems 743

power Po. Hence the secondary photocurrent I(t) generated at an APD receiver with a
multiplication factor M is given by:

I(t) = Ip M(1 + ma cos ωmt) (12.62)

where the primary photocurrent obtained with an unmodulated carrier Ip is given by

Eq. (8.8) as:

ηe
Ip = P (12.63)
hf o

The mean square signal current i 2sig which is obtained from Eq. (12.62) is given by:

i 2sig = --12 (maMIp)2 (12.64)

The total average noise in the system is composed of quantum, dark current and thermal
(circuit) noise components. The noise contribution from quantum effects and detector dark
current may be expressed as the mean square total shot noise current for the APD receiver
i 2SA given by Eq. (9.21) where the excess avalanche noise factor is written following
Eq. (9.26) as F(M) such that:

i 2SA = 2eB(Ip + Id)M2F(M) (12.65)

where B is the effective noise or post-detection bandwidth.

Thermal noise generated by the load resistance RL and the electronic amplifier noise can
be expressed in terms of the amplifier noise figure Fn referred to RL as given by Eq. (9.17).
Thus the total mean square noise current i 2N may be written as:

4KTBFn
i 2N = 2eB(Ip + Id )M2F(M) + (12.66)
RL

The SNR defined in terms of the ratio of the mean square signal current to the mean square
noise current (rms signal power to rms noise power) for the APD receiver is therefore
given by:

ASD i2 --12 (maMIp) 2

= sig = (APD) (12.67)
C N F rms i 2 2eB(Ip + Id)M2F(M) + (4KTBFn /RL)
N

It must be emphasized that the SNR given in Eq. (12.67) is defined in terms of rms
signal power rather than peak signal power used previously. When a unity gain photo-
detector is utilized in the receiver (i.e. p–i–n photodiode) Eq. (12.67) reduces to:

ASD --12 (maIp) 2

= (p–i–n) (12.68)
C N F rms 2eB(Ip + Id) + (4KTBFn /RL)

Moreover, the SNR for video transmission is often defined in terms of the peak-to-peak
picture signal power to the rms noise power and may include the ratio of luminance to
composite video b. Using this definition in the case of the unity gain detector gives:
OPTF_C12.qxd 11/6/08 11:00 Page 744

744 Optical fiber systems 1: intensity modulation/direct detection Chapter 12

ASD (2maIpb) 2
= (p–i–n) (12.69)
C N F p–p 2eB(Ip + Id) + (4KTBFn /RL)

It may be observed that, excluding b, the SNR defined in terms of the peak-to-peak
signal power given in Eq. (12.69) is a factor of 8 (or 9 dB) greater than that defined in
Eq. (12.68).

Example 12.11
A single TV channel is transmitted over an analog optical fiber link using direct
intensity modulation. The video signal which has a bandwidth of 5 MHz and a ratio
of luminance to composite video of 0.7 is transmitted with a modulation index of
0.8. The receiver contains a p–i–n photodiode with a responsivity of 0.5 A W−1 and a
preamplifier with an effective input impedance of 1 MΩ together with a noise figure
of 1.5 dB. Assuming the receiver is operating at a temperature of 20 °C and neglect-
ing the dark current in the photodiode, determine the average incident optical power
required at the receiver (i.e. receiver sensitivity) in order to maintain a peak-to-peak
signal power to rms noise power ratio of 55 dB.
Solution: Neglecting the photodiode dark current, the peak-to-peak signal rms
noise power ratio is given following Eq. (12.69) as:

ASD (2maIpb) 2
=
C N F p–p 2eBIp + (4KTBFn /RL)

The photocurrent Ip may be expressed in terms of the average incident optical power
at the receiver Po using Eq. (8.4) as:

Ip = RPo

where R is the responsivity of the photodiode. Hence:

ASD (2maRPob) 2
=
C N F p–p 2eBRPo + (4KTBFn /RL)

and:

ASD A 4KTBFnD
2eBRPo = = (2maRPob)2
C N F p–p C RL F

Rearranging:

ASD A S D 4KTBFn
(2ma Rb)2P2o − 2eBRPo − =0
C N F p–p C N F p–p RL
OPTF_C12.qxd 11/6/08 11:00 Page 745

Analog systems 745

where:

