FIBER OPTIC CIRCUITS
Application Note 649: Jun 28, 2000
HFTA-04.0: Optical/Electrical Conversion in SDH/SONET Fiber Optic Systems
This article explains the basic functions and design challenges of the optical to
electrical and electrical to optical signal conversion in SDH/SONET fiber optic
receivers and transmitters. A complete chip set solution to develop the electronic part
of OC 12/STM 4 receivers and transmitters is presented.
The advent of cheaper and more powerful personal computers has not only expanded the user
base; it is also creating a demand for greater transmission capacity among the telecom
networks by adding an increasing volume of internet and videophone connections to the
traditional phone and fax services. The following discussion of an OC 12/STM 4
receiver/transmitter chipset supports these developments and includes a description of the
electronic components required for optic/electric (O/E) conversion in SDH/SONET fiber optic
transmission systems.
Competition among network providers enables the multimedia market to grow, and the
introduction of new and improved products and services in the near future should strengthen
the demand for increased transmission capacity. This need for more data throughput can be
satisfied economically with fiber optic (FO) cables because the transmission capacity is
potentially very high (versus that of copper wires). The physical nature of the fiber cable lets
providers expand capacity by increasing the transmission bit rate or by introducing alternative
transmission techniques, without the need for further upgrades or additional cable installations.
These advantages have led many countries to build extensive fiber networks, and further
expansion of these networks can be expected.
To transmit optical data via fiber cables, signals must be converted from electrical to optical at
the transmit end, and then converted back to electrical at the receive end. These necessary
conversions are handled by receiver/transmitter units that contain electronic devices along with
the optical components.
FO transceivers
The widely used Time Division Multiplex (TDM) transmission technique now enables bit rates
up to 10Gbps and is well established in modern transport systems. Today's high-speed fiber
optic transmission systems offer the following standard bit rates:
www.maxim-ic.com/an649
Page 1 of 11
New techniques such as Wavelength Division Multiplexing (WDM) further increase the
transmission capacity by sending numerous time-multiplexed data streams over one fiber, using
a different wavelength for each data stream. Electronic components in a WDM receiver and
transmitter (compared with those in a TDM system) differ according to the behavior of the
optical sources and line amplifiers in the WDM transport system. The following section
describes the performance required for receivers and transmitters in an optical TDM
transmission system.
Optical receivers
Optical receivers detect optical signals from the fiber and convert them to electrical signals,
which must then be amplified before their data waveforms and clock can be recovered. A
serial-to-parallel conversion of the data stream may be necessary, depending on the bit rate and
the system-specific setup of the following CMOS functions. Figure 1 shows how the receiver's
output interface provides regenerated data in a serial or parallel bit stream, along with the
recovered clock.
Figure 1. A typical receiver/transmitter unit for SONET/SDH fiber-transmission
systems.
www.maxim-ic.com/an649
Page 2 of 11
A PIN or APD (avalanche photodiode) photodetector converts the received light to a signal
current. The PIN diode is relatively cheap and operates with the same supply voltage as the
electronic components, but for a given optical power it generates fewer electrons than the APD.
As a result, the APD provides a more sensitive receiver that can be placed farther away from
the transmitter. This advantage is offset by the need for an APD bias circuit, which (depending
on the APD type) must provide a reverse operating voltage in the 30V to 100V range.
Additionally, the APD adds more noise, costs more, and requires cooling.
The photodetector delivers the extracted current to a transimpedance amplifier (TIA), which
first converts the current to a voltage. This single-ended voltage is then amplified by the TIA
and (usually) converted to a differential signal as required by state-of-the-art receivers. The
TIA should provide both high overload tolerance and high input sensitivity (i.e., a large
dynamic range).
To provide the high input sensitivity necessary to receive optical signals weakened by
transmitter aging or long transmission distance (or both), the TIA noise must be reduced to a
minimum. On the other hand, a high overload tolerance is required to avoid bit errors due to
distortion in the presence of strong optical signals. Further, the TIA's maximum achievable gain
depends on the operating frequency. To ensure stable operation and the required bandwidth,
gain can be optimized only within a narrow range. This limitation may cause the output voltage
resulting from low-power optical signals to be insufficient for further processing. To amplify
small TIA voltages in the 1mV to 2mV range, the TIA function must be followed by a
postamplifier, which in most cases is a limiting amplifier (LA).
