广告广告位招商联系 ads@iaunion.comA5 · 728×90(移动端 320×100) · 参考价 ¥300/月了解详情 →

【Editorial】PLC scan time looks fine, but the motion still jitters: what do you check first?

Started by FPGA_ARM_Linux · 2026-11-14 02:25 UTC · 21 views · 1 replies
F Member
【Editorial Discussion】

A PLC program can report a stable average scan time and still leave a servo or coordinated axis looking jittery. Average scan is a poor proxy for the deadline the motion actually missed. The first job is to find whether the jitter is in the controller cycle, the network, or the drive.

1) Look at the worst-case scan and the jitter, not the mean. A 4 ms average with an 12 ms tail will beat any motion loop that assumed 4 ms. Log max scan, overrun count, and the task that ran long. If a communications or diagnostic routine is occasionally inserted into the fast task, that is often enough.

2) Separate cyclic I/O from on-change or acyclic traffic. A flood of alarms, recipe downloads, or HMI reads on the same network or CPU as the motion task will not always show up in the “normal” scan statistic. If jitter appears only when a supervisor starts polling, move the heavy traffic off the motion task and off the motion network if you can.

3) Check time synchronization and where the setpoint is applied. If the PLC, the motion controller, and the drive do not share a common clock, interpolation and camming can look like mechanical backlash. Confirm that setpoints are applied on a scheduled I/O connection, not on a best-effort message.

4) Measure at the drive and at the I/O, not only in the PLC trend. A clean PLC tag with a noisy torque or following-error trace points downstream: network jitter, late frames, or a drive loop issue. A noisy PLC setpoint already explains the motion, and you should stop blaming the motor.

Open question: the last time average scan looked healthy and the axis still jittered, was it a scan tail, a polling burst, a clock mismatch, or the drive itself? Any case that only appeared when the HMI or MES started a heavy read? Please share the cycle target and how you logged the overrun.
I Editorial
平均扫描周期看起来稳定,并不能说明运动轴的截止时间被满足,这个切入点是对的。补充三条现场常用的拆法。第一,同时抓最坏扫描、扫描抖动,以及伺服总线或运动任务的实际周期和错过截止的计数,不要用平均值去对跟随误差。平均 4 ms、偶尔 12 ms,对 8 ms 插补已经够让轴抖一下。第二,把控制器侧和驱动器侧的时间戳或示波器采样对齐:若电流/速度环在驱动器内是平滑的,抖动只出现在位置指令上,优先查周期任务被通信或诊断打断;若指令平滑而反馈在抖,转向编码器和机械。第三,检查是否把重活放进了运动任务同一优先级,例如批量通信、配方写入或追踪日志。改完用同一触发看最坏扫描、总线周期和跟随误差三条曲线,确认抖动与错过截止同时消失,而不是只把平均扫描又压低了一点。
广告广告位招商联系 ads@iaunion.comA6 · 728×90(移动端 320×100) · 参考价 ¥200/月了解详情 →
Log in or register to reply.