(2maRb)2 = 4 × 0.64 × 0.25 × 0.49

= 0.314

ASD
2eBR = 3.162 × 105 × 2 × 1.602 × 10−19 × 5 × 106 × 0.5
C N F p–p
= 2.533 × 10−7

A S D 4KTBFn 3.162 × 105 × 4 × 1.381 × 10−23 × 293 × 5 × 106 × 1.413

=
C N F p–p RL 106
= 3.616 × 10−14

Therefore:

0.314P 2o − 2.533 × 10−7 Po − 3.616 × 10−14 = 0

and:

2.533 × 10−7 ± √[(2.533 × 10−7)2 − (−4 × 0.314 × 3.616 × 10−14)]

Po =
0.628
= 0.93 μW
= −30.3 dBm

It must be noted that the low-noise preamplification depicted in Example 12.11 may not
always be obtained, and that higher thermal noise levels will adversely affect the receiver
sensitivity for a given SNR. This is especially the case with lower SNRs, as illustrated in
the peak-to-peak signal power to rms noise power ratio against average received optical
power characteristics for a video system shown in Figure 12.44 [Ref. 86]. The performance
of the system for various values of mean square thermal noise current i 2t = 4KTBFn /RL,
where i 2t is expressed as a spectral density in A2 Hz−1, is indicated. The value for the
receiver sensitivity obtained in Example 12.11 is approaching the quantum limit, also
illustrated in Figure 12.44, which is the best that could possibly be achieved with a noise-
less amplifier.
The quantum or shot noise (when ignoring the photodetector dark current) limit occurs
with large values of signal current (i.e. primary photocurrent) at the receiver. Considering
a p–i–n photodiode receiver, this limiting case which corresponds to large SNR is given
by Eq. (12.68) when neglecting the device dark current as:

ASD m2 Ip
 a (quantum noise limit) (12.70)
C N F rms 4eB
OPTF_C12.qxd 11/6/08 11:00 Page 750

750 Optical fiber systems 1: intensity modulation/direct detection Chapter 12

Example 12.14
The 10–90% rise times for possible components to be used in a D–IM analog optical
fiber link are specified below:
Source (LED) 10 ns
Fiber cable: intermodal 9 ns km−1
chromatic 2 ns km−1
Detector (APD) 3 ns
The desired link length without repeaters is 5 km and the required optical bandwidth
is 6 MHz. Determine whether the above combination of components gives an
adequate temporal response.
Solution: Equation (12.74) may be used to calculate the maximum permitted
system rise time which gives the desired bandwidth where:

0.35 0.35
Tsyst(max) = = = 58.3 ns
Bopt 6 × 106

The total system rise time using the specified components can be estimated using
Eq. (12.43) as:
1
Tsyst = 1.1(T 2S + T 2n + T 2c + T 2D)---2
1
= 1.1(102 + (9 × 5)2 + (2 × 5)2 + 32)---2
 52 ns

Therefore the specified components give a system rise time which is adequate for the
bandwidth and distance requirements of the optical fiber link. However, there is little
leeway for upgrading the system in terms of bandwidth or distance without replacing
one or more of the system components.

12.7.3 Subcarrier intensity modulation

Direct intensity modulation of the optical source is suitable for the transmission of a base-
band analog signal. However, if the wideband nature of the optical fiber medium is to be
fully utilized it is essential that a number of baseband channels are multiplexed onto a
single fiber link. This may be achieved with analog transmission through frequency divi-
sion multiplexing of the individual baseband channels. Initially, the baseband channels
must be translated onto carriers of different frequency by amplitude modulation (AM),
frequency modulation (FM) or phase modulation (PM) prior to being simultaneously
transmitted as an frequency division multiplexing signal. The frequency translation may
be performed in the electrical regime where the baseband analog signals modulate elec-
trical subcarriers and are then frequency division multiplexed to form a composite electrical
signal prior to intensity modulation of the optical source.
A block schematic of an analog system employing this technique, which is known
as subcarrier intensity modulation, is shown in Figure 12.45. The baseband signals are
OPTF_C12.qxd 11/6/08 11:00 Page 751

