Beruflich Dokumente
Kultur Dokumente
by
August–December 2003
c C. D. Cantrell (06/2003)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
20 20
0 0
-20 -20
-40 -40
-40 -20 0 20 40 -40 -20 0 20 40
40 40
20 20
0 0
-20 -20
-40 -40
-40 -20 0 20 40 -40 -20 0 20 40
Horizontal averages are time averages; vertical averages (at a fixed time
t) are ensemble averages
C. Kittel, Elementary Statistical Physics (Wiley, 1958)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
The random process x(t) itself may not have a Fourier transform that is
a well-behaved function
◦ Trivial example: x(t) = constant
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
WHITE NOISE
• For general random processes, fluctuations die away in a finite time; e.g.,
x(t)x(t + τ ) = x(t)2e−|τ |/τc
• For white noise, the autocorrelation is proportional to a delta function:
x(t)x(t + τ ) = XRMS
2
δ(τ )
The Weiner-Khinchin theorem implies that the power spectral density is
flat (independent of frequency):
∞
Sn(f ) = x(t)x(t + τ )e−2πif τ dτ = XRMS
2
−∞
It’s called white noise because its spectrum consists of equal amplitudes
of all frequencies, just as white light consists of equal amplitudes of all
colors
The total noise power due to a white noise source is proportional to the
bandwidth of the system with which you “look at” the noise
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School
PhoTEC
c C. D. Cantrell (01/1999)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
RECEIVER RESPONSE
2π∆f
ω
0
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
NOISE IN A PHOTODETECTOR
THERMAL NOISE
SHOT NOISE
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
NOISE-EQUIVALENT POWER
c C. D. Cantrell (07/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
S. W. Hinch and Christopher M. Miller, Fiber Optic Test and Measurement, Dennis Derickson, Editor, Fig. 8.13
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
Before transmission
After 1km 50MMF
H
Mode partitioning example
Four optical spectra of the same device taken about 1 minute apart
850 852 854 856 858 860 862 850 852 854 856 858 860 862
Wavelength (nm) Wavelength (nm)
850 852 854 856 858 860 862 850 852 854 856 858 860 862
Wavelength (nm) Wavelength (nm)
Note: These are time averaged measurements and do not give more than a
qualitative indication that mode partitioning is happening.
H
Measuring Mode Partitioning
laser receiver
Optical Preselector/
Filter Sampling
Oscilloscope
(e.g. HP71451B)
laser spectrum
isolated mode
all modes
individual
mode
H l=792.46nm
l=793.07nm
Mode partitioning - 1300nm
Fabry-Perot laser
l=1316.48 l=1318.76
H
Mode partitioning - 850nm
Fabry-Perot laser
l=855.79nm
850 852 854 856 858 860
Wavelength (nm)
l=854.28nm l=855.04nm
H
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
c C. D. Cantrell (06/2003)
Modal Noise: Causes 1
Spatial modes propagating in a multimode fiber interfere and create
speckle pattern:
Multimode fibre
Laser diode
vibrations
temperature changes
λ λ
High frequency effects
(laser mode partitioning)
input
side output
side
With the changing speckle pattern, the amount of power lost in the mode selective
element varies, creating amplitude noise at the receiver.
The amount of modal noise depends on:
1. Coherence of the light source
2. Fiber bandwidth
3. Amount of the mode selective loss
c C. D. Cantrell (11/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
• The magnitude of the optical electric field at the output of an optical am-
plifier is
E = Go Ein + Esp
where Ein is the input field to the amplifier and Esp is the electric field of
the light spontaneously emitted (and amplified) by the amplifier
Photocurrent produced in a detector placed at the amplifier output:
c C. D. Cantrell (11/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
c C. D. Cantrell (11/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
c C. D. Cantrell (12/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
noise
Wavelength
• The bit error rate or bit error ratio is the probability of detecting a
bit incorrectly in signaling single bits
An experimental estimate of the probability of error is the ratio
E(T )
BER(T ) =
N (T )
where E(T ) is the number of errored bits in the gating period T , and
N (T ) is the total number of bits
Basic bit-error-rate test (BERT) arrangement:
c C. D. Cantrell (08/2002)
The University of Texas at Dallas Erik Jonsson School
PhoTEC
Dennis Derickson, Ed., Optical Fiber Test and Measurement, Fig. 8.9
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
• A useful performance metric for a digital receiver is the bit error rate,
defined as the probability of detecting a bit incorrectly
For a binary line code,
BER = p(1)P (0|1) + p(0)P (1|0)
Signal
I1
1 bit level
P(1|0) ID
decision threshold
P(0|1)
I0
0 bit level
Time
sampling times
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
10 0
10 −2
10 −4
10 −6
BER
10 −8
10 −10
10 −12
10 −14
10 −16
2 3 4 5 6 7 8
Q
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
POWER PENALTIES
• The BER for a binary line code is determined by Q = (I1 − I0)/(σ1 + σ0),
where I1 is the photocurrent in a “1” bit and I0 is the current in a “0” bit
The value of Q at a distance z from the transmitter (z = 0),
I1 − I0
Q(z) = ,
σ1 + σ0 z
is degraded by attenuation, noise, dispersion, and other effects as the pulse
train travels from z = 0 to its destination at z = L
Each source of noise contributes to σ1 + σ0, thereby making
(σ1 + σ0)z=L > (σ1 + σ0)z=0
while other effects reduce I1 and I0, making Q(L) < Q(0) and therefore
BER(L) > BER(0)
The overall power penalty, ρpen, is equal to the factor by which I1
and I0 must be multiplied to make Q(L) equal to Q(0):
Q(0)
ρpen = Units: dB
Q(L)
c C. D. Cantrell (06/2003)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
EXTINCTION RATIO
• The extinction ratio is the ratio of the power in “0” bits to the power in “1”
bits:
I0
=
I1
• If
= 0, there is a power penalty equal to the power required to bring
I1 − I0
Q=
σ1 + σ0
up to the value it would have if I0 = 0
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
c C. D. Cantrell (07/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
σ1 + σ0 2 +σ
σs2 + σth th
Find the value of Q for a given BER, then solve for the average received
power:
Q σth
P rec(Q) = qFAQ∆f +
R M
P rec(Q) is called the receiver sensitivity for the BER determined by
the value of Q
◦ The receiver sensitivity depends on the bit rate through the receiver
bandwidth ∆f
For a given modulation format (RZ, NRZ), ∆f depends on the block-
encoded data rate
The higher the data rate, the higher (worse) P rec(Q) is
c C. D. Cantrell (09/2002)
The University of Texas at Dallas Erik Jonsson School of
Engineering & Computer Science
-17
Optical Sensitivity (dBm)
-19
-21
-23
-25
-27
• Typical sensitivities for a BER of 10−10 are −26 dBm at a bit rate B = 2.5
Gb/s, or −18 dBm at a bit rate B = 10 Gb/s
Agilent 5988-5927EN.pdf