VCSEL application in fiber LAN by hcj


									Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

                VCSEL application in fiber LAN
 —— The development of                            low cost, fine quality
           1.25Gb/s laser transceiver

                                Zhang Heng

 Grade 00, School of information technology, Department of electronics

    This project aims at designing a low cost, fine quality 1.25 Gb/s laser
transceiver, which is mainly used in LAN, with VCSEL (Vertical Cavity
Surface Emitting Laser).This paper gives a brief introduction of the
optical fiber system and the character of VCSEL, discusses the design
principle and structure of the transceiver and shows an experimental


  本项目的主要内容是利用 VCSEL(垂直腔面发射半导体激光器)设计并制作一
种专门用于接入网和家庭用户的、具有良好性能与较低成本的 1.25 Gb/s 光收发
一体化模块。本文先对光纤通信系统、光收发模块及 VCSEL 的特点作简要介绍,

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

1. The significance of the task
   In modern society, with the development of science and technology,
communication becomes so frequent and the information content is so huge
that it is of vital importance to find a new fast-transmitting media. For
instance, the 10/100 Mbp/s Fast Ethernet running on CAT 5E cables has
dominated LAN standards for years. However, it still takes considerable
time to transfer the content of a 2-hour long movie. It seems that LAN
system based on cable is not capable for today’s information transmission.
At the same time, with the rapid development of optical fiber
communication technology, it is already common for fiber to reach a
10Gbp/s transmitting rate, and the rate will soon soar to 40Gbp/s.
Moreover, the connecting speed of network device is speeding up, and
traditional LAN media (e g. copper) is no longer suitable for high-speed
information transmission.

Table 1.Transmitting rate vs. distance for copper cable
Transmitting rate                              Transmitting distance
2 Mb/s                                         Several kilometers
100 Mb/s                                       100 meters
2.5 Gb/s                                       Less than 5 meters

   It can be seen from table 1 that the application of fiber in LAN is
   According to the distance light travels, the optical fiber
communication can be classified as WAN (Wide Area Network) and LAN (Local
Area Network).The characters of WAN and LAN are listed in table 2.

Table2: Character of WAN and LAN
WAN                                            LAN
Long distance: 100 ~ more than 1000            Close distance: 100 meters ~ several
km                                             kilometers
High transmitting rate: it is                  Relatively low transmitting rate:2
already common for 10 Gb/s fiber               Mb/s to 2.5 Gb/s
communication, which will soon soar
to 40 Gb/s
Dense wavelength channels: 32~ up to           Low wavelength channel density:
100 channels                                   single or several channels
Vile circumstance, through rivers,             Good circumstance, mainly
swamps, mountains                              distributed in populous area
High utilization, an influx from               Low utilization: only shared by a
many area and many application                 few people

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
   Table 2 shows that the requirement of laser source of WAN and LAN is
different. WAN asks for a single-mode laser source that has a narrow
spectrum, uses indirect modulation, and has stable wavelength and small
temperature drift. Since there are many users to share the cost, it is
possible to use the high-cost DFB laser. On the other hand, LAN puts much
less stress on the laser source; therefore, it is possible to use a cheap
laser as the source.
   As the manufacture technique of VCSEL is very different from the
traditional DFB laser, it is suitable for mass production, thus costs
little, its current price is less then 10 dollar/per. Therefore, VCSEL
extremely fits for fiber LAN. As a result, this task will do some
contribution to the development of the information modernization of our
country, especially in the area of LAN.

2. Signal transmitting system in fiber communication
   2.1. Basic conception:
   Optical fiber communication: it is mainly referred to the
communication system that uses laser as the carrier signal and transmit
message via fiber. An optical fiber communication system usually contains
such device as electrical transmitter, optical transmitter, optical
receiver, electrical receiver, and the fiber circuit. Figure 1 gives a
simple model of the system:


                 Optical                                       Optical
                transmitter                                   receiver

          Electrical transmitter                            Electrical receiver
                             Figure1 Fiber communication system
          Fig10.Test      system
          for the drive circuit
    2.2. Source and fiber are acting as leading poles in optical fiber
communication. A brief introduction to them is given below:
          Fig9.The impedance
    2.2.1. Source: Typical sources are LED and LD. It must be considered
          matching circuit
synthetically to decide which one to choose. The performance of the two
          Electric transmitter
are listed in table 2

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
    From table3, we can see that the most important virtues of LD are its
large output of light power and high modulation rate. Since this task
demands that the transceiver can work at the speed of 1.25 Gbp/s, so it
is necessary to choose LD as the light source. Therefore, a power control
circuit to improve the reliability of LD is indispensable.