Analog systems 751

Figure 12.45 Subcarrier intensity modulation system for analog optical fiber
transmission

modulated onto radio-frequency (RF) subcarriers by either AM, FM or PM and multi-

plexed before being applied to the optical source drive circuit.* Hence an intensity modu-
lated (IM) optical signal is obtained which may be AM–IM, FM–IM or PM–IM. In
practice, however, system output SNR considerations dictate that generally only the latter
two modulation formats are used. Nevertheless, systems may incorporate two levels of
electrical modulation whereby the baseband channels are initially amplitude modulated
prior to FM or PM [Ref. 87]. The FM or PM signal thus obtained is then used to intensity
modulate the optical source. At the receive terminal the transmitted optical signal is
detected prior to electrical demodulation and demultiplexing (filtering) to obtain the
originally transmitted baseband signals.
A further major advantage of subcarrier intensity modulation is the possible improve-
ment in SNR that may be obtained during subcarrier demodulation. In order to investigate
this process it is necessary to obtain a general expression for the SNR of the IM optical
carrier which may then be applied to the subcarrier intensity modulation formats.
Therefore, as with D–IM, considered in the preceding section, an electrical signal m(t)
modulates the source intensity. The transmitted optical power waveform is of the same
form as Eq. (12.59), where:
Popt(t) = Pi(1 + m(t)) (12.75)

Also the secondary photon I(t) generated at an APD receiver following Eq. (12.62) is
given by:
I(t) = IpM(1 + m(t)) (12.76)

The mean square signal current i 2sig may be written as [Ref. 84]:
i 2sig = (IpM)2Pm (12.77)

* When microwave frequency rather than radiofrequency subcarriers are employed the strategy is
usually referred to as subcarrier multiplexing or SCM (see Section 12.9.2).
OPTF_C12.qxd 11/6/08 11:00 Page 752

752 Optical fiber systems 1: intensity modulation/direct detection Chapter 12

where Pm is the total power of m(t), which can be defined in terms of the spectral density
Sm(ω) of m(t) occupying a one-sided bandwidth Bm Hz as:
2πBm

Pm =
1
2π −2πBm
Sm(ω) dω (12.78)

Hence the SNR defined in terms of the mean square signal current to mean square noise
current (i.e. rms signal power to rms noise power) using Eqs (12.77) and (12.66) can now
be written as:

ASD i2 (IpM)2Pm
= sig =
C N F rms i2 2eBm(Ip + Id)M 2F(M) + (4KTBFn /RL)
N

I 2p Pm
=
2Bme(Ip + Id)F(M) + (4KTBFn/M2RL)
(RPo)2Pm
= (D–IM) (12.79)
2BmNo

where we substitute for Ip from Eq. (8.4) and for notational simplicity write:

4KTBFn
No = e(Ip + Id)F(M) + (12.80)
M 2RL

The result obtained in Eq. (12.79) gives the SNR for a direct intensity modulated
optical source where the total modulating signal power is Pm. In this context Eq. (12.79) is
simply a more general form of Eq. (12.67). However, we are now in a position to examine
the signal-to-noise performance of various subcarrier intensity modulation formats.

12.7.4 Subcarrier double-sideband modulation (DSB–IM)

A simple way to translate the spectrum of the baseband message signal a(t) is by direct
multiplication with the subcarrier waveform Ac cos ωct giving the modulated waveform
m(t) as:

m(t) = Aca(t) cos ωct (12.81)

where Ac is the amplitude and ωc the angular frequency of the subcarrier waveform. For a
cosinusoidal modulating signal (cos ωmt) the subcarrier electric field Em(t) becomes:

Ac
Em(t) = cos(ωc + ωm)t + cos(ωc − ωm)t (12.82)
2

giving the upper and lower sidebands. The time and frequency domain representations
of the modulated waveform are shown in Figure 12.46. It may be observed from the
frequency domain representation that only the two sideband components are present as
OPTF_C12.qxd 11/6/08 11:00 Page 753

Analog systems 753

Figure 12.46 Time and frequency domain representations of double-sideband

modulation

indicated in Eq. (12.82). This modulation technique is known as double-sideband modula-

tion (DSB) or double-sideband suppressed carrier (DSBSC) AM. It provides a more effi-
cient method of translating the spectrum of the baseband message signal than conventional
full AM where a large carrier component is also present in the modulated waveform.
The DSB signal shown in Figure 12.46 intensity modulates the optical source.
Therefore the transmitted optical power waveform is obtained by combining Eqs (12.75)
and (12.81) where for simplicity we set the carrier amplitude Ac to unity, giving:

