2018 IEEE Vehicular Networking Conference (VNC)

Ringing Mitigation Schemes

for Controller Area Network
Daisuke Umehara Takeyuki Shishido
Faculty of Information and Human Sciences Graduate School of Science and Technology
Kyoto Institute of Technology Kyoto Institute of Technology
Kyoto 606-8585, Japan Kyoto 606-8585, Japan
Email: [email protected]

Abstract—Controller area network (CAN) is widely distributed audio, video, voice, and data signals via POF or electrical
as an in-vehicle communications standard. CAN with flexible conductor. 100-Mbps Ethernet (100BASE-T1) and 1-Gbps
data-rate (CAN FD) is defined in the ISO standards to achieve Ethernet (1000BASE-T1) were standardized in IEEE 802.3bw
higher data rate than the legacy CAN. A number of CAN nodes
are able to connect with a single transmission medium, i.e. CAN and IEEE 802.3bp, respectively. Also a power over data
enables us to constitute cost-effective bus-topology networks. line (PoDL) for Ethernet was standardized in IEEE 802.3bu.
CAN puts carrier sense multiple access with collision resolution These Ethernets are designed for automobile point-to-point
(CSMA/CR) into practice by using bit-wise arbitration based full-duplex networks via a single twisted pair cable to satisfy
on wired logical AND in the physical layer. The prioritized strict requirements with respect to the reliability, safety, and
message is delivered without interruption if two or more CAN
nodes transmit messages at the same time thanks to the bit-wise electo-magnetic compatibility (EMC) in cars [5]. However,
arbitration. However, the scalability of CAN networks suffers they need some high-cost switching hubs and additional wiring
from ringing caused by the signaling mechanism establishing to accommodate three or more nodes. 100 or more ECUs and
the wired logical AND. We need to reduce networking materials sensors are connected via several CAN buses in a part of
in order to reduce the car weight, save the fuel and the cost, classy cars. More and more CAN devices and buses are needed
and develop a sustainable society by establishing more scalable
CAN networks. In this paper, we show the reduced wiring as the driving assistance and autonomous driving technology
technology for CAN to enhance the network scalability and the has been developing. However, we need to reduce networking
cost efficiency. materials in order to reduce the car weight, save the fuel and
the cost, and develop a sustainable society. Therefore, we need
I. I NTRODUCTION to develop the CAN technology to establish more scalable
Controller area network (CAN) is a message-based multi- CAN networks.
drop network protocol which was designed for reliable in- A message has the identifier (ID) in a CAN bus and all
vehicle control networks via electrical twisted pair cable. CAN messages are fully prioritized with their IDs. A message is
has been originally developed by Robert Bosch GmbH in 1983 broadcast to all receivers on the CAN bus from a sender
and has been extended to CAN with flexible data-rate (CAN and the message is taken in the receivers which need it
FD) by Bosch in 2012. CAN and CAN FD have been defined by acceptance filtering. Any CAN device does not have its
in the ISO 11898 standards [1] and for the implementation of address so that the CAN network is fully different from
a vehicle communication network in the SAE J2284 standards the address-based communication networks. CAN employs
with 500 kbps, 2 Mbps [2], and 5 Mbps [3]. The CAN FD a carrier-sense-based access mechanism, called carrier-sense
devices coexist on the same network with the legacy CAN multiple access with collision resolution (CSMA/CR), and
devices as long as the CAN FD devices do not use their all the nodes on a CAN bus including the sending nodes
extending functions. A cost-effective serial network protocol, listen to the bus voltage. While nodes are transmitting, any
local interconnect network (LIN), is also widely utilized for other nodes do not transmit their messages. If two or more
some automobile components via single wiring because CAN nodes transmit their messages at the same time, the bus access
costs too high to implement for every component. FlexRay conflict is resolved by content-based arbitration with the fully
was originally developed as an automotive control protocol prioritized IDs. The message with the minimum ID survives
via plastic optical fiber (POF) or electrical twisted pair cable the other simultaneous messages so that the mechanism of
by FlexRay Consortium but it costs too high to implement arbitration ensures that neither information nor time is lost
for popular cars except a part of classy cars. As a result, unlike CSMA with collision detection (CSMA/CD) defined for
CAN and LIN are mainstreams to develop automotive control the half-duplex Ethernet, and CSMA with collision avoidance
networks [4] for a number of electronic control units (ECUs) (CSMA/CA) defined for IEEE 802.11 and IEEE 802.15.4. The
and sensors. Media oriented systems transport (MOST) is a wired logical AND and the in-bit-time detection in the physical
high-rate multimedia network protocol. The MOST network layer establish the mechanism of bit-wise arbitration in the
employs daisy-chain topology or ring topology and transports medium access control (MAC) layer. The unipolar non-return-

978-1-5386-9428-2/18/$31.00 ©2018 IEEE

 2018 IEEE Vehicular Networking Conference (VNC)

to-zero (NRZ) line code is employed for the CAN signaling.

