One drive on a roughly 30-node DLR is responsible for the rapid-fault count. With the VFD in the ring (7th device downstream of the supervisor), its port increments about 10 MAC errors per second. With the drive jumped out, the rest of the ring logs about one error every 10 minutes. A full drive replacement, including the backplane and network card, left the rate unchanged. The fault is therefore in the link settings at that ring position, in the drive's two ring neighbors, or in the electrical environment around the cable. It is not in the drive hardware.
Beacon loss and the rapid-fault counter
A DLR supervisor sends beacon frames in both directions around the ring and watches for them to return. When beacons stop arriving, the supervisor declares the ring broken and unblocks its second port. When beacons resume, it restores normal operation. A link that corrupts frames intermittently produces this break-and-restore cycle over and over, and the supervisor records it as a rapid fault.
The deciding quantity is the corrupted-frame rate on one link. Every frame that fails its checksum is discarded at the receiving port. At 10 bad frames per second, beacons regularly land in the discard window, so fault/restore cycles repeat.
Duplicate IP addresses do not cause this symptom. IP conflicts break CIP connections, not Layer 2 frame integrity. Ruling them out was correct, but it points the search toward the physical and data-link layers.
Error rates that decide the case
Link-layer counters separate a bad link from a bad device. The table lists what to read, the observed or target value, and where to find it.
| Quantity | Observed / limit | Where to read |
|---|---|---|
| MAC errors, VFD port | About 10 per second observed; target is no steady increment | RSLinx node diagnostics, or the device's Ethernet port diagnostics page |
| FCS errors, VFD port | Large count observed; target is a flat counter over the test window | Same page as MAC errors |
| Background errors, rest of ring (VFD jumped out) | About 1 per 10 minutes observed | Each node's port diagnostics |
| Late collisions / half-duplex state on neighbor ports | Any nonzero late collision on a ring port indicates a duplex mismatch | Port diagnostics of ring devices 6 and 8 |
| Negotiated speed and duplex, every ring port | Must be identical around the entire ring; the PowerFlex 525 E2P module is 100 Mb | Port configuration/status on each device |
| Supervisor rapid-fault and ring-fault counts | Must stop incrementing after the correction | Supervisor DLR diagnostics in the programming software or device web page |
Symptom-to-cause mapping at the drive position
| Symptom | Mechanism | Discriminating test |
|---|---|---|
| FCS/MAC errors on one side of the link, late collisions on the other | Duplex mismatch: one end forced to full duplex, the other auto-negotiating and falling back to half duplex | Compare negotiated duplex on the VFD port and the facing neighbor port |
| Errors persist after a complete drive swap | The replacement comes up at factory auto-negotiate while the neighbors keep their forced settings, so the mismatch returns with the new hardware | Read the settings on the replacement drive versus devices 6 and 8 |
| Errors rise when the motor runs and fall when it stops | PWM output edges couple common-mode noise into Ethernet pairs routed parallel to motor leads | Keep the drive powered and stopped, then run it, and compare counter rates |
| Link is unstable or comes up at an unexpected speed | Two Gb-capable devices negotiate 1 Gb in the middle of a 100 Mb ring | Inventory the PHY capability of every ring member and read the negotiated speed |
| The same symptom appears in almost every control panel on site | A systemic cause: a site-wide port configuration standard, or a repeated panel wiring practice | Check whether affected panels share a configuration template or cable routing layout |
| Errors stop when the VFD is jumped out | The fault is local to the drive's two links or its cable run | This result is already established for this ring |
Link configuration strategies compared
Auto-negotiation sends link pulses that advertise speed and duplex. A port forced to fixed settings stops advertising. When an auto port faces a forced port, it uses parallel detection. Parallel detection identifies the speed correctly but cannot determine duplex, so the auto port defaults to half duplex.
The result is the classic mismatch. The half-duplex side logs late collisions. The full-duplex side logs FCS and runt errors. DLR beacons flow continuously in both directions, so the errors accumulate even when application traffic is light.
| Strategy | How speed/duplex is set | Failure mode | Fit for this ring |
|---|---|---|---|
| Auto-negotiate on every port (factory default) | Each link pair negotiates independently | Two Gb-capable neighbors can negotiate 1 Gb mid-ring | Good when no ring member has a Gb-capable PHY; replacement drives come up already matching |
| Forced 100 Mb full duplex on every port | Configured on every device | Any device reset to defaults or swapped in returns to auto and creates a mismatch | Good when discipline is enforced; matches the 100 Mb E2P module; blocks Gb negotiation |
| Mixed forced and auto ports | Inconsistent | Duplex mismatch at every forced/auto boundary | Never acceptable |
The rule that cannot be broken is that every port runs one speed, and both ends of every link match. Choose auto-negotiate ring-wide when no Gb-capable device sits in the ring. Choose forced 100 Mb full duplex on every port, or 2-pair cable, when one does.