Popt(t) = Pi(1 + a(t) cos ωct) (12.83)

Furthermore, in order to prevent overmodulation, the value of the message signal is nor-
malized such that | a(t) ≤ 1 | with power Pa ≤ 1. The DSB modulated electrical subcarrier
occupies a bandwidth Bm = 2Ba, and with a carrier amplitude of unity, Pm = Pa /2. Hence,
the ratio of rms signal power to rms noise power obtained within the subcarrier bandwidth
at the input to the DSB demodulator is given by Eq. (12.79) where:

ASD (RPo)2Pa /2 (RPo)2Pa

input DSB = = (12.84)
C N F rms 2 × 2Ba No 8Ba No

However, an ideal DSB demodulator gives a detection gain of 2 or 3 dB improvement

in SNR [Ref. 87]. This yields an output SNR of:

ASD ASD (RPo)2Pa

output DSB = 2 input DSB = (12.85)
C N F rms C N F rms 4Ba No

Comparison of the result obtained in Eq. (12.85) with that using D–IM of the baseband
signal given by Eq. (12.79) shows a 3 dB degradation in SNR when employing DSB–IM
under the same conditions of bandwidth (i.e. Bm = Ba), modulating signal power (i.e. Pm =
Pa), detector photocurrent and noise. For this reason DSB–IM systems (and also AM–IM
 Digital Links 307

For InGaAsP lasers, distances of 150 m can be achieved at 100-Gb/s rates in OM4 multimode fiber at
1.3 mm (see Sec.13.4). A single-mode fiber can provide significantly higher rates over longer distances.
Since laser diodes typically couple from 10 to 15 dB more optical power into a fiber than an LED,
greater repeaterless transmission distances are possible with a laser. This advantage and the lower
dispersion capability of laser diodes may be offset by cost constraints. Not only is a laser diode itself more
expensive than an LED, but also the laser transmitter circuitry is much more complex, since the lasing
threshold has to be dynamically controlled as a function of temperature and device aging. However, a
wide variety of cost-effective laser transmitters are on the market.
For the optical fiber, we have a choice between single-mode and multimode fiber, either of which
could have a step- or a graded-index core. This choice depends on the type of light source used and
on the amount of dispersion that can be tolerated. Light-emitting diodes (LEDs) tend to be used with
multimode fibers. The optical power that can be coupled into a fiber from an LED depends on the core-
cladding index difference D, which, in turn, is related to the numerical aperture of the fiber (for D = 0.01,
the numerical aperture NA ⯝ 0.21). As D increases, the fiber-coupled power increases correspondingly.
However, since dispersion also becomes greater with increasing D, a tradeoff must be made between the
optical power that can be launched into the fiber and the maximum tolerable dispersion.
When choosing the attenuation characteristics of a cabled fiber, the excess loss that results from the
cabling process must be considered in addition to the attenuation of the fiber itself. This must also include
connector and splice losses as well as environmental-induced losses that could arise from temperature
variations, radiation effects, and dust and moisture on the connectors.
8.1.2 Link Power Budget
An optical power loss model for a point-to-point link is shown in Fig. 8.2. The optical power received at
the photodetector depends on the amount of light coupled into the fiber and the losses occurring in the fiber
and at the connectors and splices. The link loss budget is derived from the sequential loss contributions
of each element in the link. Each of these loss elements is expressed in decibels (dB) as
Pout
loss = 10 log (8.1)
Pin
where Pin and Pout are the optical powers entering and leaving the loss element, respectively. The loss
value corresponding to a particular element generally is called the insertion loss for that element.
In addition to the link loss contributors shown in Fig. 8.2, a link power margin is normally provided in
the analysis to allow for component aging, temperature fluctuations, and losses arising from components