A dominant pulse with high differential voltage transmits a
bit “0” and a recessive pulse with zero differential voltage
transmits a bit “1.” The dominant pulse survives the recessive
pulse on the shared CAN medium so that the wired logical
AND is established. If a node transmitting the recessive pulse
in the arbitration field detects a dominant pulse in the bit
Fig. 1. Multiple on-board node configuration defined in SAE J2284-4 [2]
time on the CAN bus, the node terminates transmitting its with CAN FD data rate at 2 Mbps.
message. A bit is decided to be “0” or “1” by the hysteresis
comparator at the sample point in the bit time. However, if
the propagation round-trip time exceeds the sample point, the SAE J2284-4 [2]. The ringing may make us divide a large-
mechanism of bit-wise arbitration will be destroyed. The cable size CAN bus into a multiple of small-size CAN buses, i.e.
length and bit rate are restricted by establishing the in-bit- it may cause increase of communication wiring. We have
time detection. Cena and Valenzano originally proposed the tackled the reducing wiring problems for CAN networks.
concept of CAN FD in 1999 to enhance the data rate of the We proposed short pulse CAN (SP-CAN) in [16] and silent
original CAN [6]. Ziermann et al. developed CAN+ protocol, interval impedance matching (SIIM) for SP-CAN to eliminate
in which high-rate data are embedded in a dominant pulse, the ringing effects [17]. However, the data rate is limited
to achieve higher data rate than the Cenas’ extension [7]. due to shortening the dominant pulse width as compared
However the high-rate version of CAN had not been appeared with the CAN FD signaling. Mori et al. proposed a method
for a long time in the market until CAN FD was developed in where the node itself detects the transition from a dominant
2012 owing to developing higher functionality for ECUs and to a recessive based on the received voltage and reduces its
sensors. A CAN FD data frame is divided into the arbitration impedance for the first half of the recessive period in order to
phase, in which multiple nodes may transmit, and the data reduce the ringing duration [18]. The method has established
phase, in which a single node transmits. The data in the the communication up to 2 Mbps on a 8-node CAN network
arbitration phase are transmitted with the original CAN data but the communication with 4 Mbps would be failed due to the
rate and the data in the data phase are transmitted with a remaining ringing waves. We proposed a bit detection scheme
higher data rate [8]. Also the amount of data in the data phase with logical OR at multiple sample points taking the alternat-
is getting larger to enhance the data rate of CAN FD. Several ing property of ringing into consideration [19]. However, the
new research topics have appeared due to the development of logical OR detection can enhance the communication reliabil-
CAN FD. Schreiner et al. proved the robustness and reliability ity in a probabilistic manner not a deterministic manner. In this
of CAN FD from an OEM’s point of view [9]. Adamson paper, we propose three different approaches to reduce CAN
discussed a hybrid network configuration with CAN and CAN communication wiring. The CAN bus accommodates more
FD devices [10]. The higher-rate transmission with CAN FD CAN nodes by mitigating the ringing effects. Accommodating
cannot be exploited in networks where legacy CAN devices are more CAN nodes enables us to reduce the number of CAN
connected to. Cena et al. considered the improvement of the buses in a car. We discuss the advantages and disadvantages
compatibility between CAN FD and legacy CAN devices [11]. of those proposed methods through simulation results obtained
Lennnartsson developed a CAN FD filter for legacy CAN from a VHDL-AMS simulator, SaberRD.
controllers to make it possible to receive CAN FD frames [12]. The remainder is organized as follows: In Section II, we
The control commands and sensing data transmitted through describe the physical and data link layers of CAN FD. In
the CAN networks have relatively smaller size as compared Section III, we present a triple-star network to discuss the
with the maximum data amount in the extended data field advantages and disadvantages of the proposed schemes. We
of CAN FD. Garnatz and Decker proposed that a CAN FD propose three ringing mitigation schemes to reduce the com-
data frame carries a multiple number of protocol data units munication wiring. In Section IV, we summarize the results
(PDUs) [13]. Cena et al. proposed an extension of CAN FD to reduce CAN wiring and we discuss the future direction of
protocol, called CAN with extensible in-frame reply (CAN this research topic.
XR), in which multiple nodes share the data field in a CAN FD
frame [14]. The CAN XR protocol is similar to time-triggered II. C ONTROLLER A REA N ETWORK
protocols such as the FlexRay protocol and the time-triggered In this section, we describe the physical and data link layers
CAN (TTCAN) protocol. of CAN FD based on the SAE standard [2] and the ISO
Although CAN FD was originally proposed without modify- standards [1].
ing the network topology of CAN, the data rate and network
topology for CAN FD are restricted by ringing occurred in A. Physical Layer
the recessive pulse after a dominant pulse [15]. Actually, The CAN protocol is designed for multi-drop networks
CAN FD with the data rate of 5 Mbps is restricted to point- comprised of a number of nodes as illustrated in Fig. 1. The
to-point networks in SAE J2284-5 [3] although CAN with network has two terminated nodes with a resistor of RT =
the data rate of 2 Mbps is applied for 24-node networks in 120 Ω to match the characteristic impedance Z0 = 120 Ω of
 2018 IEEE Vehicular Networking Conference (VNC)

