3 The CAN Physical Layer
• the physical layer described in the
reference document and in the ISO
(International Standards Organization)
• the properties of the bit,
• the types and structures of networks,
• the problems of signal propagation,
distance and bit rate.
• What is the actual speed of the bit?
• At what moment can we say that it is a '1'
or a '0'?
• What is the best physical medium for the
• Are ordinary wires good enough for
carrying the data?
• Are optical fibres a 'must'?
• What is the optimum bit rate for a given
• What maximum operating distance can we
• Are there one or more standard socket(s)?
• How do we 'drive' the line?
• How do we protect the line?
3.1.1 The CAN and ISO physical
• CAN users often depict the physical layer in an
(over-)simplified way, talking about its 'line'
aspect only, because this is the feature we first
• The ISO has issued documents describing the
standardization of the data link layers and
physical layers for specific applications
• These documents describe the CAN architecture
in terms of hierarchical layers, according to the
definitions of the 'Open Systems Interconnection
(OSI) base reference model'
• Data link layer (layer 2), DLL,
• Logical link control sublayer, LLC,
• Medium access control sublayer, MAC, described in the
preceding chapters, and
• Physical layer (layer 1), PL:
-Physical signalling sublayer, PLS (bit coding, bit timing,
-Physical medium attachment sublayer, PMA
(characteristics of the command stages (drivers) and
-Medium dependent interface sublayer, MDI
• Transmission medium:
-bus termination networks.
3.1.2 Basic and specific qualities
required in the physical support
to be used
• the type of physical support used to carry the
CAN frames is not explicitly defined
• medium in the CAN protocol (for example: wire, RF, IR,
optical fibres, visible light, arm signals, semaphore,
telegraph or maybe smoke signals)
• the chosen medium must conform to the CAN structure
and architecture by
- being capable of representing bits having dominant and
recessive states on the transmission medium;
- being in the recessive stage when a node sends a
recessive bit or when no nodes are transmitting;
- supporting the 'wired and' function, simply as a result of
- being in the dominant state if any of the nodes sends a
dominant bit, thus suppressing the recessive bit.
3.2 The 'CAN Bit'
• OSI concerned with the 'specification of
the physical signalling sublayer' (PLS).
• consideration of the problems of the bit
(coding, duration, synchronization, etc.)
and the 'gross' and 'net' bit rates of the
3.2.1 Coding of the CAN bit
• the bit is coded according to the non return
to zero (NRZ) principle
3.2.2 Bit time
• The period of time for which the bit is actually present on
the bus is called the bit time, tb.
• The bus management functions are responsible for
managing the CAN protocol:
- carried out within the period of duration of the bit (tb, or
'bit time frame')
- including the synchronization constraints of the stations
present on the network,
- compensation of propagation delays,
- definition of the exact position of the sample point
• The bus management functions are defined by a bit
timing programmable logic circuits.
3.2.3 Nominal bit time
• The nominal bit time is identical to the nominal duration
of a bit.
• By its nature, this value can only be an ideal and
• Each station in the network must be designed so that it
'nominally' has this bit time tb,
• because of the real components (oscillators, tolerances,
temperature and time fluctuations, etc.), the
instantaneous value of the bit will never be the nominal
value at any instant
• The nominal bit time is normally constructed from a
specified number of 'system' cycle pulses, based on
what should form the internal clock of each station.
• The official CAN protocol document indicates
that the nominal bit time is divided into a number
of time segments that cannot overlap.
• the synchronization segment,
• the propagation time segment,
• segment 1, called 'phase buffer 1',
• segment 2, called 'phase buffer 2'.
• The 'synchronization segment' is the part
of the bit time used to synchronize the
various nodes present on the bus. By
definition, the leading edge of an incoming
bit should appear within this time segment.
The propagation time segment
• This part of the nominal bit time is designed to
allow for and compensate the delay time(s) due
to the physical phenomenon of propagation of
the signal on the medium used and the types of
topology used to develop the geometrical
configuration (topology) of networks.
• this value is approximately a time equal to twice
(outgoing and return) the sum of the propagation
times present due to the medium and the delays
caused by the input and output stages of the
participants which are also connected to the bus
Phase buffer segments 1 and 2
• This pair of phase buffers1 (phase buffer
segments 1 and 2) is used to compensate
for the variations, errors or changes in the
phase and/or position of the edges of the
signals which may occur during an
exchange on the medium
• These two segments are designed to be
either extended or shortened by the
3.2.4 Sampling, sample point and
• There is no point in carrying data unless they
are to be correctly received, read and
• define the precise instant when the incoming bit
on the bus is sampled and is consequently read
and interpreted and has its value defined.