Fig. 8.2 Optical power loss model for a point-to-point link. The losses occur at connectors
(lc), at splices (lsp ), and in the fiber (a).
 308 Optical Fiber Communications

that might be added at future dates. A link margin of 3 to 6 dB is generally used for systems that are not
expected to have additional components incorporated into the link in the future.
The link loss budget simply considers the total optical power loss PT that is allowed between the light
source and the photodetector, and allocates this loss to cable attenuation, connector loss, splice loss, and
system margin. Thus, if PS is the optical power emerging from the end of a fiber flylead attached to the
light source or from a source-coupled connector, and if PR is the receiver sensitivity, then
PT = PS - PR
= 2lc + aL + system margin (8.2)

Example 8.1 To illustrate how a link loss budget If a = 3.5 dB/km, then a 6.0-km transmission path is
is set up, let us carry out a specific design example. We possible.
begin by specifying a data rate of 20 Mb/s and a bit-error The link power budget can be represented graphically as
rate of 10–9 (i.e., at most one error can occur for every is shown in Fig. 8.4. The vertical axis represents the optical
109 bits sent). For the receiver, we choose a silicon pin power loss allowed between the transmitter and the receiver.
photodiode operating at 850 nm. Figure 8.3 shows that the The horizontal axis gives the transmission distance. Here,
required receiver input signal is –42 dBm (42 dB below we show a silicon pin receiver with a sensitivity of –42 dBm
1 mW). We next select a GaAlAs LED that can couple a (at 20 Mb/s) and an LED with an output power of –13 dBm
50-mW (–13-dBm) average optical power level into a coupled into a fiber flylead. We subtract a 1-dB connector
fiber flylead with a 50-mm core diameter. We thus have loss at each end, which leaves a total margin of 27 dB.
a 29-dB allowable power loss. Assume further that a Subtracting a 6-dB system safety margin leaves us with a
1-dB loss occurs when the fiber flylead is connected to tolerable loss of 21 dB that can be allocated to cable and
the cable and another 1-dB connector loss occurs at the splice loss. The slope of the line shown in Fig. 8.4 is the
cable-photodetector interface. Including a 6-dB system 3.5-dB/km cable (and splice, in this case) loss. This line
margin, the possible transmission distance for a cable with starts at the –14-dBm point (which is the optical power
an attenuation a can be found from Eq. (8.2): coupled into the cabled fiber) and ends at the –35-dBm level
PT = PS – PR = 29 dB (the receiver sensitivity minus a 1-dB connector loss and a
= 2(1 dB) + a L + 6 dB 6-dB system margin). The intersection point D then defines
the maximum possible transmission path length.
– 20
InGaAs pin
Receiver sensitivity (dBm)

– 30 (1300 nm)
Si pin
(800-900 nm)
– 40

– 50
InGaAs APD
(1550 nm)
– 60
Si APD
– 70 (800-900 nm)

– 80
.5 1 2 5 10 20 50 100 200 500 1000
Data rate (Mb/s)

Fig. 8.3 Receiver sensitivities as a function of

bit rate. The Si pin, Si APD, and InGaAs Fig. 8.4 Graphical representation of a link
pin curves are for a 10-9 BER. The loss budget for an 850-nm LED/pin
InGaAs APD curve is for a 10-11 BER. system operating at 20 Mb/s
 Digital Links 309

Here, lc is the connector loss, a is the fiber attenuation (dB/km), L is the transmission distance, and the
system margin is nominally taken as 6 dB. Here, we assume that the cable of length L has connectors
only on the ends and none in between. The splice loss is incorporated into the cable loss for simplicity.
A convenient procedure for calculating the power budget is to use a tabular or spreadsheet form. We
illustrate this by way of an example for a 2.5-Gb/s link that may be used for SONET OC-48 or SDH
STM-16.

Example 8.2 Consider a 1550-nm laser diode that assume a 1-dB connector loss occurs at each fiber joint (two
launches a +3-dBm (2-mW) optical power level into a fiber at each end because of the jumper cables).
flylead, an InGaAs APD with a –32-dBm sensitivity at Table 8.1 lists the components in column 1 and
2.5 Gb/s, and a 60-km long optical cable with a 0.3-dB/km the associated optical output, sensitivity, or loss in
attenuation. Assume that here, because of the way the column 2. Column 3 gives the power margin available
equipment is arranged, a 5-m optical jumper cable is needed after subtracting the component loss from the total optical
at each end between the end of the transmission cable and power loss that is allowed between the light source and the
the SONET equipment rack as shown in Fig. 8.5. Assume photodetector, which, in this case, is 35 dB. Adding all the
that each jumper cable introduces a loss of 3 dB. In addition, losses results in a final power margin of 7 dB.