A factory reset of the PowerFlex 525 returns its port to auto-negotiate. On a forced ring, that single reset puts the mismatch back.
Noise coupling versus configuration faults
A configuration fault is time-invariant: the error rate is roughly constant whenever the link is up, whether the motor turns or not. A noise fault tracks the power stage. The rate climbs with drive output and drops toward zero when the inverter stops switching. This one test tells the two classes apart, and it takes minutes.
The multi-panel pattern and the remaining background rate (about one error per 10 minutes with the VFD out) both point to some coupling across the site. That coupling is separate from the dominant fault at the drive. Network cables still run alongside motor cables in these panels.
Common-mode current from the inverter's fast voltage edges couples capacitively and inductively into long parallel runs. It enters twisted-pair Ethernet as common-mode noise that the receiver cannot fully reject. Separate Ethernet from motor leads, cross them at right angles where they must meet, and terminate cable shields per the drive installation manual.
Recommended correction sequence
Fix the link configuration first, because the full hardware swap already failed and configuration is the only local cause the swap cannot clear. Then remove noise coupling. Then decide whether the drive belongs on the ring at all.
- Record the configured and negotiated speed and duplex on the VFD port and on the facing ports of ring devices 6 and 8.
- Inventory every ring member for Gb-capable ports. Treat any 1 Gb negotiated link as a fault on a ring that includes 100 Mb E2P modules.
- Choose one ring-wide setting using the decision in the strategy table: auto everywhere, or forced 100 Mb full duplex everywhere.
- Apply the setting to the drive and both neighbors. If the site standard is forced, reapply it after any drive replacement or factory reset.
- Force renegotiation on each changed link by cycling the port or reseating the cable. Confirm both ends report the same speed and duplex.
- Clear the error counters on the drive and both neighbors.
- With the drive powered and stopped, watch the MAC and FCS counters over a fixed interval.
- Run the motor across its normal speed range and watch the same counters over the same interval.
- If errors appear only while running, reroute the Ethernet run away from motor leads and correct shield termination. Then repeat steps 6-8.
- Repeat the configuration audit in every affected panel. A shared template is the likely site-wide cause.
Moving drives off the ring onto a Stratix star
A DLR survives one break. Two breaks isolate every node between them. With drives embedded in the ring, powering down any two drives for maintenance drops every device between them, even though no cable failed.
Connecting drives in a star to a Stratix managed switch, with the switch itself on the ring, removes that exposure. A drive power-down then affects only that drive. The switch also gives per-port counters and fixed port configuration in one place.
A ring earns its keep in distributed architectures: many small local panels, each wired to its own island of I/O, daisy-chained instead of home-run to a central switch. Drives are effectively single I/O points, usually grouped in large panels or MCCs near the controller. In that layout a ring saves little cabling over direct switch runs.
A failed device, which is more common than a failed cable, gets no more protection from DLR than from a star. Full Ethernet redundancy is PRP, which carries a large upfront cost and needs dedicated IT staff for configuration.
Counter baselines that prove the ring is stable
A clean result is a flat counter, not merely a lower one. After the correction, the VFD port's MAC and FCS counters must not increment over a stopped-drive interval or a running-drive interval of equal length. The rate must drop from about 10 per second to the ring's background level or below.
Neighbor ports must show zero late collisions and matching duplex. The supervisor's rapid-fault and ring-fault counts must stay constant through a full production cycle, including motor starts and stops.
Finally, break the ring deliberately at one cable. Confirm that the supervisor reports a single fault and a single restore when you reconnect, with no repeated cycling. That proves beacon integrity end to end.
FAQ
Can I mix forced 100 Mb full duplex and auto-negotiate ports on a DLR?
No. An auto port facing a forced port detects the speed by parallel detection but falls back to half duplex. The result is late collisions on one end and FCS errors on the other, which break beacons and trigger rapid faults.
Does a factory reset of a PowerFlex 525 change its Ethernet link settings?
Yes. A reset returns the port to auto-negotiate. On a ring where the neighbors are forced to 100 Mb full duplex, that reset recreates a duplex mismatch until you reapply the forced settings.
Can I use 2-pair Category 5 cable to stop Gb negotiation on a DLR?
Yes. Some combinations of two Gb-capable devices fail to negotiate at all over 2-pair cable, so verify each link where both ends are Gb-capable.
Does replacing the drive fix DLR rapid faults that follow the drive position?
Rarely. A replacement comes up at factory auto-negotiate, and any mismatch with forced neighbors or any noise coupling in the cable run stays in place. If the counters still increment with speed and duplex verified identical on both ends, the drive stopped, and the Ethernet run separated from the motor leads, stop swapping hardware. Open a case with Rockwell Automation technical support and bring the port counter logs, the negotiated settings of every ring member, and the supervisor's ring-fault history.