RZ signals need new measurements
Greg LeCheminant, Agilent Technologies, Santa Rosa, CA- September 1, 2002
The world's appetite for faster communications now pushes the development of 40-Gbits/s optical-network components. At such a high data rate, the familiar non-return-to-zero (NRZ) signaling method runs into problems when carrying bits over long distances. Optical pulses spread as they move along fiber-optic lines, causing bit errors if one bit spills into another. Another signaling method, called return-to-zero (RZ), overcomes this problem and, despite having a few drawbacks, is gaining momentum for use in long-distance, 40-Gbits/s optical networks.
|Figure 1. NRZ signals (top trace) remain at high power for successive logic-1 bits while RZ pulses return to the low-power state.|
To understand the problems that RZ signaling can solve, you need to understand how NRZ and RZ signals differ. The key difference comes in how the signals handle optical power, as illustrated in Figure 1. The top trace in Figure 1 represents a 11101001 NRZ bit pattern; the bottom trace shows the same bit pattern in RZ format. In NRZ signals, a string of successive logic-1 bits keeps the optical power at its high level until a logic-0 bit occurs. With RZ signaling, the power level drops (returns to logic-level 0) even when successive logic-1 bits occur.
Because RZ pulses always return to the low-power state, they are less likely to produce bit errors caused by dispersion. Dispersion occurs when a signal's components have different propagation velocities as a light pulse moves through a fiber-optic cable. Two types of dispersion—chromatic dispersion and polarization-mode dispersion—affect a light pulse.
With chromatic dispersion, a pulse's propagation velocity depends on wavelength (Ref. 1). Optical pulses can distort as they travel, because they contain a range of wavelengths that arrive at the optical receiver at different times. Thus, an optical pulse will broaden as it travels through a cable. At 40 Gbits/s, one bit period takes just 25 ps, so even a minor amount of broadening will cause the bit to interfere with an adjacent bit. Because RZ signals use short pulses that are less likely to spread into adjacent bits, they lessen the effects of chromatic dispersion.
Polarization-mode dispersion causes optical pulses to travel at speeds that vary based on polarization (Ref. 2). Pulse components that travel in different planes can arrive at the receiver a few picoseconds apart. That time difference will cause the pulse to spread, perhaps enough to interfere with an adjacent bit. RZ pulses, with shorter pulse widths, can suffer more dispersion than NRZ pulses without causing bit errors.
RZ signals also can take advantage of the soliton effect (Ref. 3). RZ transmitters can emit soliton pulses that don't change shape as they travel through a fiber. A fiber's index of refraction controls the velocity of the light. Thus, an RZ transmitter can overcome the index of refraction by changing the shape and power level of a pulse (Ref. 4).
Although pulses in RZ signals are less likely to spill into adjacent bit periods, they have a drawback. Returning all logic-1 bits to low power produces more transitions than in NRZ signals. More transition requires more bandwidth, which increases the cost of optical fiber and network components.
While RZ signals require measurements similar to those that NRZ signals require, the definitions and the way that test equipment performs those measurements differ. Both NRZ and RZ signals require two important measurements—extinction ratio and jitter. RZ signals also need a contrast-ratio measurement. All three measurements require eye diagrams.
|Figure 2. The familiar NRZ eye diagram shows high-power levels (logic-1 bits), low power levels (logic-0 bits), and transitions between the two polarities.|
|Figure 3. The RZ eye diagram shows a line for logic-0 bits and pulses for logic-1 bits.|
|Figure 4. Extinction-ratio measurements for NRZ signals require an instrument to measure the mean of an eye diagram’s logic-1 levels and logic-0 levels.|
|Figure 5. Histograms from logic-0 bits overlap with histograms from the trough of logic-1 bits.|
|Figure 6. Jitter measurements on RZ signals require a histogram of rising or falling edges at 50% of pulse amplitude.|
NRZ and RZ waveforms produce different eye diagrams. Figure 2 shows the familiar NRZ eye diagram, and Figure 3 shows an eye diagram for an RZ signal. An optical oscilloscope, often called a digital communications analyzer (DCA) or communications signal analyzer (CSA), builds the eye diagram by overlaying the waveforms of many bits, both logic 0's and logic 1's.
Because NRZ signals can remain at either high power or low power over successive bits of the same logic level, you see straight lines across the eye diagram at both power levels. The upper line in the NRZ signal indicates that several logic-1 bits have occurred in succession during the signal acquisition. The lower straight line indicates successive logic-0 bits. The traces in between indicate bit logic-level changes.
