Current generated by photodetector is very weak and is adversely effected by random noises associated with photo detection process. When amplified, this signal further gets corrupted by amplifiers. Noise considerations are thus important in designing optical receivers.
Most meaningful criteria for measuring performance of a digital communication system is average error probability, and in analog system, it is peak signal to rms noise ratio. ...
Optical Fiber Communication Part 3 Optical Digital Receiver
1. OPTICAL FIBER
COMMUNICATION
⢠Digital Receivers
⢠Probability of error
⢠Quantum Limit
⢠Shot noise
⢠Noise Penalty
⢠Pre-amplifier types
PART III:-
DIGITAL RECEIVERS
3. DIGITAL RECEIVER
⢠Analog systemâSignal to rms noise ratio
⢠Digital systemâ Average error probability
4. DIGITAL RECEIVER-
PHOTON DETECTION QUANTUM NOISE
⢠Due to random arrival rate of signal photons.
⢠Makes primary photocurrent a time varying Poisson
Process.
⢠If detector illuminated by optical signal p(t), then
average number of electron-hole pair generated in time
Ď is --
⢠Ρ is detector quantum efficiency.
5. DIGITAL RECEIVER-
PHOTON DETECTION QUANTUM NOISE
⢠Actual number of electron-hole pair n fluctuates from
average according to Poisson distribution.
6. DIGITAL RECEIVER
ď˘ As the pulse progresses, it spreads and enters into
adjacent time slots causing ISI.
ď˘ Major part Îł in desired slot while rest spreads.
7. DIGITAL RECEIVER
ď˘ hp(t) - input pulse shape
ď˘ Hp(Ď) â Fourier transform of
ď˘ HB(Ď) â transfer function of bias circuit
ď˘ Heq(Ď) â transfer function of equalising circuit
ď˘ A â Gain of amplifier
8. DIGITAL RECEIVER
ď˘ Îˇ â Quantum efficiency of photo detector
ď˘ Cd â Photodiode capacitance
ď˘ Rb â Detector bias resistance
ď˘ Ra ĐĐ Ca â Amplifier input impedance
ď˘ Ca â Amplifier shunt capacitance
ď˘ ib(t) â Thermal noise current generated by Rb
ď˘ ia(t) â Thermal noise current generated by Ra
ď˘ va(t) â Thermal noise voltage of amplifier channel
ď˘ Input voltage develops across Ra
ď˘ Two amplifier noise sources ia(t) , va(t)
ď˘ One detector noise source ib(t) due to bias resistor.
9. DIGITAL RECEIVER
ď˘ All noises Gaussian, have flat spectral response ( white
noise), uncorrelated, statically independent .
ď˘ Occurrence of one doesnât effect occurrence of other.
ď˘ Input pulse train is -
⢠bn â Amplitude of nth pulse
⢠hp - received Pulse shape
⢠Tb â Bit period
⢠bn can be bon or boff
⢠hp(t) normalized to have unit area.
10. DIGITAL RECEIVER
ď˘ Mean output current from detector -
⢠Current amplified and filtered to produce mean
voltage at output of equalizer -
⢠hB(t) and heq(t) are impulse responses of bias and
equalizer circuits.
12. PROBABILITY OF ERROR
Ne â errors occurring in time t
Nt â total pulses transmitted in time t
B â bit rate = 1/Tb
ď˘ Assuming the noise has Gaussian probability density
function with zero mean.
ď˘ Noise voltage n(t) sampled at any arbitrary time t,
ď˘ The probability that the measured sample n(t) falls in
range n to n+dn is -
13. PROBABILITY OF ERROR- CASE I â â0â IS BEING SENT.
ď˘ Let transmitted pulses are â0â and â1â, vth = V/2
ď˘ Transmitted = â0â
ď˘ Received r(t) = n(t) = v
ď˘ Probability that it be detected as â1â is probability
that v lied between V /2and â.
14. PROBABILITY OF ERROR- CASE II â â1â IS BEING SENT.
ď˘ Transmitted = â1â
ď˘ Received r(t) = V + n(t) = v
ď˘ n(t) = v - V
ď˘ Probability that it be detected as â0â is probability that
v lied between - âand V /2.
15. TOTAL PROBABILITY OF ERROR-
ď˘ Pe = a p0(v) + b p1(v)
ď˘ Assuming â0â and â1â are equiprobable, a = b = 0.5.
ď˘ Pe = ?
ď˘ Looking at distribution p0(v) and p1(v) are
identical.
ď˘ Integrating double of one part -
19. QUANTUM LIMIT TO DETECTION
ď˘ Ideal photo detector having unity quantum efficiency and
no dark current.
ď˘ No e-h pair generated in absence of optical pulse.â0â
ď˘ Possible to find minimum received optical power
required for specific BER performance in digital system.