• This instant - the 'sample point' - must occur not
too soon and not too late during the period of
duration of the bit.
• in order to overcome the constraints of the fundamental
principles of the protocol (arbitration and
acknowledgement phases), the problems due to the
deformation of the signal representing the bit and delays
due to the phenomena of signal propagation on the
medium, the value of the bit should be read and
interpreted as late as possible in the bit time to give us
the maximum confidence in this value;
• not wait for the very last moment, for at least two
reasons: firstly, because all the tolerances allowed for in
the establishment of the bit time must be taken into
account, and, secondly, because of the time required for
calculating the value of the bit when it is sampled.
• the bit sampling should take place towards the end of the
period of duration of the bit
• the 'sample point' - forms, by definition, the exact
boundary between phase buffer segment 1 and
phase buffer segment 2.
• A start of frame (SOF) obliges all the controllers
present on the bus to carry out a 'hardware
synchronization' on the first change from a
recessive to a dominant level
• allow for any arbitration phases during which
several controllers can transmit simultaneously.
• it is also necessary to include additional
synchronization time buffers located on either
side of this instant (phase buffers 1 and 2)
• a time buffer segment positioned after the
sampling point, with the purpose of participating
in any necessary (re)synchronization of the bit
• information processing time: located and starting
immediately after the sample point is necessary
for the calculation of the value of the bit
• bit calculation. To improve the quality of the
capture and validation of the bit, it has always
been permissible to carry out digital filtering by
sampling the bit value several times following
the sample point instant.
• SJA 1000: carry out a 'multiple sampling' of the
bit value an odd number of times (3)
3.3 Nominal Bit Time
• Each of these segments plays its part in
determining the quality, the bit rate, the length,
etc., of a CAN network
• showing how the bit time is made up,
• describing the role of the different segments,
• explaining how they are interwoven,
• and, finally, describing a way of determining
their values in an example of application.
3.3.1 Constructing the bit time
• The nominal bit time of a CAN network is fixed
and specified for a project, and each station
must be able to construct its own nominal bit
time locally using its own resources.
• Each local division of nominal bit time can be
structurally different from one station to another.
• minimum time quantum: the definition of the
base for the construction of the bit time of a local
station as the smallest reference time element
usable by this station
Minimum time quantum of a
• The minimum time quantum is generally
derived from the station's clock
Time quantum of the bit time
• The time quantum is a fixed unit of time, derived
and obtained from the minimum time quantum
from which the CAN bit is to be constructed.
• The actual construction of this time unit is
carried out with the aid of programmable (pre-
)dividers (by integers m), whose division range
must extend from 1 to 32.
• Time quantum = m x Minimum time quantum
where m is the division factor of the (pre-)divider.
3.3.2 Dividing the bit time
• the values defined by the CAN protocol.
• Synchronization segment Propagation: 1 time quantum
(this value is fixed)
• Propagation time segment: from 1 to 8 time quanta
• Segment 1 of the phase buffer: from 1 to 8 time quanta
• Information processing time : <2 time quanta
• Segment 2 of the phase buffer: equal to not more than
the value of segment 1 + Information processing time
• The total number of time quanta cannot be less than 8 or
more than 25
• SJA 1000 protocol: define a new time 'segment'
- time segment 1 (TSEG 1) - characterizing the
set of segments compensating for the
'propagation times' and the synchronization
buffer 'phase segment 1'- of the CAN protocol
• It is important for the oscillators of the different
ECUs to be 'coordinated' or 'harmonized', so as
to provide the system with a time quantum
specified in the best possible way.
3.3.3 The sequence of events
• Our four segments are now in place, and
we need only to define the number of
quanta assigned to the propagation
segments and to phase buffers 1 and 2.
• Synchronization segment
• The duration of this segment is no problem;
it is permanently fixed at 1 quantum.
The so-called propagation
• the medium (and its choice),
• the propagation speed of the signal on this
• the maximum length of the network (medium),
• the topology of the network,
• the electrical input/output parameters of the line
• the quality of the received signals, etc.
