Accessing the shared! communication medium!! Media! Access! Conntrol!

Transcription

1 Accessing the shared communication medium Media Access Conntrol 1 J. Kaiser, IVS-EOS

2 Common Layering in the fieldbus area 7 7 application layer Service Access Point (SAP) 2 2 link layer 1 1 physical layer Assumption: homogeneous, closed system Not all layers are necessary (e.g. routing) Empty layers in the ISO/OSI- model Higher layers directly access the SAPs of lower layers. Efficiency improvement Direct mapping of layer 7 services to layer 2 functionality. 2 J. Kaiser, IVS-EOS

3 Media Access Control & Logical Link Layer arbitration flow control fault and error handling message re-transmissions transfer of data blocks control data transfer Logical Link Media Access LL -layer MAC -layer fault and error treatment, re-transmission flow control access control, arbitration control 3 J. Kaiser, IVS-EOS

4 What are the impairments of predicatbilty? Load Failures Predictability Network contention Arbitration conflicts transmission errors, lost messages 4 J. Kaiser, IVS-EOS

5 controlled access MAC-protocols random access Collision avoidance Collision resolution Reservation-based Token-based Time-based Master- Priority-based Slave dynamic ATM static TDMA: TTP, Maruti Token-Ring Token-Bus Timed Token Protocol CSMA/CA : Collision Avoidance IEEE P-persistent CSMA LON, VTCSMA ProfiBus DP FIP CAN-Open CSMA/CA : Consistent Arbitration CAN Ethernet probabilistic CSMA/CD : Carrier Sense Multiple Access / Collision Detection 5 J. Kaiser, IVS-EOS

6 Predictability in random access networks: Probabilistic very low overhead and latency in low load conditions very flexible wrt. extensibility thrashing in high load situations Collision avoidance (static & dynamic) Static: distinguishes planning phase and operation phase. Dynamic-1: balances the latency against the collision probability Dynamic-2: maps message priority to waiting time Consistent arbitration with Collision Resolution needs support from the physical layer maintains a constant throughput in all load conditions supports sophisticated fault handling 6 J. Kaiser, IVS-EOS

7 CSMA access modes 1-persistent Sender checks medium continuously. If medium is free, sender starts sending. In case of collision, sender waits random period of time before it checks the medium again. Non-persistent (0-persistent) Sender senses the medium. If medium is free, sender starts transmitting the data. If the channel is busy, sender waits for a certain (random or specified) amount of time and then repeats the procedure. P-persistent Sender checks the medium continuously. If the medium is free, the sender transmits a frame with a probability p. Unslotted variant: With probability 1-p the sender waits for a specified period of time. It again checks the availability of the channel and repeats the procedure. Slotted variant: With probability 1-p the sender waits until the next available time slot. It again checks the availability of the channel and repeats the procedure. 7 J. Kaiser, IVS-EOS

8 1-persistent CSMA check channel check channel check channel channel free, transmission starts channel occupied, collission wait random time t 8 J. Kaiser, IVS-EOS

9 Non-persistent CSMA (0-persistent) check channel check channel wait random time check channel again channel free, transmission starts immediately channel occupied, channel free, transmission starts immediately t 9 J. Kaiser, IVS-EOS

10 p-persistent CSMA check channel wait random time with probability 1-p check channel check channel channel free, transmit immediately with probability p channel occupied, t 10 J. Kaiser, IVS-EOS

11 Controlled Access by Collision Exclusion: Master/Slave all control information in one place maximum of control easy to change Global Time Easy temporal co-ordination Minimal communication overhead Token-based Decentralized mechanism Integration of critical and non- critical messages 11 J. Kaiser, IVS-EOS

12 CAN-Bus Controller Area Network 12 J. Kaiser, IVS-EOS

13 CAN Milestones 13 J. Kaiser, IVS-EOS

14 Controller Area Network Standards und Anwendungen ISO :2003 Road vehicles Part 1: Data link layer and physical signalling ISO :2003 Road vehicles Part 2: High-speed medium access unit ISO :2006 Road vehicles Part 3: Low-speed, fault-tolerant, medium dependent interface ISO :2004 Road vehicles Part 4: Time-triggered communication ISO :2007 Road vehicles Part 5: High-speed medium access unit with low-power mode ISO/NP Road vehicles Part 6: High-speed medium access unit with selective wake-up functionality CAN (FD) with Flexible Data-Rate, Bosch White Paper 2011 Application Areas: Automotive connecting smart ECUs, sensor units and multi-media units Industrial Automation time critical sensor-actuator control loops Elevators connecting control units of the elevator and groups of elevators Medicine MRT, CRT, lab units, e-wheel chairs Avionics connecting cabin techmology, flight control units Space parallel bus architectures Music and stage techn. control of digital power amplifiers Trains Ships DGzRS (German off shore SAR) uses CAN in their SAR ships aus: 14 J. Kaiser, IVS-EOS