Table3.Performance of LD and LED
               Laser Diode                            LED
Output optical Large power output :                   The output power only reaches
power          about 1 mW-100 mW                      1-2 mW
Modulation     Wide bandwidth and high                Narrow bandwidth, modulation
rate           modulation                             rate : 1 MHz-200 MHz
               rate:100 MHz-10 GHz
Optical beam   Good directivity, little               Poor directivity, large
character      divergence                             divergence
Coupling       High efficiency in                     The efficiency of coupling
efficiency     coupling with fiber:                   with optic-fiber is below 10%
               above 80%
Spectrum       Narrow spectrum, pure                  Wide spectrum and not so pure
characteristic light color                            light color
Temperature    Needing special                        Working normally in wide
characteristic temperature controlling                temperature range
               circuit to get a constant
               light output
Linearity      Good linearity of P-I                  Easy to saturate in large
               curve                                  current situation
Reliability    Not so good in the aspect              Good reliability and long life
and life       of reliability and life

    2.2.2. Fiber: a kind of dielectric light guide, with a waveguide
structure to constrain light to travel inside it. It is usually a thin
glass wire with a diameter of only 0.1mm.
    According to the number of the transmission modes, fiber can be divided
into two kinds: single mode fiber and multimode fiber. Another
classification of fiber is depended on the working wavelength of fiber
and can be labeled as short wavelength fiber(0.8-0.9µm),long wavelength
fiber(1.0-1.7µm) and ultra long wavelength fiber( >2µm).With the
difference of refractive index distribution on the cross section, fiber
can also be assorted as step change fiber and gradual change fiber.

    2.3. The main virtues of optical fiber communication
1) Wide transmission band and large communication capacity

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
2)   Low loss in transmission, long distance between relays
3)   Strong ability to resist electromagnetic interference
4)   Small diameter and light weight
5)   Good security without crosstalk interference

3. Principle of the light transmitter and light receiver
    3.1. Module: referred to a relatively compact structure containing
many integrated cells, it helps devices achieve better functions and
performance. It also offers users more convenience to design and use the
system. So the modularization of photoelectric device is a necessary trend
nowadays. Basic optoelectronic modules contain light transmitter module,
light receiver module, transceiver module, single fiber-two way module,
etc. A brief introduction is given to the light transmitter and the light
1) Light transmitter: the main function of this module is to convert
   electrical signal into optical signal and send it out. The module
   usually includes an LD, an LD driver, and control circuits assuring
   the normal and stable operation of LD.
2) Light receiver: this module functions by converting the optical signal
   back into electrical signal. After being amplified and equalized, the
   signal is sent to a timing decision circuit, then the original signal
   is resumed. This module usually contains a preamplifier, a limiting
   amplifier, an equalizer amplifier, a clock extracting and regenerating
   circuit and an AGC (auto-gain-circuit) circuit, etc.

     3.2. Index of Optical interface technique
      3.2.1. Index of the light transmission:
1) Average output power Pt: Since transmitter is to convert electrical
   pulse sequence into optical pulse sequence and put it into the fiber,
   and as the pulse sequence is random, it is obvious that the optical
   power meter will receive the average power of the random sequence,
   namely average output power, rather than a constant power.
2) Extinction ratio (EXT): it refers to the ratio between the average power
   that LD gives out under the condition of all “0” sequence (P0) and
   that under the condition of all “1” sequence (P1).That is:
                                    EXT         100%

(There is a coefficient “2” in the denominator because the probability
  of “0” and “1” in a random sequence is equal) .EXT should be very
  small, usually regulated as less than 5%
     3.2.2. Index of the light receiver:

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
1) Receiving sensitivity S: it is defined as the lowest optical power the
   receiver needed to ensure a certain error rate.
2) The dynamic range of receiver D: it refers to the dB value of the ratio
   between the maximum and minimum power a receiver can detect at the
   precondition that a certain error rate is ensured.

4. The performance of VCSEL
    4.1. Main parameters for LD
1) P-I curve: namely the relation curve between the output power (P) and
   input current (I).As the input current increases, LD begins to increase
   its spontaneous emission, until it begins stimulated emission. The most
   important parameter is the accurate input current that makes LD begin
   stimulated emission, which is called threshold current, presented by
   Ith. As for an LD, we always expect it to have a low threshold current.
2) Output light power (P): it is defined as LD’s output power when the
   positive current above threshold current reaches the required
   modulation current (Imod). Its unit is mW or dBm
3) Linearity of P-I curve: it is a parameter used to measure the deviation
   between the theoretical output power and the actual output power.
4) Drive current of LD: it means the total current (including threshold
   current and modulation current) the LD needed to put out certain power.
5) The slope of P-I curve: Besides a low threshold current, users also
   want to get a big output power from a small current. So an LD with good
   current to light inversion speed has a good performance. This
   performance is reflected in the slope of P-I curve above the threshold
   current, presented by △P/△I
6) Monitoring photocurrent: it is defined as the photocurrent of the
   photodiode at a certain inverse voltage and a certain LD’s output light
   power. In a digital fiber communication system, it is required that
   the LD can put out stable power at any surrounding temperature and after
   any time. So the digital pulse transmission circuit must have the
   function of auto power control, which means that the transmitter must
   equip an APC circuit.

