Recommended readings

 Dominique Paret, Multiplexed Networks for Embedded

Systems: CAN, LIN, FlexRay, Safe-by-Wire..., ISBN: 978-0-
470-03416-3, 434 pages, WILEY, UK, 2007.

 Wolfhard Lawrenz, CAN System Engineering: From, Theory to

Practical Applications, ISBN-10: 0387949399, ISBN-13: 978-
0387949390, 468 pages, Springer, 1997

 Konrad Etschberger, Controller Area Network, IXXAT

Automation GmbH (August 22, 2001), ISBN‐10:3000073760,
ISBN‐13: 978‐3000073762, 430 pages, 2001.

0
 Lecture 3.
Data Link Layer in CAN: Recessive and
Dominant bits, Arbitration, Frame Structure:
standard and extended frames (CAN 2.0A and
CAN 2.0 B), Bit Stuffing, Error Detection, Error
Management (Error passive, error active and
bus-off states)
 General Setting: Distributed Control System (DCS)
 Desired features:
 Communication is message oriented
and not target oriented (the same
message is likely to be of interest for
multiple endpoints)
 Supports broadcast, multicast: a
controller can use the sensors from
other controller as its own
 Real-time: rapidly changing information
has to be transmitted faster than other
(e.g. engine load changes rapidly than
temperature)
 Entire bus capacity is used to sent the
most relevant information at some point
(DoS due to traffic overload is not
possible)
 Does not require physical destination addresses
 Easy to add nodes without hardware or software modifications to existing nodes2
With respect to these, the CAN
Standard from Bosch specifies

3
 Comparison of CAN architecture
with the OSI/ISO
 Application (can implement in software absent layers)

 Frame setup, error control,

flow control, multiplexing

(picture c.f. Parret, 2007)

 Electrical and mechanical
specifications 4
CAN layers, more precise…

5
 Bit representation: dominant and
recessive bits
 Dominant bits can overwrite recessive bits
 In CAN: 0 – dominant, 1 – recessive
 Some physical examples:

 Trivial: Optical example

Light off – recessive Light on – dominant

 More technical: ASK

(Amplitude Shift Keying)

Signal present – dominant Signal absent – recessive

6
 NRZ - encoding

 CAN uses Non-return-to-Zero:

 “0” is represented as one physical level
 “1” is represented as another physical level
 Main deficiency
 As physical level remains constant over a bit, transmitting long
sequences of bits leads to loss of synchronization

7
 Bit Stuffing
 CAN bit stuffing rule: after 5 consecutive bits of identical value 1
bit of the opposite value is added
 De-stuffing has to be done at reception

(picture c.f. Parret, 2007) 8

 CAN Frame
 The frame is an envelope for a message (or for a part of a
message)
 As the data field in a frame is at most 8 bytes a message can
be distributed over more than 1 frame
 There are 4 different kinds of frames:
 Data frame - carries data from a transmitter to the receivers
 Remote frame - is transmitted by a bus unit to request the transmission
of the data frame with the same id
 Error frame - is transmitted by any unit on detecting a bus error
 Overload frame - is used to provide for an extra delay between the
preceding and the succeeding data or remote frame
 Data frames and remote frames are separated from preceding
frames by an interframe space
 Two frame formats CAN 2.0A – CAN standard frame and CAN
2.0B – CAN extended frame
9
 CAN 2.0A standard frame

(picture c.f. Parret, 2007) 10

 Start of Frame
 Used for synchronization
 A single dominant bit
 Can start only if bus was idle

11
 Arbitration Field
 Identifier 11 bit:
 Lower the value higher the priority
 The first 7 bits must not be all recessive => 2032 maximum combinations
 Remote Transmission Request RTR bit:
 dominant in data frames
 recessive in remote frames

12
 Control Field
 The first 2 bits reserved for upward compatibility
 Identifier Extension bit – dominant is Standard CAN frame is used
 r0 bit - reserved
 The last 4 bits indicate the data length

13
 Note that 4 bits indicate 16 possible data
lengths and only 9 lengths are used in CAN

(picture c.f. CAN

Specification v. 2.0)