15 The CAN Standard Developed by BOSCH, CAN Specification 1.2 CAN Specification 2.0 Difference between the specifications mainly is: the different lenth of message identifiers (CAN-ID) Standard CAN: 11 Bit IDs (defined in CAN 2.0 A 1.2) Extended CAN: 29 Bit IDs (defined in CAN 2.0 B) CAN-Controller Implementations: Basic CAN: 1 Transmit + 1 Receive (Shadow) Buffer Extended CAN: 16 Configurable Transmit/Receive Buf. 15 J. Kaiser, IVS-EOS

16 Basic CAN properties Prioritised messages Bounded and guaranteed message delay for the highest priority message. Constant troughput in all load situations Error detection and signalling in the nodes. Automatic re-transmission. Fail silent behaviour of nodes. Consistent message delivery. Multicast with time synchronization. 16 J. Kaiser, IVS-EOS

17 Layers defined by the CAN standard Data Link Layer LLC Acceptance Filtering Overload Notification Recovery Management MAC Data Encapsulation/ Decapsulation Frame Coding (Stuffing/Destuffing) Medium Access Management Error Detection Error Signalling Acknowledge Serialization /Deserialization Supervisor Fault Confinement Physical Layer Bit Encoding / Decoding Bit Timing Synchronization Bus Failure Driver/Receiver Characteristics LLC = Logical Link Control MAC = Medium Access Control 17 J. Kaiser, IVS-EOS

18 CAN differential transmission scheme CAN node CAN node CAN node CAN_H 128 Ω 128 Ω CAN_L CAN termination resistors 3,5 CAN_H 2,5 1,5 CAN_L recessive dominant recessive 18 J. Kaiser, IVS-EOS

19 The CAN physical layer Vcc physisal Bus connection by Wired And 0 = dominant level 1 = recessive level physisal connection S E CAN driver CC CC CC logic equivalent CC CC CC 19 J. Kaiser, IVS-EOS

20 CAN Bit Synchronisation node node CAN Bus After a certain time, all nodes have seen the value of a bit Bit rate dependend on the length of the bus Bit Monitoring 20 J. Kaiser, IVS-EOS

21 Bit-timing and bit synchronization Bitzeit = 1/nominal bit rate signal propagation- segment synchronization- segment phase buffer segment 1 phase buffer segment 2 sample (read) point Länge der Zeitsegmente werden in Vielfachen einer aus der Oszillatorperiode abgeleiteten Zeiteinheit (time quantum) spezifiziert: synch.-segment 1 time quanta sig. propag. seg time quantas phase buffer seg time quantas phase buffer seg time quantas 21 J. Kaiser, IVS-EOS

22 CAN transfer rates in relation to the bus length T d = T TT-delay + T line delay T TT-delay ~ 100 ns (driver, transceiver, comparator logic, etc.) T line delay ~ 5ns/m (0,2 m / ns) twisted pair Bitrate max. network (kbits/s) extension (m) J. Kaiser, IVS-EOS

23 CAN payload payload # of bytes Std. frame extended frame kbits/sec kbits/sec ,1 61, ,1 122, ,2 183, ,3 244, ,4 305, ,4 366, ,5 427, ,6 488,5 23 J. Kaiser, IVS-EOS

24 The CAN MAC and Logical Link Control (LLC) levels Frame types and formats: Data Frame normal data transmision initiated by the sender Remote Frame participant requests frame which is sent with the identical frame ID from some other participant. Error Frame participant signals an error which it has detected Overload Frame used for flow control. Results in a delayed sending of the subsequent frame. 24 J. Kaiser, IVS-EOS

25 CAN Standard Data Frame RTR-Bit inter frame Space 11 6 data frame inter frame space or overload- frame start of frame abitration field control field data field CRC field ACK field end of frame 25 J. Kaiser, IVS-EOS