Fig. 2. The bit format for CAN FD.

Fig. 4. The terminated node structure and the signaling for CAN FD.
Fig. 3. The FD base frame format (FBFF).

extended frame formats have 29-bit ID. In this paper, we deal

a shielded twisted pair (STP) cable. The transmission medium
with FBFF illustrated in Fig. 3. The number in the bracket
is shared with a multiple of nodes so that the CAN networking
stands for the number of bits. The region without the bracketed
provides a cost-effective solution for in-vehicle networks.
number consists of a single bit. The former arbitration phase
A unipolar NRZ line code is employed to transmit binary
is ranged from the start of frame (SOF) consisting of a single
data via a CAN bus. Both of conductors have 2.5 V, i.e. the
dominant pulse to the sample point in the bit rate switch (BRS)
differential voltage VD = 0 V if all the nodes are in the
consisting of a single recessive pulse. The latter arbitration
reception mode. A node drives one conductor, CAN H, to
phase is ranged from the sample point in the CRC delimiter
be 3.5 V and another conductor, CAN L, to be 1.5 V, i.e. the
consisting of a single recessive pulse to the end of frame
differential voltage VD = 2 V if the node transmits a dominant
(EOF). The data phase is ranged from the sample point in
pulse of bit “0.” The impedance of the node becomes about
BRS to the sample point in the CRC delimiter. A single node
RT = 120 Ω during driving the dominant pulse. The node is
survives after the former arbitration phase so that only the node
in the reception mode if a node transmits a recessive pulse
transmits in the data phase. The ACK bits are transmitted from
of bit “1.” The impedance is RR = 30 kΩ for the node in
receivers successfully receiving the data in the latter arbitration
the reception mode. As a result, ringing tends to occur in the
phase. As a result, multiple nodes may transmit at the same
recessive pulse after a dominant pulse when a larger number
time in the arbitration phase with a lower data rate and a single
of nodes are connected and the data rate gets higher.
node transmits in the data phase with a higher data rate.
Figure 2 illustrates the bit format of CAN FD. A bit is
The bit detection will be reliable in the lower-rate arbitration
composed of 9 time quanta where Tq stands for a quantum
phase because the ringing effect is relatively smaller due to
time so that the bit time is expressed as Tb = 9Tq . The
the longer pulse width. However the bit detection error due
synchronization segment (SS) is composed of 1 time quantum
to ringing in the data phase still destroys the CAN system.
and the pulse edge is expected to be detected within this
If a node sending a bit in the data phase detects a mismatch
segment. The phase buffer segments 1 and 2 (PBS1 and PBS2)
between the transmitting value and the bus level, the node
are used to compensate for edge phase errors. These segments
terminates sending the data frame and then starts transmitting
are lengthened or shortened by resynchronization. The bus
the error frame. If a node receiving a data frame suffers from
level is read at the sample point, the location of which is at the
intensive ringing, the node may fail to pass the CRC and then
end of PBS1, and the respective bit is detected. Bit stuffing is
start transmitting the error frame. Therefore, we have to avoid
employed in the CAN protocol to avoid slipping the bit timing.
any bit detection error caused by ringing in the data phase.
A bit of opposite value is inserted after 5 consecutive bits of
the same value. The threshold of differential voltage transited III. R EDUCED W IRING T ECHNOLOGY
from recessive to dominant is 0.9 V whereas the threshold
A. Simulation Configuration
transited from dominant to recessive is 0.5 V.
We employ a VHDL-AMS simulator, SaberRD, to evaluate
B. Data Link Layer the proposed and original CAN FD systems including analog
The frame format is categorized into classical base frame and mixed-signal (AMS). Figure 4 illustrates the basic struc-
format (CBFF), classical extended frame format (CEFF), FD ture of a terminated node and an example of the CAN FD
base frame format (FBFF), and FD extended frame format signaling. The HS CAN (High Speed CAN) generic model
(FEFF). The base frame formats have 11-bit ID whereas the is used as the CAN transceiver in SaberRD simulations. A
 2018 IEEE Vehicular Networking Conference (VNC)

TABLE I
L INE PARAMETERS PER UNIT LENGTH ON S ABER RD.

