Making Light of High-Speed Communications

It’s obvious that over-the-air (OTA) radio transmissions go through the air and, similarly, fiber-optic communications use optical fibers. What may not be immediately apparent are the implications of the large differences between the two media.

OTA is a well-accepted acronym, but wireless signals actually propagate well in a vacuum. Generally, the atmosphere impairs radio communications, for example by absorbing microwave energy at frequencies above about 10 GHz, depending on humidity. Conversely, the ionosphere reflects short-wave signals in the 3- to 30-MHz range, enabling very long-range reception, far beyond the usual line-of-sight limitations.

Nevertheless, although radio transmissions can be reflected multiple times and suffer frequency-dependent attenuation, the medium is fundamentally linear. Signal level doesn’t cause the medium’s characteristics to change. The concept of a linear medium is important because signals only add or subtract in a linear medium.

A common linear effect in an optical fiber is dispersion, a term that refers to a frequency-dependent change in velocity. Chromatic dispersion (CD) is caused by frequency-dependent variations in a fiber’s index of refraction. The effect restricts dense wavelength division multiplexing (DWDM) to fibers having a very small CD index. Otherwise, the various wavelengths propagate at different group velocities and data reception over a long distance becomes error-prone.

On the other hand, so-called four-wave mixing (FWM) is a nonlinear effect that can occur in DWDM systems that have essentially constant phase and group velocities across the various carrier frequencies. This is important in a DWDM system with uniform carrier spacing ?f because the FWM products of mixing signals at f, f + ?f and f + 2?f are signals at f-?f, f + ?f and f + 3?f, which directly coincide with carriers already at those frequencies. A common solution is to alternate lengths of fibers with positive and negative CD or to use nonuniform frequency spacing.

FWM is one of the problems that result from the Kerr effect, an instantaneous nonlinear polarization in the medium that becomes more obvious at high power levels. Unfortunately, as more channels are used in DWDM systems and the power of each is raised to achieve greater transmission range, the increased overall power in the fiber causes larger nonlinear effects. In addition to FWM, self-phase modulation and cross-phase modulation (XPM) arise from the Kerr effect.¹

Polarization mode dispersion (PMD) also affects fiber performance and is caused by imperfections in fiber symmetry. Because a signal’s polarization has both X and Y components, anything that causes a difference in propagation of either component will affect the signal’s polarization. Mechanical problems such as a tight bend radius can change PMD or the fiber itself may not be perfectly round or have completely uniform consistency throughout its length.

Because of the very large amount of information that an optical fiber can carry, especially when DWDM is used, most optical communications links operate with simple on-off keying (OOK). Either a pulse is present or it isn’t. More recently, complex modulation has been seen as a way to improve spectral efficiency, just as it has successfully done for wireless communications. The various impairments that exist in commercial networks as well as the very high symbol rates involved in complex optical modulation highlight the need for good measurement tools.

Testing Complex Fiber-Optic Modulation

Modulation Formats
Simply increasing the data rate won’t work. Higher data rates based on OOK require proportionately wider bandwidths and add noise in the same ratio. Improving the optical signal to noise ratio (OSNR) by increasing the laser power isn’t feasible and would also magnify CD and PMD effects. The present 50-GHz DWDM channel spacing limits the available bandwidth, and an important goal is to increase spectral efficiency while working with the existing fiber networks and channel spacing. The best solution is to adopt complex modulation that encodes phase and amplitude, achieving a reduced baud or symbol rate appropriate for the bandwidth.

Differential modulation formats are particularly attractive because an absolute reference isn’t required: The previous pulse becomes the reference for the current one. Differential phase shift keying (DPSK) compares the phase of the present pulse with that of the previous one. It only encodes one bit per symbol so isn’t much of an improvement over OOK.

Quadrature phase shift keying (QPSK), however, uses four phase values spaced at ?/2 intervals to encode two bits per symbol. Differential QPSK (DQPSK) has similar advantages, but the encoding is carried by the difference between successive states rather than by the states themselves. Both QPSK and DQPSK types of modulation also are used in wireless networks.