26 Compatibility between standard and extended frames Standard Format SF (compatible to CAN Spezifikation 1.2) arbitration field control field data S O F 11-Bit Identifier R T R I D E r 0 DLC Extended Format EF (CAN Spezifikation 2.0) arbitration field control field data S O F 11-Bit Identifier S R R I D E 18-Bit Identifier R T R r 1 r 0 DLC RTR: Remote Transmission Request. In Data Frame: RTR = dominant. In Remote Frame: RTR = recessive. IDE: Identifier Extension. In the SF this is part of the control field, has a dominant value but is not interpreted. In the EF it is part of the addressing field, has a recessive value and causes the format to be recognized as EF. SRR: Subsitute Remote Request.Always recessive, replaces RTR in the EF for compatibility reasons. DLC: Data Length Control. 0-8 Byte. r0, r1: reserved 26 J. Kaiser, IVS-EOS

27 Arbitration on a CAN-Bus 0 = dominant 1 = recessive Station C stops Station A stops A B x x x x x C 0 1 CAN-Bus x x x x x CAN enforces a global priority-based message scheduling 27 J. Kaiser, IVS-EOS

28 N arbitration all bits received? Y wait for interframe space SOF receive state netx arbitration bit lost arbitration: sent: recessive level received: dominant level compare bus level with the level of the bit put on the bus bit error: sent: dominant received: recessive error state send Y all bits put on the bus? N 28 J. Kaiser, IVS-EOS

29 CAN Standard Data Frame RTR-Bit inter frame Space 11 6 data frame inter frame space or overload- frame start of frame abitration field control field data field CRC field ACK field end of frame 29 J. Kaiser, IVS-EOS

30 Anonymous acknowledgement of a CAN message CRC field acknowledge field end-of-frame pattern 1 Bit 1 Bit 7-Bit dom. recess. recess. ACK-Slot ACK delimiter positive anonymous acknowledgement (Broadcast ) receivers that correctly received a message(a matching CRC sequence) report this in the ack-slot by superscibing the recessive bit of the sender by a dominat bit. The sender switches to a recessive level. Message is acknowledged by a single correct reception on a correct node. Systemwide data consistency requires additional signalling of local faults. 30 J. Kaiser, IVS-EOS

31 Termination sequence of a frame 11 recessive bits CRC field Ack-field end of frame inter frame space 1 bit 1 bit 7 bit min. 3 bit dom. recess. recess. Goals: 1. Detecting AND signalling the error within the actual fame in which it occured 2. Identifying the node which may have caused the error. 3. Creating a systemwide view on the reception state of the message. Approach: End of frame pattern consisting of 7 recessive bits. 1. Any error detection is signalled by putting a dominant bit on the bus. 2. An out-of-sync node, not being aware of the EOF sequence will signal an error at position "6". 31 J. Kaiser, IVS-EOS

32 Interframe Space Frame Inter-Frame Space Frame sof I1 I2 Intermission Bus Idle Intermission: no data- or remote Frame may be started Intermission 1: active overload Frame may be started Intermission 2: re-active overload frame (after detecting a dominant bit in I1) 32 J. Kaiser, IVS-EOS

33 Error Detection and Error Signalling in CAN Bit-Stuffing enforces the following rule: Violation of the Bit-Stuffing Rule: Used for Error Detection and Signalling A sequence of 5 identical bit levels is followed by a complementary bit level CAN Data Frame Arbitration ACK End Of Frame (EOF) pattern Cntrl......Data.... CRC validation point for receivers validation point for the sender 33 J. Kaiser, IVS-EOS

34 Error detection 1.) Monitoring: Sender compares the bit sent with the bit actually on the bus. Type of faults: local sender faults Error detection: sender based 2.) Cyclic Redundancy Check: Type of faults: 5 arbitrarily distributed faults in the code word, burst error max. length 15. Error detection: receiver based 3.) Bitstuffing: Type of faults: transient faults, stuck-at-faults in the sender Error detection: receiver based 4.) Format control: Type of faults: the specified sequence of fields is violated. Error detection: receiver based 5.) Acknowledgment: Type of faults: no acknowledge Error detection: sender based, sender assumes local fault. 34 J. Kaiser, IVS-EOS

