Advantages of CAN FD Error Detection Mechanisms Compared to Classical CAN

Transcription

1 Advantages of CAN FD Error Detection Mechanisms Compared to Classical CAN 15 th international CAN Conference, Vienna 27 th October 2015 Dr. Arthur Mutter, Florian Hartwich Robert Bosch GmbH FD 1 AE/EIN2 27th Oct Robert Bosch GmbH All rights reserved, also regarding any disposal, exploitation, reproduction, editing,

2 Motivation The Physical Layer may be disturbed CAN_H disturbance (e.g. EM radiation) Falsified bus signal CAN_L RX node: PCS converts the bus signal into a bit sequence CAN Frame bit flip SOF ID ACK EOF Falsified frame RX node: MAC detects falsified frames up to a given extent This Presentation Advantages of CAN FD Error detection mechanisms compared to Classical CAN From OSI Model: PCS: Physical Coding Sub-layer, MAC: Medium Access Control 2

3 Agenda 1 2 Improved Error Detection in CAN FD Fault Types 3 Classification of Error Cases 3

4 Agenda 1 2 Improved Error Detection in CAN FD Fault Types 3 Classification of Error Cases 4

5 Improved Error Detection in CAN FD CAN FD Frame SOF ID Stuff Count Field Sequence Four main improvements (compared to Classical CAN) EOF Fixed stuff bit 1 polynomials of higher order 2 Dynamic stuff bits included into calc. 3 Stuff Count 4 Fixed stuff bits in field Advantages Provides Hamming distance as in Classical CAN Eliminates Classical CAN Weakness Detects conversions: data <-> stuff bit Reduces the probability that a random bit sequence matches Improves the frame format check 5

6 Agenda 1 2 Improved Error Detection in CAN FD Fault Types 3 Classification of Error Cases 6

7 Fault Types Definition (in context of this paper) Fault = An erroneous frame modification seen by a CAN node Two Fault Types assumed during ISO Standardization Fault Type A: Bit Flip Fault Type B: Bit Drop or Bit Insertion Consequence RX node: interprets frame differently/modified TX node: may detect a bit error 7

8 Fault Type A: Bit Flip Definition A node samples a bit with the inverse (flipped) value Property A bit flip can occur in an RX node and in a TX node Example TX signal Node T RX signal Node R bit 0 bit 1 bit 2 falsified signal bit 3 bit 4 bit 5 bit flip bit boundary 8

9 Fault Type B: Bit Drop or Bit Insertion (1) Definition An RX node drops a bit from or inserts a bit into the bit sequence Property Can occur only in an RX node To occur, the RX node needs to have a large phase error Phase Error of RX node Distance (time) between the detected position and the expected position of an edge RX signal Node R View of Node R bit 1 phase error bit 2 bit 3 bit bou ndary 9

10 Fault Type B: Bit Drop or Bit Insertion (2) How to cause a Fault of Type B? 1. A disturbance (EM radiation) influences the CAN physical layer 2. Additional or shifted rec-dom edges appear in the bit sequence 3. TX node and/or RX node resynchronize on faulty edges 4. Resynchronization increases phase error of RX node If the phase error is above a critical level, the RX node drops a bit from or inserts a bit into the bit sequence Bit Drop Example Initial phase error due to f RX < f TX TX: i RX: (5 th 0 dropped) Case: f RX < f TX, i.e. BitTime RX > BitTime TX TX signal Node T RX signal Node R View of Node R bit 0 bit 1 falsified signal 0 phase error resync bit 2 bit 3 bit 4 bit 5 stuff bit bit boundary bit lengthened Sample Point (SP) 10

11 Fault Type B in a Practical Setup (1) Question How realistic is a bit drop or insertion in a practical setup? Practical Setup Focus on Data Phase Data Phase Bit Timing Sample Point = 70% SyncJumpWidth = 20% Sample Po int SJW SJW 20% 20% 70% 100% (one bit time) Facts A single resync. can change the phase error by at most 20% For a Bit Drop: RX node requires a phase error of > 30% For a Bit Insertion: RX node requires a phase error of > 70% 11

12 Fault Type B in a Practical Setup (2) Question How many consecutive resynchronizations on faulty rec-dom edges are at least necessary for one bit drop or insertion? Evaluation Summary Case For one Bit Drop For one Bit Insertion f RX f TX 2 (2*20% > 30%) 4 (4*20% > 70%) f RX < f TX 1 (1*20%+pe > 30%) 6 (6*20% pe > 70%) f RX > f TX 3 (3*20% pe > 30%) 2 (2*20%+pe > 70%) pe = phase error due to oscillator tolerance f RX and f TX differ due to the oscillator tolerance Special conditions: 1 disturbance may lead to 1 bit drop Typically conditions: 2-6 disturbances for 1 bit drop or 1 bit insertion should be considered as multi-bit error 5-7 disturbances for 1 bit drop and 1 bit insertion 12

13 Agenda 1 2 Improved Error Detection in CAN FD Fault Types 3 Classification of Error Cases 13

14 Classification of Error Cases Facts: - All error detection mechanisms have their limits - Beyond the limit, an RX node may accept a faulty frame as correct Involved Mainly Classification of Error Cases Fault Types Detected by 1. Normal Bit Errors A 2. Encoding Related Errors A and B + Stuff Count 3. Message Length Modifying Errors A and B Field, EOF, Examples TX frame ID data Class 1: RX frame ID data alig ned Class 2: RX frame ID data misaligned Class 3: RX frame ID data fault 14

15 Class 3: Message Length Modifying Errors (1) Properties RX node samples field misaligned Misalignment range: 1 bit N bytes TX frame RX frame ID ID data data alig ned misaligned fault RX node accepts frame, if Classical CAN Comparison Sampled bit sequence = Expected field Classical CAN Frame format and dynamic stuffing correct as in Classical CAN ACK bit does not disturb transmitted frame as in Classical CAN Main difference between Classical CAN and CAN FD Length of expected field Comparison 15

16 Class 3: Message Length Modifying Errors (2) field lengths Classical CAN: 16 bit CAN FD (17): 28 bit CAN FD (21): 33 bit TX frame ID RX frame ID alig ned data data misaligned fault Comparison: CAN FD <-> Classical CAN 17: 28 16=12 more bits need to match in CAN FD 21: 33 16=17 more bits need to match in CAN FD Result It is much less likely, that the field matches in a CAN FD frame 16

17 Summary Error detection in CAN FD four improvements compared to classical CAN polynomials of higher order Dynamic stuff bits included into calculation Stuff Count TX signal bit 0 bit 1 bit 2 Node T falsified RX signal signal Fixed stuff bits in field bit flip Node R Fault Types Fault Type A: Bit Flip simple Fault Type B: Bit Drop or Bit Insertion complex In practical setup: typically several coordinated disturbances needed for 1 bit drop or 1 bit insertion Case: frx < ftx, i.e. BitTimeRX > BitTimeTX Class 2: RX frame ID Classification of Error Cases Class 3: RX frame ID aligned 3 Classes specified In Class 3: accidental match of field is much less likely in CAN FD TX signal Node T RX signal Node R View of Node R bit 0 bit 1 falsified signal 0 phase error resync TX frame Class 1: RX frame bit 3 bit 4 bit 5 bit boundary bit 2 bit 3 bit 4 bit 5 stuff bit bit boundary ID ID Sample Point (SP) data data data data misaligned fault 17

18 Conformance Test according ISO/DIS M_CAN 3.2.1: pass M_TTCAN 3.2.1: pass 18