Information is associated with polarization in newer weather radar systems, but generally polarization modulation (PM) isn’t used for wireless communications. Nevertheless, in fiber-optic communications networks, it is practical to combine PM with QPSK. PM also is called dual polarization state modulation. PM-QPSK can be represented as a standard wireless QPSK phase diagram but with the two polarization states superimposed in a 3-D fashion. By encoding eight states per symbol, dual polarization QPSK (DP-QPSK) supports 100-Gb/s within a 50-GHz bandwidth.

Electrical Instruments
Several aspects of high-rate complex modulated optical signals make measurements difficult. The first is simply the high rate itself. QPSK achieves 100 Gb/s with a 25-GHz symbol rate. In practice, error correction also requires some bandwidth, so either the actual bit and symbol rates will be somewhat higher or the data rate a little less.

Although only a few QPSK systems now are being used commercially, research is under way to develop even faster schemes. Michael Schnecker, business development manager at LeCroy, described recent activity: “Our WaveMaster 830Zi DSO with 30-GHz analog bandwidth and 80-GS/s sampling rate was used by Alcatel-Lucent, Bell Labs researchers to demonstrate DP-QPSK signal transmission at 224 Gb/s, equivalent to a 56-Gbaud symbol rate. The combination of 30-GHz real-time bandwidth and the capability to emulate the coherent receiver in an intradyne system enables the detection of complex optical modulation formats such as DP-QPSK or various quadrature amplitude modulation (QAM) formats.”

For the Bell Labs experiment, four 14-Gb/s pseudo-random bit sequences (PRBS) were separately delayed and multiplexed to create I and Q 56-Gb/s bit streams. The two bit streams then drove a QPSK modulator. After splitting, one part of the signal was delayed and recombined with the other part in a polarization beamsplitter (PBS). By suitably adjusting this delay, the end result was a DP-QPSK modulated signal equivalent to a single channel rate of 224 Gb/s.

The coherent receiver is based on an intradyne technique. Heterodyne receivers mix the carrier with a local oscillator to produce an intermediate frequency (IF) higher in frequency than the bandwidth of the modulation to be recovered. With a sufficiently high IF, wide bandwidth modulation is preserved in phase and amplitude. In a homodyne receiver, the LO has exactly the same frequency as the carrier, so the modulation is recovered relative to DC. Intradyne receivers develop a relatively low-frequency IF that falls within the modulation bandwidth.

LeCroy’s Mr. Schnecker explained further, “Optical coherent receivers use an LO laser to generate a low IF because it is impractical to build an optical phase-locked loop (PLL) that would allow full heterodyne mixing as is done in RF applications. The intradyne IF is a complex signal that has a slowly varying complex center frequency causing the IQ plane to rotate at the rate of this offset.

“Once the I and Q signals are digitized in a DSO, a digital PLL recovers the bit rate, which allows the complex magnitude and phase of each symbol to be computed. The phase of a symbol is compared to the phase angle of the nearest constellation point for the given type of modulation, and the phase difference is used to correct the next symbol,” he said. “This process unwraps the phase to provide a constellation diagram from which signal errors can be measured.”

The high bandwidth and sampling rate available in Textronix DPO/DSA72000B Scopes also have proven useful in optical communications testing. According to Pavel Zivny, performance oscilloscope product engineer at Tek, “The early research simply used the scope as a raw digitizer. The DPO72004B’s combination of 20-GHz bandwidth, four channels, and 50-GS/s simultaneous sampling [made it] the only scope that could handle the multichannel signaling of 112 Gb/s on the 28-Gbaud DP-QPSK format.”

Today, several application-specific optical processing instruments have been developed that support research into even faster rates and more complex modulation schemes. In an announcement related to the 2010 OFC/NFOEC conference, Daniel van der Weide, vice president of engineering at Optametra, said, “Thanks to the flat frequency response, fast rise times, and low noise floor of Tek’s DPO72004B, we can confidently state that our OM4000 [Series of Coherent Lightwave Signal Analyzers] captures the complete electric field of the optical signal at a full 112 Gb/s.”²

Figure 1 shows various views of 44-Gb/s QPSK return to zero (RS) single-polarization modulated PRBS analyzed by the Optametra OM4106 with a Tek DSA72004B as the digitizer. In addition to the constellation and coherently demodulated IQ eye diagrams, the inset at top center displays the output of a software-simulated delay-line interferometer (DLI) differential in-phase eye, and the inset to the right shows the power eye—the degenerate eye as it would appear on a sampling scope.