• All these parameters are based on concepts of
the medium, topology, etc.
Buffer segments 1 and 2
• The parameters used to determine the value of
• the nominal values of the clock frequencies of
• the tolerances of these clocks,
• the temperature and time fluctuations of these
• the worst cases, etc.
• All these parameters affect the capacity to
successfully synchronize or resynchronize the
stream of incoming bits
3.4 CAN and Signal Propagation
• Whole volumes have been written on network
• My aim is not to reproduce these but to cast
some light on the properties of the CAN protocol
in its applications to local networks.
• The concept of a network implies structure,
topology and medium.
• We cannot diverge from these major standards
and the commentaries on them.
3.4.1 Network type, topology and
• Regardless of the types of medium used
to create networks, their topologies are
largely dictated by the possibilities of the
data transport protocol.
• In the case we have been describing over
many pages, the topology can be varied,
and I will give some examples.
• The 'bus' is the easiest topology to
implement and the most widespread for
this kind of protocol.
• the 'backbone' type: problems may arise unless
you are careful and that you should think in
terms of electrical lines with allowance for 'stubs'
connected to these lines.
• for a bit rate (NRZ coded 'square' signal wave)
of 1 Mbits-1 (i.e. an equivalent frequency of 500
kHz for square signals), propagated for example
in a differential pair at 200,000 km s-1, the
fundamental (sinusoidal) frequency of the wave
would be a wavelength of
• the other wavelengths of the base signal would
be 400/3 m, 400/5 m and 400/7 m, i.e.
approximately 133 m, 80 m and 56 m, which is
often very short!
• recommendations for the CAN low speed
mode (less than 125 kbit s-1) and CAN
high speed mode (more than 125 kbit s-1
and up to 1 Mbit s-1).
• operate on the boundary of the two modes:
a kind of 'medium speed' (a non-standard
term, invented for this occasion) of
approximately 250 kbit s-1.
• a network or partial network takes the
topological form of a star
• When systems are required to have greater operating
security and/or reliability in case of mechanical failure of
the network, the bus between two or more stations is
frequently doubled, or split
• it necessary to examine carefully the estimation and
management of propagation time, especially for the
resolution of the problem of the 'bit' arbitration phase.
3.4.2 Propagation time
• If a line exists, there is
• a phenomenon of signal propagation on the line, a
propagation speed, a delay due to the time taken for the
propagation of the signal on the medium,
• a characteristic impedance, an impedance match,
tnodes and antinodes, standing waves, line-termination
rebound and possible collisions on the bus between
incoming signals and reflected signals, etc.
• the choice of medium (for reasons of cost, EMC
• the propagation speed of the signal in the chosen
• the length of the medium (for application-specific
• the internal delay times of the transmission and
reception circuit components.
3.4.3 Estimating the value of the
• in the worst case, taking into account the paths
followed, the signal propagation delays and the
tolerances and fluctuations of the oscillators of
• In a single network, the longest distance
between two CAN controllers determines the
maximum 'outgoing' propagation delay time.
• to ensure the correct operation of the protocol
during the arbitration and acknowledgement
phases, a longer time interval must be provided
to compensate for the total propagation delay
times having an effect on an exchange on the
• total time represents the maximum necessary
'propagation time' taken by the signal to travel
from one controller to another and back again
• A recessive bit is sent by node A, for example.
• The signal from node A is propagated towards node B.
• Some 'micro-instants' before the signal reaches its
destination, node B, which has not yet noticed the
presence of the incoming bit from A, starts to send a
dominant bit, for example.
• The signal sent from B is propagated towards A.
• The collision occurs at a point very near B and makes
the state of the line dominant (and wired) at this point.
• The resulting signal (dominant) is then propagated
towards A, taking a certain time, approximately equal
to the time taken to travel from A, to arrive.
• Station A, receiving a dominant bit while it is sending a
recessive bit, can then determine, according to its
instantaneous state, whether this represents a loss of
arbitration for it (if the bit that it sent formed part of the
identifier) or a bit error, etc.
3.4.4 Precise definition of the
propagation segment time
• for a given medium, the time interval to be
assigned to the CAN bit propagation segment
plays a considerable part in the evaluation of the
maximum length of the network and vice versa
• To precisely define the minimum period to be
assigned to the propagation segment
(segTprop), we must consider the totality of the
time contributions of all the elements in the
• the delay time required by the transmission
controller to output the signal on its terminals;
• the delay time required by the transmission
interface to generate the signal on the medium;
• the time taken to transport the signal along the
• the time taken by the reception interface for
transferring the signal to the reception controller;
• the time taken by the reception controller to
process the incoming signal.
