The requirement, from both businesses and consumers, for 'data-on-demand' is forcing communications companies around the world to deliver more and more bandwidth. Copper connectivity, having been given a new lease of life through the likes of various digital subscriber line (DSL) technologies, can now provide relatively high bandwidth but only over short distances. It also falls far short of the bandwidth obtainable through air. The problem with transmission through air is that, at a fundamental level, it is very much like trying to hold a conversation in a very crowded and noisy room.
Optical communications seem to hold all the answers and multi-fiber optics (again at an extremely fundamental level) can be thought of as having access to several 'airs' through which to create a very high bandwidth communications systems.
As with (copper) wired and wireless communications systems there are essentially three configurations an optical-based system can take: point-to-point; broadcast; or multicasting (from multiple sources to multiple users - as we have with phone systems and the Internet for example). However, while a general optical communications system will typically support many users, we shall, for the purpose of this article, only flesh out the basics of a physical system with a single user: and not tackle the multiplexing of signals for multiple users.
In tackling a single user system, we need only model a point-to-point system and once the basic underlying physical mechanisms have been established the system limitations can be easily explored. These limitations will include range, bit error-rate (BER) performance, maximum sustainable and peak data rates, timing jitter, sensitivity, and Signal to Noise Ratio (SNR). Only when the limitations have been identified is it possible to optimize a system in order to conform to a particular standard. For optical communications such a standard might be SONET/SDH.
As the source information (such as audio, video or sonar) comes in many forms, the first step is to convert the information into a bit-stream. This can be done with, as a minimum, traditional analog to digital converters (ADCs), but will often require the ADCs to work in conjunction with other conversion and conditioning techniques.
Essentially though, once the information is stored electronically and whatever the form, data can be converted into raw binary data prior to transmission. Typical file formats for given types of information include: WAV, MID, MP3 for audio; MPEG and AVI for video; JPG, BMP and TIF for graphics; and BIN, MAT and DAT for raw data.
It is worth noting that according to information theory, the maximum amount of information that can be transferred across a binary symmetric channel occurs when 0s and 1s occur with an equal probability. For this reason, information is usually encoded in a scheme that compresses data and represents it in the source-encoding step. While this step removes redundancy, the next step (channel encoding) introduces redundancy to combat channel errors later in the system. There are two different types of channel errors; burst and random. Fortunately, interleaving and forward error correction (respectively) help alleviate these problems.
The next step is to modulate the baseband and this is where, in constructing an optical communications system, we start making compromises. Different modulation schemes make varying use of the bandwidth, but the trade-offs will include distance, compression, BER and power.
Due to the lack of 'coherent' optical systems for extracting phase information, optical communication systems use pulse code modulation (PCM) schemes to modulate the baseband. Coherent optical systems, while under development in numerous research labs around the world, are not yet commercially available. When they are available though we hope to have access to modulation schemes not too dissimilar from those currently employed in the world of pure electronics and RF.
The most common PCM scheme employed in optical communication systems is the non-return to zero (NRZ) format. Most RF engineers will be familiar with NRZ, but for those unfamiliar with it, it is essentially nothing more complex than sending a signal during the bit period for a '1' and almost nothing during a '0'. While common, NRZ is not particularly bandwidth efficient and is prone to potential problems with dispersion, where dispersion can be thought of as pulse spreading.
Return to zero (RZ) coding is more resilient towards pulse dispersion as it pulses high during half of the bit period for a '1' and almost nothing for a '0'. Another benefit of RZ is that it transmission requires only half the power of NRZ. However, the price to be paid here is that twice the bandwidth is needed than with NRZ.
An alternative baseband modulation technique, called Duobinary Coding (available with or without pre-coding) is available (see Figures 1a and 1b).
Figure 1a: Duobinary baseband modulation shown in the time domain.
Figure 1b: Duobinary baseband modulation shown in the frequency domain.
Duobinary coding deliberately introduces 'controlled' Inter-Symbol Interference (ISI) by adding the input sample with the previous one. This, in effect, reduces the spectrum by a factor of 2, but again at a price: more power is needed for transmission to achieve the same BER performance as NRZ (see Figure 2). Pre-coding boosts Duobinary Coding's performance, by summing (modulo-2) the past pre-coded input with the input prior to the final summing.
Figure 2: BER versus channel SNR for NRZ (green) and Duobinary with and without pre-coding (blue and red respectively). While there is little difference between Duobinary with pre-coding and Duobinary without pre-coding, the difference between NRZ and Duobinary is significant and it takes about 4.77dB extra power to obtain the same BER performance as that for NRZ.
Baseband filtering (pulse shaping)
Following baseband modulation comes baseband filtering or pulse shaping. Filtering is typically achieved through the root raised cosine (RRC) method as it exhibits good spectral containment. The spectral plot (shown in Figure 3) shows the optical spectrum with and without baseband filtering: emphasizing how critical filtering is to designs that do not have the luxury of infinite bandwidth.
Figure 3. Spectral containment through baseband filtering.