14
 Data Field
 At most 8 bytes of actual data

15
 CRC Field
 15 bits with Hamming distance 6
 1 bit CRC delimiter (always recessive)
 Must detect:
 up to 5 randomly distributed (independent) errors in a message are detected
 burst errors of length less than 15 in a message are detected
 errors of any odd number in a message are detected
 Total residual error probability for undetected corrupted messages: less than
message error rate 4.7 * 10-11

16
 BCH (Bose and Ray-Chaudhuri)

 Also known as CRC-15

 Close to optimal
 The following polynomial is used:

P  x   x15  x14  x10  x8  x7  x 4  x3  1

17
 ACK Field
 1 ACK slot recessive, overwritten by a dominant bit
 1 ACK delimiter, always recessive

(picture c.f. Parret, 2007) 18

 ACK management
 Sender writes a recessive bit in the slot
 Each receiver that correctly received the message writes a dominant bit
 Consequence: the sender knows if there is at least one receiver had
correctly received it (does this imply that the message was correctly
transmitted?)
 Still, this does not mean the receiver actually uses this information !
 How to decide if the message was received by the intended node ? Use
layer 7

19
 End of Frame Field (EOF)
 Just 7 recessive bits

(picture c.f. Parret, 2007) 20

 Interframe Space (IFS)
 IFS - 3 recessive bits
 A successfully ACK is followed by 11 recessive bits

(picture c.f. Parret, 2007) 21

More?
 Idle, recessive bits >= 0

22
See the big picture

(picture c.f. Parret, 2007)

23
 Remote Frame
 RTR bit is recessive
 Data is 0 bytes

(picture c.f. Parret, 2007)

24
 Error Frame
 Error flag 6 bits:
 Active error – 6 dominant bits (no bit stuffing => violates bit stuffing=>chain
reaction)
 Passive error (may overlap with active error)– 6 recessive bits (no bit stuffing)

 Error delimiter 8 bits

 Transmission of error frame starts simultaneously with the next
bit after the detection of an error
 CRC error frames start after the ACK delimiter

(picture c.f. Parret, 2007) 25

 Overload Frame
 Used to indicate that a unit cannot receive additional frames due to overload
 Identical to an active error frame (again 6 consecutive bits => chain
reaction)
 At most 2 consecutive overload frames can be used
 According to CANSpec there are 2 situation for OFs:
1) The internal conditions of a receiver, which requires a delay of the next
DATA FRAME or REMOTE FRAME.
2) Detection of a ’dominant’ bit during INTERMISSION.
 CANSpec: The start of an OVERLOAD FRAME due to OVERLOAD
condition 1 is only allowed to be started at the first bit time of an expected
INTERMISSION, whereas OVERLOAD FRAMEs due to OVERLOAD
condition 2 start one bit after detecting the ’dominant’ bit
 Remark CANSpec: In case that there is a ’dominant’ bit detected during the
3rd bit of INTERMISSION locally at some node, the other nodes will not
interpret the OVERLOAD FLAG correctly, but interpret the first of these six
’dominant’ bits as START OF FRAME. The sixth ’dominant’ bit violates the
rule of bit stuffing causing an error condition

26
 CAN 2.0 B frame
 Extended CAN, uses 29 bit identifiers
 SRR – substitute remote request bit
 IDE is recessive
 Remark: Since the SRR bit is received before the IDE bit, a receiver cannot decide
instantly whether it receives a RTR or a SRR bit. That means only the IDE bit
decides whether the frame is a Standard Frame or an Extended Frame.
 … see pictures below

(picture c.f. Parret, 2007)

27
CAN 2.0 A – CAN 2.0 B interoperability
 CAN 2.0 B controllers must support standard CAN frames
 CAN 2.0 B Active – can read/write CAN 2.0B frames
 CAN 2.0 B Passive – can read CANB 2.0 frames
 CAN 2.0 B can work with standard or extended frames at the same time

(picture c.f. Parret, 2007)

28
 CAN bit stuffing Revisited

(picture c.f. Parret, 2007)

 The maximum size of the stuffing area: 98 bits => maximum 19 stuff bits

29
 CAN frame overhead
 Overhead: Data size / Frame size (consider also bit stuffing)

70

60
50

40

30

20
10

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

CAN 2.0 A CAN 2.0 A CAN 2.0 B CAN 2.0 B

(without stuffed bits) (with stuffed bits) (without stuffed bits) (with stuffed bits)
30
 Arbitration procedure
 High priority messages must
have low identifiers !!!