As the name implies, a limiting amplifier delivers a certain output-voltage swing whose
maximum is independent of the input signal strength. Also included is a loss-of-power indicator
(LOP) that warns when the incoming signal falls below a user-defined threshold. As a systemdependent parameter, this threshold must be adjusted externally. A comparator with hysteresis
ensures chatter-free operation for the LOP flag when the signal is close to the threshold level.
A key component that follows the limiting amplifier in a receiver unit is the clock and data
recovery (CDR) circuit. The CDR performs timing and amplitude-level decisions on the
incoming signal, which leads to a time- and amplitude-regenerated data stream. First to be
recovered from the received signal is the clock. Several possibilities can support this clockrecovery function (external SAW filter, external reference clock, etc.), but only the fully
integrated approach can save both cost and effort.
The challenge for an integrated clock-recovery circuit is to meet the jitter specification
recommended by the International Telecommunication Union-Telecom Standards Sector (ITUT). Jitter refers to the effect in which individual bit transitions ("0" to "1" and vice-versa) are
not exactly in phase. The effect becomes visual in an "eye diagram," in which several pseudorandom bit-pattern sequences are superimposed. An eye diagram illustrates the quality of a data
stream in terms of the eye opening, measured using the "eye mask" (Figure 2).
www.maxim-ic.com/an649
Page 3 of 11
Figure 2. An "eye diagram" illustrates the signal quality of a data stream.
ITU-T recommendations specify limits on the tolerance, transfer, and generation of jitter.
Signal quality at the LA output (as represented by the eye opening) is usually low, mostly as a
consequence of nonideal components in the optical transmission system. Because the CDR
must accept a certain amount of input data jitter to achieve normal error-free operation, all
receiver units in line-termination and regenerator applications must comply with the ITU-T
recommendations for jitter tolerance.
Jitter transfer refers to the portion of jitter allowed to transfer from input to output of the CDR,
and jitter generation is that produced by the CDR itself. The ITU-T specs for these two
parameters must be met for regenerators in a long-haul system, because at each stage the
recovered clock enables transmission to the next regenerator, allowing jitter contributions to
accumulate from regenerator to regenerator. Conversely, for line-termination receivers (which
are in the majority of applications) the jitter transfer and jitter generation need not meet ITU-T
recommendations. In those applications, the regenerated data is synchronized to the system
clock.
Aside from jitter effects, noise and pulse distortion both reduce the phase margin in which
received bits can be clocked for the purpose of sensing their logic level. The use of a phaselocked loop (PLL) is essential in synchronizing the clock with the data stream, to ensure
alignment of the clock with the middle of a data word. To further optimize the bit error rate
(BER) in the presence of asymmetrical rise and fall transitions of the received data signal, the
system should include an option to adjust the phase relation between clock and data.
The CDR often includes a loss-of-lock (LOL) alarm, which monitors whether the PLL is
locked to the received data stream. The CDR's serial stream of regenerated data and the
recovered clock signal are usually fed to a deserializer, whose conversion ratio depends on the
data's bit rate and the interface capability (speed) of the CMOS system components. The
www.maxim-ic.com/an649
Page 4 of 11
deserializer must also provide a CMOS-compatible interface. To support bit alignment of the
serial data stream to the different deserializer outputs, the deserializer should include bitsynchronization capability.
Optical transmitter
The optical transmitter in a fiber optic system converts the electrical bit sequence delivered
from the CMOS system components to an optical data stream. As shown in Figure 1, it contains
a serializer with clock synthesizer (which depends on the system setup and transmission bit
rate), a driver, and an optical source.
