In-car networking technology
Using Ethernet in advanced automotive systems
There are a broad variety of data busses available for use in today’s automotive communication infrastructure: from the cost efficient dedicated control message systems, such as LIN, CAN and FlexRay, up to optimized media streaming networks like MOST. In addition, specific solutions for high speed peer to peer connections in display and camera applications are available. So, why would we need Ethernet in a car at all?
Advanced driver assistant systems (ADAS) with cameras, real time image processing, and many other high speed sensors are major trends for the near future. Furthermore, there is a rising demand for integration of smart phones and other consumer mobile devices within the infotainment systems. Future vehicles will be connected to all types of wireless data services including new applications like car to car communication. Cost effective system solutions will be required to be able to compensate for the resulting massive increase in complexity and number of nodes of the car internal network. Package oriented switched data transmission protocols can provide the necessary flexibility and scalability for such applications. The development of specific physical layers for automotive Ethernet based on single unshielded twisted pair (UTP) cabling for 100Mbit/s, and possibly even for 1Gbit/s in the future, addresses this demand. From the cable harness point of view, there are specific automotive grade interconnection solutions available. These solutions provide excellent shielding and signal integrity performance up to a couple of gigahertz, but are also expensive. With regards to a cost effective system solution, it is necessary to determine whether automotive standard terminals with tab sizes in the range of 0.5 - 0.64mm are similarly suitable. This can be done by analyzing the signal integrity of the transmission channel; i.e., insertion loss and return loss. Additionally, parameters for mode conversion and cross-talk to adjacent pairs are of interest with regards to the electromagnetic compatibility (EMC) in an unshielded differential system. Measurements and simulations with cable assemblies constructed using standard components show promising results (Figure 1). The results are well within the limits for one pair automotive 100Mbit/s Ethernet as defined by the OPEN Alliance Special Interest group. Additionally, the proposed limits for the IEEE802.3bp 1Gbit/s solution can be met.
Signal integrity within the link
The link insertion loss is dominated by the frequency dependent cable attenuation. Use of low loss cable types is recommended for this reason especially for 1Gbit/s applications. The short connector structure has only a minor contribution to the system loss. This is different for the return loss since signal reflections are caused by impedance discontinuities. The magnitude of the reflected signal depends on the length of the discontinuity and on the mismatch between cable and connector impedance. Figure 2 shows the impedance profile of a contact pair with 0.64mm pin size. A tolerance of ±10 ohms can be met even for signal rise times of 350ps corresponding to 1GHz frequency bandwidth. For 100Mbit/s Ethernet signals with frequency bandwidths below 100MHz, the discontinuity is not noticeable.
The untwisted area of the cable (directly behind the contact crimp zone) is another potential cause of impedance discontinuities (see Figure 3). The length of this area depends on the actual cable type and termination process. The general recommendation is to keep the length of twist disruption as short as possible. Tolerance analyses can be done by modelling this zone as transmission lines using length and impedance as parameters. Worst case topologies for return loss are short links with low loss cable and a maximum number of discontinuities, i.e. number of inline connectors. Furthermore, all cable segments should have the same length and a characteristic impedance at the tolerance maximum or minimum, respectively. Figure 3 shows the link return loss for such scenarios with different termination zone parameters. The limit for 100Mbit/s data rate can be met with a termination zone length of 20mm at all contacts measured from the connector housing end with an assumed impedance of 150 ohms (see Figure 3, case 4). Making cable terminations with such values is feasible with a certain degree of process control. The ‘worst case' analysis also demonstrates that in general, Gigabit performance can be reached with those contact systems; but improvements regarding tolerance would be required. This is shown in the statistical analysis in Figure 4.
Inference with environment and within harness – EMI emission & immunity
The electromagnetic compatibility in an unshielded system is determined by the balance of the differential pair. Imbalances caused by different wire lengths or asymmetric coupling to an adjacent ground, may cause partial conversion of the differential mode data transmission energy to common mode signals relative to ground. Neighboring lines in the harness or RF antenna applications in the car can be disturbed by this effect. The Ethernet system itself is immune to common mode noise coupled from external sources into the twisted pair due to filter circuits and differential signal inputs at the transceivers. However, the data transmission might be affected if part of the common mode noise is converted to differential signals by such harness imbalances. The mode conversion limits for link segment and components proposed by the IEEE802.3bp and OPEN Alliance working groups are based on automotive component level test scenarios and appropriate EMC models. Figure 5 illustrates the effect of an imbalance on a two position automotive connector. There is a remarkable increase of mode conversion, especially for higher frequencies, due to the asymmetric coupling to ground by a metal block close to the contact pair. These effects may also occur in the wire termination zone, which is another reason to minimize and shorten untwist length. The proposed limits for 100Mbit/s Ethernet can still be met with existing designs, but connector design and cable optimization will be required for use at 1Gbit/s in the future.
Cross-talk between adjacent pairs caused by coupling of electrical and magnetic fields from one pair to the other is another important EMC parameter. Mode conversion can occur if the coupling effects are asymmetric, i.e. a common mode disturbance on one pair might be converted to a differential mode noise on the second pair. These cross-talk effects can be controlled by appropriate pair to pair arrangement in multi pin connectors.
Summary & conclusion
All these analyses demonstrate the suitability of miniaturized automotive connector systems with pin sizes in the range of 0.5 - 0.64mm for 100Mbit/s Ethernet applications. Established harness manufacturing processes can further be used if the cable termination zone is controlled regarding length, impedance and symmetry. System modelling methods can be used to perform tolerance analyses and to determine geometric limits. While 100Mbit/s Ethernet technology based on single UTP is being introduced in mass production cars (figure 6), the challenges for 1Gbit/s are much higher. A high degree of balance is essential to meet automotive EMC requirements with an unshielded system at this data rate. The discussed contact systems are suitable for these requirements, but optimized connector designs, cables and termination processes will be necessary.
Finally, it should be mentioned that the independence of a specific physical layer is a major advantage of Ethernet. Alternative solutions, such as coaxial cabling or Plastic Optic Fiber (POF), could be an interesting approach for application where UTP is not sufficient, e.g. for EMC reasons at high data rates in very sensitive areas.