Bits battle noise
You must measure OSNR as well as BER to fully characterize optical receivers.
Martin Rowe, Senior Technical Editor -- Test & Measurement World, 3/1/2005
Optical receivers face the same battle with noise when they try to distinguish 1's and 0's when signals loom close to the noise floor. As light travels through fibers, it distorts and loses energy. Long-haul networks use optical amplifiers to boost weak signals, but these amplifiers add noise to the original signal. The more amplifiers used in an optical link, the more noise a receiver will see relative to received signal power. The result: a decrease in optical signal-to-noise ratio (OSNR), which can increase the link's bit-error rate (BER).
Figure 1 illustrates the problem. At the transmitter's output, the signal is clean and its amplitude is high. At the amplifier's input, the signal has lost amplitude. The amplifier returns the signal to its original amplitude, but with added noise. At the receiver, the signal is again diminished in amplitude, and the added noise may cause the receiver to misinterpret a bit.
 |
| Figure 1. Optical signals need amplifiers to boost them for the long haul, but amplifiers add noise. |
You can determine whether a receiver maintains the required 10–12 BER by measuring BER vs. OSNR. To learn the best ways to make these measurements, I asked engineers from several companies to describe their measurement setups. Some hold a signal's noise level constant and vary signal power while measuring BER. Others prefer to hold the signal power constant and vary the noise level while measuring BER. In a third method, called "IsoBER," the engineers adjust both signal and noise levels to attain a constant BER.
From the results of their measurements, equipment manufacturers and service providers can
- verify conformance of optical receivers used in long-haul networks,
- develop a BER vs. OSNR range by which manufacturing can guarantee system performance with a margin of safety, and
- determine system limitations based on the performance of various receivers and optical links.
At Ciena's Linthicum, MD, optical lab, senior principal engineer Doyle Nichols evaluates optical receivers for use in the company's 10-Gbps long-haul transport systems. To measure BER vs. OSNR, Nichols uses a setup similar to that illustrated in Figure 2. He uses a communications test set to generate a pseudorandom bit sequence (PRBS) and measure BER, and he uses an optical spectrum analyzer (OSA) to measure OSNR.
 |
| Figure 2. Ciena’s Doyle Nichols uses this test setup to evaluate optical receivers. Courtesy of Ciena. |
Nichols starts by measuring the noise produced by the OSA with no input signal; that establishes the measurement system's noise floor. He then characterizes the DUT's dynamic range by measuring BER with a clean signal.
When he tests a receiver that uses a PIN diode, he does so at an input power that ranges from 0 dBm to –2 dBm for high power and –18 dBm to –20 dBm for low power. The high-power range often distorts the input signals, testing the receiver's ability to tolerate distortion. To measure the distortion, Nichols views an eye diagram of the receiver's electrical output on a digital communications analyzer.
Next, Nichols introduces noise into the system by coupling a noise source into the PRBS signal through a 50/50 optical combiner. A 95/5 splitter lets him connect the OSA and communications analyzer so he can measure OSNR and distortion. After running the signal through an erbium-doped fiber amplifier (EDFA) and an optical band-pass filter, Nichols employs another splitter to tap off 10% of the remaining signal power into an optical power meter. With the power meter, he monitors the total power going into the receiver under test. The filter removes noise that occurs at frequencies outside those of interest. Without the filter, the power meter's measurement would read excessively high.
To perform the BER vs. OSNR measurement, Nichols chooses three values of OSNR. At each OSNR value, he sweeps the receiver's input power by adjusting the level of the signal and noise in concert across the receiver's dynamic range. Thus, he holds OSNR constant during those power sweeps. For each of the three OSNR values, Nichols sets the receiver's 0/1 threshold level to the point that produces the best BER, which is the number he reports at each OSNR value. Each sweep takes about 4 min to complete.
High-level inputsYou could also perform the BER vs. OSNR measurement by adjusting an amplifier's input power with the amplifier running in saturation. Wolfgang Moench, senior marketing manager for fiber-optics at Acterna, explained that some engineers use optical amplifiers to adjust OSNR. "With an amplifier in saturation, you'll always get the same output power for a certain range of input power," noted Moench.
 |
| Figure 3. The amplifier adds noise to the attenuated signal and the second attenuator controls the receiver’s input power. Courtesy of Acterna |
A high-level input signal will produce a high OSNR because the amplifier's output power will be relatively constant. An attenuator at the amplifier's input signal (Figure 3) will decrease OSNR because the noise produced by the amplifier will be greater relative to the amplifier's signal output because the attenuator decreases signal power.
The system in Figure 3 uses a second attenuator in front of the device under test (DUT) that can decrease the DUT's input power without changing OSNR, because it reduces both signal and noise power proportionally. By varying the DUT's input power, you can measure BER vs. input power with a constant OSNR at numerous power levels. Moench noted that his staff might measure BER vs. input power at OSNR levels of 15 dB, 20 dB, and 25 dB, depending on the DUT's specifications. With the two attenuators, they can hold either OSNR or input power constant while varying the other.
Hold steady
 |
| Figure 4. The procedure for performing an IsoBER test consists of finding a constant BER and varying OSNR. Courtesy of Circadiant Systems. |
Figure 5, which came from an IsoBER measurement, shows a plot of optical power vs. OSNR for a constant 10–12BER.The power required to maintain that BER level drops when OSNR reaches a "knee," which for this receiver occurs at about 22 dB OSNR. Below that OSNR level, the receiver requires a more powerful signal to maintain the 10–12BER.
The long haul to RomeNot all BER vs. OSNR measurements are made in an engineering lab or a production environment. Service providers measure BER vs. OSNR when they bring new links or equipment on line, and researchers working on the latest optical links also need to make these measurements. For example, a consortium of engineers working on the ATLAS (All-optical Terabit per second LAmbda-Shifted transmission) project built and tested 40-Gbps optical links between Rome and Pomezia. The group consisted of engineers from Pirelli Labs (Milan) as well as from several universities. "Testing 40 Gbit/s multi-channel systems with frequency conversion: 1st ATLAS results" describes the test in detail (exp.telecomitalialab.com/upload/articoli/V03N02Art336.pdf).
 |
| Figure 5. The signal power needed to maintain 10–12 BER falls until OSNR reaches a threshold. Courtesy of Circadiant Systems. |
Pirelli explained that the team held the power at the receiver constant while they varied the noise level. The test used two amplifiers. The first ran in the linear range, which let the engineers vary the noise level and measure BER. The second amplifier, closer to the receiver, ran in saturation mode, thus providing a constant power at the receiver's input. They varied the noise by adjusting the gain of the first amplifier, which was closer to the transmitter.
"We decided to use the varying OSNR method because it is more realistic than varying optical power," said Schiffini. "In a real system, in case of degradation of some devices placed in the line—EDFA degradation, fiber bending, or OADM VOA degradation—the system experienced a decrease of OSNR, with constant power at the receiver."
At the time of the experiment, which ended in December 2002, no 40-Gbps BER tester was commercially available. Schiffini and the team of engineers had to develop their own equipment to generate a test pattern and measure BER.
Regardless of whether you make BER vs. OSNR measurements on the bench or in the field, you need to adjust signal power, OSNR, or BER. The adjustments for each measurement often take a few minutes, but they provide valuable information about component, equipment, or network performance.
