The Uses of New Test Methods and Test Equipment
The time-sensitive networking (TSN) extension of the standard Ethernet protocol is gaining widespread support in the automotive sector as a backbone network technology. Offering a scalable platform for transmitting data in vehicles at a current rate of 100Mbits/s, and 1Gbit/s in the future, it is supported by a wide choice of off-the-shelf hardware and software components. Since Ethernet uses a switch-based architecture, it enables manufacturers to increase the number of nodes in new vehicle designs while reducing the number and length of their data cables.
Unlike older automotive network technologies such as controller area network (CAN) and FlexRay buses, which used a balanced electrical interface, Ethernet components and systems typically use a similar balanced interface but employ a pulse amplitude modulation versus digital modulation schema. In pulse amplitude modulation, different voltage levels represent digital information which differs from CAN signaling. This slight difference introduces new challenges for susceptibility and can increase the vulnerability to interference from electro-magnetic noise sources in the analog domain. While the TSN extensions to the Ethernet protocol are working to provide standards for fail-safe operation and deterministic delivery of high-priority signals, it is crucial for manufacturers of electronic control and infotainment modules, and for the vehicle manufacturers which use them, to carefully verify that the Ethernet networking interface performs as specified under all operating conditions.
Development and testing practices deployed elsewhere, such as in the telecom industry and in data centers, provide a proven and effective model for the verification of Ethernet systems in the presence of strong and varied sources of external interference. This article describes how the principles of laboratory testing of electrical noise impairment can be followed in the automotive sector to dramatically reduce the manufacturer’s development time and effort, while also providing comprehensive verification of the safe and predictable performance of Ethernet networks under any realistic real-world conditions.
How Ethernet Supports the Bandwidth-Hungry Vehicles of the Future
The TSN extension of the standard Ethernet protocol provides automotive manufacturers with the assurance they need that Ethernet can provide predictable, deterministic performance and very low latency. The introduction of TSN has thus made Ethernet the next-generation backbone networking technology of choice for passenger cars and other vehicles.
With the introduction of ever more active safety, infotainment and connectivity applications in vehicles, the amount of data carried over automotive networks and the transmission speed required are both set to increase very rapidly. Ethernet provides headroom to support this increasing hunger for bandwidth. By contrast, legacy automotive networking technologies such as the CAN bus have largely reached the limit of their networking capability.
The Ethernet TSN standards lay down a set of protocols which guarantee real-time, deterministic transmission of mission-critical signals. Since vehicles are subject to stringent functional safety standards, this ability to guarantee transmission within a defined time window is an essential feature of Ethernet TSN systems, and enables Ethernet to replace or to work alongside existing deterministic network technologies such as the CAN and FlexRay buses.
However, the implementation of Ethernet also provides automotive system designers with some unfamiliar implementation challenges. The familiar and well-understood bus communication technologies, CAN and FlexRay, are inherently robust and deterministic. This is because:
- The digital modulation schemes offer only binary signaling which offer higher signal-to-noise ratio (SNR) than pulse amplitude modulation. This means that they are highly immune to the many sources of electro-magnetic noise in vehicles because the differential of the bus will always be binary; and
- The operation of CAN and FlexRay buses is designed to ensure that collisions between two simultaneous signals are handled safely, and that priority is also maintained during collision arbitration. Each packet is acknowledged on the bus, instead of the way in which the Ethernet handles it, using the transmission control protocol (TCP) which employs timeouts for retransmissions.
In consequence, system designers developing modules, sub-systems or entire vehicles have been able to verify the performance of CAN or FlexRay communication system designs relatively quickly and easily, drawing on many years of development experience and using traditional approaches to embedded systems design. These earlier designs provided robust operation of CAN and FlexRay buses at the physical layer, benefiting in the case of large car manufacturers from a history of usage in multiple previous vehicle platforms. The development of a new communications bus for the physical layer will be a more robust version of Ethernet PAM encoding, but will require additional testing at the physical layer to ensure they do not fail because of interference radiated by or conducted from other systems in the vehicle.
