IMDHMIControls LibraryShop main store

How-to

How a SCADA system reads EtherNet/IP adapter sensor data

SCADA and data logging Published 9 min read

Short answer

A tag-based SCADA client cannot browse an EtherNet/IP adapter, because an adapter publishes fixed I/O assemblies and has no tag server. Run a scanner-class client on the gateway and configure the EDS-declared assembly instances and RPI, or let a PLC own the Class 1 connection and read that PLC's tags.

On this page
  1. Why can't a tag-based SCADA driver read an EtherNet/IP adapter sensor?
  2. Which path gets sensor values into SCADA: a scanner client or a PLC in the middle?
  3. What does the EDS file tell you before you configure anything?
  4. How do you configure the scanner connection step by step?
  5. What RPI and timeout settings keep the link stable?
  6. How do you verify the values and prove a dead link goes bad?
  7. Which network settings quietly break an adapter connection?
  8. So which setup do you run for your sensor?
  9. Questions asked

An EtherNet/IP sensor controller is an adapter, not a tag server: it publishes its measurements in fixed I/O assemblies and waits for a scanner to open a Class 1 connection. A tag-based SCADA driver browses a PLC's symbol table for named tags, finds none, and reads nothing. Two paths work. Run a scanner-class client on the SCADA gateway and configure the input, output and configuration assembly instances and the RPI exactly as the device's EDS file declares them, or let a PLC own the Class 1 connection and read that PLC's tags with the driver you already have. Prove either path by comparing SCADA values against the sensor controller's own display and by confirming that every tag reads bad quality within the connection timeout.

Why can't a tag-based SCADA driver read an EtherNet/IP adapter sensor? #

Because the sensor controller is an adapter: it holds its measurements in fixed Assembly object instances and answers only a scanner's Class 1 connection, so a driver that browses for named tags finds nothing to list and nothing to read.

EtherNet/IP carries two kinds of traffic. Implicit, Class 1 messaging opens a connection with a Forward_Open request, then exchanges cyclic I/O packets over UDP at the RPI. Explicit messaging is an unconnected request-and-response exchange, and the tag and symbol services that let a driver browse live in PLC-class devices, not in a sensor adapter.

That is why the driver family matters more than the driver version. A tag-based driver is an explicit client: it lists a PLC's tags, then reads and writes them by name. The adapter instead implements Assembly objects — instance numbers with fixed byte layouts — and no listing service. Nothing in a software update changes that, so the fix is a client that speaks assemblies, or a PLC that speaks for the sensor.

Which path gets sensor values into SCADA: a scanner client or a PLC in the middle? #

Use the PLC path whenever a controller already scans the sensor for machine control, and reserve the scanner path for a standalone monitoring sensor, because an input assembly accepts one Exclusive Owner connection and the adapter refuses a second Forward_Open while the first is open.

Ask one question before opening any software: is a controller already scanning this sensor for machine control?

If yes, read the owning PLC's tags with the driver you already use, exactly as in how a SCADA client reads a small PLC over EtherNet/IP. Owning an adapter's connection is routine controller work — the same pattern as configuring a PowerFlex 525 on EtherNet/IP from an Omron NJ301. A second Exclusive Owner Forward_Open from the gateway would be refused, or worse, fight the controller for the outputs.

If no controller scans it, let the SCADA gateway own the connection through a scanner-class client. The gateway then sits in the I/O path, and a gateway restart drops the connection until it reopens. That makes the host choice real: when the client runs on a computer in the cabinet at the machine, pick an industrial mini PC with its own interface on the device network, not a virtual machine behind NAT.

One limit on both paths: treat the link as monitoring data, because no interlock or stop function may depend on a SCADA read that a restart or timeout can stall for seconds.

What does the EDS file tell you before you configure anything? #

Everything the scanner asks for comes from the device's EDS file: the input (T→O), output (O→T) and configuration assembly instances, their exact byte sizes, the supported connection types, and the RPI range — take these numbers from the file, never from another device's working configuration.

The EDS is a text file the device manufacturer supplies; it lists the CIP objects the device implements, including every Assembly instance with its size. The user manual's EtherNet/IP chapter gives the byte-by-byte meaning of each buffer — which offsets hold measured values, which hold status flags, and where any command fields sit.