(picture c.f. Lawrenz, 2007)

31
 Example …

(picture c.f. Lawrenz, 2007)

32
 … and another example

(picture c.f. Parret, 2007)

33
 Arbitration: Data Frame vs. Remote
Frame
 Data frame wins

34
 Arbitration: Standard Frame vs.
Extended Frame
 Standard frame wins

35
 Arbitration type
ND + CSMA/CA + AMP

i.e. Non-Destructive + Carrier Sense Multiple

Access / Collision Avoidance + Arbitration
on Message Priority

36
 Timing Considerations
 Can we compute the maximum delay at which a particular
message will arrive ?

37
 Timing Considerations
 Can we compute the maximum delay at which a particular
message will arrive ?
 Only for the highest priority message

38
 Questions
 Which is the worst time interval at which
an interrupt for a receive routine is
triggered ?
 How fast will the highest priority message
arrive ?

39
 Data rate revisited
 Data rate = useful data rate from the entire frame at 1Mbps

40
 Error Detection
 Five types of errors are defined on CAN Standard:
1) BIT ERROR - A unit that is sending a bit on the bus also monitors the bus. A
BIT ERROR has to be detected at that bit time, when the bit value that is
monitored is different from the bit value that is sent. An exception is the
sending of a ’recessive’ bit during the stuffed bit stream of the ARBITRATION
FIELD or during the ACK SLOT. Then no BIT ERROR occurs when a
’dominant’ bit is monitored. A TRANSMITTER sending a PASSIVE ERROR
FLAG and detecting a ’dominant’ bit does not interpret this as a BIT ERROR.
2) STUFF ERROR - A STUFF ERROR has to be detected at the bit time of the
6th consecutive equal bit level in a message field that should be coded by the
method of bit stuffing.
3) CRC ERROR - The CRC sequence consists of the result of the CRC
calculation by the transmitter The receivers calculate the CRC in the same way
as the transmitter. A CRC ERROR has to be detected, if the calculated result is
not the same as that received in the CRC sequence.
4) FORM ERROR - A FORM ERROR has to be detected when a fixed-form bit
field contains one or more illegal bits.
5) ACKNOWLEDGMENT ERROR - An ACKNOWLEDGMENT ERROR has to be
detected by a transmitter whenever it does not monitor a ’dominant’ bit during
the ACK SLOT.

41
 When is a message valid ?
 The point of time at which a message is taken to be valid, is
different for the transmitter and the receivers of the message:
 Transmitter:
 The message is valid for the transmitter, if there is no error until the end
of END OF FRAME. If a message is corrupted, retransmission will follow
automatically and according to prioritization. In order to be able to
compete for bus access with other messages, retransmission has to start
as soon as the bus is idle.
 Receivers:
 The message is valid for the receivers, if there is no error until the last
but one bit of END OF FRAME.

 Question: Why are these rules different for senders and

receivers ? 42
 Error Handling
 The following steps are done:
1) Error detected
2) Error frame transmitted
3) Message discarded by all nodes
4) Error counters incremented
5) Message retransmitted

43
 Error Limitation
 To prevent a permanently disturbed bus each CAN
controller has 3 states:
1) Error active - an ’error active’ unit can normally take part in
bus communication and sends an ACTIVE ERROR FLAG
when an error has been detected.
2) Error passive - an ’error passive’ unit must not send an
ACTIVE ERROR FLAG. It takes part in bus communication
but when an error has been detected only a PASSIVE
ERROR FLAG is sent. Also after a transmission, an ’error
passive’ unit will wait before initiating a further transmission.
(See SUSPEND TRANSMISSION)
3) Bus off - a ’bus off’ unit is not allowed to have any influence
on the bus. (E.g. output drivers switched off.)
44
 CAN node states
 Error passive and error active

45
 Evolution of error counters (example)

46
 Tx Error counter modifications
(brief and incomplete)

 + 8 when an error in the transmitted message is detected

 - 1 when a message is correctly transmitted and acknowledged
(not applied when counter is 0)

47
 Rx error counter modifications (in
brief)

 + 1 when it detects an error

 + 8 when it is the only unit that detected the error
 - 1 when a frame is successfully received

48
 Efficiency of Error Management

 Main idea: locate faulty nodes => faulty nodes always

accumulate more points then other nodes

49
 Error Counters Modification Rules
 According to CAN Spec (note that more than 1 rule can be applied)

50
51
52
 Example 1

 What happens if a node A gets a receive error due to

a local disturbance ?

