Design Features March 3, 1997

The right test equipment
simplifies measuring BER in burst-mode systems

Lutz Kristen, Hewlett-Packard

Without the right equipment, experimentally verifying burst-mode receiver designs can be difficult. An example demonstrates a new method of determining such receivers' error performance. The discussion explains the equipment setup and the measured results.

The use of burst-mode data transmission is increasing in digital-communication systems. These systems use point-to-point and point-to-multipoint connections over optical, wireless, and coaxial media. Designers worldwide are working on local-access approaches using time-division multiple-access (TDMA) technology. Examples include passive optical networks (PONs), fiber to the home, fiber to the curb, and hybrid fiber/coax. Optical computer interconnects, digital-satellite communication systems, and military communication systems also use burst-mode data streams. For example, military aircraft and ships employ an optical MILbus.

Residential access--the combination of voice, data, and video communication to and from the home--is one of today's hottest research-and-development areas. Interactive services increase traffic from the home to the central office (CO). This traffic demands higher bandwidth. A PON is one of the technologies that can supply the needed bandwidth. PONs use TDMA for upstream (home-to-office) communication.

Burst-mode receivers are among TDMA networks' most critical parts. Design criteria include cost, transmission distance, performance quality, and network reliability.

The key performance measurement of transmission links is the bit-error rate (BER). BER is the number of bits that exhibit errors divided by the number of bits sent. This measurement indicates the probability of errors occurring over arbitrary numbers of bits. Although BER measurements on continuous data are common, such measurements on burst-mode data are more difficult. BER measurements are even more complicated on TDMA networks. One reason for this difficulty is that clock recovery requires resynchronizing the analyzer. Measurements must account for the different propagation delays that result from different path lengths. Receivers must also handle wide ranges of input power, which result from different path lengths.

This example explains a laboratory procedure for characterizing the physical-layer transmission performance of a PON network's TDMA receiver. Figure 1 shows a simplified PON with only four optical network units (ONUs). A passive optical coupler connects the ONUs and the CO. This passive coupler transmits data broadcast downstream from the CO to all ONUs. The passive optical star sums the time-multiplexed data, which the ONUs transmit upstream to the CO. Each ONU uses a certain time slot for transmitting data. Such a point-to-multipoint network, using time-division multiplexing, is called a TDMA network.

Figure 2 shows the ONUs' upstream signal at the CO input. In this example, each time frame consists of two cells broadcast by the ONUs: cell A from ONU 1 and cell B from ONU 2. The cells repeat once per frame. Each cell may consist of several segments with different lengths. For example, segments may include a preamble (used in Figure 2's upstream signal for bit synchronization); a delimiter used for byte synchronization; and information identification, which shows that the cell function is operation, cell administration, or maintenance. The physical-layer devices can use the other segments. The payload contains the data destined for the CO. The time between cell A and cell B is called "guard time" (Figure 1).

Method and equipment

The best test environment for the receiver is a model of its real-life usage, so the complete device under test (DUT) not only consists of the receiver to be characterized but also includes the ONUs and the passive optical coupler. In the test environment, unlike the real network, the ONUs and the coupler are physically close to the receiver; optical attenuators simulate the different branch distances. To characterize TDMA and PON devices accurately and reliably, you need measurement equipment that can do the following:

• Generate proprietary cell formats: No standard exists; changing technologies require flexibility in data generation.
• Generate and analyze cells and data packets that contain user-defined patterns and pseudorandom bit sequences (PRBSs); cell segments that perform different functions should be available: For example, you might have a preamble of variable length for synchronization; a header for cell identification; a payload with either a user-defined pattern or a PRBS (pure or with predefined errors); and a trailer.
• Vary the guard time to test the receiver's threshold-adjust circuit: This ability allows stressing the clock-recovery circuit by applying odd guard times.
• Synchronize the receiver and the transmitter (a requirement for BER measurements): In time-synchronized systems, the delays associated with the received data must be stable through the DUT. In a system with clock recovery and retimed data, the analyzer should be able to synchronize to a unique pattern within the cells or to an external bursted clock that is synchronous with the bursted data.
• Measure results either as error count or as BER.
• Vary bit rates: This capability is necessary because upstream and downstream data rates are often different, and different access technologies use different bit rates.
• Generate control signals for the receivers and transmitters: You should be able to easily relate the signal timing to the burst-mode data, that is, to a receiver-reset pulse, to the transmitter's laser on/off signal, or to oscilloscope trigger signals.