   4.2. Virtues of VCSEL
   Vertical Cavity Surface Emitting Laser (see fig 2) and its array           is
a new kind of laser diode. It is a breakthrough in the integration            of
optical device and has excellent performance that can not replaced            by
common LD. Among various laser diodes, VCSEL has a unique advantage           in
the application of fiber LAN, its main virtues are:

1) High radiation efficiency: This can help to simplify the design of

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
   receiver circuit. Because there is more power sent out from VCSEL, the
   demand on the sensitivity of the receiver circuit is not so high.
2) Low threshold and working current: Threshold current can be as low as
   between 1mA to 1µA, and the working current is also very low, usually
   5mA to 15 mA. So it can be driven directly by PECL or ECL logic circuit,
   which helps to simplify the design of drive circuit.
3) Stability under different temperature: VCSEL needs no feedback control
   from photodiode under the working speed of 200 Mb/s, and it seldom needs
   any ATC circuit.
4) High working speed: utmost working speed can be as high as 3 Gb/s
5) Long life, low price (1 dollar/per), high yield. So VCSEL is suitable
   to be used widely for cut in network and domestic consumer
6) VCSEL can work not only in single mode situation, but also in multi-mode
   situation. Due to its small divergence angle, VCSEL has a high
   efficiency in coupling with fiber (The coupling efficiency between
   VCSEL and multi-mode fiber can be as high as 90%).So it can be widely
   used in LAN where multi-mode fiber (62.5mm core diameter) is the main
   transmission media.
7) VCSEL focuses on 850nm short wavelength technique, so it is apt to meet
    the working speed of nowadays and future network and it can accord with
    the fiber widely laid in constructions nowadays.

                                    Fig2.A model of VCSEL
5. The design of the optical transceiver
   In order to acquire a low cost for the transceiver, it should be assured
that every circuit is simple enough and costs low to produce besides a
proper LD. Figure 3 is the functional frame of the module:

       Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

            Data                                                                 Fiber
                             Driver                LD             Light
                                                                                 Light out

                           APC circuit              PD

                         Figure3. A functional frame of the light transmitter

          5.1. This experiment uses the HFE438x-521 VCSEL from Honeywell Co. It
       works at the wavelength 850nm and is suitable for 50/125µm,62.5/125µm
       multi-mode fibers. It is mainly used in LAN
          The main technical parameters of HFE438x-521 are listed in table4:

       Table4.     Main technical parameters of HFE438x-521
Symbol             Technical           Testing           Minimum        Typica   Maximum     Unit
                   parameter           condition         value          l        value
Poc                Coupling            IF = 12 mA                       350                  µW
                   light power
                   of fiber at
                   peak current
I th               Threshold                                            3.5      6           mA
△ I th             Variation of        Ta =              -1.5                    1.5         mA
                   Ith with            0 to 70 0C
VF                 Forward             IF = 12 mA 1.6                   1.8      2.2         V
                   voltage of LD
tR                 Rising time                                          150      300         ps
tF                 Falling time                                         200      300
y                  Slope of P-I        Poc =             0.02           0.04     0.1         mW/mA
                   curve               0.35mW
I PD               Photodiode          Poc =             0.09                    0.58        mA
                   current             0.35mW
△ I PD /△T         Variation of        Poc =                            0.0                  %/ 0C
                   I PD with           0.35mW

          From table4, we can see that this VCSEL has a stable working
       performance and the variation of parameter with temperature is very small.

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
Therefore, the optical transmitter does not need an ATC circuit. In
addition, its photodiode is packed with the LD and can be used in APC
feedback control circuit. We test of the threshold current of LD and the
relation between the current on the LD and the PIN diode, and the data
is shown in figure4:

         Figure4. The relation between the current on the LD and the PIN

    From figure4, we can see that the threshold current of LD is 3.4 mA,
the relation between the current change of the LD and the current change
of the PIN is:
                            I LD
                                    19.4
                            I PIN