Name Value
DC inductance, LH and LL 325 nH/m
DC mutual inductance, LHL 8.72049 nH/m
DC capacitance, CHG and CLG 87.4 pF/m
DC capacitance, CHL 3.11211 pF/m
DC resistance for conductors 50√mΩ/m
Skin-effect resistances for conductors 180 µΩ/m/ MHz

Fig. 6. A triple-star network for the SaberRD simulations.

TABLE II
T HE CHANNEL NUMBER CORRESPONDING TO THE PAIR OF THE TX AND
RX NODES

Channel TX node RX node

Fig. 5. The STP cable model on SaberRD.
Channel 1 Node 1 Node 1
Channel 2 Node 1 Node 2
Channel 3 Node 1 Node 3
Channel 4 Node 1 Node 5
non-terminated node removes the termination composed of RT Channel 5 Node 1 Node 7
and CT shown in Fig. 4. Table I shows the main parameters Channel 6 Node 3 Node 1
of a STP cable model on SaberRD illustrated in Fig. 5. The Channel 7 Node 3 Node 2
Channel 8 Node 3 Node 3
characteristic impedance is approximately derived as Channel 9 Node 3 Node 4
√ Channel 10 Node 3 Node 5
LH + LL + LHL Channel 11 Node 3 Node 7
Z0 ≈ −1 −1 −1 ≈ 119.4 Ω. (1) Channel 12 Node 5 Node 1
CHL + (CHG + CLG )
Channel 13 Node 5 Node 3
Channel 14 Node 5 Node 5
A triple-star network illustrated in Fig. 6 is commonly used Channel 15 Node 5 Node 6
to evaluate the proposed schemes in the SaberRD simulations.
There are 15 different channels on the triple-star network as
numbered in Table II. The received waveforms for two bit
sequences are observed in the simulations. One is BR = (CAN/IMDP) in this paper. In CAN/IMDP, the impedance of
010101010111 by which we can analyze ringing waves in the the transmit node is electrically matched to the characteristic
last three recessives and the other is BD = 010101010000 by impedance of the STP cable while transmitting a recessive
which we can analyze ringing waves in the last four dominants, pulse within the data phase by using a switch. Figure 7
following alternating bit sequence with high frequency. The illustrates the non-terminated node structure and an example
data rate is ranged from 1 Mbps to 10 Mbps every 1 Mbps of the signaling for the proposed CAN/IMDP. In the data
to evaluate the ringing mitigation schemes. Notice that the phase, the logical signal TXD1 normally inputs to the CAN
triple-star network has 32 nodes. The number of nodes exceeds transceiver whereas the logical signal TXD2 inversely inputs
24 which is the permissible number of nodes in the SAE to the switch. The transmit node does not reflect the receiving
standard with CAN FD data rate at 2 Mbps [2]. The received waves in the data phase because its impedance keeps up
waveforms are bit-synchronized with the timing exceeding RT = 120 Ω during the data phase as shown in Fig. 7.
0.9 V at the rising edge of the first dominant pulse and Figure 8 illustrates the received waveforms of the bit
the received bits are detected from the received waveforms sequence BR = 010101010111 by CAN FD and CAN/IMDP
assuming an ideal hysteresis comparator by off-line analysis. via Channel 8 between TX node 3 and RX node 3 with the data
The success ratio is defined as the percentage of successes rate of Tb−1 = 8 Mbps, where the sample points are drawn by
among 30 communication conditions composed of 15 channels circles. We observe high-amplitude and long-duration ringing
and two 12-bit sequences for a given bit rate over the triple- for CAN FD and the ringing effect for CAN/IMDP does not
star CAN network. Our aim is to discuss the advantages and affect the bit detection while transmitting the bit sequence
disadvantages for the proposed schemes through the received of BR . The proposed CAN/IMDP has the effect of low-
waveforms obtained from the SaberRD simulations not to pass filter (LPF) for the ringing observed in the CAN FD
eliminate the ringing effects for any given network conditions. signaling as shown in Fig. 8. The 11th received bit in CAN
FD is incorrectly detected as dominant whereas the received bit
B. Reduced Communication Wiring sequence in CAN/IMDP is correctly detected. It is interesting
1) Impedance Matching in the Data Phase: We pro- to pass the one-bit later recessive bit and fail the two-bit
pose CAN with impedance matching in the data phase later recessive bit following a dominant pulse in CAN FD.
 2018 IEEE Vehicular Networking Conference (VNC)

2
1

VD [V]
0
-1
-2
0 1 2 3 4 5 6 7 8 9 10 11 12
Nomalized bit time [1/Tb]
TX: CAN FD RX: CAN FD

2
1

VD [V]
0
-1
-2
0 1 2 3 4 5 6 7 8 9 10 11 12
Nomalized bit time [1/Tb]
TX: CAN/IMDP RX: CAN/IMDP

Fig. 9. The received waveforms of the bit sequence BR by CAN FD and

CAN/IMDP via Channel 14 between TX node 5 and RX node 5 with the
data rate of Tb−1 = 6 Mbps.