RZ signals never stay at high power, even when successive logic-1 bits occur. So, you see the flat line at the low-power (logic-0) level only. In Figure 3, the low-level straight line indicates occurrences of logic-0 bits, and the pulses indicate logic-1 bits.
An analyzer such as a DCA or a CSA can derive information about an optical network's performance from eye diagrams. Extinction-ratio measurements tell you the vertical distribution of the waveform, which indicates how efficiently a transmitting laser converts its input power to output power. To calculate extinction ratio, an analyzer divides a signal's mean high power by its mean low power. The analyzer calculates the mean high-power (logic 1) and low-power (logic 0) values based upon histograms. But how the analyzer sets up the histograms differs from NRZ to RZ signals.
For NRZ signals, an analyzer will take an amplitude histogram of the central 20% of an eye's width. Figure 4 shows the width of the area (between the markers) that the instrument uses. The instrument takes histograms at the logic-1 level and the logic-0 level within the 20% window. From the two mean values, the instrument can calculate the signal's extinction ratio.
For RZ signals, an analyzer will take a histogram from a smaller horizontal area because of the narrow RZ pulses; the equipment takes measurements based on a window that's just 5% of the eye's total width. The narrow window for taking the histogram ensures that the instrument measures just the peak of the logic-1 pulses. The analyzer also creates a histogram of the logic-0 levels at a 5% width centered directly below the peaks. From the two histograms, the instrument calculates extinction ratio.
RZ signals require measurements that NRZ signals don't require. One such measurement, called contrast ratio or suppression ratio, compares the peak of the logic-1 pulse to the trough or "pit" of the signal. If you measure contrast ratio where a pulse enters a fiber and where it exits the fiber, you'll get an indication of how much the fiber alters the pulse's shape. Contrast ratio doesn't give you a direct measure of communication system performance, but it provides you with a numeric value for how pulse shapes change as they travel through a fiber.
To make a contrast-ratio measurement, an analyzer will measure the lowest point of a logic-1 pulse that's followed by another logic-1 pulse. An ideal logic-1 pulse would return to the same power level that the receiver sees for logic-0 signals. But the eye diagrams of the RZ signals show that the power level between successive logic-1 pulses doesn't get that low, because the next logic-1 pulse causes the power level to rise again.
An analyzer builds an amplitude histogram at the low point between two logic-1 pulses. If you look at a histogram that covers the lowest 5% of an eye diagram, you'll find that the histogram contains hits from the logic-0 bits, too (Figure 5). An analyzer must compensate for that condition before it can proceed with its measurement calculations.
To compensate for the overlap, a DCA or CSA must first take the histogram of the logic-0 levels at the center of the eye, just like it did for extinction ratio. That procedure provides a baseline logic-0 measurement. Next, the analyzer takes the aggregate histogram of the logic-0 signals and the logic-1 pulses where the logic-1 pulses reach their minimum power level. In software, the instrument can subtract the baseline histogram from the aggregate histogram of the low points of the logic-1 pulses and the overlapping logic-0 levels. This yields the true histogram of the low points of the logic-1 pulses, so the instrument can perform the ratio calculation.
While contrast ratio or suppression ratio applies to RZ signals only, jitter occurs in all digital signals. You must take jitter measurements in both NRZ and RZ signals. Jitter measurements tell you the horizontal (time) differences among pulses. As jitter increases, eye openings close. At some point, the eye will close to where a receiver will misinterpret bits.
The definitions of the measurements differ and are more difficult to get from an RZ eye diagram. For NRZ signals, an analyzer measures jitter at the crossing points of an eye diagram, which an analyzer can find easily . To measure rms jitter, the instrument will find the standard deviation of a histogram at the crossing points. To measure peak jitter, the analyzer will find the width of the signal at a crossing point. When making jitter measurements on NRZ signals, you can use any crossing point, because all crossing points include both rising edges and falling edges.
RZ eye diagrams, though, lack a crossing point. To measure the eye closure caused by jitter in RZ signals, an analyzer measures the jitter on both the rising and falling edges, which can differ. To measure the line (trace) width, the instrument finds the point that's halfway between the mean logic-0 power level and the mean logic-1 power level. Figure 6 shows a jitter-measurement histogram from a rising edge.
Extinction ratio, contrast ratio (suppression ratio), and jitter provide important information about an RZ signal. Communications test equipment will have to evolve as the communications industry adopts standards for the signaling method. For example, no standard masks exist that define the limits for the signal's shape. The communications industry needs those standards before 40-Gbits/s networks can move data around the world.