Packet Loss in Serial Communication of Industrial Computers: Optimization Approaches from Baud Rate Matching to Flow Control Settings
In the factory control room at 3 a.m., Engineer Lao Zhang stared at the "Communication Timeout" alerts flashing on the monitoring screen, beads of sweat forming on his forehead. He had just completed the renovation of an automated production line, but the serial communication between the newly deployed industrial computer and the PLC kept experiencing packet loss. Temperature sensor data would go missing every few minutes, and mechanical arm movement instructions occasionally failed to reach their destination, causing frequent production line shutdowns. To make matters worse, the failures were completely unpredictable: sometimes the system ran smoothly for hours, while other times, issues arose as soon as it started up.
Lao Zhang's predicament is not unique. In the field of industrial automation, serial communication remains the mainstream method for device interconnection due to its simplicity, reliability, and low cost. However, according to statistics from the State Grid, over 60% of industrial sites experience serial communication packet loss issues, resulting in annual equipment downtime losses amounting to billions of yuan. These packet loss failures often lurk beneath seemingly normal communication appearances, acting like time bombs ready to trigger production accidents at any moment.
In Lao Zhang's case, the root cause turned out to be a seemingly insignificant parameter—the baud rate. Although both the industrial computer and the PLC he used were labeled as supporting a baud rate of 115200 bps, actual testing revealed that the computer's baud rate was 116000 bps, while the PLC's was 114000 bps. This 1.7% deviation may seem minor, but it was enough to cause the communication system to collapse.
Technical Analysis:
UART asynchronous communication lacks a clock line for synchronization and relies entirely on the internal clocks of both parties to send and sample data at the same rate "tacitly." Taking 115200 bps as an example, each bit lasts for 8.68 μs, and the receiving party typically adopts a 16-times oversampling strategy, sampling near the center point of each bit. If the baud rate deviation between the two parties exceeds ±2%, the sampling points will gradually shift, eventually leading to frame errors. In Lao Zhang's case, the deviation was close to the tolerance limit, and the problem worsened under high-temperature or electromagnetic interference environments.
Solutions:
Prioritize the use of external crystals (such as 16 MHz) with an accuracy of ±50 ppm, far superior to the ±2% accuracy of internal RC oscillators.
Select UART controllers that support fractional frequency division (such as the USART_BRR register in STM32).
In critical scenarios, use automatic baud rate detection functions (such as those supported by some STM32L4 series, which can deduce the baud rate through specific bytes).
At another automotive component factory, Engineer Xiao Li found that the newly deployed serial communication system frequently experienced packet loss on rainy days. After three days of troubleshooting, it was discovered that the grounding wire between the control cabinet and the equipment was not reliably connected. When the damp environment increased the grounding resistance, the common-mode interference voltage exceeded the anti-interference threshold of the UART chip, causing data errors.
Technical Analysis:
Serial communication is essentially the transmission of voltage level signals. If the two parties do not share a common ground, the signal reference levels will be inconsistent. For example, what the sending party considers to be "0 V" may appear as +2 V at the receiving party, leading to logical judgment errors. In industrial settings, motor startups, relay operations, and other activities generate common-mode interference. Without proper grounding, this interference will directly couple into the communication lines.
Solutions:
Use shielded twisted-pair (STP) cables and ensure that the shielding layer is reliably grounded.
Install independent grounding electrodes near both communication parties, with a grounding resistance of ≤4 Ω.
Use isolated serial communication modules (such as ADI's ADM3251E) to break the ground loop.
In Lao Zhang's production line renovation, in addition to the baud rate issue, there was another fatal flaw—hardware flow control was not enabled. When the industrial computer sent a large amount of configuration data to the PLC, the PLC's receive buffer quickly filled up, but the computer continued to send data, causing the buffer to overflow and subsequent data to be discarded.
Technical Analysis:
Hardware flow control coordinates the communication rates of both parties through RTS/CTS (Request to Send/Clear to Send) signals. When the receiving party's buffer is nearly full, it pulls the CTS signal low, and the sending party detects this and pauses sending. Once there is space in the buffer, the CTS signal is pulled high again, allowing transmission to resume. This mechanism can completely avoid buffer overflow and does not rely on software implementation, offering extremely high reliability.
Solutions:
Prioritize the selection of industrial computers that support hardware flow control (such as the USR-EG628, which comes standard with two RS485 ports, both supporting RTS/CTS).
Correctly connect the RTS/CTS signal lines (note the polarity; usually, RTS is connected to the other party's CTS, and CTS is connected to the other party's RTS).
