Overcoming CAN Design Challenges: How to Easily Terminate CAN Signals
This article explores and provides answers to some common questions associated with CAN design and implementation challenges — in particular, signal termination in CAN systems.
Although a Controller Area Network (CAN) system seems like a common and simple interface, many questions and problems do arise during its design and implementation. Fortunately, many engineers have done the hard work for you. This article is the first installment in a new frequently asked questions article series designed to help you navigate the challenges of this popular interface standard.
This first installment of the series focuses on signal termination in CAN systems. CAN bus signals rely heavily on proper termination techniques to propagate signals to all CAN nodes in a network. Without correct termination, physically long conductors like the CAN bus can experience signal reflections – potentially limiting communication between all of its parts.
A quick note before diving in. While most applications involving CAN pertain to in-vehicle networks, these questions and answers also apply to industrial applications.
Question 1: How Many CAN Nodes are Allowed on a Bus and What are the Limiting Factors?
The maximum number of CAN nodes on a bus will depend heavily on system design requirements, but the main limiting factor is typically the bus input impedance on the CAN transceiver. The maximum number of nodes used to be 32, but with more modern CAN transceivers, this number is now much greater. That is because of the increased input impedance on the bus and the ability to limit leakage current to and from the bus.
When the bus has too many nodes, the transceivers can sink or source excessive current to or from the bus. When any individual node sends a message to the bus, it has to sink or source all of the leakage current on the bus, as well as drive the standard voltage signal across the termination resistance. One of two things will happen: the transceiver will output additional current, causing a reduction in the signal voltage that may push the voltage down below acceptable communication levels, with the current output limited in more extreme cases. Or the transceiver will go into thermal shutdown or become damaged while attempting to service the excessive current demand.
A modern CAN transceiver like the TCAN1042 has minimum 30-kΩ differential input impedance, adhering to the minimum 12-kΩ and maximum 100-kΩ specification requirements in International Organization for Standardization (ISO) 11898-2. The input impedance theoretically allows for 100 nodes on the bus: 100 transceivers with 30-kΩ input impedance in parallel is equivalent to 300 Ω, which in parallel with a 60-Ω bus impedance is equivalent to 50 Ω. With signal loss across the physical bus, ground offsets and parasitic loadings (along with other factors), the actual number of possible nodes tends to be lower.
Question 2: Is Termination Only Necessary at the CAN Bus End Nodes, or Do You Also Need to Place Termination Resistors at Each Node Between the End Points?
Termination at both end nodes of a CAN bus is a necessity. Without a 120-Ω termination at both ends, signal reflections caused by an impedance mismatch between the CAN bus and the driver will threaten the communication integrity.
Figure 1 shows a simple CAN bus topology with the end nodes terminated, while the in-between nodes have no termination. With nodes not at the end points but in between, termination isn’t required, but it will help if there’s a possibility that the stubs created from these nodes could affect signal integrity. Previous versions of ISO 11898 defined the maximum stub length in an automotive CAN bus as 0.3 m at 1 Mbps. This rule-of-thumb length is useful, but it doesn’t mean that you won’t see the effects of the stub even if you stay within these bounds.
Figure 1. Diagram showing simplified CAN bus topology with end nodes terminated.
If there are signal integrity issues, adding termination at these offshoot nodes helps attenuate some of the signal reflection caused by the impedance mismatch from the stub. This termination requires higher resistance than the standard 120 Ω so as to not reduce the effective bus impedance. If the bus impedance doesn’t meet the requirement for the CAN transceiver, the CAN driver will not be able to drive the correct voltage levels. Typical values for these termination resistances are in the 1kΩ to 2kΩ range.
Question 3: Is a Common-Mode Choke Required for a CAN Transceiver to Function Correctly? Why is the Common-Mode Choke Used?
A common-mode choke is not required for a CAN transceiver to translate signals from the controller to the CAN bus. Although the majority of automotive CAN applications – such as advanced driver assistance systems (ADAS), gateway and infotainment– have a common-mode choke in place on their printed circuit board (PCB), having one installed on the PCB is not required for proper CAN transceiver functionality. The common-mode choke does, however, help with CAN transceiver electromagnetic emissions, while also improving the transceiver’s electromagnetic immunity against high-frequency noise coming from the CAN bus.