    5.2. Drive circuit design:
   This circuit adopts direct modulation method. Direct modulation is
simple, economical and easy to realize. It is widely used in fiber
communication, and suitable for laser diode (LD). Because the output
optical power of LD (namely the linear part beyond the threshold current)
generally proportionate the input current, and the conversion of current
change into light intensity modulation is also linear, so it is possible
to realize light intensity modulation by changing input current.
   Light modulation can be sorted as analog signal modulation and digital
signal modulation. Analog signal modulation uses continuous analog signal
(e.g. sound or TV signal) to modulate light source, while digital
modulation mainly refers to PCM code modulation, that is, by sampling,
quantizing and coding continuous analog signal, we can use“1” and “0”
pulses to represent signal. After many years’ experiment, it has already
been verified that digital communication has much more advantage than
analog communication. Digitalization is what communication will develop
into. This experiment uses digital signal modulation.
   The choice of bias current has a direct impact on the
high-speed-modulation character of LD. The choice of bias current must

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
follow some rules: (1) It should be near the threshold current, which can
result in a great reduction in electric-optic delay time, and a relaxation
oscillation suppression; (2) The direct bias current should not be too
much, or else it will worsen the LD’s character of extinction rate (Since
extinction rate directly affects the sensitivity of light receiver, it
should not be less than 10dB).
   The choice of the amplitude of modulation current should accord to the
P-I curve of LD. It must both consider ample amplitude of the output light
pulse and LD’s burden; it should also avoid the region where the LD’s
self-pulsation may occur.
   In the high speed modulation process of LD, the modulation circuit
should have a fast enough switch speed, maintaining good current pulse
waveform at the same time. The rising and falling edges of the current
pulse can affect the responding speed of light pulse, and the overshoot
in the rising edge can worsen the relaxation oscillation of light pulse.
To reach these two requirements, the design and manufacture of modulation
circuit are both important.
   Some practical high speed modulation circuits that can be chosen are
bias network and bias triplet; bias triplet is a coaxial device that can
add AC signal or RF signal into DC bias driving current. Although it is
very convenient to use, its price is very high, thus can not be widely
used. What we need to design is a drive circuit that can match VCSEL; it
must have a low price, a high modulation speed and a simple structure.
Combined with the character of VCSEL, the circuit should provide a
modulation current between 5mA and 15mA, and the current amplitude can
be adjusted to satisfy the demand of different LD and to counteract the
disadvantage brought about by the individual difference among the same
type of LD. The working speed of VCSEL can reach as high as 10Gb/s, but
considering the limit of external devices and fiber, it is uneasy for the
LD to work to the speed of 3Gb/s.What’s more, the problems of radiation
shield and the transfer velocity of electrons in transistor in high-speed
circuit hold the continuous high price of high speed, stable LD driver,
which seriously affects the popularization of fiber LAN. So it becomes
meaningful to develop a high speed but low cost LD driver.
   Current 2.488 Gb/s LD driver on the market can be assorted as GaAs FET
device or silicon bipolar transistor. Compared with Silicon bipolar
transistor device, GaAs FET device is expensive to produce, has poor
antistatic performance and requires rigorous bonding technology, which
cause it not suitable for the low costing fiber LAN. So we chose silicon
bipolar transistor to develop a high speed (more than 2.488 Gb /s) LD
driver. There are two ways to improve the working speed of this device,
one is to improve the structure, material and fabricating technique of
the device; another method is to change the topological property of
circuit and parameter optimization. A traditional LD driver usually

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
consists of several differential amplifier stages, has a complex
topological structure, uses many elements, has a poor low-frequency
performance and can not put out all “0” or all “1” signals. In the
experiment, we adopted the latter method and developed a new kind of driver.
It employs four compound polycrystalline silicon transistors and a common
silicon bipolar transistor as the active device, and it has a simple
circuit structure, low price and an excellent performance.
   The drive circuit can be seen in figure 5.What makes if different from
traditional LD drivers is that this driver has a flip-flop circuit as the
input buffer stage, and the output stage that drives LD is a differential
amplifier. In addition, the RC accelerating and equalizing network
improves the speed silicon bipolar transistor drops out saturation state,
enabling the driver to work at a very high switching speed.

                                                          R8               C2

                                                                                 C3         R9     C4              D-
                  0               L
                              1                  2                         R13                   R15
                                                     R6        R7


                                       R1                            R3



                       Figure5. Schematic diagram of the drive circuit

   5.2.1. Input buffer stage: it is a flip-flop circuit consisting of two
RF silicon bipolar transistors T1,T2 and eight resistances from R1 to R8,
its virtues are: 1) it can accurately control the transistors’ saturation
and cut-off level, hence effectively improve the switching speed of the
transistor; 2) it ensures that the drive circuit can put out stably all
‘0’ or all ‘1’ drive current and voltage, and avoid the problem of
output voltage shift, hence can get a wide working speed range and fine
eye pattern; 3) as the drive circuit is triggered by the rising or falling
edge of input signal, it can reshape the input signal; 4) the drive circuit
can adopt both single-end triggering and double-ended triggering.

