Optical Time-Division Multiplexing (OTDM)
TDM is commonly performed in the electrical domain to obtain digital hierarchies for telecommunication systems. In this sense, even single-channel lightwave systems carry multiple TDM channels. The electrical TDM becomes difficult to implement at bit rates above 40 Gb/s because of the limitations imposed by high-speed electronics. A solution is offered by the optical TDM (OTDM), a scheme that can increase the bit rate of a single optical carrier to values above 1 Tb/s. The OTDM technique was studied extensively during the 1990s, and further research has continued in recent years in the context of WDM systems with channel bit rates of 100 Gb/s or more. Its deployment requires new types of optical transmitters and receivers based on all-optical multiplexing and demultiplexing techniques. In this section we first discuss these new techniques and then focus on the design and performance issues related to OTDM lightwave systems.
1. Channel Multiplexing
In OTDM lightwave systems, several optical signals at a bit rate B share the same carrier frequency and are multiplexed optically to form a composite bit stream at the bit rate NB, where N is the number of channels. Several multiplexing techniques can be used for this purpose. The figure below shows the design of an OTDM transmitter based on the delay-line technique.
It requires a laser capable of generating a periodic pulse train at the repetition rate equal to the single-channel bit rate B. Moreover, the laser should produce pulses of width Tp such that Tp < TB = (NB)-1 to ensure that each pulse will fit within its allocated time slot TB. The laser output is split equally into N branches, after amplification if necessary. A modulator in each branch blocks the pulses representing 0 bits and creates N independent bit streams at the bit rate B.
Multiplexing of N bit streams is achieved by a delay technique that can be implemented optically in a simple manner. In this scheme, the bit stream in the nth branch is delayed by an amount (n-1)/(NB), where n = 1, ..., N. The output of all branches is then combined to form a composite signal. It should be clear that the multiplexed bit stream produced using such a scheme has a bit slot corresponding to the bit rate NB. Furthermore, N consecutive bits in each interval of during B-1 belong to N different channels, as required by the TDM scheme.
The entire OTDM multiplexer (except for modulators which require LiNbO3 or semiconductor waveguides) can be built using single-mode fibers. Splitting and recombining of signals in N branches can be accomplished with 1 x N fused fiber couplers. The optical delay lines can be implemented using fiber segments of controlled lengths. As an example, a 1-mm length difference introduces a delay of about 5 ps. Note that the delay lines can be relatively large (10 cm or more) because only the length difference has to be matched precisely. For a precision of 0.1 ps, typically required for a 40-Gb/s OTDM signal, the delay lengths should be controlled to within 20 μm. Such precision is hard to realize using optical fibers.
An alternative approach makes use of planar lightwave circuits fabricated using the silica-on-silicon technology. Such devices can be made polarization insensitive while providing a precise control of the delay lengths. However, the entire multiplexer cannot be built in the form of a planar lightwave circuit as modulators cannot be integrated with this technology. A simple approach consists of inserting an InP chip containing an array of electroabsorption modulators in between the silica waveguides that are used for splitting, delaying and combining the multiple channels. The main problem with this approach is the spot-size mismatch as the optical signal passes from Si to InP waveguide (and vice versa). This problem can be solved by integrating spot-size converters with the modulators. Such an integrated OTDM multiplexer was used in a 160-Gb/s experiment in which 16 channels, each operating at 10 Gb/s were multiplexed. In a different approach, a cascaded nonlinear process inside periodically poled LiNbO3 waveguides (resulting in FWM) is employed.
An important difference between the OTDM and WDM techniques should be apparent from the figure above: The OTDM technique requires the use of the RZ format. Historically, the NRZ format used before the advent of lightwave technology was retained even for optical communication systems. Starting in the later 1990s, the RZ format began to appear in dispersion-managed WDM systems. The use of OTDM requires optical sources emitting a train of short optical pulses at a repetition rate as high as 40 GHz. Two types of lasers are commonly used for this purpose. In one approach, gain switching or mode locking of a semiconductor laser provides 10-20 ps pulses at high repetition rate, which can be compressed using a variety of techniques. In another approach, a fiber laser is harmonically mode locked using an intracavity LiNbO3 modulator. Such lasers can provide pulse widths ~ 1 ps at a repetition rate of up to 40 GHz.