53
 Example 1

 Consider a node A gets a receive error due to a local

disturbance:
 A increments its Rx counter by 1 with rule 1)
 A sends an active error flag
 Other nodes detect a bit stuffing error on the 6th bit of the error flag and
send an error flag
 A detects a dominant after its error flag and increments Rx counter by 8
according to rule 2)

54
 Example 2

 What happens if node A transmits when he is the only

node on-line ?

55
 Example 2

 Consider a node A transmits when he is the only node

on-line:
 A does not get an acknowledgment
 A increment its Tx counter and get in error passive
 A cannot get to BUS OFF due to exception 1 in rule 3

56
 Lecture 4.
Physical Layer in CAN: Implementation
aspects and bit timing procedures, hardware
synchronization and resynchronization, high
speed and fault tolerant CAN, theoretical
limitations of CAN speed, oscillator tolerance
for CAN networks
 ISO 11898-5 specifies the CAN
physical layer
 Recessive condition: if CAN-H is not higher than CAN-L plus
0.5 V
 Dominant condition: if CAN-H is at least 0.9 V higher than
CAN-L
 Each node must be capable to produce output voltage between
1.5V and 3.0V

58
Parret, 2007:

59
The ZK-S12-B Starter Kit

60
 Fault-tolerant CAN transceiver
 Philips Double Wire TJA1054 CAN Transceiver

61
Fault tolerant transceivers specifications
 Low speed, data rates up to 125 kbit/s and uo to 32 nodes
 Detect and handle the following bus error conditions:
 Interruption on CAN-H
 Interruption on CAN-L
 Short circuit of CAN-H with VCC
 Short circuit of CAN-H with GND
 Short circuit of CAN-L with VCC
 Short circuit of CAN-L with GND

62
 Bus wiring
 120 ohm termination resistor (not needed for
TJA1054)

63
CAN Connector according to CIA recommendation
 Can-in-Automation (CIA) DS-102 – CAN physical layer for
industrial applications
 9 pin D-SUB connector

64
 Bit Time and Nominal Bit Time
 Bits are encoded with NRZ – main issue: lack of
synchronization
 Bit Time = the period of time for which the bit is present on the
bus
 Nominal Bit Time = the theoretical period of time for which the
bit is present on the bus
1
NominalBitTime=
NominalBitRate
 Nominal bit time is divided in 4 parts

65
 Synchronization segment
 Used to synchronize all nodes, the leading edge of an
incoming bit should appear within this segment

66
 Propagation Time Segment
 Used to compensate the delays on the network (twice the
transmission plus propagation time)

67
 Phase Buffer Segment 1 and 2
 Used to compensate for edge phase errors (mostly due to small frequency
variations in the oscilators)
 Designed to be extended or shortened by resynchronization procedure
 Sample Point = the point at which the bus level is read (located at the
boundary between TSGE1 and TSEG2), two relevant issues:
 Must be as late as possible - to overcome delays and have maximum
confidence on the value
 Must not be too late – time is also needed to compute the actual value
 => The Sampling Point must be towards the end of a bit period
 Information Processing Time = the time segment starting with the Sample
Point (TSEG 2)
 Multiple sampling is available on some circuits: a but can be sampled more
than once (e.g., odd number of time and use majority logic)

68
 Computing the Nominal Bit Time
 Minimum Time Quantum – generally derived from the
station’s clock
 Time Quantum of the Bit Time – derived from the minimum
time quantum by the use of dividers (division in the range
from 1 to 32)

TimeQuantum=m  MinimumTimeQuantum

69
Length of Time Segments

70
An example from [Parret, 2007]

71
 And another example …
 Note that on some microcontrollers one can set only TSEG1
and TSEG2, in this case TSEG1 must include the
propagation time

72
 Exact computation for time segments
length
 Synchronization Segment – 1 time quanta
 Propagation Speed Segment – depends on the transmission delay caused
by:
 Medium

 Signal speed on the medium

 Length of the network

 Topology

 Electrical parameters

 Quality of signal

 TSEG1 and TSEG2 – depend on clock frequencies and their tolerances,

fluctuations, temperature (remember that these segments are used to
compensate for errors, etc.)
73
 Computing Propagation Time
 The longest distance between 2 CAN nodes determines the propagation
delay
 To assure correct arbitration the bit time must be at least twice the delay
required to send a bit from A to B. Why ?

