How-to
Can one S7 CPU PN port run Modbus TCP and S7 comms together
Short answer
Yes. A 414-5H's integrated PN interface runs a Modbus TCP server for the DCS and S7 communication for the operator station on the same port, because both are TCP-based services on one IP stack. What decides the project is the CPU's connection-resource total, the station's transport type, the loaded response delay against each timeout — and on an H pair, the fact that Modbus sessions break at switchover unless you use the redundant option on a CP 443-1 pair.
On this page
- Can one PN port carry Modbus TCP and S7 communication together?
- What limits the shared port if the protocols coexist?
- How do you tell whether the workstation uses S7 communication or ISO?
- Is the Modbus TCP server redundant on a 414-5H?
- How much delay does sharing the port add, and what limit applies?
- How do you add the DCS connection without disturbing the operator station?
- How do you verify both protocols under simultaneous load?
- When is a second interface the right decision?
- Questions asked
Yes. The integrated PROFINET interface of an S7-400 CPU — including the 414-5H — carries Modbus TCP for the DCS and S7 communication for the operator or engineering station on the same port, because both are TCP-based application protocols sharing one IP stack. Siemens builds the combination deliberately: the MODBUSPN instruction runs a Modbus TCP client or server on a CPU with an integrated PN interface under the S7 Open Modbus TCP PN option, while the same interface keeps serving the stations.
What decides whether your project works is not the protocol pair but three numbers and one behavior: the CPU's connection resources, the station's transport type, the response delay under combined load, and the fact that on an H system the Modbus sessions do not survive a CPU switchover. Work the checks below in order; each ends in a decision.
Can one PN port carry Modbus TCP and S7 communication together? #
Yes — the integrated PN interface of a 414-5H runs Modbus TCP server connections for the DCS and S7 communication for the operator or engineering station at the same time, because both are TCP-based services on one interface, and the port does not separate them.
Siemens documents the MODBUSPN instruction for exactly this use: communication between a CPU with an integrated PN interface and a partner that speaks Modbus TCP, as client or as server. Your DCS is the client; the 414-5H is the server, exposing its data in registers the DCS polls. The operator and engineering stations reach the same interface with S7 communication over TCP.
One naming trap: "PN" is the interface's name, not a restriction. Calling the port a PROFINET interface does not mean it carries only PROFINET I/O, and calling the station link "PROFINET" in conversation does not make it so — that link is an S7 connection, and the section on transport type shows where to verify it. Adding actual PROFINET I/O devices to the same interface is a separate capacity question outside this arrangement.
What limits the shared port if the protocols coexist? #
Connection resources do. Every Modbus TCP server connection you configure in MODBUSPN and every operator, engineering or other S7 connection consumes one of the CPU's fixed communication resources, so that total — not the protocol pair — is what stops the design.
Count like this. Write down every connection that terminates on the CPU: one per server connection configured for the DCS or any other Modbus partner, one per operator station, one for the engineering session, plus any existing S7 connections. Open the connection-resources table in the CPU's hardware configuration and compare the sum against the total documented for your CPU and firmware in the manual. Leave headroom for an engineering connection at commissioning time.
The failure is easy to misread: a full table refuses the next connection, whichever partner connects last, and the refused station drops off exactly like one that lost the network. A panel in that state shows the behavior described in why HMI values show question marks — blank values, not a distinct resource alarm.
Decision: the sum plus headroom fits the documented total, and the port sharing stands on resources.
How do you tell whether the workstation uses S7 communication or ISO? #
Open the station's connection in the project and read the transport: an S7 connection to the CPU's PN interface is ISO-on-TCP over TCP and coexists with Modbus TCP, while classic ISO transport belongs on a CP 443-1, not on the integrated port.
The concern behind this check is real but narrow. Stations fitted with Industrial Ethernet cards were sometimes configured for classic ISO transport, which does not ride on TCP. Read the connection object and sort it:
- S7 connection to the CPU's PN address: ISO-on-TCP, rides on TCP, coexists with Modbus TCP. This is the common case, and the shared port supports it.
- ISO transport, no TCP: that station belongs behind a CP 443-1, and whether a CP carries ISO alongside its other services is answered in that CP's own documentation.
Record the type next to the connection count from the previous section; together they say whether both partners belong on the integrated interface. Do not relabel an S7 connection as PROFINET or Modbus in the project to make the picture look simpler — the transport is what the CPU acts on.
Decision: TCP-based transport, proceed; classic ISO, move the station to a CP before touching the DCS side.
Is the Modbus TCP server redundant on a 414-5H? #
Not on the integrated port. A switchover between the H pair breaks the TCP sessions, so the DCS must reconnect and data exchange pauses; a switchover-tolerant Modbus TCP server requires the redundant Modbus option and a CP 443-1 in each of the two racks.
The H in the type name protects the controller, not the TCP sessions. Each CPU of the pair presents its own integrated interface, and the DCS's connections terminate on the CPU it reached; when that CPU leaves the pair, its sessions close with it. The DCS then reconnects — to the standby's interface address, or by whatever path your network provides — and reads and writes resume from the standby's data. Plan for a pause of seconds, driven by the DCS's reconnect behavior, not milliseconds.
The documented route to a Modbus server that holds its sessions through a switchover is the redundant Modbus TCP option for the S7-400H: the redundancy license plus a CP 443-1 in each rack, with the DCS pointed at the CP pair.
Decision: a reconnect pause the process tolerates, keep the integrated port; Modbus that must not pause, the CP 443-1 pair with the redundant option, and the DCS moves off the integrated port.
How much delay does sharing the port add, and what limit applies? #
Judge it against each partner's own timeout, never a universal figure: in the worked example, a 1000 ms DCS timeout against a 45 ms loaded response leaves 955 ms of headroom — 4.5 percent of the timeout — and the shared port passes.