100

Fig. 7. The non-terminated node structure and the signaling for CAN/IMDP.
80

Success ratio [%]

60
2
1
VD [V]

0 40
-1
-2
20
0 1 2 3 4 5 6 7 8 9 10 11 12
Nomalized bit time [1/Tb]
TX: CAN FD RX: CAN FD 0
1 2 3 4 5 6 7 8 9 10
Data-phase bit rate [Mbps]
2 CAN FD CAN/IMDP
1
VD [V]

0
-1 Fig. 10. The success ratio versus the bit rate in the data phase for CAN FD
-2
and CAN/IMDP.
0 1 2 3 4 5 6 7 8 9 10 11 12
Nomalized bit time [1/Tb]
TX: CAN/IMDP RX: CAN/IMDP
as 000000000011. The proposed CAN/IMDP suffers more bit
Fig. 8. The received waveforms of the bit sequence BR by CAN FD and errors than the original CAN FD in the case of Channel 14
CAN/IMDP via Channel 8 between TX node 3 and RX node 3 with the data and the date rate 6 Mbps.
rate of Tb−1 = 8 Mbps.
Figure 10 illustrates the success ratio versus the bit rate in
the data phase for CAN FD and CAN/IMDP. The proposed
CAN/IMDP achieves the same or higher success ratios than
We need to analyze ringing within consecutive recessives and the original CAN FD for all the bit rates. The proposed
dominants. CAN/IMDP devices can coexist on the network with the
Figure 9 illustrates the received waveforms of the bit existing CAN FD devices because the transmit waveforms
sequence BR = 010101010111 by CAN FD and CAN/IMDP of CAN/IMDP devices are the same as those of CAN FD
via Channel 14 between TX node 5 and RX node 5 with the devices. Replacing an unreliable CAN FD device with a
data rate of Tb−1 = 6 Mbps, where the sample points are CAN/IMDP device can enhance the reliability of the existing
drawn by circles. We can see the effect of LPF for the ringing CAN networks due to reducing self-reflection waves at the
observed in the CAN FD signaling by using the proposed node.
CAN/IMDP. However, this ringing in the CAN FD signaling 2) Bipolar Coding in the Data Phase: We propose CAN
has a large low-frequency component so that the bit detection with bipolar coding in the data phase (CAN/BCDP) as an
is destructively affected by the low-frequency component extension of CAN/IMDP in this paper. Figure 11 illustrates the
for the proposed CAN/IMDP although the oscillation of the non-terminated node structure and an example of the signaling
ringing is mitigated by the impedance matching. The 6th and for the proposed CAN/BCDP. CAN/BCDP employs two CAN
10th received bits in CAN FD are incorrectly detected whereas transceivers as shown in Fig. 11 and this architecture provides
the received bit sequence in CAN/IMDP is incorrectly detected us a bipolar NRZ line code. The CAN transceiver 1 is normally
 2018 IEEE Vehicular Networking Conference (VNC)

2
1

VD [V]
0
-1
-2
0 1 2 3 4 5 6 7 8 9 10 11 12
Nomalized bit time [1/Tb]
TX: CAN FD RX: CAN FD

2
1

VD [V]
0
-1
-2
0 1 2 3 4 5 6 7 8 9 10 11 12
Nomalized bit time [1/Tb]
TX: CAN/BCDP RX: CAN/BCDP

Fig. 12. The received waveforms of the bit sequence BR by CAN FD and
CAN/BCDP via Channel 14 between TX node 5 and RX node 5 with the
data rate of Tb−1 = 6 Mbps.

2
1

VD [V]
0
-1
-2
0 1 2 3 4 5 6 7 8 9 10 11 12
Fig. 11. The non-terminated node structure and the signaling for CAN/BCDP. Nomalized bit time [1/Tb]
TX: CAN FD RX: CAN FD

