The most exhausting fault in PROFINET is the one that isn’t permanent. An IO device drops out a few times a week and comes back on its own; maintenance replaces the cable, the problem disappears for a while, then starts again. You do not escape this cycle by swapping parts. The order of checks we use for symptoms such as intermittent device dropouts, watchdog errors and “device not found” is this: a layered procedure that goes down from the engineering software to the physical layer and from there to packet analysis, eliminating one possibility at each step.
First classify the symptom: permanent, intermittent, load-dependent
Before starting diagnosis, write down the character of the fault. This classification determines which tool you will use:
- Permanent fault: The device never connects or constantly reports “device not found”. This is usually configuration (device name, IP, GSDML version, topology assignment) or a definite physical break.
- Intermittent fault: The device drops out and comes back within a few seconds. Typical causes are a loose connector, a damaged cable, corrupted frames caused by shielding/grounding problems, or traffic bursts that occur from time to time on the network.
- Load- or event-dependent fault: The problem appears only when a particular motor runs, at the start of a shift, at backup time or when an engineering PC is connected to the network. EMI or non-PROFINET traffic (IT traffic, multicast, broadcast storms) are strong suspects.
For each event, record the timestamp, the affected devices and what was happening on site at that moment. If several devices drop out at the same time, the problem is usually in the common path: the same switch, the same uplink or the link in front of the first device to drop out in the chain.
Online diagnostics and the alarm buffer in the engineering software
The first step is to read what the controller sees. Go online in TIA Portal and examine the diagnostic buffer; every device failure and return is listed here with a timestamp. If the controller clock is not synchronised with an NTP source, you cannot compare these timestamps with switch logs. This is one of the most frequently overlooked settings.
Make the following distinctions in the diagnostic buffer:
- Station failure: The controller received no data from the device for the watchdog time. The communication path is broken; the cause lies in the network or in the device’s power.
- Device-originated diagnostic alarms: The device sends alarms such as module failure, channel fault or missing supply using PROFINET alarm frames. These are not communication problems but states reported by the device.
- Port and synchronisation diagnostics: Many devices report per-port diagnostics such as link loss, neighbour mismatch and exceeded transmission error thresholds. These are direct warnings of physical layer problems.
Siemens’ free tool PRONETA scans the network without access to the engineering project and lists device names, IPs and topology information. In the field, it is the fastest answer to the question “which name does this device really respond to?”.
Switch port statistics and physical layer checks
Most intermittent faults in industrial network infrastructure start at the physical layer and leave traces in switch port counters. Reset the counters on the managed switch and read them again after a set period; this reveals which link is generating errors. Example commands on Cisco IOS-based industrial switches:
! Reset the counters, note the start of the observation period
clear counters GigabitEthernet1/0/5
! List error counters per port
show interfaces counters errors
! Single-port detail: CRC, input errors, runts, giants, late collision
show interfaces GigabitEthernet1/0/5
! Link state changes appear in the log
show logging | include GigabitEthernet1/0/5
When reading the counters, the following relationships are a guide. CRC and FCS errors point to a faulty cable, damaged connector, EMI or a shield continuity problem. Late collisions and an error count rising on one side only are the classic signs of a duplex mismatch. Both ends of a link must run in the same mode: on a typical PROFINET port, autonegotiation is enabled; where a fixed setting is needed, such as Fast Start-Up or some fiber links, both ends are fixed to 100 Mbit/s full duplex. Fixing only one side is the most common cause of this mismatch. Packets dropped on the output side (output drops) show that the uplink is congested or that QoS prioritisation is not working.
If you can access the counters of unmanaged switches or devices’ embedded switches via SNMP or the engineering tool, do the same analysis. Cable measurement is essential on a link whose counter shows errors. A continuity test is not enough. On a PROFINET line, NEXT, return loss and shield continuity should be checked with a cable certification tester. Grounding shielded cable at both ends and equipotential bonding are basic requirements of the PROFINET installation guidelines; PROFINET cables running in the same duct as frequency converter cables are a common source of EMI.
Compare the actual and planned topology with LLDP
PROFINET devices recognise their neighbours via LLDP. This information serves two purposes: comparing the planned topology in the engineering software with the actual connections on site, and automatically assigning the name to a new device during a replacement. The online comparison in TIA Portal’s topology view shows, side by side with the plan, which port is connected to which port of which device.
The findings to look for are: a switch or device not defined in the project inserted in between, cables that have changed ports (e.g. plugged into the third port instead of a ring port) and a chain order different from the plan. In a project with a defined topology and “neighbour check” enabled, a port change produces a diagnostic; the device may still work, but automatic device replacement does not work as expected in this case. On the switch side, neighbours are verified with this command:
show lldp neighbors
show lldp neighbors GigabitEthernet1/0/5 detail
Every connection that does not match the plan is either corrected or the project is updated. There is no middle ground. An undocumented change is what slows down diagnosis most at the next fault.
