To handle this challenge, design engineers are going back to the drawing board to develop modems and headend equipment that takes advantage of the data over cable service interface specification (DOCSIS) version 1.1. The benefit here is that DOCSIS 1.1 provides a series of quality of service (QoS) techniques for prioritizing voice, video, and data services in a cable modem system design.

Simply meeting DOCSIS 1.1, however, is not enough. Impairments in the cable modem network can be perilous for the transmission of voice/video/data on the pipe. To counteract this problem, designers must emulate as closely as possible how their system architecture will operate when implemented in a real-life environment.

The answer lies in proper testing. Through the development of a strong testing plan, designers can actively emulate the impact of noise, interference, and other impairments on cable modem system and equipment design. In turn, they can account for these impairments early in the system design process.

The HFC network

Multiple service operators, (MSOs) have completed extensive upgrades to their cable network transmission facilities to support high-speed, two-way data, voice, and video communications. Today's upgraded, two-way cable network is capable of delivering broadcast TV, high-speed data, and voice services over a combination of fiber-optic and coaxial cable transmission facilities.

Fiber-optic transmission lines are used to carry packets from the cable modem termination system (CMTS) deep into the local neighborhood to take advantage of fiber's superior transmission characteristics. The fiber transmission lines terminate at the fiber node, from which coaxial cable lines carry the packets over the remaining distance to the home. Because both fiber and coaxial cable transmission facilities are used, this type of cable network is commonly referred to as a hybrid-fiber coax (HFC) cable network.

Fiber-optic cable is primarily used to carry the data between the headend and the subscriber for a number of reasons. Fiber requires fewer amplifiers than coaxial cable and optical amplifiers have better noise and distortion performance than coaxial amplifiers.

Cable modem systems must deal with a variety of impairments. The two most common, however, are noise/interference and signal distortion.

Noise and interference from a variety of sources distort packet transmission on the HFC network. Many devices, including common household appliances such as garbage disposals and blenders, emit signals at frequencies in the upstream band (5 to 42 MHz in the US).

These undesired signals enter the cable system through poorly shielded cables or through the communication devices attached to the cable network within the home. This type of interference is commonly referred to as ingress. Ingress interference is typically impulsive and bandlimited in nature as the interfering signals appear for short intervals at significant power levels within a small frequency range.

Besides ingress, additional traffic on the HFC network, such as signals from other cable modem subscribers and television signals, can interfere with packet transmission. These signals can be present in both the downstream and upstream channels and appear as interfering signals at the same frequency as the data channel (co-channel interference) or at nearby frequencies (adjacent channel interference).

The combination of ingress and other interfering signals reduces the carrier-to-noise ratio (CNR) in the user channel. A lower CNR reduces throughput as forward error correction (FEC) techniques are more heavily utilized to deal with transmission errors.

Regardless of how well an HFC network is designed, some level of ingress and other interfering signals will be present and will have some effect on the performance of the cable modem system.

Some equipment manufacturers are adding specialized features to the CMTS to gain a performance advantage by minimizing the effect of ingress. One example of this type of feature is the ability of a CMTS to detect ingress in a channel and then change upstream channels. Both equipment manufacturers and MSOs should evaluate these types of advanced features before deployment to measure the impact on performance levels.

Newly-developed DOCSIS 1.1 tests provide examples of the types of noise, ingress, and interfering signals that should be used to measure the performance of the cable modem system. The DOCSIS RF-PHY-22 upstream packet error rate test details a frame loss characterization test, which finds the carrier signal-to-impairment ratio that produces a target packet loss value. PHY-22 specifies a variety of noise and interference types, including background noise, quadrature amplitude modulation (QAM)-16 and quadrature phase shift keying (QPSK) interferers (DOCSIS upstream modulation techniques), continuous wave (CW) interferers, and AM signals, that model typical conditions on a HFC network.

Distorting signals

Amplifiers, which are used to compensate for the loss that occurs as signals travel through the HFC network, produce an impairment called intermodulation distortion (IMD). Figure 1 shows a constellation diagram of an unimpaired cable modem signal. The constellation diagram displays the complex amplitude and phase components of the received cable modem signal on a two-dimensional (X-Y coordinate) display, where each defined point represents multiple bits of user data.