As a result, automotive system suppliers and vehicle manufacturers will have to implement dedicated noise-impairment tests to verify the performance of the Ethernet bus network under the influence of electro-magnetic interference. The industry is currently rolling out this new standard to infotainment systems, and development of Ethernet-based safety-critical system designs is already underway.
Sensitivity of Ethernet Systems to External Noise
The behavior of a CAN or FlexRay bus is in marked contrast to that of Ethernet technology as used, for instance, for desktop computer networking. Ethernet is by origin a “best effort” technology. The Ethernet TSN extension overlays a set of deterministic behaviors on top of the basic Ethernet protocol.
This TSN extension also defines how the network handles communication failures and impairment. This is necessary because the Ethernet’s pulse amplitude modulation (PAM3) modulation scheme produces a much lower signal-to-noise ratio than a CAN bus or FlexRay bus enjoys. This makes Ethernet signals in a vehicle much more sensitive to noise events than signals carried over CAN or FlexRay connections. Impairment caused by noise will have safety implications if the network is not designed and tested properly.
Figure 1: The PAM3 modulation scheme used by Ethernet networking equipment
This is already true of existing vehicle system architectures which, for example, might have electronic connections between the brake pedal and the brakes or between the steering wheel and the steering rack. The increasing deployment of active safety systems will see a vast increase in the amount of safety-critical data traffic carried over vehicle networks. For instance, collision-avoidance systems continually pull information from radar proximity sensors at the front of the vehicle to an electronic control unit (ECU). When the ECU decides that it has come dangerously close to the vehicle in front, it can automatically apply the brakes. The network connections linking the sensors to the ECU and the ECU to the braking system are therefore part of a safety-critical system. Their operation must be known and predictable under all potential real-world noise conditions.
By specifying Ethernet network equipment in conformance to the appropriate Ethernet TSN standards, automotive OEMs will expect to be able to achieve the safe, known and predictable network behavior that safety regulations require.
But they will, of course, need to verify that their network design in practice achieves what they expected it to, even in the presence of multiple and severe noise sources. Fortunately, proven testing methods can provide the required level of verification. Since laboratory test equipment is used, manufacturers can capture essential technical data on the performance of Ethernet components and equipment at all stages of the product development process, from the earliest prototypes up to volume production units.
The Types of Noise Sources Which Affect Ethernet Systems
To test the effect of noise on an automotive Ethernet network, the system manufacturer must know what type of noise the network will be exposed to. Real-world noise measurements captured by Spirent in road tests illustrate the wide range of noise sources, and the types of noise that they generate. Automotive Ethernet equipment will need to measure the performance of a system when exposed to all these types of noise, and at a wide range of amplitudes.
The noise sources captured by Spirent’s road tests include those depicted in Figures 2a-2e, represented by a graph showing the amplitude of the noise signal in mV plotted over time. This series of graphs illustrates clearly the differences between the energy signatures of each type of noise source. The effect on an Ethernet network of, for instance, noise characterized by a series of low-energy pulses at short intervals, such as that generated by a running engine, might be quite different from the effect of a higher transient voltage, such as that generated by the horn. These differences raise design concerns for the high-speed analog PMA receiver input functions, as it relates to signal equalization and echo cancellations adaptive algorithms.
Figure 2a: Flashing the car’s headlights—a short-lived, high-energy burst of noise, followed by an elevated constant white noise floor
Figure 2b: Starting up the cabin ventilation fan—the noise starts up, briefly peaks then fades due to the fan’s electric motor drawing a high inrush current before settling at a steady operating current
Figure 2c: The internal combustion engine running—a repeating series of transient voltage events
Figure 2d: The internal combustion engine stops—a complex and random pattern involving many voltage spikes
Figure 2e: The car’s horn sounds—widely-spaced voltage peaks of extremely high amplitude
These noise effects might also differ depending on the nature of the traffic carried by the Ethernet network. For instance, transmissions of time-sensitive data from active safety equipment, such as the collision avoidance radar described previously, have very short windows within which they must travel between sender and receiver. What happens if an unwanted burst of noise energy happens during this short window? Will the systems recover? Will the systems use bad data, or fall back on a redundant signal path? How much noise will it take for the system to fall back onto the redundant system? How long will it take to qualify the link quality on the primary system before returning to it? These are questions that systems engineers must address to ensure a network operates in a safe and robust manner.