PROFINET packet analysis with port mirroring and Wireshark
If the previous steps have not narrowed down the problem, it is time for packet analysis. Set up port mirroring (SPAN) on the managed switch closest to the faulty device, or insert a network TAP between the device and the switch. Example SPAN configuration on Cisco IOS:
monitor session 1 source interface GigabitEthernet1/0/5 both
monitor session 1 destination interface GigabitEthernet1/0/24
If the mirror port cannot carry the total traffic of the source ports, it loses frames and leads you to a wrong diagnosis; prefer a TAP on high-traffic ports. PROFINET RT frames can be transmitted with a VLAN tag carrying priority information; because some network card drivers strip this tag, you may need to configure the capturing PC’s card to preserve tags.
Useful display filters in Wireshark (everything after the # sign is a comment and is not typed into the filter):
pn_dcp # Device name and IP assignment/identification (DCP)
eth.type == 0x8892 # All PROFINET RT frames
pn_rt.frame_id >= 0x0100 && pn_rt.frame_id < 0xfc00 # Cyclic data (RT/IRT)
pn_rt.frame_id == 0xfc01 || pn_rt.frame_id == 0xfe01 # High/low priority alarms
lldp # Neighbour information
pn_io # Connection setup and record reads (RPC)
Look at four things in the capture. Cyclic data timing: When you filter a single device’s cyclic frames by source MAC and examine the time between consecutive frames, significant deviations from the update time and gaps indicate network-related delay. DCP traffic: Unexpected DCP Set requests on a running line show that someone is changing a device name or IP; two devices responding to the same name is a name conflict. Alarms: Whether the device sent an alarm just before dropping out distinguishes whether the problem is inside the device or in the transmission path. Non-PROFINET traffic: If you see heavy broadcast, multicast or large file transfers at the same time, the problem lies in network separation. For general packet capture techniques, the network packet analysis with Wireshark guide and the Wireshark documentation are good starting points.
Symptom, possible cause and check table
| Symptom | Likely cause | Confirming finding / check |
|---|---|---|
| Device not found | Device name or IP mismatch | Read the actual name with PRONETA/DCP |
| Two devices dropping out alternately | Duplicate device name or IP conflict | Compare the responding MACs with pn_dcp |
| All devices further down the chain drop out | The link in front of the first device to drop out in the chain | Port counters and cable test of that link |
| CRC/FCS counter rising | Cable, connector, EMI, shielding problem | Certification measurement, routing and grounding |
| Late collision | Duplex mismatch | Autonegotiation setting on both sides |
| Dropouts at certain times | Backup, IT traffic, multicast | Output drops and packet capture at that time |
| Dropout when a cable is pulled despite the ring | Watchdog < ring recovery time | Watchdog and MRP settings |
For findings related to ring configuration, repeat the break test in the article PROFINET MRP ring topology. Design-related root causes usually coincide with PROFINET network design mistakes.
After the fault is closed: permanent monitoring
The job is not done once the root cause has been found and fixed. To catch the same problem faster if it recurs, make monitoring permanent:
- Periodically collect the port error counters and link state changes of managed switches via SNMP; define thresholds for rising trends.
- Send switch and controller logs to a central server with synchronised time.
- Leave a permanent TAP or mirroring point on critical uplinks; at the next event you can capture without unplugging cables.
- After every change, compare the actual topology, device names and IPs with the project and update it.
- Archive the symptom, root cause and solution of each fault in a short record; for intermittent faults, past records reveal the pattern.
For the design of this monitoring setup and ongoing fault analysis, see our industrial network monitoring and fault analysis service page, and for reorganising PROFINET projects, the PROFINET and PROFIBUS communication page.
Frequently asked questions
Where should PROFINET fault diagnosis start?
First, read the time of the fault and the affected devices from the controller’s diagnostic buffer. Then move on, in this order, to the switch port counters along the common transmission path, the LLDP topology comparison and, if necessary, packet capture.
Why is PROFINET traffic not visible in Wireshark?
The capture point may be wrong; a switch forwards traffic only to the relevant ports. Use port mirroring or a TAP. In addition, some network card drivers strip the VLAN tag; you may need to configure the card to preserve tags.
What do CRC errors indicate?
A CRC or FCS error shows that the frame was corrupted in transit. Typical causes are a damaged cable or connector, shielding and grounding problems, and electromagnetic interference from routes close to power cables.
Is a “device not found” error always a cable fault?
No. The most common cause is that the device name or IP address does not match the project. A name not being assigned after a device replacement, or two devices carrying the same name, also produces this error.