Table 8.1 Example of a spreadsheet for calculating an optical-link power budget

Component/loss parameter Output/sensitivity/loss Power margin (dB)

Laser output 3 dBm
APD sensitivity at 2.5 Gb/s –32 dBm
Allowed loss [3 – (–32)] 35
Source connector loss 1 dB 34
Jumper + connector loss 3 + 1 dB 30
Cable attenuation (60 km) 18 dB 12
Jumper + connector loss 3 + 1 dB 8
Receiver connector loss 1 dB 7 (final margin)

SONET OC-48
equipment with
optical transceivers

3-dBm input –25-dBm output

5-m jumper Optical connectors

5-m jumper

60-km optical transmission cable

Patch panel Patch panel

Fig. 8.5 A 2.5-Gb/s 60-km optical fiber link with 5-m optical jumper cables at each end
 310 Optical Fiber Communications

8.1.3 Rise-Time Budget

A rise-time budget analysis is a convenient method for determining the dispersion limitation of an optical
fiber link. This is particularly useful for digital systems. In this approach, the total rise time tsys of the link
is the root sum square of the rise times from each contributor ti to the pulse rise-time degradation:
12
⎛ N ⎞
tsys = ⎜ ∑ ti2 ⎟ (8.3)
⎜ ⎟
⎝ i =1 ⎠

The four basic elements that may significantly limit system speed are the transmitter rise time ttx, the
group-velocity dispersion (GVD) rise time tGVD of the fiber, the modal dispersion rise time tmod of the
fiber, and the receiver rise time trx. Single-mode fibers do not experience modal dispersion, so in these
fibers the rise time is related only to GVD. Generally, the total transition-time degradation of a digital
link should not exceed 70 percent of an NRZ (non-return-to-zero) bit period or 35 percent of a bit period
for RZ (return-to-zero) data, where one bit period is defined as the reciprocal of the data rate (NRZ and
RZ data formats are discussed in more detail in Sec. 4.4).
The rise times of transmitters and receivers are generally known to the designer. The transmitter rise
time is attributable primarily to the light source and its drive circuitry. The receiver rise time results from
the photodetector response and the 3-dB electrical bandwidth of the receiver front end. The response of
the receiver front end can be modeled by a first-order lowpass filter having a step response11–12
g(t) = [1 - exp (-2p Bet)]u(t)

where Be is the 3-dB electrical bandwidth of the receiver and u(t) is the unit step function which is 1
for t ≥ 0 and 0 for t < 0. The rise time trx of the receiver is usually defined as the time interval between
g(t) = 0.1 and g(t) = 0.9. This is known as the 10- to 90-percent rise time. Thus, if Be is given in megahertz,
then the receiver front-end rise time in nanoseconds is (see Prob. 8.3)

350
trx = (8.4)
Be

In practice, an optical fiber link seldom consists of a uniform, continuous, jointless fiber. Instead, a
transmission link nominally is formed from several concatenated (tandemly joined) fibers that may have
different dispersion characteristics. This is especially true for dispersion-compensated links operating
at 10 Gb/s and higher (see Chapter 13). In addition, multimode fibers experience modal distributions at
fiber-to-fiber joints owing to misaligned joints, different core index profiles in each fiber, and/or different
degrees of mode mixing in individual fibers. Determining the fiber rise times resulting from GVD and
modal dispersion then becomes more complex than for the case of a single uniform fiber.
The fiber rise time tGVD resulting from GVD over a length L can be approximated by Eq. (3.44) as

tGVD ≈ D Lσλ (8.5)

where sl is the half-power spectral width of the source, and the dispersion D is given by Eq. (3.47) for
a non-dispersion-shifted fiber and by Eq. (3.49) for a dispersion-shifted fiber. Since the dispersion value
generally changes from fiber section to section in a long link, an average value should be used for D in
Eq. (8.5).
 Digital Links 311