2
1
VD [V]

connected to the CAN bus whereas the CAN transceiver 2 is 0

inversely connected to the CAN bus. In the data phase, the -1
-2
logical signal TXD1 normally inputs to the CAN transceiver
0 1 2 3 4 5 6 7 8 9 10 11 12
1 whereas the logical signal TXD2 inversely inputs to the Nomalized bit time [1/Tb]
CAN transceiver 2. The CAN/BCDP node drives CAN H of TX: CAN/BCDP RX: CAN/BCDP
3.5 V and CAN L of 1.5 V for a dominant pulse and the
CAN H of 1.5 V and CAN L of 3.5 V for a recessive pulse in Fig. 13. The received waveforms of the bit sequence BR by CAN FD and
the data phase. Also the CAN/BCDP node has the impedance CAN/BCDP via Channel 14 between TX node 5 and RX node 5 with the
data rate of Tb−1 = 5 Mbps.
of RT = 120 Ω in the data phase so that the self-ringing
can be reduced as with the CAN/IMDP node. The output
logical signal RXD1 of the CAN transceiver is utilized for
the detection at the sample point. the differential voltage VD drops more deeply, the sample
Figure 12 illustrates the received waveforms of the bit points for received dominant pulses are under the threshold
sequence BR = 010101010111 by CAN FD and CAN/BCDP of 0.5 V, and the received bits corresponding to dominant
via Channel 14 between TX node 5 and RX node 5 with pulses are not correctly detected. The combination of RXD1
the data rate of Tb−1 = 6 Mbps, where the sample points and RXD2 makes it possible to establish a virtual hysteresis
are drawn by circles. The combination of Channel 14 and comparator, the voltage threshold of which transited from
the data rate Tb−1 = 6 Mbps makes it difficult to establish recessive to dominant is 0.9 V and the voltage threshold of
reliable communication for CAN/IMDP whereas the received which transited from dominant to recessive is −0.9 V. The
bit sequence in CAN/BCDP is correctly detected. In the received bit sequence with the dual hysteresis comparator is
proposed CAN/BCDP, in addition to the LPF effect like correctly detected for the bipolar NRZ signaling even in the
CAN/IMDP, the differential voltage VD moves to the region case of Channel 14 and the data rate Tb−1 = 5 Mbps.
of minus voltage so that the bit detection is quite reliable as Figure 14 illustrates the success ratio versus the bit rate in
compared with the other schemes. the data phase for CAN FD, CAN/BCDP, and CAN/BCDP2,
Figure 13 illustrates the received waveforms of the bit where CAN/BCDP2 stands for CAN/BCDP with the dual
sequence BR = 010101010111 by CAN FD and CAN/BCDP hysteresis comparator. The proposed CAN/BCDP achieves
via Channel 14 between TX node 5 and RX node 5 with the same or higher success ratios than the original CAN
the data rate of Tb−1 = 5 Mbps, where the sample points FD for all the bit rates but 10 Mbps. On the other hand,
are drawn by circles. Although the data rate in Fig. 13 is the proposed CAN/BCDP2 establishes a remarkable result
lower than the previous data rate in Fig. 12, we observe that that it achieves 100% success ratio for any given bit rate
 2018 IEEE Vehicular Networking Conference (VNC)

100
2
1

VD [V]
80 0
-1
-2
Success ratio [%]

60 0 1 2 3 4 5 6 7 8 9 10 11 12
Nomalized bit time [1/Tb]

40 TX: CAN FD RX: CAN FD

2
1

VD [V]
20
0
-1
0 -2
1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10 11 12
Data-phase bit rate [Mbps]
CAN FD CAN/BCDP CAN/BCDP2 Nomalized bit time [1/Tb]
TX: SSTL-CAN RX: SSTL-CAN
Fig. 14. The success ratio versus the bit rate in the data phase for CAN FD,
CAN/BCDP, and CAN/BCDP2. Fig. 16. The received waveforms of the bit sequence BR by CAN FD and
SSTL-CAN via Channel 8 between TX node 3 and RX node 3 with the data
rate of Tb−1 = 8 Mbps.

2
1

VD [V]
0
-1
-2
0 1 2 3 4 5 6 7 8 9 10 11 12
Fig. 15. The insertion of stub resistors on the stub lines. Nomalized bit time [1/Tb]
TX: CAN FD RX: CAN FD

2
on the triple-star CAN network with 32 nodes. As with 1
VD [V]