ď˘ Called Quantum limit.
20. QUANTUM LIMIT TO DETECTION
ď˘ Optical pulse of energy E falls on photo detector in
time interval Ď.
ď˘ During transmission signal if too low to generate
any e-h pair and detected as 0.
ď˘ Then for error probability Pr(0), there exists a
minimum energy E at wavelength Îť, to be detected
as 1.
ď˘ Probability that n=0 electrons are emitted in
interval Ď-
21. QUANTUM LIMIT TO DETECTION - PROBLEM
ď˘ Digital fiber optic link operating at wavelength 850nm
requires maximum BER of 10-9. Find quantum limit and
minimum incident power Po that must fall on photo
detector , to achieve this BER at data rate of 10Mbps for
simple binary level signaling system. Quantum
efficiency is 1.
ď˘ Solution âfor maximum BER,--
22. QUANTUM LIMIT TO DETECTION - PROBLEM
ď˘ Minimum incident power that must fall on photo
detector Po --- E= Po Ď
ď˘ Assuming equal number of 0 and 1, 1/ Ď = B/2
23. RECEIVER NOISES
ď˘ Noise voltage vN(t) causes vout to deviate from
mean or average < vout >
⢠vs(t) â Quantum or shot noise due to random
multiplied poisson nature of photocurrent is(t).
⢠vR(t) â thermal noise due to bias resistor Rb.
⢠vI(t) ânoise due to amplifier input noise
⢠vE(t) ânoise due to amplifier due to ea(t).
24. Bbae â Noise equivalent bandwidth of bias ckt, amplifier
and equalizer
⢠Calculating the three thermal noise currents at the
output of equalizer :-
25.
26. SHOT NOISES-
â˘Shot noise in bit period Tb is shot noise contribution from
a pulse within that period as well as from all other pulses
outside that period.
â˘Worst case shot noise when all neighboring pulses are â1â.
â˘Greatest ISI.
â˘Hence mean unity gain photocurrent over Tb for 1 pulse -
28. SHOT NOISES â â0â WITH ALL NEIGHBOR â1âS â
BOFF = 0
â˘Substitute <io>o and <io>1 in vs
2(t) to find worst case shot
noise for pulse â1â and â0â.
29. SNR REQUIRED TO ACHIEVE MIN BER
ď˘ Assuming output voltage is approximately Gaussian.
ď˘ Mean and variance of Gaussian output for â1â and â0â
are bon, Ďon
2 and boff and Ďoff
2.
ď˘ Decision threshold vth set for equal error probability
for â1â and â0â.
30. SNR REQUIRED TO ACHIEVE MIN BER
ď˘ Error probability for â1â and â0â are
-
⢠Defining Q related to SNR to achieve desired min BER-
⢠Putting Q/â2 = x, change integral and limits.
31. SNR REQUIRED TO ACHIEVE MIN BER
ď˘ Relative to noise at boff, threshold vth must be
ATLEAST Q standard deviation above boff.
ď˘ Or, vth should be above boff by rms value Q.
ď˘ Relative to noise at bon, threshold vth must not be
below bon by more than Q sandard deviation to have
min BER
32. NOISE PENALTY IN PRACTICAL SYSTEM- POWER
PENALTY
ď˘ In practical system, many factors degrade the
performance.
ď˘ We assumed that â
ď Optical energy of each bit is impulse response h(t).
ď Zero energy sent during â0â.
ď Receiver amplifier sharply band limited.
ď No random variation in amplitude and arrival time of
bit.
ď˘ In practical systems, each violation demands increase in
received signal power to ensure given error probability.
ď˘ This additional excess power âP required in practical
system is called power penalty â in dB
33. 1. NON-ZERO EXTINCTION RATIO
ď˘ Assumed boff = 0 during â0â.
ď˘ In actual system, light source biased slightly ON at
all times to obtain shorter turn-on time in LED or
keep it above threshold in LASER.
ď˘ Extinction ratio Đ is optical energy emitted in the â0â
pulse to that during â1â pulse.
ď˘ Đ = boff / bon
ď˘ Varies between 0 and 1.
ď˘ Any dark current in photodiode appears to increase
Đ.
ď˘ With equally probable â0â and â1â, minimum received
power ( sensitivity) Pr min is given by average energy
detected per pulse times the pulse rate 1 / Tb.
34. 1. NON-ZERO EXTINCTION RATIO
ď˘ The extinction ratio penalty i.e., the penalty in
receiver sensitivity as a function of extinction ratio
is-
35. 2. FINITE PULSE WIDTH AND TIMING JITTER
ď˘ Received optical pulse has a finite pulse width.
ď˘ Some timing jitter is present.