Two important wavelength ranges (windows 2 and 3) are in use for transmitting information
over a fiber cable in telecommunication networks. Within an optical window, the signals
benefit from a lower impact on quality (less dispersion) and less attenuation per unit of fiber
length. The range between 1000nm and 1300nm, called the second optical window, is known
for low dispersion-as low as 0dB. The range from 1500nm to 1800nm, known as the third
optical window, offers the lowest attenuation per unit of fiber length (Figure 3).
www.maxim-ic.com/an649
Page 5 of 11
Figure 3. Variations of attenuation and dispersion vs. wavelength for the first, second,
and third optical windows.
Several optical sources are available for today's optical transmission systems. Light-emitting
diodes (LEDs), for example, are often used for low-cost, short-distance local area network
(LAN) connections. Disadvantages, however, preclude use of the LED as a transmitter for
telecommunications systems: its broad spectral bandwidth allows the coexistence of many
optical modes, and it cannot operate at wavelengths of the second and third optical windows.
Unlike the LED, the optical-modulated laser transmitter (the electro-absorption and MachZehnder types, for instance) is an optical source with high spectral purity that can operate in the
www.maxim-ic.com/an649
Page 6 of 11
third optical window. It is preferred, therefore, for ultra-long-distance or WDM transmission
systems in which high performance is mandatory and cost is not a major consideration. For
optical links in the majority of telecommunication trunk lines, various types of directmodulated semiconductor laser diode offer an optimum cost/performance ratio for short,
intermediate, and long-haul transmissions. Devices are available for operation in both the
second and third optical windows.
All semiconductor laser diodes used for direct modulation have in common the need for a DCbias current to set the operating point and a modulation current for signal transmission. The
values for DC-bias and modulation current depend on characteristics of the laser diode, which
can differ from type to type and version to version. The drift of these characteristics with time
and temperature should be evaluated carefully when designing a transmitter unit, especially
with regard to the more cost-effective, uncooled types of semiconductor laser. The laser driver
must therefore offer bias and modulation currents with sufficient range to support the
development of optical transmitters with a wide choice of laser diodes.
To compensate for the drift of laser characteristics over time and temperature, the laser driver
must maintain the initially adjusted DC operating point. The best way to realize this
compensation is to introduce automatic power control (APC). To detect the actual laser power,
a photodiode converts the laser light to a proportional current and feeds it to the laser driver,
where the actual value is compared with a previous fixed value. Any difference causes the DCbias current to increase or decrease as required to reach the initially defined laser power.
Often, the APC includes an alarm function that warns if the laser diode's optical power can no
longer be sustained due to aging. Like the operating point, optical signal strength is affected by
the drift of laser-diode characteristics over time and temperature. To maintain the optical
"amplitude," it is necessary to compensate for a decreasing slope in these characteristics caused
by time and temperature. The problem is solved either with additional external circuitry or with
an integrated Automatic Modulation Control (AMC), which may employ the photodiode
already present in the APC loop.
In addition to these fundamental functions, the system must be capable of stopping laser
transmissions by disabling the driver without interrupting data reception at the input. By adding
a flip-flop or latch (as part of the laser driver or the serializer), jitter performance can be
improved by retiming this data stream before it reaches the laser driver's output stage.
Residing between the laser-diode driver and the lower-speed CMOS system components, the
serializer converts parallel data to a serial stream for the laser driver. Like the receiver unit's
deserializer, the serializer's conversion ratio depends on the transmission bit rate and the speed
of the CMOS system interface. The retiming and serialization function requires a transmission
clock, which must to be synthesized. This clock synthesizer can be integrated with the
serializer, and usually incorporates a PLL. The challenge for the synthesizer is to ensure data
transmission with the lowest possible jitter. As a result, the synthesizer plays a key role in the
transmitter of an optical transmission system.
Complete chipset for STM 4 Rx/Tx units
All components of an optical transmission system for telecommunications must comply with
the relevant ITU-T recommendations. Provided this basic requirement is met, the next most
www.maxim-ic.com/an649
Page 7 of 11
important criteria in designing an O/E unit are power dissipation, supply voltage, integration
level, and margin of performance. The following section describes a complete chipset that
allows designers to optimize the above criteria while developing competitive STM 4
receiver/transmitter units (Figures 4 and 5).
Figure 4. Three packages from Maxim form an STM 4 receiver.