The constellation diagram in Figure 2

Compression of these points - where transmitted power is at its peak - causes the cable modem receiver to misinterpret one constellation point as another. This results in the cable modem exhibiting an increased error rate as the information represented by a constellation point is altered during transmission over the HFC network.

Each amplifier also has associated diplex filters that block out-of-band signals before and after the amplifier. When multiple amplifiers are used on a cable network, the group delay and amplitude distortion responses of the diplex filters combine to produce an overall group delay and amplitude distortion shape. This phenomenon occurs primarily in the upstream direction, where the low-pass filters typically have a corner frequency of approximately 42 MHz.

The combined response of multiple filters produces increasing group delay and amplitude distortion near the edge of the filter passband. Severe group delay and amplitude distortion characteristics in the upstream channel lead to increased error rates during cable modem data transmission. The architecture of an HFC network may include as few as five or as many as 20 amplifiers, meaning that cable modem systems must be designed to operate over a wide range of group delay and amplitude distortion characteristics.

Micro-reflections

Impedance mismatches between items in the coaxial cable plant, such as amplifiers, couplers, and the coax cable, produce another impairment called micro-reflections. When micro-reflections occur, additional copies of the desired signal arrive at the receiver delayed and attenuated by certain amounts. For cable modem transmission techniques, the presence of multiple copies of the desired signal results in intersymbol interference (ISI) that may cause the receiver to improperly detect the amplitude and phase of the incoming signal.

DOCSIS 1.1 cable modem systems are designed to include a pre-equalization function that counteracts the effects of micro-reflections. The CMTS determines the pre-equalization coefficients by characterizing the micro-reflections present on the HFC network and then transmits the coefficients to the cable modem to use in subsequent transmissions.

Both the CMTS and the cable modem must operate properly for the pre-equalization to maintain reliable packet transmission in the presence of micro-reflections. The DOCSIS RF-PHY-20 pre-equalizer test provides seven different micro-reflections test cases to analyze cable modem/CMTS performance. Modeling different micro-reflection scenarios is critical for a system designer to understand how a cable modem system performs in a variety of conditions.

Test system requirements

It is critical to have an understanding of the key impairments encountered in a cable modem network. The key, however, is accounting for these impairments in the modem design. To achieve this task, designers must employ a proper test system that allows them to emulate noise, interference, micro-reflections, and other impairments during the system development process.

A typical physical layer (PHY) cable modem test system includes two main components: a packet generator/analyzer and an HFC network impairment emulator. The data generator/analyzer sends/receives IP packets to/from cable modems and CMTS and uses performance metrics such as packet loss and latency to evaluate performance. Packet loss is a common metric defined as the ratio of the number of packets lost in transmission between the CMTS and cable modem to the total number of transmitted packets. The cable network impairment emulator recreates real-world HFC network impairments including noise, interference, IMD, and micro-reflections in a controlled, repeatable environment.

By combining these instruments into an integrated test system, designers can perform a variety of tests that help track down the impact of impairments on cable modem equipment designs.

Additive impairment testing approaches

One of the most common impairment measurements is the packet loss versus carrier noise ratio (CNR) test. This test examines the transmission performance of a cable modem system in the presence of variable noise and interference conditions. Various types of noise and interference, including wideband noise, bandlimited noise, or modulated interference, can be used to model the different conditions that are present on the HFC network.

Wideband noise best represents the background noise present on an HFC network that remains at a relatively constant level. Bandlimited noise and modulated interferers best model transient ingress conditions that can be present at severe levels for a given set of frequencies.

An example set from a packet loss versus CNR test is shown Figure 3. In this example, the upstream performance of the cable modem system is being characterized. The downstream channel attenuation is configured in the cable network impairment emulator to produce a 0-dBmV receive signal level at the cable modem while the upstream signal attenuation was set to 30 dB.

During each step of the test, packets were transmitted from the cable modem to the CMTS for a period of 90 seconds and the packet loss was measured by the data generator/analyzer. The upstream CNR ratio in the cable network impairment emulator was varied from 25 to 10 dB in steps of 0.5 dB to model a range of high transmission quality and low transmission quality on HFC networks. Three DOCSIS 1.0 cable modems were individually tested with a DOCSIS 1.0 CMTS and the test results for each of the three cable modem systems are displayed on the same graph in Figure 3 to simplify comparisons.