Noise Impairment Tests for the Automotive Industry
Developers and users of automotive Ethernet network equipment not only need to model the various noise sources found in vehicles, but also the varieties of impairment that can degrade or disable an Ethernet network’s operation. A comprehensive noise impairment test plan will reveal the robustness of a network when stressed at any of the physical, protocol or application layers. This calls for the implementation of the following types of test:
Link quality testing—While injecting noise into the physical layer, tests can measure the effect on the physical signals over an Ethernet link. The link quality may be rated numerically, in a range from 5 (very high quality) to 0 (poor quality). This test enables the user to build a model, for each type of noise, of the noise amplitude and duration at which the physical link will suffer some level of impairment.
Noise profile testing—As shown previously, the characteristics of different noise sources vary widely. By injecting noise from each source in turn, the user can model the effects on system performance of each noise type. This will show the effect on network bandwidth of the occurrence of noise spikes on the Ethernet interface generated by, for instance, a horn or a fan’s motor.
Packet loss testing—Interference from external noise can cause packets to be lost. The TSN extensions to the Ethernet protocol define the network’s behavior in the event of packet loss. By injecting noise into the network, the user can verify that these packet-loss behaviors are performed as specified in the standards.
Forced error testing—While packet loss testing is a form of protocol test, the user will also want to implement performance testing to verify that lost packets are reassembled correctly at switches or end nodes.
Golden sample testing—In volume production, golden sample testing provides confirmation that the manufactured unit is a functionally identical instance of the final product design in the form of a reference board. To do so, the user injects test signals and known noise samples into the production device under test, and measures the number of packets received. This number should match that received by the reference board. This method can help ensure that all reactive and non-reactive components are populated on a board in production.
Redundancy testing—The IEEE 802.1cb standards for the TSN extension to the Ethernet protocol implement a second redundant channel, and specify a zero switch-over time between the channels to give the user a seamless service even when the primary channel fails. In redundancy testing, the user deliberately overloads the primary channel with noise to the point at which it fails, and then measures the switch-over time to check that it is within the tolerance acceptable to the control system functionally. This test will also check that the output at the receiver node is fault-free and that no packets are dropped during the seamless switch-over process.
Arbitrary waveform testing—In noise profile testing described above, the user tests the network’s response to known automotive noise sources such as fans, lights and the engine. In the real world, it is possible that the network could be affected by noise characterized by an unpredictable or random profile. This phenomenon can be tested by means of an arbitrary waveform generator, which generates noise events which have a custom profile designed by the user.
All of the tests described here must be performed in the presence of a known noise source. So what is the best way to subject an automotive device under test to noise?
Repeatable and Reproducible Noise Generation for Automotive Ethernet Network Testing
A common, and essential, method for testing automotive systems is the drive test. This has the advantage of offering real-world conditions in which to test a vehicle’s systems. Whereas a simulation-based test is only as good as the reference models on which the simulations are based, drive tests put the device under test into exactly the operating conditions in which it will be used.
This makes drive testing an essential element in proving the reliability and safety of a complete automobile. But it has serious drawbacks for the testing of the performance of Ethernet network systems and components in the presence of noise. In a road test, the type, severity and duration of noise events cannot be precisely controlled and cannot be repeated at will. Furthermore, instances of noise cannot readily be isolated from one another; a certain level of background electro-magnetic noise will always be present, and multiple noise events may coincide. This makes it practically impossible to link the cause of adverse network performance or even network failure to a specific noise event.