35 Risk of undetected errors Bit monitoring: An error will not be detected if - the sender is correct and monitoring doesn't detect an error - all other nodes receive the same bit pattern which is different from that of the sender and contains a non-detectable error. Bit-stuffing: double errors within 6 bits will not be detected CRC: difference between frame sent and received is a multiple of the generator polynome. Frame errors: the frame is shortened or additional bits are added. At the same time a correct end-of-frame sequence is generated. Unruh, Mathony und Kaiser: Error Detection Analysis of Automotive Communication Protocols, SAE International Congress, Nr , Detroit, USA, 1990 Scenario: nodes: 10, Bit error rate: , message error rate: 10-3 risk of undetected errors: 4, When the number of nodes increase, the probability of undetected errors decreases. 35 J. Kaiser, IVS-EOS

36 CAN error frame data error field IFS or overload frame dominant recessive 6-Bit error flag 8-Bit error frame EOF sequence error flag superposition 6-12 Bits 36 J. Kaiser, IVS-EOS

37 Error frame resulting from a sender fault SP X (sender) SP Y (receiver) SP Z (receiver) bus level SL: send level 6 Bit 6 Bit 8 Bit 3 Bit error frame IFS S O F time to re-transmit a faulty message frame: min. error recovery time: 23 bit times 37 J. Kaiser, IVS-EOS

38 Error frame resulting from a receiver fault SP X (sender) SL: send level SP Y (receiver) SP Z (receiver) bus level CRC-field Ack- field 7 Bit 8 Bit 3 Bit IFS reeor frame S O F time to re-transmit: min. error recovery time: 20 bit times 38 J. Kaiser, IVS-EOS

39 Enforcing fault confinement and a Fail Silent behaviour Problem: Assumption: Faulty component may block the entire message transfer on the CAN-Bus. 1. A faulty node detects the error first. 2. frequently being the first which detects an error --> local fault in the node approach: error counter for receive and transmit errors. If error was first detected by the node, the counter is increased by J. Kaiser, IVS-EOS

40 Enforcing fault confinement and a Fail Silent behaviour States of a CAN node: error active error passive bus off RxCNT: Value of the receive counter TxCNT: Value of the transmit counter RxCNT > 127 OR TxCNT > 127 error active error passive TxCNT > 255 bus off RxCNT 127 AND TxCNT 127 Reset + configuration + reception of 128 x 11 recessive Bits (i.e. 128 correct EOF recognitions) 40 J. Kaiser, IVS-EOS

41 CAN bus Error Handling - Transmit Error Counter error passive state error active state 41 J. Kaiser, IVS-EOS

42 Analysis of CAN inaccessibility CAN Data Frame Interframe Space 2 data frame Interframe Space or overload frame Start of Frame arbitration field control field data field CRC field ACK field End of Frame longest possible message: Format-Overhead: 67 bit times Data: 64 bit times Bitstuffing (max): 23 bit times total: 154 bit times 42 J. Kaiser, IVS-EOS

43 CAN Inaccessibility Times* Data Rate 1 Mbps, Standard Format Scenario t inacc (µs) Bit Errors Bit Stuffing Errors CRC Errors Form Errors Ack. Errors Overload Errors 40.0 Reactive Overload Errors 23.0 Overload Form Errors 60.0 Transmitter Failure Receiver Failure worst case single worst case multiple P. Verissimo, J. Ruffino, L. Ming: How hard is hard real-time communication on field-busses? 43 J. Kaiser, IVS-EOS

44 Predictability of various Networks* Worst Case Times of Inaccessibility* t inacc (ms) ISO 8002/4 Token Bus (5 Mbps) ISO 8002/5 Token Ring (4 Mbps) ISO 9314 FDDI (100 Mbps) Profibus (500 kbps) CSMA/CD unbounded CSMA/CA stochastic CAN-Bus(1Mbps) 2.48 Token-based Protocols CSMA Protocols * P. Verissimo, J. Ruffino, L. Ming: How hard is hard real-time communication on field-busses? 44 J. Kaiser, IVS-EOS

45 CAN-Bus Properties (summary) Event-triggered communication with low latency Priority-based arbitration with collision resolution for guaranteed throughput error handling: anonymous positive acknowledge negative ack. in case of an error (system wide messaging) identification of faulty nodes immediate synchronization and retransmission content-based addressing with a high flexibilitx (system elasticity) 45 J. Kaiser, IVS-EOS