ď˘ Hence noise penalty is required forâ
ď Non-optical filtering is needed to provide equalization
against pulse distortion or to minimize ISI.
ď Some ISI remains and degrades SNR.
ď˘ To calculate magnitude of these effects, it is
necessary to define
ď Shape of received pulse.
ď The distribution of the jitter
ď˘ We deal with only the former.
36. 2. FINITE PULSE WIDTH AND TIMING JITTER
ď˘ Power penalty Vs Ď/T for Gaussian shaped pulse shown.
ď˘ Ď is rms width of the pulse (due to changes in pulse shape.)
T is basic pulse width.
37. 2. FINITE PULSE WIDTH AND TIMING JITTER
ď˘ It demonstrates possible trade-off between bit rate and
signal power.
ď˘ Relates effects of fiber attenuation and fiber disperssion.
ď˘ Power penalty < 1db if Ď remains less than T/5.
ď˘ Result independent of pulse width.
ď˘ But if Ď > 1dB, PP increases sharply.
ď˘ It becomes more sensitive to pulse shape.
ď˘ In practice, system is
ď either limited by fiber dispersion (BW Limited)
ď or by fiber attenuation(Power Limited) .
ď Possible trade-off between two is quite small.
38. PREAMPLIFIER TYPES
ď˘ Sensitivity and bandwidth of a receiver are effected by noise
sources at the front end, i.e. at preamplifier.
ď˘ Preamplifier should give maximum receiver sensitivity with
desired bandwidth.
ď˘ Three main types. But intermediate types can also be used.
39. PREAMPLIFIER TYPES â LOW IMPEDANCE (LZ)
ď˘ Simplest, but not optimum design.
ď˘ Photodiode operates into a low impedance amplifier
(appox.50âŚ)
ď˘ Bias or load resister Rb used to match amplifier impedance
by suppressing standing waves and give uniform frequency
response.
ď˘ Rb with amplifier capacitance gives BW equal to or greater
than signal BW.
ď˘ LZ amplifier can operate over a wide BW.
ď˘ Gives low resistivity as a small voltage develops across
amplifier input and Rb .
ď˘ Hence used only for short distance applications where
sensitivity not of concern.
41. PREAMPLIFIER TYPES â HIGH IMPEDANCE
(HZ)
ď˘ Noise reduced by reducing input capacitance.
ď˘ By selecting.
ď Low capacitance, high frequency devices.
ď Detector of low dark current.
ď Bias resistor of minimum thermal noise.
ď˘ Thermal noise can be reduced by high impedance amplifier
with large photo detector Rb , hence HZ ampr.
ď˘ But causes large RC time constant â low front-end BW.
ď˘ Signal gets integrated.
ď˘ Equalization technique required. (Differentiator).
ď˘ Integrator-differentiator is core of HZ design.
ď˘ Gives low noise but low dynamic range of signal.
45. HIGH IMPEDANCE (HZ)- ANALYSIS
ď˘ SNR â K can be improved by increasing M until shot
noise term (iii) increases by F(M).
ď˘ F(M) increases with M, gets comparable to other terms.
Has optimum value for M.
ď˘ K improves by increasing front end R till (i) and (iv) are
significant. But increases RC. Hence requires more
equalization and low C.
ď˘ If equalization required, (ii) dominates at high freq.
Noise increases as C2. Hence low C required.
ď˘ Shot noise (iii) depends on input signal level.
ď˘ Assumption of all noises statistical not true in reality.
ď˘ All noises including F(M) not purely Gaussian.
46. HIGH IMPEDANCE (HZ)- ANALYSIS
ď˘ Hence Actual SNR may be lesser.
ď˘ Two limitations due to Integrator-differentiator
ď˘ Equalization required for broadband applications.
ď˘ limited dynamic range.
47. PREAMPLIFIER TYPES
â TRANS IMPEDANCE
ď˘ High gain and high impedance amplifier with
feedback.
ď˘ Gives low noise and large dynamic range.
ď˘ 1/R = 1/Ra + 1/Rb + 1/Rf
ď˘ C = Ca + Cd
54. PREAMPLIFIER TYPESâ TRANS IMPEDANCE -
ANALYSIS
ď˘ R and Rf can be increased to reduce SNR, (i) and (iv)
without equalization provided A>>2ĎC Rf B.
ď˘ It has wide dynamic range.
ď˘ Output resistance is small so that amplifier is less
susceptible to pickup noise.
ď˘ Transfer characteristic is actually trans impedance
feedback resister. Hence amplifier stable and easily
controlled.
ď˘ Although TZ ampr is less sensitive than HZ ampr as
S/NTZ > S/NHZ, the difference is usually only 2 to 3 dB
for most practical wideband design.