The difficulty in predicting the bandwidth (and hence the modal rise time) of a series of concatenated
multimode fibers arises from the observation that the total route bandwidth can be a function of the order
in which fibers are joined. For example, instead of randomly joining together arbitrary (but very similar)
fibers, an improved total link bandwidth can be obtained by selecting adjoining fibers with alternating
over- and undercompensated refractive-index profiles to provide some modal delay equalization. Although
the ultimate concatenated fiber bandwidth can be obtained by judiciously selecting adjoining fibers for
optimum modal delay equalization, in practice this is unwieldy and time-consuming, particularly since
the initial fiber in the link appears to control the final link characteristics.
A variety of empirical expressions for modal dispersion have thus been developed.13–16 From practical
field experience, it has been found that the bandwidth BM in a link of length L can be expressed to a
reasonable approximation by the empirical relation
B0
BM ( L) = (8.6)
Lq
where the parameter q ranges between 0.5 and 1, and B0 is the bandwidth of a 1-km length of cable. A
value of q = 0.5 indicates that a steady-state modal equilibrium has been reached, whereas q = 1 indicates
little mode mixing. Based on field experience, a reasonable estimate is q = 0.7.
Another expression that has been proposed for BM based on curve fitting of experimental data, is
q
⎡N ⎛ ⎞
1 q⎤
1 1
= ⎢ ∑ ⎜⎜ ⎟⎟ ⎥ (8.7)
BM ⎢ n =1 ⎝ Bn ⎠ ⎥
⎣ ⎦
where the parameter q ranges between 0.5 (quadrature addition) and 1.0 (linear addition), and Bn is the
bandwidth of the nth fiber section. Alternatively, Eq. (8.7) can be written as
q
⎡N ⎤
tM ( N ) = ⎢ ∑ (tn )1 q ⎥ (8.8)
⎢⎣ n =1 ⎥⎦
where tM (N) is the pulse broadening occurring over N cable sections in which the individual pulse
broadenings are given by tn.
We now need to find the relation between the fiber rise time and the 3-dB bandwidth. We assume that
the optical power emerging from the fiber has a gaussian temporal response described by
1 2σ 2
e− t
2
g (t ) = (8.9)
2πσ
where s is the rms pulse width.
The Fourier transform of this function is
1
e−ω σ2 2
2
G (ω ) = (8.10)
2π
From Eq. (8.9) the time t1/2 required for the pulse to reach its half-maximum value, that is, the time
required to have
g(t1/2) = 0.5 g(0) (8.11)
 312 Optical Fiber Communications

is given by

t1/2 = (2 ln 2)1/2s (8.12)

If we define the time tFWHM as the full width of the pulse at its half-maximum value, then

tFWHM = 2t1/2 = 2s (2 ln 2)1/2 (8.13)

The 3-dB optical bandwidth B3dB is defined as the modulation frequency f3dB at which the received
optical power has fallen to 0.5 of the zero frequency value. Thus, from Eqs. (8.10) and (8.13), we find that
the relation between the full-width half-maximum rise time tFWHM and the 3-dB optical bandwidth is
0.44
f3dB = B3dB =
tFWHM
(8.14)
Using Eq. (8.6) for the 3-dB optical bandwidth of the fiber link and letting tFWHM be the rise time
resulting from modal dispersion, then, from Eq. (8.14),
0.44 0.44 Lq
tmod = =
BM B0
(8.15)
If tmod is expressed in nanoseconds and BM is given in megahertz, then
440 440 Lq
tmod = =
BM B0
(8.16)
Substituting Eqs. (3.20), (8.4), and (8.16) into Eq. (8.3) gives a total system rise time of
12
tsys = ⎡ttx2 + tmod
2 2
+ tGVD + trx2 ⎤
⎣ ⎦
12
⎡ ⎛ 440 Lq ⎞
2 2⎤
⎛ 350 ⎞ ⎥
= ⎢ttx2 + ⎜ 2 2 2
⎟ + D σλ L + ⎜⎜ ⎟⎟ ⎥ (8.17)
⎢ ⎜ B ⎟
⎢⎣ ⎝ 0 ⎠ ⎝ Be ⎠ ⎥
⎦