74
 Detailed computation of propagation time
 The following delays must be added:
 The delay required by the controller to output the signal
 The delay required by the transceiver to generate the signal
 The time to transport the signal
 The time required by the reception transceiver to send the signal to the controller
 The time required by the controller to process the signal

75
 Can be further synthesized in two types of delays

Tres = Tmed +Telec

76
 The delayed caused by the electronic components has the
following components:
 Tsd – the delay caused by the CAN controller
 Trx, Ttx – the delay caused by the transceiver
 Tqual – the delay caused by imperfections in the signal

Telec  Tsd  Ttx  Trx  Tqual

77
 The delayed caused by the medium is determined by the
length of the medium and the propagation speed:

L
Tmed 
v prop

 Therefore the minimum value of the propagation segment is:

 seg 
prop min  2 Tmed  Tsd  Ttx  Trx  Tqual 

78
 Relations between maximum bit-
rate and length of the network

 Question 1: what is the maximum distance at a particular bit-

rate on some medium ?

 Question 2: what is the maximum bit-rate on a given network

length and medium ?

79
 We now that the propagation segment must be at least twice the delay,
assume that this segment is only x percents of the bit time, then we have:

 seg 
prop min  xTbit  2Tprop  2 Tmed  Tsd  Ttx  Trx  Tqual 

 

 L  0.2mns 1  x
 2 BaudRate
 Telec 


80
 Bit Synchronization
 Hardware Synchronization
 Internal bit-time started in SYNC_SEG
 Forces the edge that caused Hard Synchronization
to lie in SYNC_SEG

 Resynchronization

81
 Synchronization rules

 According to CAN standard:

1. Only one synchronization allowed during one bit time
2. An edge is used only if the value after the edge differs from
the previous sample point
3. HARD Synchronization is performed only after recessive-
dominant transition during BUS IDLE
4. All other recessive-dominant transitions are used for
RESYNCHRONIZATION

82
 Resynchronization
 PHASE_SEG1 may be lengthen or PHASE_SEG1 may be
shortened
 The amount of lengthening/shortening is given by
RESYNCHRONIZATION JUMP WIDTH
 RESYNCHRONIZATION JUMP WIDTH = min(4,
PHASE_SEG1)
 The maximum length between 2 transitions that can be used for
resynchronization is 29 bit times

83
Phase Error
 Phase Error of an Edge = difference in
time quanta between the position of the
edge and SYNC_SEG:
 e=0 lies in SYNC_SEG
 e>0 lies before the SAMPLEPOINT
 e<0 lies after the SAMPLEPOINT

84
 Resynchronization Rules

 Two cases
i) If PHASE ERROR is smaller (or equal) then
RSJW the effect is the same of a HARD
Synchronization
ii) If PHASE ERROR greater then RSJW then
 If PE > 0 then PHASE_SEG1 lengthened by RSJW
 If PE < 0 then PHASE_SEG2 is shortened by RSJW

85
 Effects of the oscillator frequencies

 The actual time on each node is (here od = oscillator drifts)

Tmin  Tnom 1  od /100   Tnom  Tmax  Tnom 1  od /100 

86
 Estimation of RJW
 If x is the oscillator tolerance then for one bit the
synchronization jump must be (explain why 2):
x
RJW  2  Tbit 
100
 CAN standard specifies that resynchronization must be feasible
after 29 bits, therefore we have:
x
RJW  58  Tbit 
100
 The minimum bit-time is:
Tbit min  1  segProp  1tq  RJWmax

 Therefore we have
 58  58
RJWmax 1  x  x  2tq  segProp 
 100  100
87
 Question

 What is the maximum oscillator tolerance for a

node in a CAN network?

88