Code-Division Multiplexing (CDM)
A multiplexing scheme, well known in the domain of wireless communications, makes use of the spread-spectrum technique. It is referred to as code-division multiplexing (CDM) because each channel is coded in such a way that its spectrum spreads over a much wider region than occupied by the original signal. Although spectrum spreading may appear counterintuitive from a spectral point of view, this is not the case because all users share the same spectrum. In fact, CDM is used often for cell phones as it provides the most flexibility in a multiuser environment. It is also relatively secure because it is difficult to jam or intercept the signal in view of its coded nature. The term code-division multiple access (CDMA) is often employed in place of CDM to emphasize the asynchronous and random nature of multiuser connections.
Even though the use of CDMA for fiber-optic communications attracted attention during the 1980s, it was only after 1995 that optical CDM (OCDM) was pursued seriously. It can be easily combined with the WDM technique. Conceptually, the difference between the WDM, TDM, and CDM can be understood as follows. The WDM and TDM techniques partition the channel bandwidth or the time slots among users. In contrast, all users share the entire bandwidth and all time slots in a random fashion in the case of CDM. The transmitted data can still be recovered because of the orthogonal nature of the codes employed. In this respect, CDM is similar to the OFDM (Orthogonal Frequency-Division Multiplexing) technique discussed earlier.
The new components needed for any CDM system are the encoder and decoder located at the transmitter and receiver ends, respectively. The encoder spreads the signal spectrum over a much wider range than the minimum bandwidth necessary for transmission by means of a unique code. The decoder uses the same code for compressing the signal spectrum and recovering the data. Several methods can be used for encoding depending on whether it is done in the time domain, spectral domain, or both. The codes employed are referred to as being two dimensional when both time and frequency are involved. The time-domain codes include direct-sequence encoding and time hopping. The spectral codes can be implemented using the amplitude or the phase of various spectral components. In this section we discuss several encoding schemes used in recent experiments.
1. Time-Domain Encoding
The following figure shows an example of time-domain coding for optical CDMA systems.
Each bit of data is coded using a signature sequence consisting of a large number, say M, of shorter bits, called "chips" borrowing the terminology used for wireless (M = 7 in the example shown). The effective bit rate (or the chip rate) increases b the factor of M because of coding. The signal spectrum is spread over a much wider region related to the bandwidth of individual chips. For example, the signal spectrum becomes broader by a factor of 64 if M = 64. Of course, the same spectral bandwidth is used by many users distinguished on the basis of different signature sequences assigned to them. The recovery of individual signals sharing the same bandwidth requires that the signature sequences come from a family of the orthogonal codes. The orthogonal nature of such codes ensures that each signal can be decoded accurately at the receiver end. Transmitters are allowed to transmit messages at arbitrary times. The receiver recovers messages by decoding the received signal using the same signature sequence that was used at the transmitter. The decoding is accomplished using an optical correlation technique.
The encoders for signature-sequence coding typically use a delay-line scheme that looks superficially similar to that shown here (which is for multiplexing several OTDM channels).
The main difference is that a single modulator, placed after the laser, imposes the data on the pulse train. The resulting pulse train is split into several branches (equal to the number of code chips), and optical delay lines are used to encode the channel. At the receiver end, the decoder consists of the delay lines in the reverse order (matched-filter detection) such that it produces a peak in the correlation output whenever the user's code matches with a sequence of time chips in the received signal. Chip patterns of other users also produce a peak through cross-correlation but the amplitude of this peak is lower than the autocorrelation peak produced when the chip pattern matches precisely. An array of fiber Bragg gratings, designed with identical stop bands but different reflectivities, can also act as encoders and decoders. Different gratings introduce different delays depending on their relative locations and produce a coded version of the signal. Such grating-based devices provide encoders and decoders in the form of a compact all-fiber device (except for the optical circulator needed to put the reflected coded signal back onto the transmission line).