46 High level issues Routing: How does a message reach a receiver? CAN: Broadcast, message subjects Filtering: CAN: message filters How can the receiver only receive those messages selectively in which it is interested in? 46 J. Kaiser, IVS-EOS

47 The CAN communication modell CPU CPU CAN-ID write read CAN-ID Send/ Receive Register CAN-ID CAN-ID Comm. Contrl. CPU CAN-ID Producer/Consumer, Model of a distributed memory CAN-ID : Transmit Request 47 J. Kaiser, IVS-EOS

48 The CAN communication modell CAN-ID 29/11 0 node a node b node n 48 J. Kaiser, IVS-EOS

49 CAN Components (Micro-) Controller CAN Communication Controller Stand-alone Microcontroller with integrated CAN I/O-components (SLIO) Transceiver chips Full CAN Basic CAN Transceiver 49 J. Kaiser, IVS-EOS

50 Tasks of the CAN Controller Arbitration Serialisation of message frames to transmit Assembling of received frames Calculating and testing the checksum Filtering messages Error detetction and signalling Building CAN-message formats Implementing Bit-Stuffing Generation and check of the acknowledge-bits Synchronization of the received bitstreams 50 J. Kaiser, IVS-EOS

51 Basic CAN CAN-Bus bus-interface protocol automaton acceptance filter global status/control register transmit-buffer receive buffer (shadow buffer) processor-interface processor 51 J. Kaiser, IVS-EOS

52 SJA1000 (Philips) 52 J. Kaiser, IVS-EOS

53 SJA1000 (Philips) Register and Memory Layout 53 J. Kaiser, IVS-EOS

54 SJA1000 (Philips) 54 J. Kaiser, IVS-EOS

55 message filtering message register filter mask x 1 x x x x x x x x The number of registers to store messages, configuration options and the ways to filter messages are dependent on the communication controller 55 J. Kaiser, IVS-EOS

56 SJA1000 (Philips) single mask option ACR: defines the pattern of CAN message IDs which are accepted. AMR: defines a mask of "don't care" positions. 56 J. Kaiser, IVS-EOS

57 Full CAN CAN-Bus bus-interface protocol- automaton acceptance filter global status/control register send/receive - buffer send/receive- buffer processor-interface processor... send/receive- buffer 57 J. Kaiser, IVS-EOS

58 AVR AT90CAN128 CAN Controller From Data Sheet: ATMEL: 8-Bit AVR Microcontroller with 128K Bytes of ISP Flash and CAN Controller AT90CAN J. Kaiser, IVS-EOS

59 AVR AT90CAN128 Acceptance Filtering Identifier tag (IDT) Identifier mask (IDMSK) Identifier extension (IDE) Identifier extension mask (IDEMSK) Remote transmission request (RTRTAG) Remote transmission request mask (RTRMSK) Data length code (DLC) Reserved bit(s) (RB) Examples: To accept only ID = 0x317 in part A. To accept ID from 0x310 up to 0x317 in part A. - ID MSK = ID MSK = ID TAG = ID TAG = xxx 59 J. Kaiser, IVS-EOS

60 What CAN can t All-or-nothing property under all single (crash/omission) fault conditions Consistent order of messages Temporal guarantees for message transmissions TTCAN FTTCAN 60 J. Kaiser, IVS-EOS

61 Error Detection and Error Signalling in CAN The Case for Inconsistencies Bit-Stuffing enforces the following rule: Violation of the Bit-Stuffing Rule: Used for Error Detection and Signalling A sequence of 5 identical bit levels is followed by a complementary bit level CAN Data Frame Arbitration ACK End Of Frame (EOF) pattern Cntrl......Data.... CRC validation point for receivers validation point for the sender 61 J. Kaiser, IVS-EOS

62 Consequences from the validition protocol J. Rufino, P. Veríssimo, C. Almeida, L. Rodrigues: Fault-Tolerant Broadcasts in CAN, Proc. FTCS-28, Munich, Germany, June J. Kaiser, Mohammad Ali Livani: Achieving Fault-Tolerant Ordered Broadcasts in CAN Proc. of the 3 rd European Dependable Computing Conference, (EDCC-3), Prague, Sept inconsistent message duplicates inconsistent omission faults (potentially) unbounded latencies 62 J. Kaiser, IVS-EOS