   5.2.2. RC equalization and acceleration network: it consists of 4
capacitances from C1 to C4 and five resistances from R12 to R16.Since T1,
T2, T3, T4 work at the state of continuously converting between saturation
and cut-off, and according to the dynamic switching characteristic of
transistors, the time transistors needed to quit saturation state is much
longer than that it needed to quit cut-off state because the collector

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
deposits charges in saturation state. The key to improving the switching
speed of the flip-flop circuit is to take away the charges stored in the
collector quickly. The four accelerating cells R12 and C1, R13 and C2,
R14 and C3, R15 and C4 can take away the charges in the collector at proper
time, consequently improving the switching speed of the flip-flop
circuit.R12 series with C1 to form the accelerating cell for T1, R13 and
C2 for T2, R14 and C3 for T3, R15 and C4 for T4.R16 and L ensure the balance
of the switching speed of T1, T2.

   5.3. Auto Power Control (APC) circuit
   5.3.1. Necessity of APC
1) The output characteristic of LD is largely affected by the circumstance
  temperature and aging effect, which is mainly embodied in the variation
  of threshold current and external differential quantum efficiency.
  The threshold current ( I th ) change is exponential to the change of
  circumstance temperature, that is I th  I th 0eT / T 0 (Ith0 is the threshold
  current at the temperature of T0, T is the working temperature), it is
  obvious that I th will increase as the working temperature increases.
  So if the drive current (DC I th plus modulation current) is small, it
  is possible that temperature change will make drive current smaller
  than I th , resulting in no laser put out and normal communication
2) The relationship between external differential quantum efficiency of
   LD (Yd) and temperature is: Yd  Y0 e T / T 0 (Y0 is the external differential
   quantum efficiency at the temperature T0), which means that LD’s
   external differential quantum efficiency drops as temperature
   increases. This can be exhibited on the change of LD’s P-I curve. The
   higher the temperature is, the smaller the slope is, and the lower the
   current to light conversion efficiency of LD is.
   In addition, from the test curve on the life of LD, it is clear that
even though the temperature is constant, if LD works to a certain time
and enters the loss failure period, the increase of the threshold current
and the decrease of the slope of P-I curve can also make LD unable to work
   So in digital fiber communication, the APC circuit in optical
transmitter is as important as the drive circuit.

   5.3.2. The design method of APC circuit:
   In order to control the output light power accurately, two aspects must
be considered: 1) control the bias current of LD and make it automatically
track the change of the threshold current of LD, so that LD will always

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
bias at the best condition; 2) control the amplitude of the current pulse
modulating LD and make it automatically track the change of the external
differential quantum efficiency of LD, hence keeping the amplitude of
output light pulse constant. Since the external differential quantum
efficiency of LD is not very sensitive to the change of temperature,
considering that our aim is to bring down the cost and simplify the control
circuit, we can only consider the change of threshold current.
   A simple and efficient method to stabilize the output power of LD is
to adjust the DC bias current on the LD, that is, if the threshold current
of LD changes, the DC bias current should accordingly change to ensure
that the output power of LD will keep stable under the same modulation
   There are numerous APC methods, but they can all be assorted as
open-loop circuit and closed-loop circuit. Open-loop temperature
compensation APC uses the temperature characteristic of the voltage in
the emitter — base of silicon transistor to compensate that of the
threshold current of LD. And this experiment adopted closed-loop
temperature compensation APC by using optical feedback to stabilize
average output light power: the PIN diode packed with the LD detects some
light power that linearly reflects the change of output light power and
convert it into electrical signal, by this the negative feedback loop
automatically adjust the bias current ( I b ) on the LD.

   See the APC circuit schematic diagram in figure6:
                                                    0                    0


                                                        PD                   LD


                                                                                  0                    0



                           +                                            Q2
                                               R2           R3                        D34
                                           1                                                R7
                                     OUT                                                                   Vref


                     R8                                                      R5       R6


                                                        0                    R9


                                Figure6. Schematic diagram of APC circuit

1) The obtain of feedback signal:
   The feedback signal should be obtained from the photodiode in

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
HFE438x-521.The photodiode works normally only when an inverse voltage
is added on it. This inverse voltage is provided by the LD’s forward
voltage (In this device, the anode of photodiode and the cathode of LD
is connected together to form pin2). The feedback current from the
photodiode converts into feedback voltage through R1, and is sent to the
positive input terminal of the operational amplifier. Since the forward
voltage on the LD is above 1.6V, the choice of R1 must observe two points
to ensure the back-biased state of photodiode: 1) When the LD has the
largest current, the photodiode should still remain back-biased, so R1
can not be too large; 2) The choice of R1 should ensure a sensitivity as
high as possible for APC, that is, when the current on the LD changes,
there must be a big enough voltage change on R1, so R1 can not be too small.
Since the largest current admitted on VCSEL is 15mA, and the ratio between
                                                             I LD
the current on the photodiode and that on the LD is:                19.4 , when
                                                            I PIN
ILD exceeds the threshold current. The value of R1 should be 2.7 kΩ,then
when the current on the LD is 15 mA (experiments have shown that the forward
voltage of LD at this time is 2V,and the threshold current is 3.4 mA),
the          reverse           voltage     on      the      photodiode        is
    15  3.4
2            2.7  0.49(V ) ,furthermore, when R1 = 2.7k we can get a large
feedback voltage.

