Troubleshooting
Fix robot Profinet output writes over OPC UA
Short answer
Set the Profinet Internal Device Input Size and Output Size to 64 bytes each, under Controller tab, Configuration, I/O System, Profinet Internal Device, and robot output writes reach the PLC array. The failing direction, not the OPC UA status number, identifies the cause: a Bad-class write with the other direction working means the output byte image is missing.
On this page
Set the virtual controller's Profinet Internal Device to a 64-byte image in both directions and the output writes will land in the PLC array. The failing setup is easy to recognize: writing the PLC array drives the robot's Profinet input signals, but setting a robot output signal leaves the PLC array flat while the RobotStudio OPC UA client reports Bad result writing OPC UA data for node 2 (status: 2155085824). That one-way split tells you the connection, session and node resolution all work — the output byte image behind the mapping does not exist yet. Two system parameters create it.
Why one working direction rules out the usual suspects #
A rejected write alongside a live read path narrows the fault immediately. If the server were unreachable, the session dead or security misconfigured, both directions would fail together and nothing would move. Here a PLC-side array write reached the robot's input signals, so the Siemens OPC UA server accepted a mapped node and the client authenticated against it. The only difference between the two paths is the data image behind each. The internal device PN_Internal_Device carries exactly two system parameters, Input Size and Output Size, as ABB's PROFINET Controller/Device application manual (document 3HAC065546-001) defines it, and each direction gets the bytes its parameter declares. Map signals into a direction whose size declares nothing, and the write is aimed at a node with no image behind it.
What status 2155085824 does and does not tell you #
Do the arithmetic before trusting any decode of that number: 2155085824 − 2147483648 = 7602176, and 7602176 = 0x740000, so the full status word is 0x80740000. IEC 62541, the OPC UA specification, puts severity in the top bits of the status word, and anything from 0x80000000 upward is a Bad code: the server rejected that one write; it is not a timeout and not a link fault. A public issue on the node-opcua client library reports the same numeric status returning a type-mismatch description, which fits a write aimed at a node whose backing data does not match the request. The number will not name the missing image for you — the direction test above is what does. That same first cut, dead link versus dead data path, is the one why HMI values show question marks: PLC comms timeouts applies to clients showing placeholders.
Set both sizes to 64 bytes #
- In RobotStudio, select the virtual controller and open Controller tab → Configuration → I/O System → Profinet Internal Device.
- Set Input Size to 64 and Output Size to 64. The parameters are byte counts, so each direction gets an image of 64 bytes; that is the example image the manual itself uses, and it is the value that restored both directions in this configuration.
- Apply the configuration the way the workflow requires so the controller loads the new image, then reopen the OPC UA client mapping and confirm the input array is assigned to the input direction and the output array to the output direction.
- Set one output signal and watch the mapped DB element before calling it fixed. A follow-on message,
Unable to write robot input data on device: PN_Internal_Device, appeared in the working case while the data moved, so judge the repair on the array, not on that text.
On a production IRC5 rather than a virtual controller, the same edit changes the telegram length the PLC hardware configuration expects, so make it as a coordinated change with the cell in a safe state.
A worked check from signal address to array element #
The image set above carries 64 × 8 = 512 signal addresses, numbered 0 through 511, and maps onto a USINT array of 64 elements at 1 byte each, indices 0 to 63. Addresses in the robot's I/O configuration count bits from the start of the device image. Take an output signal at address 20: 20 ÷ 8 = 2 remainder 4, so it is bit 4 of byte 2, which is array element [2], bit 4. Switch it on and element [2] should read 0 + 2^4 = 16; switch it off and it drops back to 0. Repeat at a distant address: 100 ÷ 8 = 12 remainder 4, element [12] bit 4, value 16 again. Two elements moving at opposite ends of the image prove the mapping spans it. Neither moving after the size change points back at the mapping rows, not at the sizes.
Why array type changes were never going to fix it #
The reported attempts to use other PLC array types never even entered the RobotStudio mapping. That is expected: the Siemens OPC UA server exposes a mapped array as one node covering the whole array, read or written in a single operation, so element type and array bounds must line up with the byte image the client maps. A type change reshapes the node; it cannot conjure the missing bytes. A link carries exactly what its mapping exposes and nothing more — the same rule why Modbus cannot transfer a PLC ladder program over TCP states for a different protocol.
The values that decided this case #
| Item | Value in the working configuration |
|---|---|
| Input Size | 64 (bytes) |
| Output Size | 64 (bytes) |
| Failing write status before the change | 2155085824 (0x80740000) |
| Diagnostic after the change | Unable to write robot input data on device: PN_Internal_Device |
| Mapped PLC data | two USINT arrays, one per direction, in one DB |
The decision #
PLC-to-robot data flowing while robot-to-PLC writes fail with a Bad status: set both internal device sizes to the value above and retest, because it is the manual's example image and the value reported working in this exact setup. Both directions dead: the sizes are not your first suspect; check the mapping rows and the server node access first, since the size fix presupposes one live path. Array updating while the robot-input diagnostic persists: log the message, keep the verified data path in service, and take the recorded set to the vendor whose side rejects the write.
Questions asked
How do I size the internal device when my signals need more than 64 bytes?
The sizes are byte counts. Take the worst-case signal count per direction, divide by eight and round up: 120 digital signals need 120 ÷ 8 = 15 bytes, so any image of 15 bytes or more carries them. Stay at the example value unless your byte count demands more, because the PLC array and the client mapping grow with the image.
Which data type should the mapped PLC arrays use?
Use one USINT array per direction, as the working configuration did. A USINT is one byte, so the element index and the byte offset are the same number and the mapping arithmetic stays direct. Wider element types break that one-to-one offset match, and in the reported case alternative types could not be added to the RobotStudio mapping at all. Keep input and output in separate arrays in one DB.
Do these size settings apply to a real IRC5 controller as well?
Yes. Input Size and Output Size are documented system parameters of PN_Internal_Device in ABB's PROFINET Controller/Device application manual for IRC5 controllers, not a virtual-controller-only setting. On a production cell the same edit changes the telegram length the PLC hardware configuration expects, so coordinate it with the PLC side and test it with the cell in a safe state.
What should I collect for support if the array still will not update?
Send the exact diagnostic text including the status number, both configured sizes, the DB array definitions with bounds and element type, the client mapping rows showing which array carries each direction, and a repeatable test: one signal address, the element index you expect it to hit, and the element value before and after. Route it to the robot vendor or the PLC vendor depending on which side reports the rejection.