• These parameters are frequently divided
into two classes: those relating more
specifically to the physical properties of
the medium used (Tmed) and those
concerned with the electronic
characteristics found in the network (Telec)
• The sum of these two components (Tres)
characterizes the total time taken by a bit
to actually move from one station to the
• Tmed represents the time required for transporting the
signal along the medium, in the case of the greatest
distance between two CAN controllers in the network.
• If we call this speed vprop, and if the length of the
medium is L, this time will be
• a differential pair, the propagation speed is substantially
equal to 0.2 mns-1, i.e. 200 m μs-1 (or, to put it another
way, the medium has a propagation time of 5 ns m-1)
• Telec: This time interval is made up of three elements.
• Tsd represents the sum of the delay times due to the
signal processing carried out in the output stages (on
departure) and input stages (on arrival) of the CAN
controllers of the departure and arrival stations.
• Ttx represents the delay time required by the
transmission interface to generate the signal on the bus.
• Trx represents the time taken by the reception line
interface to transfer the signal to the reception controller.
• T qual_sign: At the receiver, the input interface
of an integrated circuit generally consists of a
differential amplifier, provided, followed or
associated with a comparator.
• this comparator has its own threshold and
• These elements, plotted as voltages, must be
translated into 'equivalent times' (Figure 3.14)
according to the shape and quality of the
• the asymmetry between the 'rise time' and the 'decay
time remind you two problems which are sometimes
• These new electrical 'delay times' (Tqua]_sign) can
then be considered equivalent to propagation or
supplementary distances of the medium.
• In view of all these factors, the overall value of
Telec can be fairly well estimated at about a 100
• the required minimum total value for the duration
of the 'propagation time segment':
• (segT-prop)min≧ 2(Tmed + TSd + Ttx + Trx + Tqual_Sign)
• When this value has been determined, it is easy
to determine the minimum number of time
quanta required to construct the 'propagation
segment' (from 1 to 8) of the nominal bit time.
3.4.5 Corollaries: relations between
the medium, bit rate and length of
• What is the maximum distance that can be
covered by CAN for a bit rate X and a
• What is the maximum bit rate that can be
used for a network of length X using a
• For a given bit rate and medium, CAN makes it clear that
the distance between two nodes is largely dependent on
• the intrinsic propagation speed of the physical medium
used to construct the line;
• the delay introduced by the output stage of the
• the delay introduced by the input stage of the receiving
• the desired nominal bit rate (and therefore the maximum
value of the minimum duration of the propagation
• the precise instant when the signal is sampled and the
way in which the value of the bit is measured during its
• the frequencies and tolerances of the various
oscillators of the CAN controllers
(micro-controllers or 'stand-alone' devices)
present on the network;
• the quality of the signals, etc.
• 'The maximum distance of a CAN network is
primarily determined by the time characteristics
of the medium used and the consequences of
the principle adopted for the operation of the
arbitration procedure defined in the non-
destructive protocol (at the bit level) present on
the line (and see all the preceding sections)'.
Relationship between bit rate
and maximum length for a given
• The value called the 'propagation time segment' of the
nominal bit time of a CAN network must therefore be
greater than (or at least equal to) approximately twice
• segT_prog ≧ 2 Tres
• If x is the percentage represented by this segment,
• segT_prog = x ． Tbit
• Telec = 100 ns, vprop = 0.2 m ns-1 (wire and optical
medium), x = 0.66 = 66%.
Example SJA 1000 circuit
Bit rate Maximum Bus timing
distance BTR0 BTR1
1.6 Mbit s-1 10m 00h 11h
1 Mbit s-1 40 m 00h 14h
500 kbit s-1 130 m 00h 1Ch
250 kbit s-1 270 m 01h 1Ch
125 kbit s-1 530 m 03h 1Ch
100 kbit s-1 620 m 43h 2Fh
50 kbit s-1 1.3 km 47h 2Fh
20 kbit s-1 3.3 km 53h 2Fh
10 kbit s-1 6.7 km 67h 2Fh
5 kbit s-1 10 km 7Fh 7Fh