2) The setting of bias current:
   APC circuit also includes the bias current generating pares in order
to make it easier to adjust bias current. The bias current is produced
by the base and emitter of transistor Q2 and has nothing to do with the
collector. Thus, the bias current depends solely on the output voltage
of the operational amplifier, while the change of the forward voltage of
LD located at the collector of Q2 does not affect the bias current. We
demand that when there is no modulation current, the bias current should
be 4.5mA (above threshold current and modulation current added on it can
bring a great change of the output light power) and the average current
on the LD at normal working condition is about 8mA, as the operational
amplifier should work at the linear amplification state when the LD works
normally(or else it can not adjust the bias current effectively), the
operational amplifier must work at the comparator state and put out
positive voltage when there is no modulation current on the LD. Thus we
can get the value of the protective resistance R5. R9 in series with R5
is a trimmer potentiometer aiming to modulate the DC bias circuit on LD.
To ensure the APC circuit of a wide adjusting range, the supply voltage
of the operational amplifier is +5V and -5V (If the supply voltage is 0
and -5V, the input voltage can not reach 0V,but when there is no current
on the LD ,the voltage at the positive input terminal of operational

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
amplifier is 0,then the operational amplifier will not work properly),and
we designed a clamp circuit to make the output voltage above 0V.

3) How to adjust bias current:
   APC circuit assures the stability of average output power of the LD
by adjusting the bias current on the LD. Considering that the practical
modulation signal is at a rate of gigabytes per second, we expect the APC
circuit only response to the much slower change of the average output power
rather than the fast one caused by the modulation current. So we use a
capacitance C1 between the output terminal and negative input terminal
of the operational amplifier to construct an integrator.

4) The setting of average output power:
   Vref at the negative input pole of the operational amplifier is used
to set the average output light power. When the operational amplifier work
at linear amplification state, the voltages at the positive and the
negative input terminals are equal. Therefore if the voltage at positive
input terminal is not equal to Vref, there is a current between the
negative input voltage and Vref, which will change the output voltage
because C1 will store or release charges, and then the bias current on
the LD and the output light power will also change, since the positive
input voltage is a feedback voltage, it will change with output light power,
until the positive input voltage is equal to Vref. In this way, Vref can
decide the average output light power. Vref is given out by a circuit
composed of a voltage regulator, R7, R6 and R10, and can be adjusted by
the potentiometer R10.The adjusting process of APC is as below:
   When the average output light power increases, the photocurrent
through the photodiode also increases, reducing the positive input
voltage of the operational amplifier. Since the negative input voltage
follows the change of positive input voltage, C1 will release charges and
the output voltage of the operational amplifier will decrease. So the bias
current on the LD also decreases, which will decrease the average output
light power. In this way APC circuit keeps the average output light power
stable. Similarly, when the output power decreases, the APC circuit can
return it to normal by increasing the DC bias circuit of LD.

6. Design of the Optical receiver
   Figure7 shows the frame of this module:

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

               APD                                                                  Coaxial
 fiber                                                                              cable
                                  Pre-                 Main             Impedance
                                  amplifier           amplifier         matching

                               ntrol (AGC)

                      Figure7. Frame of the optical receiver

   We choose HFD3381-002 from Honeywell Co. as the light receiver device;
it integrates a Ga-As photoelectric converter and a BiCOMS preamplifier,
and it is designed specially for the 1.25 Gb/s Ethernet. It works at the
wavelength of 850nm and can match with 50/100 or 62.5/125 mm multi-mode
   MAX3269 is chosen as the main amplifier; it is designed specially for
the light receiver system with a working rate at Gb/s. It can provide a
high gain and amplify the mV dimension signal from preamplifier to the
voltage lever which the subsequent judge circuit can recognize.
   MAX3269 requires that the network at the output terminal should be
equivalent to circuit at figure8.This is because high rate circuit must
consider the problem of reflection. If the resistances at the source and
the load do not match, some voltage will be reflected from the load back
to the source. If the resistances at the source and the load are equal,
then the reflection will not happen. Since this design requires that the
resistance at the load is 50Ω, and the node voltage should meet certain
requirement, we get the impedance matching circuit as figure9:



                           50                      50 

                  Figure8. Equivalent circuit at the output terminal of MAX3269

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)


                           82                    82 


                          125                    125 

                        Figure9. Impedance matching circuit of MAX3269

   7.The design and manufacture of the transceiver
    Transceiver is a new photoelectric device that packs together the
transmission and receiver elements that are traditionally separated into
a single shell. It has many advantages:
   1) Miniaturization: The module uses highly integrated IC to accomplish
the functions like APC, driver in traditional transmitter and
preamplifier, limiting amplifier in traditional receiver, and it has a
size only equal to traditional transmitter or receiver.
   2) Low cost: Since the light transmission and receiving system are
integrated together, transceiver uses less material and can be packaged
with plastic shell. So it is very suitable for fiber LAN
   3) Excellent performance: The transmission and receiving system in the
transceiver are completely independent, and the power supply and ground
are also isolated, minimizing the crosstalk between the two systems.
   Our transceiver consists of light device that can be inserted or drew
out, electrical functional circuit and optical interface. Since the
working speed is 1.25 Gb/s, some key technology in high-speed circuit
design must be applied:
   1) The area cooperation problem should be solved at the structure
                 Figure9.Impedance matching circuit
design as a whole; the usage of elements with 0603 package can not only
meet the demand of high working speed, but also reduce the area of PCB
   2) To improve the sensitivity and stability of the receiving part, the
integrated preamplifier technique is used to minimize the influence of
parasitical parameters and enforce the circuit’s ability to resist outer

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)
    The design of PCB board should follow some rules:
    1) Loop number and area should be minimized to the furthest in order
to decrease the oscillation, crosstalk and radiation caused by the
inductance at the current loop at high working speed;
    2) All the connections must be the shortest mode;
    3) All the wire width is set as 30mil, in this way, the resistance on
the line can be reduced and the circuit can work more effectively at high
    4) Power supply should be decoupled before connecting to the circuit,
reducing the effect of transient process, and the decoupling capacitance
should be as near to the power supply as possible so that the current loop
is smallest. And a small capacitance should be paralleled connected with
the big capacitance to minimize both low frequent and high frequent system
noise, widening the filtering range.
    5) Several measures are taken to minimize crosstalk:
    i. Any of two lines should keep a certain distance from each other and
minimize parallel lines.
    ii. Use GND to isolate signal lines to minimize the capacitive coupling
energy between high-speed signal lines
    The careful layout design assures that the whole system has an
excellent electromagnetic compatibility performance. And we put a light
isolation between the LD and fiber so as to minimize the light transmission
    Figure 10 is the photo of the transceiver. It has an area of 6.5 cm4

                 Figure10. Photo of the 1.25 Gb/s transceiver

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

8. The test and result of the experiment
    8.1. Test of APC circuit
    The curve of power vs. temperature is shown in figure 11:

                         Figure11. The curve of power vs. temperature
   From figure11, we can see that the light power only decrease 0.41 dBm
as the temperature increases from 30℃ to 49℃ this demonstrates that the
APC circuit has a fine performance.

   8.2. Test of eye pattern:
  Observing the baseband signal waveform at the receiver end is a
qualitative and convenient way to evaluate a transmission system. Put the
signal into the vertical amplifier of the oscillograph, make the toothed
wave that generates horizontal scan and the code synchronous, then an
eye-like diagram will be observed on the oscillograph screen. This diagram
is called eye pattern. The degree of the opening of the eye can reflect
the condition of crosstalk and channel noise. The bigger the opening is,
the smaller of the system crosstalk and channel noise are.
  8.2.1. The test of the electrical characteristic of the driver:
  The test system we use to test the electrical characteristic of the
driver is as figure 12:
                                   Imod    Ibias

                                                      Vout+         HP54120B
        Advantest                 LD/EA                             20 GHz digital
        D3186                     driver                            oscillograph
                             Figure12. Test system of the drive circuit

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

    When the code is 223  1 NRZ pseudonoise and the bit rate is 1.25 Gb/s,
the eye patterns of the silicon bipolar transistor LD driver under
different voltage supplies are given at figure13 to figure16:

                Voltage supply = -5.2V NRZ pseudonoise code 223  1

            Figure13. Eye pattern of LD driver at -5.2V voltage supply

               Voltage supply = -4.45V NRZ pseudonoise code 2  1

           Figure14. Eye pattern of LD driver at -4.45V voltage supply

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

                Voltage supply = -4.3V NRZ pseudonoise code 223  1

            Figure15. Eye pattern of LD driver at -4.3V voltage supply

               Voltage supply = -3.85V NRZ pseudonoise code 2 -1
           Figure16. Eye pattern of LD driver at -3.85V voltage supply