Figure 5. Two packages from Maxim form an STM 4 transmitter.
The chipset is based on Maxim's state-of-the-art, high-performance bipolar technologies: CB-2
and GST-2. CB-2 is a fast, complementary-bipolar process whose transit frequencies are
6.4GHz for pnp transistors and 8.7GHz for npn transistors. GST-2 is a very-high-speed,
submicron bipolar process with a transit frequency of 27GHz for npn transistors.
The combination of modern high-performance manufacturing and extensive IC design
experience has produced a highly integrated, flexible, and powerful STM 4 chipset consisting
of five ICs including the serializer and deserializer. In serial-I/O modules the chipset consists of
only three ICs, and they can be delivered in die form to accommodate "chip-on-board"
mounting technology.
www.maxim-ic.com/an649
Page 8 of 11
Power dissipation is an important consideration because system cooling requirements usually
allow only a limited power budget in the O/E units. Maxim's STM 4 chipset makes extensive
use of 27GHz, high-speed technology in reducing power dissipation. It can further reduce
power dissipation by operating on +3.3V instead of today's more common +5V. Rather than
require an additional source of +5V, the O/E unit can use the +3.3V available for CMOS
system components. Or, to be flexible, it can share an existing +5V supply with the front-end
ICs. In addition to these features, which pertain to the chipset as a whole, features specific to
the individual components are described in the following sections.
Preamplifier (i.e., transimpedance amplifier)
The transimpedance amplifier (MAX3664) converts a single-ended current from the detector
diode to a single-ended voltage, which is amplified and converted to a differential signal.
Typical amplification is 6kW. This gain level can be increased by 6dB if the data outputs
(back-terminated internally with 60W) are not externally terminated as well. For input currents
beyond 100µAp-p, the high gain leads to a limited differential output-voltage swing of
900mVp-p. A DC-cancellation circuit helps to deliver differential output voltages with low
pulse-width distortion over a wide range of input-current levels.
Low input-related noise is achieved by careful circuit design and by limiting the bandwidth to
590MHz at an input capacitance of 1.1pF. Assuming a simple PIN detector diode is used, the
low noise enables a typical input sensitivity of -32dBm optical power. Power dissipation is less
than 85mW at +3.3V. Small size and an optimal bondpad configuration make this component
suitable for use in PIN-TIA modules, which combine a PIN diode and transimpedance
amplifier in one package (a TO package, for instance).
Clock and Data Recovery (CDR)
The main functions of the clock and data recovery IC (MAX3675) are to recover the clock
signal from the received data stream and to regenerate the data's timing and amplitude
characteristics. Because the chip integrates an offset-compensated limiting amplifier as well,
two standard products (MAX3664 and MAX3675) contain all the electronics necessary for an
O/E receiver unit.
The MAX3675 offers a high-sensitivity differential analog input (3mVp-p) and a differential
PECL digital input, providing flexibility that supports a wide range of receiver applications.
The MAX3675's power dissipation depends on the input in use: 215mW with analog inputs, or
155mW with digital inputs. Total power consumption for a complete receiver based on the
MAX3664 and MAX3675 is less than 300mW at +3.3V.
An LOP alarm function and input-power detector are integrated with the limiting amplifier. The
LOP alarm warns if the input signal falls below a user-defined threshold. The reference for this
threshold is an internal bandgap circuit that is independent of the supply voltage. To ensure
chatter-free operation for input signals near the threshold, the LOP's TTL-monitor output
includes hysteresis. The power detector provides a receive signal-strength indicator (RSSI pin)
whose output voltage is proportional to input power and is linear in decibels.
The PLL necessary for clock recovery is fully integrated and does not require an external
reference clock. It consists of a phase/frequency detector, a loop-filter amplifier with external
RC network, and a 622MHz voltage-controlled oscillator. The PLL provides an LOL signal
www.maxim-ic.com/an649
Page 9 of 11
(LOL pin) and a TTL-monitor output that flags when the PLL loses lock. To improve the
system's bit error rate as described in the Optical Receiverssection, users can adjust clock phase
relative to the data signal by accessing the pins PHADJ+ and PHADJ-. Finally, a decision
circuit supported by the recovered clock signal (from the PLL) regenerates timing and
amplitude characteristics for the incoming data stream.