Gaussian pulse shaping on the other hand has more consistent group delay than RRC filtering and results in less dispersion across the pulse. In addition, Gaussian pulses in the time domain have the desirable property of a Gaussian pulse shape in the frequency domain; in fact the Fourier Transforms are given by:
Also available to us is Hyperbolic Secant Pulse Shaping. This is ideal for so called 'soliton' communication systems, which (though not discussed here) do show considerable potential for high-speed communication with pulse widths in the order of 10fs and maximum transmission lengths orders of magnitude longer than normal light wave packets.
The next step (or component) in the Electro-Optic Transmitter is the all-important light source, with the most common source being the Laser. LEDs can be, and often are, used for optical communications but they produce incoherent light that severely reduces the obtainable bandwidth. LEDs are therefore typically restricted to use in low-cost, low-bandwidth applications.
Lasers produce coherent light and deliver far higher performance than LEDs - but they are not without flaws. Firstly, the wider the spectral width of a laser the more dispersion will occur over a fiber. And secondly, the narrower the spectral width, the more the signal is prone to Stimulated Brillouin Scattering (SBS), which produces a backward reflecting wave Doppler shifted down by about 11.1GHz.
At the component level, there are a number of LASER types and it is worth looking at the more common ones.
Fabry-Perot (FP) lasers are essentially LEDs with mirrors (cleaved facets) on the sides (see Figure 4a). Unwanted amplified spontaneous emission (ASE) passes out of the cavity and the desired stimulated emissions resonate at a whole number of wavelengths relative to the cavity length.
Figure 4 Part 1: Laser structures (a) Fabry-Perot (b) Distributed Feedback (c) Distributed Bragg Reflector.
The cavity length is given by:
Where m is an integer, _ph is the photonic wavelength and n is the refractive index of the material.
There are a number of ways of physically implementing FP lasers in order to localize the light for fiber penetration, and here are just two:
Gain Guided FP lasers inject localized power to help confine the radiated emissions (spectral width = 5 to 8nm, line width = 0.005nm approximately.); and
Index Guided FP lasers, which have a cavity formed physically in the semiconductor material to provide the best optical containment.
Another type of laser is the Distributed Feedback (DFB) laser (Figure 4b). It typically produces a very narrow line width, which is accomplished by placing a Bragg grating internal to the lasing cavity itself. Bragg gratings are periodic variations in the index of refraction and are placed just outside the active region of the lasing cavity where the evanescent field can still interact, but where minimal loss is incurred.
Mirrors are optional for DFBs and govern the strength of the grating required. Of note is that since the operation of a DFB Laser is based on minor differences in the index of refraction, they are sensitive to temperature variations.
Operating under the same principle as the DFB, but with the Bragg grating external to the lasing cavity and less sensitive to temperature variations, is the Distributed Bragg Reflector (DBR) laser. Both DFBs and DBRs are relatively high priced devices.
Vertical Cavity Surface Emitting Lasers (shown in Figure 4d), on the other hand are less expensive and exhibit relatively good performance. Structure wise, whereas most semiconductor lasers emit light through their edges, VCSELs have mirrors in the horizontal plane and light is emitted vertically. Since the cavity is in the vertical direction, small cavity lengths can be created, which in turn produces a small spectral width.
Figure 4 Part 2: Laser structures (d) Vertical Cavity Surface Emitting Laser (e) and the Multi-Quantum Well.
The last type of laser to be mentioned here is the Multi-Quantum Well (MQW) laser (Figure 4e). This type provides excellent confinement and can be constructed by stacking either FP or DFB structures. And the benefit of the ultra thin layer 'stacking' is that multiple coherent lasing cavities can be formed, thus providing more gain to the structure. In return, less power is needed for the production of light.
The last step in placing a signal onto an optical fibre is to modulate the carrier. One option is to modulate the Laser's bias directly, but switching a Laser at speed for high-bandwidth communications is very difficult. A more common approach for high bandwidth applications is to allow the Laser to 'free-run' and employ an external electro-optic component to modulate the light.
One such electro-optic component is the Mach-Zehnder (MZ) Modulator. The MZ is comprised of two 3dB couplers with connecting paths of different time delays or phase changes. The different time delays can be accomplished by either having different physical lengths or different indices of refraction. By applying an electric field across one or both of the paths, the index of refraction can be altered just enough to change the speed of propagation relative to the parallel path. This has the same effect as if different lengths were present, but provide the ability to manipulate this directly with an electronic signal.
This article has touched on how the demand for bandwidth is driving high-speed communications onto fiber, and has introduced the Electro-Optic Transmitter. Transmitter functions (source and channel encoding; baseband modulation and filtering; and optical carrier modulation) have been described paying particular attention to light source generation (lasers).
On balance, optical transmission shares much with microwave and RF transmission -- only the medium and frequencies/ wavelengths have changed. In part two of this article we will address fibre distortion (attenuation and pulse dispersion) and how to retrieve the information after it is transformed into light.