    The experiment result shows that at the bit rate of 1.25 Gb/s, the
driver has good eye patterns under different voltage supplies. The
modulation voltage on the 50 Ω load is 2.275V, 2V, 1.5V, 1.5V in turn
from figure 13 to figure 16, which means that the driver has fine driving

    8.2.2. Test of the optical transceiver module:
    Using a 223-1 NRZ pseudonoise code signal at a rate of 1.24 Gb/s,
we did the test of the optical transceiver module. The voltage supply of
the driver is -4.5V and the voltage supply of the light receiver is 3.91V.
The testing system is as figure 17:

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

               Coaxial cable                Fiber               Coaxial cable

 Advantest              Optical                     Optical                HP54120B
 D3186                  transmitter                 receiver               20GHz
 Signal                                                                    digital
 generator                                                                 oscillograp

                        Figure17. Testing system of optical transceiver

    The eye pattern of the optical transmitter is shown in figure18:

              Figure18. The eye pattern of the optical transmitter

    All the experiment results above are tested on an experimental board,
and we haven’t test the characteristic of the PCB board of the transceiver,
but we can surly predict that much better result will be got with that

9. Conclusion
   This paper gives a brief introduction to the optical fiber
communication system, and explains in detail the design principle and
structure of the transceiver. We have tested the qualities of the circuits
and got fine eye diagram at the speed of 1.25 Gb/s. Better result is
expected with the PCB board.

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

    I am grateful to my advisor, Prof. Wang, an erudite and vigorous man.
He is not only enthusiastic to me but also strict with me. In the past
one year, he gives me unreserved help and wise guidance from almost all
aspects, including scientific research manner, investigation method,
study orientation, etc. He always answers patiently various questions
    I encountered during my research, and he also shares his precious
experience with me. Thanks to him, I have such a chance to shoot a glance
at the research field of the optical fiber communication.
    I am grateful to Hui-Chun Chin and Tsung-Dao Lee Chinese Undergraduate
Research Endowment (CURE), which provides me such a valuable opportunity
to do some scientific research during my undergraduate years, helps me
take my first step on the scientific road and encourages me to stride
    I appreciate my instructor Mao-xinyu. He provides me with much
facility and gives me lots of good suggestions on study. I hope he would
not mind my bothering him during the past few months.
    I must also thank to my parents. They give me selflessly love and
concern, and they never complain that I didn’t go home during the past
two summer vocations, but encourage me to concentrate on my study and
research work.
    I always hold the view that I couldn’t have finished my task without
the help of all others. Thanks all!


[1] Djafar K.Mynbaev, Lowell L.Scheiner , “Fiber-Optic Communications
Technology”,科学技术出版社, 2002
[2] M.Banu, B.Jalali, et al., Electron.Lett.27,p278,1991.
[3] 黄章勇,   《光纤通信用光电子器件和组件》               ,北京邮电大学出版社,2000
[4] 顾畹仪,李国瑞,       《光纤通信系统》       ,北京邮电大学出版社,1999
[5] 李学林等,    《光纤数字通信系统及测试》            ,中国铁道出版社,1998
[6] 高炜烈,张金菊,       《光纤通信》     ,人民邮电出版社,1993

  张恒,女,1981 年 12 月出生于广东省广州市,2000 年从广东实验中学考入
京大学新生奖学金,  “明德”奖学金,学习优秀单项奖,    “新生杯”乒乓球赛第三

Series of Selected Papers from Chun-Tsung Scholars, Peking University(2003)

  我很荣幸能参加第五届“ 政基金”活动,在本科期间就可以进行科研工作,
  “ 政基金”的研究工作是很有挑战性的,我学会了更有效地利用时间,也
体验到了学习与研究之间思维方式的差异; 政基金”的研究经历是令人难忘

    王子宇,男,1954 年 12 月出生于北京,教授,博士生导师。1982 年 2 月获
成都电讯工程学院工学学士学位;1985 年 2 月获成都电讯工程学院工学硕士学
位。中国电子学会高级会员,中国通信学会青年工作委员会委员,       《电子学报》
第 7 届编委(2001~2005),教育部科学技术委员会信息学部委员,北京大学信息
与工程学部 副主任,     区域光纤通信网与新型光通信系统国家重点实验室副主任。
    1988 年获机电部科技进步二等奖;1989 年获国家科技进步三等奖;1991 年
获国防科工委光华科技二等奖;1991 年获四川省首届青年科技奖;1991 年被评
为四川省有突出贡献的硕士学位获得者;1992 年获机电工业部优秀科技青年奖;
1992 年 10 月起享受政府特殊津贴;1995 年获北京大学安泰奖;1998 年获教育
部科技进步二等奖;1999 年获北京大学华为奖;1999 年获国家科技进步三等奖;
2002 年获北京大学 JDS 奖。


To top