2. Channel Demultiplexing
Demultiplexing of individual channels from an OTDM signal requires electro-optic or all-topical techniques. Several schemes have been developed, each having its own merits and drawbacks. The figure below shows three schemes discussed in this section. All demultiplexing techniques require a clock signal - a periodic pulse train at the single-channel bit rate. The clock signal is in the electric form for electro-optic demultiplexing but consists of an optical pulse train for all-optical demultiplexing.
The electro-optic technique sues several MZ-type LiNbO3 modulators in series. Each modulator halves the bit rate by rejecting alternate bits in the incoming signal. Thus, an 8-channel OTDM system requires three modulators, driven by the same electrical clock signal [see figure (a) above], but with different voltages equal to 4V0, 2V0, and V0, where V0 is the voltage required for π phase shift in one arm of the MZ interferometer. Different channels can be selected by changing the phase of the clock signal. The main advantage of this technique is that it uses commercially available components. However, it has several disadvantages, the most important being that it is limited by the speed of modulators. The electro-optic technique also requires a large number of expensive components, some of which need high drive voltage.
Several all-optical techniques make sure of a nonlinear optical loop mirror (NOLM) constructed using a fiber loop whose ends are connected to the two output ports of a 3-dB fiber coupler as shown in figure (b) above. Such a device is also referred to as the Sagnac interferometer. The NOLM is called a mirror because it reflects its input entirely when the counterpropagating waves experience the same phase shift over one round trip. However, if the symmetry is broken by introducing a relative phase shift of π between them, the signal is fully transmitted by the NOLM. The demultiplexing operation of an NOLM is based on the XPM, the same nonlinear phenomenon that can lead to crosstalk in WDM systems.
Demultiplexing of an OTDM signal by an NOLM can be understood as follows. The clock signal consisting of a train of optical pulses at the single-channel bit rate is injected into the loop such that it propagates only in the clockwise direction. The OTDM signal enters the NOLM after being equally split into counterpropagating directions by the 3-dB coupler. The clock signal introduces a phase shift through XPM for pulses belong to a specific channel within the OTDM signal. In the simplest case, optical fiber itself introduces XPM. The power of the optical signal and the loop length are made large enough to introduce a relative phase shift of π. As a result, a single channel is demultiplexed by the NOLM. In this sense, an NOLM is the TDM counterpart of the WDM add-drop multiplexers. All channels can be demultiplexed simultaneously by using several NOLMs in parallel. Fiber nonlinearity is fat enough that such a device can respond at femtosecond time scales. Demultiplexing of a 6.3-Gb/s channel from a 100-Gb/s OTDM signal was demonstrated in 1993. By 1998, the NOLM was used to demultiplex a 640-Gb/s OTDM signal.
The third scheme for demultiplexing in figure (c) above makes use of FWM in a nonlinear medium. The OTDM signal is launched together with the clock signal (at a different wavelength) into a nonlinear medium. The clock signal plays the role of the pump for the FWM process. FWM produces a pulse at the idler wavelength only in time slots in which a clock pulse overlaps with the signal pulses of the channel that needs to be demultiplexed. As a result, the pulse train at the new wavelength is an exact replica of the channel that needs to be demultiplexed. An optical filter is used to separate the demultiplexed channel from the OTDM and clock signals. A polarization-preserving fiber is often used as the nonlinear medium for FWM because of the ultrafast nature of its nonlinearity and its ability to preserve the state of polarization despite environmental fluctuations. As early as 1996, error-free demultiplexing of 10-Gb/s channels from a 500-Gb/s OTDM signal was demonstrated by using clock pulses of about 1 ps duration. This scheme can also amplify the demultiplexed channel (by up to 40 dB) through parametric amplification inside the same fiber.