The CDM pulse trains consisting of 0 and 1 chips suffer from two problems. First, only unipolar codes can be used simply because optical intensity or power cannot be negative. The number of such codes in a family of orthogonal codes is often not very large until the code length is increased to beyond 100 chips. Second, the cross-correlation function of the unipolar codes is relatively high, making the probability of an error also large. Both of these problems can be solved if the optical phase is used for coding in place of the amplitude. Such schemes are being pursued and belong to coherent CDMA. Ad advantage of coherent CDMA is that many families of bipolar orthogonal codes, developed for wireless systems and consisting of 1 and -1 chips, can be employed in the optical domain. When a CW laser source is used in combination with a phase modulator, another CW laser (local oscillator) is required at the receiver for coherent detection. On the other hand, if ultrashort optical pulses are used as individual chips, whose phase is shifted by π in chip slots corresponding to a -1 in the code, it is possible to decode the signal without using coherent detection.
In a 2001 experiment, a coherent CDMA system was able to recover the 2.5 Gb/s signal transmitted using a 64-chip code. A sampled fiber grating was used for coding and decoding the data. Such a grating consists of an array of equally spaced smaller gratings so that a single pulse is split into multiple chips during reflection. Moreover, the phase of preselected chips can be changed by π so that each reflected pulse is converted into a phase-encoded train of chips. The decoder consists of a matched grating such that the reflected signal is converted into a single pulse through autocorrelation (constructive interference) for the signal bit while the cross-correlation or destructive interference produces no signal for signals belonging to other channels. The experiment used a NOLM (the same device used for demultiplexing of OTDM channels) for improving the system performance. The NOLM passed the high-intensity autocorrelation peak but blocked the low-intensity cross-correlation peaks. The receiver was able to decode the 2.5-Gb/s bit stream from the 160-Gchip/s pulse train with less than -3 dB penalty at a BER of less than 10-9. In 2002, this approach was used to demonstrate a four-channel WDM system, employing OCDM with 255 chips and quaternary phase coding at a chip rate of 320 Gchip/s.
2. Frequency-Domain Encoding
Spectral encoding involves modifications of the amplitude or the phase of various spectral components of a short pulse according to a preassigned code. Phase encoding has attracted most attention and has been implemented in several experiments and field trials. It can be implemented using several different schemes. A bulk-optics approach, shown schematically in the figure below, employs a diffraction grating with a reflective, liquid crystal, spatial light-phase modulator (SLPM).
The grating diffracts spectral components in different directions, and the SLPM changes their phases using a preset code. If binary coding with phase values of 0 and π is employed, SLPM simply changes the phase by π of some code-selected spectral components. The same grating combines all spectral components during the return path, and a circulator directs the resulting, temporally broadened, spectrally-coded, optical pulse to its output port.
The encoder shown above is not practical for real systems because of its bulky nature. For this reason, several integrated versions have been developed. In one experiment, the spectral phase encoder consisted of multiple microring resonators coupled to two input and output waveguides (or buses), as shown schematically in the figure (a) below.
Each set of four microring resonators (diameter ~0.1 mm) is designed to transfer one specific wavelength from input to output bus. Multiple thermo-optic phase shifters are used to change the phase of various spectral components from 0 to π, depending on the code employed; they also serve as bandpass filters. The 2006 experiment employed 8 frequencies on a 10-GHz frequency grid for implementing an 8-chip code and distributed 5-Gb/s signals to six users with a spectral efficiency of 0.375 b/s/Hz. In a 2007 field trial, the spectral phase encoder, shown schematically in figure (b) above, employed phase modulators between two AWGs, which divided the spectrum of 0.7-ps pulse into 63 parts and then combined these parts back after phase shifts imposed on them as required by the CDMA code. The same device was used as a decoder at the receiver end with complimentary phase shifts that made the optical phase uniform across the entire pulse spectrum.
3. Frequency Hopping
Spectrum spreading can also be accomplished using the technique of frequency hopping in which the carrier frequency is shifted periodically according to a preassigned code. The situation differs from WDM in the sense that a fixed frequency is not assigned to a given channel. Rather, all channels share the entire bandwidth by using different carrier frequencies at different times according to a two-dimensional code. Such a spectrally encoded signal can be represented in the form of a matrix shown schematically in the figure below.
