Troubleshooting
Connecting a CompactLogix PLC to an HMI over EtherNet/IP
Short answer
A CIP transport unsupported fault on a Generic Ethernet/IP module in a CompactLogix project means the connection was built the wrong way around. For HMI tag access, delete the module, select the panel's Rockwell PLC driver, point it at the controller IP, and rebuild the tags — no assembly instances, I/O sizes, or EDS file.
On this page
A CIP transport unsupported fault on a Generic Ethernet/IP module is not a settings problem — it is the wrong architecture. A CompactLogix L16ER exchanging tags with a Pro-face GP4000 needs no module in the controller's I/O tree, no assembly instances, no I/O sizes and no EDS file. You select the Rockwell PLC family driver in the HMI software, point it at the controller's IP address, delete the generic module, rebuild the tag list, and the connection works; that exact sequence cleared this fault in the field. The module path exists for a different job, and this article separates the two, gives the fix in order, and shows the one arithmetic check that tells a path problem from a tag problem.
What the fault actually says #
In RSLogix 5000, the Generic Ethernet/IP module entry makes the controller the client of an I/O connection: it tries to open a connection to the configured target using the Input instance, Output instance and I/O sizes you typed. The GP4000 on the other end does not serve that connection type at those instances, so the CIP request is refused and the module reports transport unsupported. The fault describes the module's request, not the panel's ability to exchange data — proven by the tag connection working the moment the module is gone.
There is a second cost. Every ETHERNET-MODULE or ETHERNET-BRIDGE entry in the I/O tree has the controller open a connection to that target, successful or not, so a dead module entry is never free.
Two architectures and one decision #
| Path | Who is the client | What you configure | Right for |
|---|---|---|---|
| Generic Ethernet/IP module (ETHERNET-MODULE profile) | Controller | Input instance, Output instance, I/O sizes, taken from the target device's documentation | Remote I/O and adapters that publish those numbers |
| HMI tag driver (Rockwell PLC family in the panel software) | Panel | Controller IP address, tag names and data types | Screens that read and write controller tags |
The decision for a GP4000 displaying and writing L16ER data is the second row, always. The panel is a CIP client doing tag reads and writes; nothing in the controller needs to represent it. The same job on the Siemens side is a connection to the controller's IP in the engineering software, not a PLC-side module either — see connecting an HMI panel to an S7-1500 PLC. Direction confusion is not unique to EtherNet/IP; if you later push data upward over OPC UA, settle HMI OPC UA server vs client: which direction works before configuring anything.
The fix in order #
- Back up both projects first — the controller program and the panel project — because the next steps delete controller configuration and rebuild panel tags. Follow the discipline in how to back up a PLC program the right way.
- In the panel software's communication setup, select the Rockwell PLC family for the CompactLogix and enter the L16ER's IP address. No instance numbers appear on this path, which is correct.
- Delete the Generic Ethernet/IP module from the controller's I/O tree, unless another documented I/O device depends on it.
- Rebuild the panel tag list against the real controller tags: same names, same data types, same array indices.
- Download both projects and test representative reads and writes, including the tag intended for the cycle-stop request. You can exercise the tag list against the live controller without involving the machine, the way testing HMI simulation against a live PLC describes.
- Watch runtime: values track, writes land, and the transport unsupported fault no longer appears because the module that generated it no longer exists.
One arithmetic check: same subnet or not #
Before blaming tags, prove the path. Example addresses, illustrative and not from your network: panel at 192.168.1.31, controller at 192.168.1.87, mask 255.255.255.0.
192.168.1.31 AND 255.255.255.0 = 192.168.1.0
192.168.1.87 AND 255.255.255.0 = 192.168.1.0
Both hosts land on the same network, so a plain switch connects them and no router is involved. If the panel still cannot reach the controller, the fault is in the path — VLAN assignment, an isolating firewall, the wrong physical port, an address conflict — and no tag setting will touch it. Move the controller to 192.168.2.87 and the second line becomes 192.168.2.87 AND 255.255.255.0 = 192.168.2.0: the networks now differ, so the panel needs a corrected address or a routed path, and a ping test will say which. Ten seconds of AND-ing octets prevents an hour of pointless tag editing.
If the driver connects but data does not move #
- A driver that connects with unresolved tags is a naming problem: check spelling, case, array indices and data types against the controller's tag database.
- Tags that read but never change point at the value being written elsewhere or not scanned: confirm the controller is in Run and the logic actually moves the value.
- A driver that will not connect at all sends you back to the subnet and path check above.
- The panel's communication status display separates "never connected" from "connected, tag errors"; use that split before changing anything else.
Where the module path and the EDS file do belong #
Use a Generic Ethernet/IP module when the design genuinely needs a controller-owned I/O connection — cyclic I/O to a remote device — and that device's documentation lists its Input instance, Output instance and sizes. The numbers come from that document and nowhere else; inventing them either fails again or exchanges the wrong bytes with a device that did accept the connection. An EDS file is the vendor's machine-readable version of that description, which is why configuration tools ask for one on the module path and never on the tag path. The ODVA catalog turning up empty for this panel was never the blocker it looked like: the working architecture does not consult it.
One panel, several controllers #
Nothing in the tag path limits you to one controller: the panel software takes one driver connection per controller IP, each polling its own tag list. Vendor guidance for EtherNet/IP clients draws one line worth respecting — when a single panel must serve several PLCs, another protocol is the better carrier, and EtherNet/IP tag access belongs on the panel-to-controller links that need its speed. One GP4000 to one L16ER is squarely the supported use.
Keeping the cycle-stop request honest #
The plan behind this connection — the panel writes a request tag, the L16ER decides and asks the machine for a graceful stop — is sound, with one boundary. The HMI tag write is a request that verified L16ER logic acts on; it is never the safety function itself, and safety stopping stays with the dedicated safety controller that already owns it. Map every displayed and commanded value to the controller that owns it before commissioning, and prove the stop path end to end through the machine's approved procedure.
Questions asked
Where do the assembly instance and I/O size numbers come from?
From the documentation of a device that actually serves an I/O connection: the vendor lists the input instance, output instance and byte sizes the device supports. A GP4000 used for tag access publishes none, which is why the search comes up empty. The tag path consumes no instance numbers at all, so an empty search means you are on the wrong path, not that a manual is missing.
Do I need an EDS file for the panel or the controller?
No, not on the tag path. An EDS file describes a device's I/O connection options to configuration tools, which is why the module route asks for one. The driver route needs the controller's IP address and correct tag names, nothing else. A missing EDS entry in the ODVA catalog is irrelevant to the architecture that works here.
Can I keep the module and fix it by trying different instance values?
No. The controller is requesting an I/O-class connection the panel does not serve, so the transport unsupported fault repeats regardless of the numbers. Remove the module if it exists only for the panel. On a device that does accept an I/O connection, guessed instances can silently exchange the wrong bytes, which is worse than a clean fault.
Will the L16ER pass data between the panel and the machine controller I cannot program?
Only through logic you write in the L16ER. The panel connection gives read and write access to L16ER controller tags and nothing more. Map each displayed value to the controller that owns it, and put any exchange with the rest of the cell into L16ER code you can inspect and test. The graceful stop request belongs in that verified logic; safety stopping stays with the safety controller.