3. System Performance
The transmission distance of OTDM signals is limited in practice by fiber dispersion because of the use of short optical pulses (~ 1 ps) dictated by a relatively high bit rate. In fact, an OTDM signal carrying N channels at the bit rate B is equivalent to transmitting a single channel at the composite bit rate of NB, and the bit rate-distance product NBL is restricted by the dispersion limits. As an example, it is evident that a 200-Gb/s system is limited to L < 50 km even when the system is designed to operate exactly at the zero-dispersion wavelength of the fiber. Thus, OTDM systems require not only dispersion-shifted fibers but also the use of dispersion-management techniques capable of reducing the impact of both the second- and third-order dispersive effects. Even then, PMD becomes a limiting factor for long fiber lengths and its compensation is often necessary. The intrachannel nonlinear effects also limit the performance of OTDM systems because the use of intense pulses is often necessary for OTDM systems.
In spite of the difficulties inherent in designing OTDM systems operating at bit rates exceeding 100 Gb/s, many laboratory experiments have realized high-speed transmission using the OTDM technique. In a 1996 experiment, a 100-Gb/s OTDM signal, consisting of 16 channels at 6.3 Gb/s, was transmitted over 560 km by using optical amplifiers (80-km spacing) together with dispersion management. The laser source in this experiment was a mode-locked fiber laser producing 3.5-ps pulses at a repetition rate of 6.3 GHz (the bit rate of each multiplexed channel). A multiplexing scheme was used to generate the 100-Gb/s OTDM signal. The total bit rate was later extended to 400 Gb/s (forty 10-Gb/s channels) by using a supercontinuum pulse source producing 1-ps pulses. Such short pulses are needed since the bit slot is only 2.5-ps wide at 400 Gb/s. It was necessary to compensate for the dispersion slope (third-order dispersion β3) as 1-ps pulses were severely distorted and exhibited oscillatory tails extending to beyond 5 ps (typical characteristic of the third-order dispersion) in the absence of such compensation. Even then, the transmission distance was limited to 40 km.
OTDM transmission at a bit rate of 160 Gb/s drew considerable attention after 2000 because it was considered a natural update for 40-Gb/s systems. In a 2001 field trial, a 160-Gb/s OTDM signal was transmitted over 116 km. By 2006, transmission over 4320 km has been demonstrated using a recirculating loop. This experiment employed the DPSK format and also demonstrated the long-term stability of OTDM systems with properly designed components. In another set of experiments the objective was to realize a single-channel bit rate of 1 Tb/s or more. In a 2000 experiment, a 1.28-Tb/s OTDM signal could be transmitted over 70 km but it required compensation of second-, third-, and fourth-order dispersions simultaneously. More recently, the DQPSK format was employed to show OTDM transmission over 240 km at 1.28-Tb/s and over 160 km at a bit rate of 2.56 Tb/s.
A simple method for realizing high bit rates exceeding 1 Tb/s consists of combining the OTDM and WDM techniques. For example, a WDM signal consisting of M separate optical carriers such that each carrier carries N OTDM channels at the bit rate B has the total capacity Btot = MNB. The dispersion limitations of such a system are set by the OTDM-signal bit rate of NB. In a 1999 experiment, this approach was used to realize a total capacity of 3 Tb/s by using M = 19, N = 16, and B = 10 Gb/s. The channels were spaced 450 GHz apart to avoid overlap between neighboring WDM channels, and the 70-nm WDM signal occupied both the C and L bands. The OTDM bit rate was extended to 320 Gb/s in a 2004 experiment that transmitted 10 such channels over a limited distance of 40 km. By 2009, five channels, each operating at 320 Gb/s, were transmitted over 525 km by using a time-domain optical Fourier transformation technique. With the use of new modulation formats and coherent detection, the total capacity of such OTDM/WDM systems should exceed 10 Tb/s. However, many factors such as various nonlinear effects in fibers and the practicality of dispersion compensation over a wide bandwidth are likely to limit the performance of such systems.
OTDM has also been used for designing transparent optical networks capable of connecting multiple nodes for random bidirectional access. Its use is especially practical for pack-based networks employing the ATM or TCP-IP protocol. Similar to the case of WDM networks, both single-hop and multihop architectures have been considered. Single-hop OTDM networks use passive star couplers to distribute the signal from one node to all other nodes. In contrast, multihop OTDM networks require signal processing at each node to route the traffic. A packet-switching technique is commonly used for such networks.