Enter the instances and sizes into the scanner's generic connection form, the same exercise as sizing a Generic Ethernet Module connection for WAGO 750 I/O. The adapter checks the declared size against the assembly it owns, so a one-byte mismatch refuses the connection outright. The connection-type list in the EDS is also the authority on whether Exclusive Owner, Input Only or Listen Only exists on this device. If you cannot obtain the EDS from the manufacturer, the manual's communication chapter usually repeats the instance numbers, and a commissioning read of the device identity confirms you have the right file.

How do you configure the scanner connection step by step? #

Create one device entry per sensor controller, enter the EDS numbers exactly, set an RPI inside the EDS range, and map tags onto the input buffer offsets from the manual — then prove quality behavior before trusting any number on screen.

  1. Install the scanner-class client on the SCADA gateway and enable its implicit Class 1 option; an explicit-only client cannot open the connection at all.
  2. Read the sensor controller's IP address, subnet mask and gateway from its own settings, then ping it from the SCADA gateway host itself — a ping from an engineering laptop proves nothing about that host's route.
  3. Open TCP 44818 for connection setup and UDP 2222 for the I/O packets in the host firewall and in any firewall between subnets.
  4. Create the device entry for the controller's IP address and enter the input, output and configuration assembly instances and byte sizes exactly as the EDS lists them.
  5. Set the RPI inside the range the EDS declares, starting slower than the maximum; shorten it only after the link has held steady.
  6. Create tags at the input-buffer offsets from the manual, setting data type, signedness and the scaling that matches the controller's decimal-point setting.
  7. Add a tag-quality or connection-status alarm so a dead link shows bad quality instead of the last good number.
  8. Leave command writes until reads are proven; commands ride the output assembly and need the controller's handshake.

Keep the RPI inside the EDS range and start slow, and plan for the connection watchdog to fire at multiplier × 4 × RPI — with a 20 ms RPI and a multiplier of 32, tags go bad about 2.6 s after the last packet.

CIP defines the Class 1 inactivity timeout as the connection timeout multiplier times four RPIs, and the scanner connection form exposes the multiplier. The numbers below are illustrative settings; the RPI range itself comes from your device's EDS:

  • Packet rate at a 20 ms RPI: 1000 ms ÷ 20 ms = 50 packets per second.
  • Connection timeout at multiplier 32: 4 × 20 ms × 32 = 2560 ms, so about 2.6 s of bad quality after a cable pull.
  • Payload traffic with a 32-byte input assembly: 50 packets/s × 32 B = 1600 B/s, about 12.8 kbit/s — trivial for any network, which is why the multicast flooding described below, not bandwidth, is the network risk.

The same arithmetic warns against a fast RPI: at 5 ms with multiplier 32 the watchdog falls to 4 × 5 ms × 32 = 640 ms, so a snappy update rate shortens the failure window unless you raise the multiplier. Start slow, prove the link, then tighten deliberately. The update-timing logic is the same on any Class 1 link, as explained in how Logix5000 RPI sets input and output update timing.

Compare the SCADA value against the sensor controller's own display at a known input and to its displayed resolution, then unplug the network cable and confirm every tag flips to bad quality inside the connection timeout and recovers on reconnect.

Three checks cover the behavior that matters in production:

  • Steady state: hold the sensor at a known input and compare the SCADA tag against the controller's own display; they must agree to the displayed resolution. Move the input and confirm the tag tracks with no lag beyond one or two RPIs plus the tag poll period.
  • Failure: unplug the cable and watch quality. Tags that freeze while staying good mean the poll detects nothing — the same mechanism behind why HMI values show question marks after communication timeouts, where a client keeps serving the last value past a dead connection.
  • Invalid data: force an out-of-range reading and confirm the controller's status flags mark it, so the SCADA filters the alarm code instead of trending it as a measurement.

Reconnect after the failure test and confirm the values return on their own; a connection that needs a manual gateway restart will bite during the first power dip.

Which network settings quietly break an adapter connection? #

The I/O packets are plain UDP on the device network, so a gateway host behind NAT or in a container accepts the connection request and then never receives the data — give the host its own interface on the device network instead.

