Test VSR optical interfaces

A telecom switching facility such as a central office may include dozens of network components that communicate at speeds up to 10 Gbits/s. Unfortunately, installing serial 10-Gbits/s optical interfaces between the network components gets expensive, mostly because of the high cost of optical components such as laser diodes, optical modulators, and photo detectors. These costs will continue to rise as transmission speeds increase to 40 Gbits/s.

To reduce the costs of optical interfaces, the Optical Internetworking Forum (OIF, www.oiforum.com) developed the Very Short Reach (VSR) optical interface (Ref. 1). A VSR interface replaces a typical serial communications link with a 12-fiber parallel optical bus in which 10 fibers carry data. A VSR interface uses ten 1-Gbit/s paths to create a 10-Gbits/s optical link. The eleventh channel is a spare data channel, and the twelfth channel carries error-correction information. Figure 1 shows where VSR interfaces connect to network equipment in a central office. 

Figure 1. VSR interfaces send data between network components over 12 parallel optical lines.

VSR interfaces carry data distances of 300 m or shorter. Because VSR carries data for such short distances, a VSR interface can use multimode optical fibers, which can use less expensive lasers and photodetectors than a serial 10-Gbits/s link needs. A 12-channel, 1-Gbit/s/channel parallel bus costs less to manufacture than one serial 10-Gbits/s optical link.

Gigabit Ethernet forms the basis for each VSR channel (Ref. 2). To get the approximately 1-Gbit/s capacity, a channel moves data at 1.24416 Gbits/s with 8B/10B encoding. The additional two bits for every eight ensure that enough bit transitions occur for a receiver to extract a clock signal. As a result of the encoding, the effective data rate becomes 1.24416 Gbits/s times 8/10, or 0.995328 Gbits/s—just under 1 Gbit/s.

The VSR interface uses 12 channels to carry data between routers, dense-wavelength-division multiplexer (DWDM) terminals, and synchronous optical network (SONET) add-drop multiplexers. Internally, though, the network components use a 16-bit parallel electrical bus known as Serdes-to-Framer Interface 4 (SFI-4). A VSR converter IC multiplexes data from 16 lines inside the switch or router to 12 lines for transmission over a VSR interface.

Each VSR channel requires a transmitter and a receiver. Transmitters in VSR interfaces use a vertical-cavity surface-emitting laser (VCSEL) array that produces light with an 850-nm wavelength. Because VCSELs emit light vertically, VSR transmitter manufacturers can easily build laser-diode arrays suitable for the parallel optical interface. VSR receivers typically consist of a parallel array of top-illuminated pin photodiodes.

Because of the short working distance, VSR interfaces can use a ribbon cable of 62.5-µm-diameter multimode fibers. Although multimode fibers cost more than the 10-µm-diameter single-mode fibers used in serial optical interfaces, component manufacturers can still use multimode fibers to build less costly optical components. Multimode fibers don't require complex, expensive optical-alignment techniques to couple light to and from VCSELs and photodiodes.

To test the 12 channels on a VSR interface, you must perform 12 sets of Gigabit Ethernet-like tests. You need to measure average power, spectral width, timing, jitter, and bit-error rate (BER). You also must measure optical and electrical crosstalk, a measurement you don't need for a serial interface. Then, you must test the performance of the entire VSR interface for its ability to reliably send SONET frames. Figure 2 shows the test setup. 

Figure 2. VSR transmitter tests require an optical spectrum analyzer to look at optical spectrums and an oscilloscope to measure rise time, fall time, extinction ratio, and jitter. A SONET test set generates test data, and an optical power meter measures transmitter power.

A VSR interface uses a transmitter-receiver module at each end of the fiber. To test an optical transmitter, connect a VSR SONET test set to a VSR interface's optical receiver. A VSR receiver converts 12 optical signals from the test set into 16 electrical channels that loop back to drive the optical transmitter (the UUT). You also need to connect 300 m of 62.5-µm-diameter multimode fiber to each transmit channel.

In a transmitter power test, an optical power meter measures the average power emitted by each channel's transmitter. Configure the test set to produce a pseudorandom bit sequence (PRBS) that repeats every 127 (2⁷–1) bits. At the transmitter, you must verify that the average output power is strong enough to overcome losses in the optical fiber. You should also verify that the optical power from all 12 lasers falls within eye-safety limits (Ref. 3).

Next, you'll need to use an optical spectrum analyzer (OSA) to measure the optical spectrum in each channel after the light has passed through 300 m of fiber. Because VCSELs emit light in multiple optical modes, the optical spectrum always contains multiple peaks. With the OSA, verify that the peak emission wavelength falls between 830 nm and 860 nm. Measure the transmitted signal's spectral width, which should be less than 0.85 nm rms.

For a communications interface to function properly, its waveforms must conform to specific limits. You must measure a signal's rise time, fall time, extinction ratio, and jitter. To make the first three measurements, use a digital oscilloscope with a 62.5-µm diameter multimode optical input. For jitter measurements, you'll need to use the oscilloscope's electrical input. For all four tests, you need an oscilloscope and optical input that, when used together, provide a bandwidth of at least 2.5 GHz.