0
CAN/IMDP, the proposed CAN/BCDP2 devices can coexist -1
-2
on the network with the existing CAN FD devices because
0 1 2 3 4 5 6 7 8 9 10 11 12
the transmit signals of CAN/BCDP2 devices can be decoded Nomalized bit time [1/Tb]
at the CAN FD devices. Replacing an unreliable CAN FD TX: SSTL-CAN RX: SSTL-CAN
device with a CAN/BCDP2 device may enhance the reliability
of the existing CAN networks due to reducing self-reflection Fig. 17. The received waveforms of the bit sequence BR by CAN FD and
waves at the node. SSTL-CAN via Channel 10 between TX node 3 and RX node 5 with the data
rate of Tb−1 = 8 Mbps.
3) Stub Series Terminated Logic for CAN: The original
node structure for CAN FD is modified in the proposed
schemes described above. In this section, although the node
itself is not modified, the CAN bus is modified to reduce ring- Figure 17 illustrates the received waveforms of the bit
ing. Stub series terminated logic (SSTL) has been originally sequence BR = 010101010111 by CAN FD and SSTL-CAN
applied to high-rate memory bus to mitigate the effect of signal via Channel 10 between TX node 3 and RX node 5 with the
reflection at stubs [20]. A stub resistor is inserted at each stub data rate of Tb−1 = 8 Mbps, where the sample points are
on the memory bus with SSTL. The high-rate CAN FD bus drawn by circles. If the transmit node is on a branch line
should be considered as a distributed constant circuit as with and it is not the received node, the dominant pulse amplitude
the high-rate memory bus. Applying SSTL to CAN FD buses significantly attenuates as shown in Fig. 17. Although the
is proposed as shown in Fig. 15. received bit sequence for SSTL-CAN is correctly detected,
Figure 16 illustrates the received waveforms of the bit the voltage margin for dominant pulses is quite small, so that
sequence BR = 010101010111 by CAN FD and SSTL- SSTL-CAN will not be robust in reference voltage variation
CAN via Channel 8 between TX node 3 and RX node 3 due to temperature fluctuation and supply voltage fluctuation
with the data rate of Tb−1 = 7 Mbps, where the sample in a car. We should optimize the value of the stab resistor RS
points are drawn by circles. The stab resistors can make the so as to maximize the dominant voltage margin and reduce
ringing amplitude smaller but the high-frequency component the ringing amplitude.
in the ringing remains unlike CAN/IMDP. Also the dominant Figure 18 illustrates the success ratio versus the bit rate in
pulse amplitude in the same branch line is enlarged by the the data phase for CAN FD and SSTL-CAN. The proposed
stab resistors. The received bit sequence is correctly detected SSTL-CAN achieves the same or higher success ratios than the
because the voltage VD at the sample point does not exceed original CAN FD for all the bit rates. Although SSTL-CAN
the threshold of 0.9 V in any recessive pulse. does not achieve 100% success ratio for 3 or more Mbps as
 2018 IEEE Vehicular Networking Conference (VNC)

100 CAN transceivers are utilized for a PoCAN node as with a

CAN/BCDP2 node. The connection of a PoCAN node to the
80 CAN bus is the same as a CAN/BCDP2 node. The received
differential signal is reliable for the data rate of 500 kbps and
Success ratio [%]