Three settings account for most connections that configure correctly and still fail:

  • Multicast: many adapters produce their input data as multicast, and a switch without IGMP snooping floods those frames to every port in the VLAN. Enable IGMP snooping with a querier, or select point-to-point unicast transport for the target-to-origin connection if the device and scanner both support it.
  • Short events: a scan class slower than the sensor's own event window misses pulses entirely, even though the RPI carried them to the gateway. Use the controller's hold or timing functions to latch short events, and poll the latched flag.
  • Duplicate addresses: two devices sharing an IP interleave their replies, and both connections drop and re-establish in bursts. Fix the address list before blaming the driver.

So which setup do you run for your sensor? #

If a PLC already scans the sensor, read that PLC's tags and skip the scanner entirely; if the sensor feeds only monitoring, run the scanner client with EDS instances and sizes, an RPI inside the EDS range, and prove quality behavior within the connection timeout before anyone trusts the screen.

Between those poles: when the SCADA must send commands and no PLC exists, the scanner client takes them on the output assembly; when uptime of the monitoring view matters more than one less device, add a controller even for monitoring, because the SCADA view then survives gateway restarts. In both cases the acceptance test is the same — values match the controller display, and a pulled cable shows bad quality, never a frozen number.

Questions asked

Can the SCADA send commands such as zeroing or a program change to the sensor?

Yes, through the output assembly with the controller's request-and-completion handshake: set the request bit, wait for the completion bit in the input assembly, then clear the request bit. A command that fires once and never again almost always has a request bit that was never cleared. Run the handshake in a gateway script or in the owning PLC, not in a screen event, so it still completes when an operator closes the display.

Can explicit messaging read the adapter's assemblies without a Class 1 connection?

Often yes for commissioning: the Assembly object can answer an unconnected read of its raw bytes. Do not build production reads on it. Some adapters restrict Assembly object services or gate values behind the handshake, and an unconnected poll has no connection watchdog, so a dead link looks like a slow value instead of bad quality.

When do Input Only and Listen Only connections work, and when do they fail?

Input Only opens its own read-only connection, so it runs beside an existing owner. Listen Only mirrors an owner's produced data and yields nothing until that owner is running, so it dies whenever the owner's connection drops. Both must appear in the EDS connection list; a device that lists only Exclusive Owner cannot share its input assembly, and the second client should read a PLC instead.

Why do measured values look correct while the status flags never change?

A byte offset or assembly instance that is one position off can land the measurements correctly while the status bytes map somewhere else, so the flags hold still. Recheck the status-byte offsets against the manual's buffer map, confirm the connection uses the input assembly instance the manual documents, then force a condition such as an out-of-range reading and watch the bit change.

Keep reading

  • Troubleshooting

    Why CitectSCADA cannot read MicroLogix data-table tags with ABCLX

    ABCLX talks to the Logix tag database, and a MicroLogix 1400 programmed with RSLogix 500 has none: its bits live in numbered data files, so an address like PROGRAM:MainProgram.B33{0} can never resolve. Change the I/O device to the ABMLXEIP driver, address the bit by file/element/bit as B33:0/0, and enter the controller's IP as a bare host address (192.168.0.100), never with a /24 suffix.

    SCADA and data logging

  • How-to

    How to read CompactLogix tags from a C# SCADA application

    Yes: a C# application reads a CompactLogix by acting as an EtherNet/IP client that requests the controller's symbolic tag read and write services over TCP port 44818. For ten tags, build that client directly and skip the OPC server; choose the documented Modbus TCP sample code for 5370 processors only when you need a register interface.

    Industrial networks

  • Troubleshooting

    Fix EtherNet/IP I/O between a ctrlX adapter and a Rockwell scanner

    Start at the Studio 5000 module status: a connection that never establishes points to the IP path or the assembly fields, while a connection with wrong or frozen data points to swapped input/output instances or the task interval. Match instance numbers and byte counts at both ends, instance 100 for consumed data with configuration instance 102 at length 0 in the documented example, and set the adapter I/O task interval equal to the scanner's RPI, 4 ms at priority 20 to 39.

    Industrial networks