Based on the results from Figure 3, its clear that achieving several dBs more of noise immunity is important in a cable modem design. This may make the difference between error-free data transmission and failed data transmission on certain networks.

Testing IMD

Another test for evaluating cable modem performance is packet loss versus IMD test. This test examines performance in the presence of variable IMD conditions, with IMD expressed in terms of the level below the carrier signal.

The downstream performance of the same cable modem system tested above was also characterized for packet loss versus IMD. The downstream channel attenuation was configured in the cable network impairment emulator to produce a 0-dBmV receive signal level at the cable modem while the up-stream signal attenuation was set to 30 dB.

Packets were again transmitted from CMTS to the cable modem for a period of 90 seconds for each step of the test and the packet loss was measured by the data generator/analyzer. The downstream IMD ratio in the cable network impairment emulator varied from -50 to -30 dBc in steps of 1 dB to model a range of IMD conditions typically present on HFC networks.

Figure 4 shows the test results for packet loss versus IMD test, with the y-axis displaying the percentage packet loss and the x-axis displaying the carrier-to-IMD ratio. As this figure shows, the presence of IMD on a cable network produces a wide disparity of performance among the cable modems. Depending on the number of amplifiers used in the design of a particular operator's network, a different range of IMD conditions will exist. Developing a modem that can maintain reliable data transmission in the presence of a higher IMD level will reduce service issues for an operator.

Handling micro-reflections

The upstream performance of a cable modem system can be characterized using a modulation error rate (MER) versus micro-reflections test. MER is a measure of total error on the signal constellation. Packets are transmitted from the cable modem to the CMTS by the data generator/ analyzer. Rather than looking at the number of lost packets as the performance metric, this test measures the MER of the upstream transmission bursts.

As outlined in the DOCSIS PHY-20 test, a baseline condition with no micro-reflections and seven different configurations of micro-reflections are tested. Each test condition has a single path with no delay or attenuation and one to three additional paths that are delayed and attenuated (see Table 1).

In this series of tests, the cable modem system is allowed to adjust the pre-equalization coefficients using the pre-equalization algorithm for a period of 60 seconds after a test condition has been enabled. PHY-20 specifies that the measurement taken for each of the seven test conditions be compared to a pass/fail criteria of a 27-dB MER.

The delay and attenuation values for each of the reflected paths can be modified to model more or less severe micro-reflection configurations. In general, smaller attenuation and delay settings produce more severe performance degradation.

Characterizing packet loss

A packet loss characterization test can be used to evaluate the upstream or downstream performance of the cable modem system. When running this measurement, the test condition is defined as an impairment such as wideband noise, bandlimited noise, modulated interference, or IMD. Rather than varying the impairment from a fixed start value to a fixed stop value as is done in the packet loss versus CNR test, the packet loss characterization test attempts to find the impairment level that produces a predefined pass/fail value for packet loss.

During the packet loss characterization test, the impairment level is increased until the packet loss exceeds the pass/fail criteria. Then the impairment level is slightly decreased to narrow down to the desired value. Target packet loss for this test is typically 1%.

Some observations can be made about the performance of cable modem systems in the presence of common HFC network impairments from the sample test results presented in this article. The most important observation is that not all cable modem systems perform equally. Even though a cable modem or CMTS may be DOCSIS certified, a great deal of variability in real-world performance still exists. This point emphasizes the importance of performance testing for both the operator and the manufacturer.

A large number of vendors are developing cable modems, hoping to capitalize on consumer and corporate demand for faster access to communication networks. With so many vendors competing in this marketplace, it is important that vendors focus on product performance in the presence of real-

world HFC network conditions. By designing a test plan that incorporates the tests described in this article, vendors will be better equipped to produce cable modems that provide the customer with reliable, high-speed performance. Waiting until a cable modem is nearing field trials or deployment to examine real-world performance could lead to costly product delays.

David Garrison is a director of product development in the TAS Division of Spirent Communications. Prior to joining the company in 1993, he received a BSEE and MSEE from Rensselear Polytechnic University. He can be reached at dave.garrison@spirent.com.