With differential signals, there are components of differential- and common-mode noise that make up the overall electromagnetic emissions profile from the CAN transceiver. Differential signals are designed to be equal and opposite in magnitude, so the noise emitted from these signals is also equal and opposite. As a result, the differential noise cancels itself out – for the most part.
It’s not as easy to deal with the common-mode portion of the noise; this is where the common-mode choke comes in. Acting like a filter inductor to choke out high-frequency noise related to changing current, the common-mode choke creates a magnetic field that opposes the common-mode noise in either direction: the noise coming from the CAN transceiver as emissions, or the noise coming from the CAN bus into the CAN transceiver externally.
So while the common-mode choke isn’t a requirement for CAN transceiver functionality, it helps in environments where susceptibility and sensitivity to electromagnetic noise is an issue. These noisy environments can be in any part of the vehicle, but typically exist in the automotive gateway, where several communication interfaces and their buses connect into one system.
Question 4: What is the Function of the SPLIT Pin on a CAN Transceiver?
The SPLIT pin’s function is to drive a stronger recessive voltage level to the CAN bus, minimizing and thus lowering the potential for common-mode emitted noise. This pin is connected between two 60-Ω resistors in a split termination scheme. Because the CAN symmetry of the bus is not optimal with older design processes, there is more deviation in the common-mode signal. Older CAN transceivers have the SPLIT pin to help with that deviation and thus mitigate electromagnetic emissions.
While the SPLIT pin may not be necessary, it is still beneficial to use split termination to help with electromagnetic interference. A split termination creates a low-pass filter for both the CANH and CANL lines, mitigating much of the high-frequency noise that the transceiver emits to the bus. Figure 2 illustrates a standard termination vs. a split termination scheme. Typical values for CSPLIT range from 4.7 nF to 100 nF.
Figure 2. Comparison diagram of standard vs. split termination scheme in a CAN transceiver.
Newer CAN transceivers like TI’s TCAN4550, TCAN1042-Q1 and TCAN1043-Q1 don’t need the SPLIT pin to function correctly. TI designed these devices with modern processes to help increase symmetry for CANH and CANL signals. As a result, there is little deviation in the differential signals, which helps lower common-mode emissions.
Correctly Terminating the CAN Bus is Necessary for Successful Communication
Whether the goal is to help with electromagnetic emissions performance, reduce or remove signal reflections through impedance matching, or determine the size and amount of nodes for your CAN system, correctly terminating the CAN bus is crucial for a successful communication scheme. Having answers to these questions should help get you started on that path.
The second installment of this series will explore how to measure CAN power consumption for multiple types of CAN transceivers, as well as why and how 3.3V power rails are used in CAN systems.
Industry Articles are a form of content that allows industry partners to share useful news, messages, and technology with All About Circuits readers in a way editorial content is not well suited to. All Industry Articles are subject to strict editorial guidelines with the intention of offering readers useful news, technical expertise, or stories. The viewpoints and opinions expressed in Industry Articles are those of the partner and not necessarily those of All About Circuits or its writers.
I’ve been working with CAN bus starting in 1992. Years ago I had a discussion with Steve Corrigan at TI with respect to CAN bus grounding. He designed some of the CAN devices TI supplies. I had set up an experiment with two battery powered transceivers and a pulse generator into one and a scope on the other. There was no physical ground connection between the two. I asked Steve, as the designer of the devices, why this even worked. His explanation was that although the receivers do behave as differential receivers and can determine dominant and recessive signals the noise immunity was very poor and the potential of common mode noise between the two nodes electrical grounds can easily exceed the device specifications. I always insist that that CAN bus systems include the ground. In other words it’s a three wire network; not two wire. Isolating the CAN signal ground is also critical if the DC electrical ground carries large amounts of current for motors or inverters. Your articles should probably also show how to wire the CAN grounds between nodes when the common DC ground connection between those nodes carries noisy 40A current resulting in spikes on the DC return bus of more what the device is rated for. And an optically isolated CAN driver is not possible. Here’s a link to a CAN system with CAN bridges and two controllers for a total of more than 1500 nodes. http://www.autoartisans.com/rings/Barge1a.jpg Although we used Thick and Thin DeviceNet cable we saw as much as a 7V drop across our 500kbps network power. On both the 48V Supply side and also back on the ground return.