Profibus faults have a reputation for mysticism, mostly because the network hides its physical state. Method beats mysticism.
Physics first
Profibus termination is powered — the classic DP connector’s switch connects termination resistors fed from the device’s 5 V. Consequences people forget: a terminating device that is powered down un-terminates the bus (the “everything died when we switched off one panel” case), and the two ends — exactly two — must have switches ON with every in-between connector OFF. Stubs (dangling programming ports included) count against you at high baud rates; 12 Mbit/s tolerates nearly nothing.
The method
- Read the master’s diagnostics before touching hardware: which stations fail, all-or-some, permanent-or-flapping. A block of failing addresses maps to a cable segment; random singles map to local problems.
- Split the segment: terminate mid-bus (a spare connector with the switch ON), isolating halves. The fault stays with one half; repeat — binary search finds the bad joint in minutes on long trunks.
- Inspect the classic offenders: connectors after panel work (screwed onto insulation, swapped A/B), cable crushed by a new cable tray, water in an outdoor junction, and the repeater everyone forgot exists.
- Counters over anecdotes: rising retry/error counters on one station’s ports indict its neighborhood; a bus-wide error rate that tracks production activity smells of EMC coupling from drives.
FAQ
Can I mix baud rates on segments? Not within one DP system — all stations follow the master. Repeaters extend, they do not translate speed.
Handheld bus testers worth owning? For plants living on DP for years yet: yes — signal-quality numbers per station turn arguments into work orders.
Zone Otomasyon still speaks fluent Profibus — someone has to, for another decade at least. Legacy bus support.