What moves the number is the CPU: its communication processing time and scan load, the DCS request rate, and bursts where several requests land together. The Ethernet wire at DCS polling rates is not the constraint. So measure, in this order:
- Baseline each service alone: Modbus response time with the station idle, station update time with the DCS disconnected.
- Repeat with both fully active at the DCS's real poll rate.
- Record the two protocols separately — a quick operator screen proves nothing about the DCS deadline, and the reverse holds too.
Worked example, illustrative numbers — replace them with your measurements:
- DCS polls holding registers every 500 ms, DCS timeout 1000 ms.
- Modbus response, DCS alone: 25 ms.
- Modbus response, both services active: 45 ms.
- Headroom: 1000 ms − 45 ms = 955 ms; loaded response is 45/1000 = 4.5 percent of the timeout.
- Operator station requirement: 1000 ms update. Measured under full Modbus load: 220 ms. Margin: 1000 ms − 220 ms = 780 ms.
- Decision: both margins are wide, so the shared port stands — re-measure after any poll-rate increase.
Decision rule: loaded delay inside each partner's timeout with margin, keep the shared port. Loaded delay near or past a timeout, cut the DCS poll rate first, since halving the rate roughly halves the Modbus load on the CPU; if it still misses, the split in the last section applies.
How do you add the DCS connection without disturbing the operator station? #
One change at a time, on a saved project: leave the operator-station connection untouched, add the Modbus TCP server on the integrated PN interface, confirm the connection-resource count, then download in a controlled commissioning window with both partners aware.
- Save a known-good copy of the project and record the PN interface addressing and every existing station connection setting.
- Count the connections that terminate on the CPU today: operator stations, engineering access, existing S7 partners.
- Configure the Modbus TCP server in MODBUSPN for the DCS: server role, the CPU's PN interface as endpoint, the TCP port the DCS is set to use, and the data area the DCS will read and write.
- Re-open the CPU's connection-resources table with the new Modbus connections included and confirm the sum plus headroom still fits the documented total.
- Download in your normal controlled commissioning window, then read the CPU's communication diagnostics before connecting anything new.
- Bring the DCS online with reads only, the operator station connected throughout, and compare DCS values against the PLC data they come from.
- Enable DCS writes only after the read test passes, with the process in a state where a wrong command is safe — a mapped register can drive real output logic.
How do you verify both protocols under simultaneous load? #
Exchange real data on both paths at once. Run the DCS poll and the operator station together, confirm values move on both sides, and record each response or update time against its own requirement — link lights and open sockets prove nothing.
- Run the loaded test long enough to cross the DCS's slowest cycle; an hour of combined traffic catches burst effects a two-minute check misses.
- Watch values, not indicators: DCS readings track the PLC data and the operator screen updates through the same period.
- Repeat the delay measurements with both services at full rate and compare each against its own timeout or update requirement.
- Connect the engineering station during the test as well; it is one more load, and commissioning day is when it gets used.
- Close by reading the CPU's communication diagnostics: no connection errors, no resource complaints, nothing retried into silence.
If the DCS writes setpoints, confirm what those writes reach: whether a Modbus write can change PLC outputs depends entirely on what you map into the server's registers.
When is a second interface the right decision? #
Add a CP 443-1 for exactly three conditions: the CPU's connection resources are exhausted, the loaded delay misses a timeout you cannot relax, or the H system must hold Modbus sessions through a switchover — otherwise the shared integrated port is the right design.
- Resources exhausted: the documented total cannot cover Modbus plus stations even after trimming. A CP 443-1 adds its own connection capacity and moves the DCS onto its own interface. Split.
- Delay: after cutting the DCS poll rate and staggering requests, a loaded delay still misses a timeout the DCS owner will not relax. Split, and put the DCS on the CP.
- H redundancy: the process cannot tolerate the switchover reconnect. The CP 443-1 pair with the redundant Modbus option is the documented answer.
None of the three applies, and the shared integrated port is the correct design — adding hardware there is spending without a failure mode to fix. If scope grows to a second Modbus network on this CPU later, plan its capacity the way how a PLC runs two Modbus networks at the same time plans it, instead of stacking everything on one interface.
Questions asked
Will the operator station slow down because the DCS polls the same port?
Slower screen updates show up long before lost connections do, and only through CPU communication load. Measure the station's update time with the DCS poll stopped, then running, and compare both against the station's required update period. A rise of a few tens of milliseconds is normal; a rise that approaches the update period is not, and the first fix is a lower DCS poll rate, not new hardware.
Is the S7 link to the operator station the same thing as PROFINET?
No. The port carries PROFINET in its name, but an operator or engineering station talks to the CPU with S7 communication, which rides on TCP — a different service from PROFINET I/O, the device-level network. That distinction matters when you read documentation: limits for S7 connections and limits for PROFINET I/O devices are stated separately, and adding actual PROFINET I/O to the same interface is its own capacity check.
What happens to the DCS connection when the H pair switches over?
The TCP sessions drop. The DCS was talking to one CPU's interface; that CPU leaves the pair, the sessions close, and the DCS must reconnect — typically to the standby's own interface address — after which reads and writes resume from the standby's data. Plan for a pause of seconds. If the process cannot tolerate that, the redundant Modbus TCP option on a CP 443-1 pair is the documented route.
Does a non-redundant S7-400 need a CP 443-1 for Modbus TCP?
No. The MODBUSPN instruction and the S7 Open Modbus TCP PN option exist precisely to run Modbus TCP on the CPU's integrated PN interface, and they serve an ordinary S7-400 exactly as they serve the H variant. You add a CP 443-1 for capacity the CPU cannot host, traffic you want isolated, or Modbus that must survive an H switchover — not because the integrated port refuses the protocol.