Deserializers (DEMUX)
To support the various CMOS system-interface circuits available today, Maxim offers the
MAX3680 and MAX3681 deserializers. The MAX3680 converts a 622Mbps serial data stream
to a 78Mbps stream of 8-bit words. Data and clock outputs are TTL compatible, and the power
consumption is 165mW at +3.3V. The MAX3681 converts a 622Mbps serial data stream to a
155Mbps stream of 4-bit words. Its differential data and clock outputs support an LVDS
interface for CMOS system components, and its power consumption is 265mW at +3.3V. Both
parts offer serial differential-PECL inputs for data and clock, and a synchronization function
(SYNC pin) that enables a bit realignment of the deserializer's data outputs.
Serializer (MUX)
The MAX3691 serializer converts four LVDS data streams at 155Mbps to a serial stream at
622Mbps. The necessary transmission clock is synthesized using a fully integrated PLL
comprising a voltage-controlled oscillator, a loop-filter amplifier, and a phase/frequency
detector that requires only an external reference clock. All the data- and clock-input buffers are
LVDS-compatible, and the serial data output delivers differential-PECL signals. Power
dissipation is 215mW at +3.3V.
Laser Driver (LD)
The main task of the laser driver (MAX3667) is to deliver the bias (IBIAS) and modulation
current (IMOD) for a direct-modulated laser diode. For flexibility, the differential inputs accept
PECL data streams and also differential voltage swings as small as 320mVp-p, with DC levels
in the range 1V to (VCC - 0.75V). Connecting an external resistor between BIASSET and
ground lets you adjust the bias current between 5mA and 90mA, and a resistor between
MODSET and ground lets you adjust the modulation current between 5mA and 60mA.
An integrated, temperature-stabilized reference voltage ensures stable bias and modulation
currents. To avoid laser damage, a protection circuit disables the MAX3667 when any of the
pins BIASSET, MODSET, or APCSET are short-circuited to ground. To avoid excessive
current that could alter the laser's performance, an internal circuit also limits the sum of output
currents IMOD and IBIAS to approximately 150mA. As described in the Optical Transmitter
section, an integrated APC circuit, supported by an external detector diode, maintains the initial
user-defined average laser power constant over time and temperature.
The detector diode's average current value is established by applying an external resistor
between the APCSET and GND pins. Two monitor outputs (BIASMON and MODMON)
deliver output currents directly proportional to the bias and modulation currents. The bias,
modulation, and APCSET currents can be disabled via the DISABLE pin, but all other
functions including the reference voltage remain active to allow a fast and predictable wake-up.
In addition, an integrated slow-start function provides a 50ns minimum turn-on time that
reduces laser stress. In contrast to other laser drivers available in today's market, the MAX3667
can operate from a single +3.3V supply.
www.maxim-ic.com/an649
Page 10 of 11
As an alternative to the 622Mbps MAX3667, the MAX3766 laser driver can be used for STM 4
transmitter units supporting data rates from 155Mbps to 1.25Gbps. Designed to operate on a
single +5V supply, the MAX3766 incorporates all attributes mentioned for the MAX3667 plus
the larger bandwidth (to 1.25Gbps). Other features include extensive laser-safety provisions
and the option to add a single external resistor that maintains "optical amplitude" by
compensating for the effect of temperature on the slope of the characteristic laser curve. The
resistor's value depends on the laser diode's temperature characteristic.
MORE INFORMATION
MAX3664: QuickView -- Full (PDF) Data Sheet (248k)
MAX3667: QuickView -- Full (PDF) Data Sheet (648k) -- Free Sample
MAX3675: QuickView -- Full (PDF) Data Sheet (568k) -- Free Sample
MAX3681: QuickView -- Full (PDF) Data Sheet (64k)
-- Free Sample
MAX3691: QuickView -- Full (PDF) Data Sheet (90k)
-- Free Sample
www.maxim-ic.com/an649
Page 11 of 11