60 the received power is of good quality to establish supplying the

power with two CAN transceivers. Bipolar NRZ line code for
40
CAN/BCDP2 does not have DC component as with alternate
mark inversion (AMI) line code used for PoCAN. The node
20
structure for signal transmission in the PoCAN is the same as
that in the CAN/BCDP2. We should investigate the integration
0
1 2 3 4 5 6 7 8 9 10 of PoCAN in the arbitration phase and CAN/BCDP2 in the
Data-phase bit rate [Mbps]
CAN FD SSTL-CAN data phase to reduce power and communication wiring for
automobile networks as a future work.
Fig. 18. The success ratio versus the bit rate in the data phase for CAN FD
and SSTL-CAN. R EFERENCES
[1] ISO 11898-1, Road vehicles — Controller area network (CAN) — Part
1: Data link layer and physical signalling, ISO Std., Dec. 2015.
[2] SAE J2284-4, High-Speed CAN (HSC) for Vehicle Applications at
with the original CAN FD, it has an advantage of the simplified 500 kbps with CAN FD Data at 2 Mbps, SAE International Std., Jun.
implementation since the node architecture does not change 2016.
unlike the other ringing mitigation schemes. [3] SAE J2284-5, High-Speed CAN (HSC) for Vehicle Applications at
500 kbps with CAN FD Data at 5 Mbps, SAE International Std., Sep.
2016.
IV. C ONCLUSION [4] W. Lawrenz, Ed., CAN System Engineering: From Theory to Practical
Applications, Second Edition. Springer, 2013.
We have described our reduced wiring technology for CAN [5] K. Matheus and T. Königseder, Automotive Ethernet, Second Edition.
to enhance the network scalability and the cost efficiency in Cambridge University Press, 2017.
this paper. A triple-star network with 32 nodes is examined [6] G. Cena and A. Valenzano, “Overclocking of controller area networks,”
Electron. Lett., vol. 35, no. 22, pp. 1923–1925, Oct. 1999.
to discuss the advantages and disadvantages of the proposed [7] T. Ziermann, S. Wildermann, and J. Teich, “CAN+: A new backward-
schemes. compatible Controller Area Network (CAN) protocol with up to 16x
We have proposed three ringing mitigation schemes in order higher data rates,” in Proc. DATE 2009, Nice, France, Apr. 2009, pp.
1088–1093.
to reduce the communication wiring. Based on the single- [8] F. Hartwich, “CAN with flexible data-rate,” in Proc. iCC 2012, Ham-
node transmission in the data phase, we have developed bach, Germany, Mar. 2012, pp. 14-1–14-9.
CAN/IMDP. The CAN/IMDP architecture enables us to per- [9] M. Schreiner, H. E. Mahmoud, M. Huber, and S. Koç, “CAN FD from
an OEM’s point of view,” CAN Newsletter, no. 2, pp. 26–33, Jun. 2014.
form impedance matching at the transmit node in the data [10] T. Adamson, “Hybridization of CAN and CAN FD networks,” in Proc.
phase. Large-amplitude and high-frequency ringing observed iCC 2015, Vienna, Austria, Oct. 2015, pp. 03-9–03-12.
in the CAN FD network looks like passing through LPF in the [11] G. Cena, I. C. Bertolotti, T. Hu, and A. Valenzano, “Improving com-
patibility between CAN FD and legacy CAN devices,” in Proc. IEEE
CAN/IMDP network. However, if ringing has low-frequency RTSI 2015, Turin, Italy, Sep. 2015, pp. 1–8.
component, CAN/IMDP makes it difficult to eliminate the [12] K. Lennartsson, “CAN FD filter for classical CAN controllers,” in Proc.
ringing effects. CAN/BCDP and CAN/BCDP2 are defined as iCC 2015, Vienna, Austria, Oct. 2015, pp. 03-13–03-20.
[13] O. Garnatz and P. Decker, “CAN FD with dynamic multi-PDU-to-frame
an extension of CAN/IMDP because they perform impedance mapping,” in Proc. iCC 2015, Vienna, Austria, Oct. 2015, pp. 05-8–05-
matching at the transmit node in the data phase. Bipolar NRZ 13.
line coding is conducted by the CAN/BCDP and CAN/BCDP2 [14] G. Cena, I. C. Bertolotti, T. Hu, and A. Valenzano, “CAN with eXtensi-
ble in-frame Reply: Protocol definition and prototype implementation,”
architectures using two CAN transceivers. The CAN/BCDP2 IEEE Trans. Ind. Informat., vol. 13, no. 2, pp. 2436–2446, Oct. 2017.
architecture with the dual hysteresis comparator establishes [15] M.-M. Hell, “The physical layer in the CAN FD world,” in Proc. iCC
a remarkable result that it achieves 100% success ratio for 2013, Paris, France, Nov. 2013, pp. 05-1–05-5.
[16] S. Nitta, D. Umehara, K. Wakasugi, S. Ishiko, and T. Tsubouchi, “High
any given bit rate over the triple-star network with 32 nodes. rate and high multiplexing CAN by short pulse line codes,” in Proc.
In the SSTL-CAN scheme, the CAN bus is improved with IEEE VTC Spring 2013, Dresden, Germany, Jun. 2013, pp. 1–5.
the insertion of stab resistors without modifying the node [17] M. Mizuma, D. Umehara, T. Takahashi, and K. Wakasugi, “Silent-
interval impedance matching for short pulse CAN to enhance scala-
configuration. However, if the received node is not the transmit bility,” in Proc. APCC 2015, Kyoto, Japan, Oct. 2015, pp. 686–691.
node and the transmit node connects to a stab line, the [18] H. Mori, N. Maeda, Y. Mori, and H. Obata, “Novel ringing suppression
node observes low-amplitude dominant pulses. We need a circuit to achieve higher data rates in a linear passive star CAN FD,” in
Proc. EMC Europe 2014, Gothenburg, Sweden, Sep. 2014, pp. 402–407.
voltage margin on automobile networks so that SSTL-CAN [19] N. Sato, D. Umehara, and K. Wakasugi, “Poster: Symbol detection with
is not preferable in a current form. As a result, we can see OR rule for CAN FD,” in Proc. IEEE VNC 2016, Columbus, OH, USA,
that CAN/BCDP2 is one of the most promising and scalable Dec. 2016, pp. 1–2.
[20] R. Crisp, “Direct RAMbus technology: The new main memory stan-
solutions to reduce communication wiring in cars. dard,” IEEE Micro, vol. 17, no. 6, pp. 18–28, Nov. 1997.
As a method to reduce power wiring, we proposed the [21] T. Matsushita, D. Umehara, and K. Wakasugi, “Poster: Power over Data
power over CAN (PoCAN) architecture [21]. The DC power Lines for CAN using AMI code,” in Proc. IEEE VNC 2016, Columbus,
OH, USA, Dec. 2016, pp. 1–2.
is superposed on the CAN bus in the PoCAN system. Two