Drive testing therefore fails to provide the controlled, repeatable and reproducible test conditions in which Ethernet equipment manufacturers can verify that their products meet the performance levels promised by the Ethernet TSN standard.
Automotive OEMs also commonly use EMC chambers for component and system testing. An EMC chamber provides a highly-controlled environment in which to measure the performance of Ethernet equipment under the influence of radiated or induced noise. However, this approach to noise impairment testing has two drawbacks:
- First, an EMC chamber provides an environment only for testing the effect of radiated noise. This involves a complicated test set-up, since it requires the tester to measure the distance of the radiator from the device under test, and its angle to the device, and to specify the antenna through which the emissions are radiated;
- Second, the use of an EMC chamber carries a very high hourly cost. In practice, this limits the availability of the test environment, inhibiting product developers’ freedom to test and refine prototypes.
By contrast, dedicated noise impairment test equipment for use in the laboratory or factory avoids all of these drawbacks. Unlike an EMC chamber, a noise impairment test instrument allows test set-ups to be quickly and easily arranged and modified. It also injects noise electrically into the device under test via reference cables and connectors, providing the user with precise control over the amplitude of the noise to which the device under test is subjected.
In addition, the use of a modern impairment noise generator means that:
- Real-world noise events can be captured and injected into the network in a controlled fashion. The amplitude of each noise event may be modulated to give the test engineer precise insight into the potential failure modes of network devices in the presence of various noise profiles. This also enables the engineer to model the effect of ageing on the noise output of components such as horns and electric motors. The harsh automotive operating environment, which is subject to temperature swings, contamination, vibration and shock, can markedly change the operating characteristics of vehicle components over time, and a sophisticated noise generator will enable the test engineer to model the changing profile of devices’ noise output over time;
- Network problems or failures caused by noise can be repeated and analyzed in precise detail in order to perform root-cause analysis;
- Arbitrary waveform can be easily generated and modified, to enable users to extend the test conditions beyond the set of recorded noise events, and to set benchmarks for the intended performance of the network;
- Severity and duration of any noise event can be precisely increased or reduced to identify the point at which Ethernet communication fails;
- Test process can be easily configured, modified and viewed via the noise generator’s PC-hosted GUI.
By using a dedicated noise impairment generator, automotive network component and equipment manufacturers can readily measure the performance of their products at both the development and production stages, pinpoint the timing and causes of problems, and verify their product’s conformance to the Ethernet TSN set of standards.
Conclusion
Electrical noise has the potential to corrupt Ethernet transmissions. How can you be sure your system design is robust against these noise phenomena? Will your network recover properly? Will your network behavior be acceptable for safety-critical applications?
Vehicles are inherently noisy environments. As the previous examples illustrate, devices found in every vehicle, such as horns, fans and motors, produce asynchronous noise spikes with varied noise peaks effecting the overall signal quality of the link. This asynchronous noise and varying voltage can negatively affect the performance of the adaptive front end receivers to perform at full line rate.
By designing their products in conformance to the specifications of the Ethernet TSN protocol, OEMs expect to ensure that safety and mission-critical transmissions always reach their destination. But until an Ethernet-based design is tested rigorously for its immunity to noise, there can be no guarantee that the system will work as designed in real-world conditions. Automotive system manufacturers should be asking important questions about the effect of noise:
- At which amplitude does a noise event cause an unacceptable level of impairment to Ethernet signals?
- How long can any given type of noise continue before signal impairment becomes unacceptable?
- Are there noise conditions in which switch-over to the redundant channel in an Ethernet TSN systems fails to occur seamlessly?
- Is it possible to verify the performance of an Ethernet system in the presence of multiple noise sources simultaneously?
Electrical noise is a fact of life in vehicles, and will always be present. Testing for it early in the development cycle will ensure a robust design to the unforeseen 11th hour production pitfalls that are typical when developing a new communications bus. The only way to be sure that an Ethernet-based system can withstand the effects of noise is to test it under various noise conditions that are known, repeatable and reproducible.