Example 8.3 As an example of a rise-time budget the modal-dispersion-induced fiber rise time is 3.9 ns.
for a multimode link, let us continue the analysis of the Substituting all these values back into Eq. (8.17) results
link we started to examine in Sec. 8.1.2. We assume that in a link rise time of
the LED together with its drive circuit has a rise time of
( )
12
tsys = ttx2 + tmat
2 2
+ tmod 2
+ trx
15 ns. Taking a typical LED spectral width of 40 nm, we
have a material-dispersion-related rise-time degradation = [(15 ns)2 + ( 21 ns)2 + (3.9 ns)2 + (14 ns)2 ]1/ 2
of 21 ns over the 6-km link. Assuming the receiver has a
= 30 ns
25-MHz bandwidth, then from Eq. (8.4) the contribution
to the rise-time degradation from the receiver is 14 ns. If This value falls below the maximum allowable 35-ns
the fiber we select has a 400-MHz · km bandwidth-distance rise-time degradation for our 20-Mb/s NRZ data stream
product and with q = 0.7 in Eq. (8.6), then from Eq. (8.15) (0.70/bit rate). The choice of components was thus
adequate to meet our system design criteria.
 Digital Links 313

where all the times are given in nanoseconds, sl is the half-power spectral width of the source, and the
dispersion D [expressed in ns/(nm · km)] is given by Eq. (3.47) for a non-dispersion-shifted fiber and
by Eq. (3.49) for a dispersion-shifted fiber. As indicated by the curves in Fig. 3.28 for G.652 single-
mode fiber, the dispersion D is less than +3.5 ps/(nm · km) in the O-band and about +17 ps/(nm · km)
at 1550 nm. For G.655 fiber the dispersion values range from –10 to –3 ps/(nm · km) across the O-band
and from +5 to +10 ps/(nm-km) in the C-band.
Analogous to power budget calculations, a convenient procedure for keeping track of the various
rise-time values in the rise-time budget is to use a tabular or spreadsheet form. We illustrate this by way
of an example for the SONET OC-48 (2.5 Gb/s) link we looked at in Example 8.2.

Example 8.4 Assume that the laser diode together Table 8.2 lists the components in column 1 and
with its drive circuit has a rise time of 0.025 ns (25 ps). the associated rise times in column 2. Column 3 gives
Taking a 1550-nm laser diode spectral width of 0.1 nm the allowed system rise-time budget of 0.28 ns for a
and an average dispersion of 2 ps/(nm · km) for the fiber, 2.5-Gb/s NRZ data stream at the top. This is found from
we have a GVD-related rise-time degradation of 12 ps the expression 0.7/BNRZ where BNRZ is the bit rate for the
(0.012 ns) over a 60-km long optical cable. Assuming the NRZ signal. The calculated system rise time of 0.14 ns is
InGaAs-APD-based receiver has a 2.5-GHz bandwidth, shown at the bottom. The system rise time, in this case, is
then from Eq. (8.4) the receiver rise time is 0.14 ns. Using dominated by the receiver and is well within the required
Eq. (8.17) to add up the various contributions, we have a limits.
total rise time of 0.14 ns.

Table 8.2 Example for a tabular form for keeping track of component contributions to
an optical-link rise-time budget

Component Rise time Rise-time budget

Allowed rise-time budget tsys = 0.7/BNRZ = 0.28 ns
Laser transmitter 25 ps
GVD in fiber 12 ps
Receiver rise time 0.14 ns
System rise time [Eq. (8.17)] 0.14 ns

8.1.4 Short-Wavelength Band

Figure 8.6 shows the attenuation and dispersion limitation on the repeaterless transmission distance as a
function of data rate for the short-wavelength (770–910-nm) LED/pin combination. The BER was taken
as 10–9 for all data rates. The fiber-coupled LED output power was assumed to be a constant –13 dBm
for all data rates up to 200 Mb/s. The attenuation limit curve was then derived by using a fiber loss of
3.5 dB/km and the receiver sensitivities shown in Fig. 8.3. Since the minimum optical power required at
the receiver for a given BER becomes higher for increasing data rates, the attenuation limit curve slopes
downward to the right. We have also included a 1-dB connector-coupling loss at each end and a 6-dB
system operating margin.
The dispersion limit depends on material and modal dispersion. Material dispersion at 800 nm is
taken as 0.07 ns/(nm · km) or 3.5 ns/km for an LED with a 50-nm spectral width. The curve shown is
the material dispersion limit in the absence of modal dispersion. This limit was taken to be the distance