You can use Figure 3's setup to perform the measurements. The test setup consists of two serial-cell generators to stimulate the transmitters with burst-mode data. Two attenuators simulate different fiber lengths; that is, they generate cells that have different optical power. The passive optical coupler combines the optical datapaths.

The receiver's electrical output connects to the input of a serial-cell analyzer's analyzer channel. A second serial-cell-generator channel provides the reset pulse for the receiver under test.

Automatic capabilities

You must be able to adjust the cell-transfer delays from generators A and B to the analyzer input. Some analyzers can automatically perform this tedious task. Other analyzers require oscilloscope measurements and manual adjustment. To start the measurement, make the following calibrations and measurements: Calibrate the attenuators with a power meter. You then know the optical power at the receiver input for each attenuation factor. Supply the software with the propagation delay between the reset-pulse generator output and the receiver's reset input. After that, the instrument takes all delays and cell and frame lengths into account, so you can directly program parameters such as guard time and distance from the reset pulse to cell A. If the transmitter modules' optical power depends on temperature, long-term measurements must monitor the analyzed cell power and regulate it to a constant level. (The measurement setup does not show this function.)

The receiver under test has an automatic threshold control, which requires a reset pulse prior to every cell to reset an internal control circuit. Without the reset pulse, the receiver may fail to detect a low optical-power-level cell that closely follows a high cell. Figure 4's optical input-signal receiver detects data from the first bit. Therefore, it needs no preamble to characterize the data transfer. The traffic cells consist of only one segment, which contains a pure PRBS.

Figure 4's upper trace shows the receiver's optical input signal, for example, the end of cell A and the beginning of cell B. The lower trace displays the electrical reset pulse.

The initial setup of the measurement used manufacturer-specified settings: ­17 dBm for the generator generating the "leading" cell and ­34 for the "trailing" cell (Table 1). The most challenging scenario for the receiver's gain control occurs when a low-power burst follows a high-power burst.

Performing measurements

Characterizing the receiver involves measuring BER vs: optical power, guard time, and the reset pulse's width and location and measuring how the distribution varies vs bit location. Figure 5 shows how the BER of cell B depends on the optical power of cell B and the power of leading cell A. The power of leading cell A influences the BER of following cell B.

BER Specifications are 10^­8 at ­34 dBm, so the receiver is within specifications, but there is not much margin. The next group of measured results includes a discussion of possibilities for improvement.

The second measurement used two reset-pulse widths and varied the guard time. Because the influence on the BER is insignificant, narrowing the guard time could increase the throughput (Figure 6).

The third measurement held the guard time constant but varied the position of the reset pulse between the two cells for two pulse widths. Again, the BER was almost independent of these timing parameters (Figure 7).

Segment mode

In the final measurement, the ability to measure the BER of individual parts of a cell (the segment) made possible a BER measurement of bits vs location. Figure 7 also shows that the error rate varies with respect to the bit location within the cell. A more detailed measurement would reveal whether ones or zeros would fail more often. If the threshold is too high, failed ones predominate; if it's too low, failed zeros predominate. This measurement of BER vs reset pulse gives the valuable hint that optimizing the receiver's threshold-adjust circuit to BER distribution vs bit location would improve the overall BER performance.

BER is the most important index of system integrity and quality of service. Burst-mode BER measurements are extremely difficult without flexible test equipment specifically designed for this purpose. Classic BER instruments have difficulty with bursted, multichannel measurements because they perform the BER test on all data, not just specified segments.

With proper test equipment, you can optimize a system to provide the best trade-off between throughput, length of transmission, and transmitter power. Moreover, you can thoroughly analyze the receiver's dynamic range, optimize threshold and power levels, and estimate recovery speed.

Author's biography Lutz Kristen has worked at Hewlett-Packard's Boeblingen (Germany) Instrument Division for 22 years and has helped to develop pulse, data, and function generators, and analyzers. In his current position as sales-development engineer, he is responsible for supporting field engineers and customers. He holds a Dipl-Ing degree from Technische Universitat, Karlsruhe, Germany, and is a member of VDE.

| EDN Access | Feedback | Subscribe to EDN | Table of Contents |