The matrix row correspond to assigned frequencies and the columns correspond to time slots. The matrix element mij equals 1 if and only if the frequency ωi is transmitted in the interval tj. Different users are assigned different frequency-hop patterns (or codes) to ensure that two users do not transmit at the same frequency during the same time slot. The code sequences that satisfy this property are said to be orthogonal codes. In the case of asynchronous transmission, complete orthogonality cannot be ensured. Such systems make use of pseudo-orthogonal codes with maximum autocorrelation and minimum cross-correlation to ensure a BER as low as possible. In general, the BER of such CDMA systems remains relatively high (typically > 10-6), but it can be improved using a forward-error correction scheme.
Frequency hopping in CDM lightwave systems requires a rapid change in the carrier frequency. It is difficult to make tunable semiconductor lasers whose wavelength can be changed over a wide range in a sub-nanosecond time scale. One possibility consists of hopping the frequency of a microwave subcarrier and then use the SCM technique for transmitting the CDM signal. This approach has the advantage that coding and decoding is done in the electrical domain, where the exiting commercial microwave components can be used.
Several all-optical techniques have been developed for frequency hopping. They can be classified as coherent or incoherent deeding on the type of optical source used for the CDMA system. In the case of incoherent CDMA, a broadband optical source such as an LED (or spontaneous emission from a fiber amplifier) is used in combination with a multi-peak optical filter (such as an AWG) to create multiwavelength output. Optical switches are then used to select different wavelengths for different chip slots. This technique can also be used to make CDMA add-drop multiplexers. An array of fiber gratings having different Bragg wavelengths can also be used for spectral encoding and decoding. A single chirped Moire grating can replace the grating array because several gratings are written at the same location in such fiber gratings. In a 2000 experiment, several Moire gratings were used to demonstrate recovery of 622-Mb/s CDMA channels. An integrated version of CDMA encoders, based on silica-on-silicon AWGs, has also been developed. In this device, variable delay lines are incorporated between two AWGs.
In another approach called coherence multiplexing, a broadband optical source is used in combination with an unbalanced MZ interferometer that introduces a delay longer than the coherence time in one is its branches. Such CDMA systems rely on coherence to discriminate among channels and are affected severely by the optical beat noise. In a demonstration of this technique, four 1-Gb/s channels were multiplexed. The optical source was an SOA operating below the laser threshold so that its output had a bandwidth of 17 nm. A differential-detection technique was used to reduce the impact of optical beat noise. Indeed, bit-error rates below 10-9 could be achieved by using differential detection even when all four channels were operating simultaneously.
The coherent CDMA systems designed with spectral encoding have a distinct advantage that the CDMA signal can be overlaid over a WDM signal such that both signals occupy the same wavelength range. The figure below shows schematically how such a hybrid scheme works.
The spectrum of the received signal consists of a broadband CDMA background and multiple sharp narrowband peaks that correspond to various WDM channels. The CDMA background does not affect the detection of WDM channels much because of its low amplitude. The CDMA receiver employs a notch filter that remove the WDM signal before decoding it. The hybrid WDM-CDMA scheme is spectrally efficient as it makes use of the unused extra bandwidth around each WDM channel.
WDM systems in which each channel is transmitted using CDM are of considerable interest. IN this case, spectral efficiency is at a premium because the CDM-signal bandwidth should not exceed exceed channel spacing. In a 2002 experiment, spectral efficiency of 1.6 (b/s)/Hz and a capacity of 6.4 Tb/s were realized in the C band alone using the combination of CDMA and WDM techniques. This system employed the QPSK format for optical encoding and decoding with ultrafast optical time-gating. By 2009, a field trial demonstrated successful operating of a WDM-CDMA system capable of distributing 10-Gb/s signals to eight users simultaneously over 100 km with the 16-chip encoders and decoders based on sampled fiber gratings.