Enable the hardware flow control option in the device driver (such as setting the crtscts flag in Linux).
For some older devices that do not support hardware flow control, software flow control (XON/XOFF) is a viable alternative. At a textile machinery factory, engineers modified the device firmware to implement XON/XOFF-based flow control: when the receiving party's buffer has less than 20% of its remaining space, it sends the XOFF character (0x13) to pause sending; once the buffer has more than 50% of its space available, it sends the XON character (0x11) to resume transmission.
Technical Analysis:
Software flow control uses special characters to control the data flow, eliminating the need for additional signal lines but occupying communication bandwidth (each flow control interaction requires sending 2 bytes). Its reliability is lower than that of hardware flow control, but it performs well in low-speed communication (≤9600 bps) or simple systems.
Solutions:
Reserve flow control characters in the communication protocol (to avoid conflicts with data).
Set reasonable buffer thresholds (usually 20%-80% of the buffer size).
Implement flow control character filtering at the sending party (to prevent data containing XON/XOFF from being misidentified).
In Lao Zhang's case, all communication issues were ultimately resolved by replacing the industrial computer with the USR-EG628. This device has been deeply optimized for industrial communication pain points, from hardware design to software architecture:
Accurate Clock System:
Utilizes the industrial-grade RK3562J chip with a built-in high-precision crystal (±50 ppm) and supports UART controllers with fractional frequency division, ensuring a baud rate error of ≤0.1% and completely eliminating packet loss caused by clock deviations.
Full-Featured Flow Control Support:
Comes standard with two RS485 ports, both supporting hardware flow control (RTS/CTS) and software flow control (XON/XOFF), allowing for flexible adaptation to different devices. In Lao Zhang's production line, enabling hardware flow control prevented buffer overflow even under high loads.
Anti-Interference Design:
Employs magnetic isolation technology to isolate serial port signals and features a shielded twisted-pair interface, capable of withstanding 10 kV/m of electromagnetic interference. Communication remains stable and reliable even on rainy days or during motor startups and stops.
Intelligent Protocol Processing:
Built-in WukongEdge edge intelligence platform supports over 100 industrial protocols, including Modbus RTU/TCP, Profinet, and EtherNet/IP. It can automatically identify device protocols and optimize communication parameters, reducing manual configuration errors.
Visual Operation and Maintenance Tools:
Through an HDMI interface, an external screen can be connected to monitor the serial communication status in real-time (such as baud rate, flow control status, and error counts), enabling quick fault location. Now, Lao Zhang can determine whether the issue is a baud rate mismatch or unenabled flow control with just a glance at the screen.
In the three months since replacing the industrial computer with the USR-EG628, Lao Zhang's production line has not experienced any serial communication packet loss issues. The delivery time of mechanical arm movement instructions has been shortened from an average of 50 ms to 15 ms, and the temperature sensor data collection cycle has been stabilized at 200 ms from 1 second, resulting in a 12% increase in production line efficiency. To his surprise, through the protocol analysis function of the WukongEdge platform, he also discovered a previously unnoticed communication conflict issue—an older device with a protocol implementation flaw would periodically send abnormal frames that interfered with other devices. By adjusting the communication time slots of this device, this hidden danger was completely eliminated.
This revolution in communication quality has not only solved the immediate packet loss problem but also shown Lao Zhang the infinite possibilities of the Industrial Internet of Things. Now, he is planning to use the edge computing capabilities of the USR-EG628 for predictive maintenance of the production line—by analyzing tiny fluctuations in device communication data, he can detect potential failures in advance and reduce unplanned downtime by another 50%.
In the field of industrial automation, communication quality directly determines production efficiency and equipment safety. A seemingly simple issue of serial communication packet loss may hide multiple pitfalls, such as baud rate mismatches, poor common grounding, and missing flow control. To solve these problems, it is not only necessary to have a deep understanding of communication principles but also to choose reliable equipment that has been rigorously tested in harsh industrial environments.
The emergence of the USR-EG628 industrial computer provides engineers with a key to unlock communication challenges. With its accurate clock system, full-featured flow control support, and powerful anti-interference capabilities, it has redefined the stability standards of industrial communication. When communication no longer becomes a bottleneck for production, engineers can devote more energy to innovation and optimization, driving industrial automation to a higher level.
As Lao Zhang said at the project summary meeting, "In the past, we were fighting against communication failures; now, we can finally focus on production itself." This is perhaps the most authentic testament to industrial communication optimization—making technology invisible and letting production shine.