Use the oscilloscope to build an eye diagram of each VSR channel's test signal. From the diagram, measure rise time, fall time, and extinction ratio. Acceptable values of these link parameters are specified by OIF for the maximum fiber length. OIF specifications set the maximum fiber length at 300 m. If your transmitter's signal parameters fall within the test limits for 300-m fiber, you'll get an acceptable BER for the optical data link.

Jitter measurements for VSR signals also follow the methodology of Gigabit Ethernet. As with Gigabit Ethernet signals, you must measure two forms of jitter: random jitter and deterministic jitter. Random jitter occurs on all data patterns, whether repeatable or random. Deterministic jitter, though, depends on the bit pattern. But you can't measure both jitter components from an eye diagram.

All digital signals contain random jitter. In simple repeating patterns such as 0000011111, you'll measure random jitter only. A histogram of that signal's jitter will produce a Gaussian curve. But live bit patterns constantly change. The response of the transmitter and receiver changes with bit pattern and those changes produce deterministic jitter.

To measure a transmitter's random jitter, use a test set to produce a repeating 0000011111 pattern (called a K28.7 pattern, Ref. 4). This pattern produces a waveform at one-fifth of the optical link's bit rate. Because a K28.7 signal lacks deterministic jitter, you can measure random jitter by simply taking a histogram at the zero-crossing point of a rising edge or falling edge.

Figure 3. A histogram reveals the amount of random jitter in a bit’s rising edge.

You must also use the photodetector and blocking capacitor to measure deterministic jitter. This time, use a K28.5 pattern, a 20-bit repeating pattern of 00111110101100000101 that produces a worst-case deterministic-jitter measurement. To find this value, you must measure the range of variation at the 10 zero-crossing points that occur in the bit pattern. Set the oscilloscope to trigger at the start of each 20-bit pattern. Use your oscilloscope to average the signals, which removes the random jitter component from the signal. The time of each zero-crossing is then compared to the mean expected time of crossing and a set of ten timing variations are determined. The deterministic jitter is the range of timing variations. 

Figure 4. In a receiver test, a SONET test set sends data through the receiver under test and reads back the data to make BER measurements.

The final transmitter test requires you to measure a transmitter's output waveform against a Gigabit Ethernet eye mask. Use the oscilloscope's 62.5-µm-diameter optical input for this measurement. Most oscilloscopes that perform eye-mask tests include a Gigabit Ethernet mask in their standard setup menus. To perform the test, pass the transmitter's output signal through a fourth order Bessel-Thompson filter (usually built into the oscilloscope) to remove any high-frequency noise from the test signal. Set the filter's cutoff frequency to 0.75 times the bit rate. For a bit rate of about 1.25 Gbits/s, you should set the filter's cutoff frequency to 937.5 MHz.

To test a VSR receiver, you need to send the received signals back to the test set. Because the receiver outputs a 16-channel electrical signal, you need to loop those signals to a VSR transmitter to send 12 optical channels back to the test set. You also need an optical power meter and 12 variable optical attenuators (VOAs) to attenuate optical power in all channels except the channel under test.

You can measure a receiver's performance at known power levels by measuring BER with the SONET test set. Set the optical attenuators so the optical power meter reads –10 dBm on all channels except the channel under test. Drive those channels with valid data, which adds the effects from any channel-to-channel crosstalk to the test. Turn the UUT's protection mechanism off so the VSR converter won't correct bit errors. Vary the power in the channel under test, monitor the power with the power meter, and measure BER with the test set. Find the power level where BER reaches 10^–13 (one error in 10¹³ bits). A power level of less than –16 dBm should achieve better than 10^–13 BER.

The test set can't measure BER on each VSR channel, but you can infer the BERs for a single channel from a test on the entire VSR interface. Remember that a VSR interface uses 10 channels to carry data. In this test case, you provide sufficient power to produce no errors on 9 out of 10 channels. Because you force errors on just 1 of 10 channels, the entire link's BER becomes 10 times less than that of a single channel. So, when you measure a link BER of 10^–13 with the test set, you can infer that the channel under test's BER is 10^–12 because you know that the other nine channels run error free.

Because the VSR interface forms a parallel bus, you might expect to measure channel-to-channel skew, the timing difference among channels. Most parallel buses require this measurement, but not VSR. The VSR converter IC can tolerate channel-to-channel skew of as much as 80 ns. Because of that wide tolerance, you don't need to measure channel-to-channel skew in a VSR interface.

Following the transmitter and receiver tests, you must perform SONET compliance tests on the entire VSR interface. A SONET test set will perform these tests for you. The test set sends data organized into frames as opposed to sending raw bits. The data travels as a payload of a SONET frame. Testing with SONET frames tells you if the VSR interface properly multiplexes and demultiplexes bits into and out of the 10 data lines that the interface uses. The VSR component under test must prove that it won't corrupt any SONET frames.