When the data is wrong during transmission, and the wrong frame is missed, it means that the wrong data is sent to the application layer. Unless the application layer has additional data recognition measures, this data may cause unpredictable results. The CAN protocol claims to have a very low rate of missed frame errors (4.7 & TImes; 10-11 & TImes; error rate). Some publicity materials are launched under certain conditions and it will take only one missed detection in 1000 years. Frame miss rate is a very important indicator. Many applications choose CAN only after seeing the instructions on the Bosch CAN2.0 specification. However, the source of this indicator has very little public information and little discussion, making it difficult for users to confirm or verify it, which poses risks to users. This paper adopts the method of reconstructing the example of missed error detection, and derives the lower limit of CAN's missed error detection frame probability, which is several orders of magnitude larger than what CAN claims. In many applications, CAN is the only choice under the balance of reliability and price, or has been produced and used for a long time. In the face of this newly discovered problem, before CAN itself is not improved, a "patch" is urgently needed. To improve. Due to the limited space, we can only briefly introduce the derivation process of the error detection rate of the wrong frame, with the focus on providing solutions. 1 Discussion on CAN missed error detection frame probability literature The Bosch CAN2.0 specification says that the probability of missed error frames is less than the message error rate & TImes; 4.7 & TImes; 10-11. Its source can be found in the reference, which does not provide an analysis algorithm for generating missed detections, only mentioning that a large number of simulations have been used to obtain the formula. To determine whether a frame will be missed after an error occurs, at least two CRC calculations are required. For each bit, only a few instructions are required for the assembly language. With the 80-90 bit frame considered in this article, the CRC covers 58-66 bit It is necessary to cycle 58 to 66 times. With the PDP11 or VAX machine commonly used in 1989, a machine instruction needs about 0.1 μs, and a frame judgment needs 0.07 ms. Even if the machine is not shut down for one year, it can be 2.20 × 1011 frames. 58 bits can constitute 258 = 2.88 × 1017 different frames, plus 58 × 57 different combinations of positions with 2 bit errors added, so the simulation that can be done is only a tiny part of the possible situation (one millionth ). Because the sample is too small, it is difficult to consider the influencing factors completely. In 1999, Tran also studied the error detection rate of wrong frames. In view of the difficulty of analysis, he also used a lot of computer simulation methods. For 11-bit ID and 8-byte data frames, he used a 600 MB Alpha server. As discussed above, although the amount of simulation is large, it is still a very small part of the possible situation. Another CAN-related standard, CANopen Draft Standard 304 (2005), gives an error detection rate of (7.2 × 10-9). Different data from the CAN Automation Association also makes people uncomfortable. 2 Derivation of the rate of missed detection of new error frames The research method in this paper is to construct an example of missed detection, determine the probability that the instance occupies a possible frame, multiply the probability of multiple dislocations corresponding to the instance, and then find all possible instances to get the wrong frame of CAN Missed detection rate. This article analyzes the two dislocations most likely to cause missed detections, and then expands to multiple dislocations. The data field takes 8 bytes, and it is assumed that all errors occur in the data field. It does not take into account the scattered multi-bit error-missing detection rate when the CRC check capability is exceeded, so what is obtained is the lower bound of the probability of missed-error frame detection. 2.1 The bit order is staggered when there is an error in CAN bit stuffing When there is a bit error in the bit stream that may cause padding, it is possible that only one of the sender and the receiver executes the padding rule, resulting in confusion in the understanding of padding bits and information bits. An error occurred in the third bit transmission in Figure 1 (a). As a result, the sender's padding bit 1 was misread as data 1 by the receiver, and the entire received data was 1 bit longer than the transmitted data. The error in the third bit transmission of Fig. 1 (b) causes the receiver to delete the padding bit, so it deletes the transmitted data 1 and the received data stream is 1 bit shorter. Figure 1 CAN's bit stuffing rules make the received bit stream change after an error It can be known from the change of the bit stream that if the two bit errors occur exactly once in the type of Figure 1 (a) and once in the type of Figure 1 (b), then the length of the transmitted data stream and the received data stream will still be equal, if Both errors occurred in the data domain, and other CAN inspections could not find them. 2.2 Conditions for missing inspection The transmitted bit stream and the received bit stream can be written as polynomial forms Tx (x) and Rx (x), CRC check is to use CAN's generator polynomial G (x) to divide these two formulas, the remainder is called CRC value If the two remainders are the same, the CRC check is passed. When a transmission error occurs, Rx (x) = Tx (x) + U (x) × G (x), the remainder obtained for Tx (x) and Rx (x) is the same, then an error occurs Missed frame detection. So as long as U (x) is found, an instance of missed detection can be constructed. On Grid Inverter,On Grid Tie Inverter,Wind Turbine Inverter,Wind Grid Tie Inverter Jinan Xinyuhua Energy Technology Co.,Ltd , https://www.xyhenergy.com