IMDHMIControls LibraryShop main store

Troubleshooting

Troubleshooting a SCADA PC station that reaches only one S7-1500

SCADA and data logging Published 9 min read

Short answer

When every S7-1500 answers ping but SCADA updates only one PLC, ping has proven nothing about the S7 link. Work the checks in order: four distinct MAC addresses on the PLC subnet, the S7ONLINE access point bound to the connected network card, one HMI connection per CPU, and every tag's Connection pointing at its own CPU's connection.

On this page
  1. Why does SCADA update only one PLC when all four S7-1500 CPUs ping?
  2. How do I check that each PLC IP address belongs to a different CPU?
  3. Which network card does the S7ONLINE access point need?
  4. Does each S7-1500 CPU need its own HMI connection?
  5. Why do tags for the other machines still read PLC 1?
  6. What if a CPU refuses the connection or its program does not match?
  7. What is the order of checks that fixes SCADA reaching only one PLC?
  8. Questions asked

Four S7-1500 CPUs, one SCADA PC station, and only the first machine's status moves on screen. All four addresses answer ping, the project downloaded without errors, and the single-PLC test project worked — which is exactly why the fault is not in the network. Ping proves reachability; the runtime needs a working S7 session; and the only things that differ between the working project and the four-PLC project are the per-connection settings. Work the checks in the order below, verify each before moving to the next, and expect the fix in the HMI connections or the tag bindings, not in the switches.

Why does SCADA update only one PLC when all four S7-1500 CPUs ping? #

Ping proves only that some device answers ICMP echo at each address; the runtime speaks S7 over ISO-on-TCP on TCP port 102, so the fault is per-connection setup — duplicated addresses, copied HMI connections, or tag bindings left on the first CPU — and the checks below run in that order.

An address can answer echo while the S7 service behind it is closed, or while two devices claim that one address between them; the gap between an address answering and a session working is the same one described in why a WAN ping is reachability, not LAN connectivity. If a single-PLC test project reached its CPU, that project already proved the PC hardware, the network card, the access point and the runtime itself. The only things that differ in the four-PLC project are settings that exist per connection, so those are the only things left to check: start at the addresses and end at the tags.

How do I check that each PLC IP address belongs to a different CPU? #

Run arp -a on the PC station after pinging each CPU once; four PLC addresses must resolve to four distinct MAC addresses, each matching the label on its CPU, because a duplicated address makes two IPs answer with one MAC and no S7 project can survive that.

  1. From a command prompt on the PC station, ping each of the four CPU addresses once, then run arp -a.
  2. Compare the MAC listed against each PLC address with the MAC printed on that CPU's PROFINET interface.
  3. Mask the PC NIC address and all four CPU addresses with the same subnet mask; all five must land in one network, or one network card cannot serve them together.
  4. Test the session port, not just echo: run Test-NetConnection to each CPU with -Port 102 and require TcpTestSucceeded : True on all four addresses.

If the cell ever grows a second subnet, a single network card cannot serve both, and the symptom turns into churn rather than silence — the pattern behind a SCADA gateway that cycles between Connecting and Disconnected.

Worked example with illustrative addresses: PC NIC 192.168.10.11, mask 255.255.255.0, CPUs 192.168.10.101 to 192.168.10.104. AND each address with the mask: 192.168.10.11 AND 255.255.255.0 = 192.168.10.0, and the same AND returns 192.168.10.0 for every CPU address, so all five sit in the 192.168.10.0/24 network and the subnet condition is met. The arp -a excerpt then reads:

192.168.10.101   08-00-06-2a-11-01
192.168.10.102   08-00-06-2a-11-02
192.168.10.103   08-00-06-2a-11-02
192.168.10.104   08-00-06-2a-11-04

Addresses .102 and .103 both resolve to MAC 08-00-06-2a-11-02: two devices claim one address, and the runtime cannot hold two separate S7 sessions to a single address. Decision: any duplicated MAC, or any MAC that matches no CPU label, is corrected in the CPU's address settings before the HMI project is touched.

Which network card does the S7ONLINE access point need? #

Bind S7ONLINE, the access point every S7 online connection uses, to the network card that physically reaches the four PLCs: open Windows Control Panel in small-icon view, open Communication Settings, select that card, and take its entry ending in .TCP — never .ISO and never .TCP(auto).

Windows ping follows the Windows routing table and can leave through any adapter; S7 traffic goes only through the card the access point names. On a station with several Ethernet ports, all four CPUs can answer ping through one port while S7ONLINE points at a port that reaches only part of the cell. The .ISO entry speaks ISO transport instead of TCP/IP, and the .TCP(auto) entry lets the interface take a temporary address instead of the fixed one, so neither belongs on a PLC network with fixed addresses. One binding serves one adapter: if the CPUs sit behind two different station ports, bring the cell onto one physical network and subnet behind a single card.

Verify before opening the HMI project: with the same interface the access point uses, open Accessible devices in the engineering application. All four CPUs must be listed with the names and addresses you expect. Decision: a CPU missing there cannot be recovered by any tag or connection work — fix the binding or the cabling first.

Does each S7-1500 CPU need its own HMI connection? #

Every CPU needs its own HMI connection, configured as a separate communication relationship with that CPU as partner; when you expand a one-PLC project by duplicating the first connection, the copies keep the first CPU as partner, so three connections poll PLC 1 and the project still compiles cleanly.

A connection created with both partners inside the same project is integrated, and integration is what lets HMI tags address PLC tags symbolically, which optimized data blocks require. Check every connection row of the PC station against this table:

Connection row field Required state on each of the four connections What you see when it is wrong
Communication partner A different S7-1500 CPU per connection Two or more rows name the same CPU, and only that CPU's data moves
Partner IP address The address of that specific CPU Every row carries the first CPU's address, so the other CPUs are never polled
Access point S7ONLINE A different, unbound access point, and the connection never establishes
HMI access password The password set in that CPU's protection settings The CPU rejects the session whenever its protection level demands one

Verification: in the network view with connections displayed, four separate connection lines must run from the PC station to four different CPUs, and each partner address must match the arp -a results from the addressing check. Decision: two rows naming one CPU — the classic result of duplicating the first connection and renaming it — are fixed by repointing each copy's partner to its own CPU or by deleting the copies and creating fresh connections, then recompiling.

Why do tags for the other machines still read PLC 1? #

Each external HMI tag carries a Connection property, and a tag table copied from machine 1 keeps pointing at machine 1's connection even after you rename the tags, so the tag for machine 2 must carry the connection of machine 2 and nothing else.

The compiler lets this pass because every referenced tag exists somewhere in the project — in the first CPU. Tags for machines 2 to 4 then return plausible values from the wrong CPU, or nothing at all where the symbol does not exist in that CPU, and renaming a tag never rebinds it: the Connection property and the PLC tag path both have to point at the new machine.

Check the chain, not only the table: open one screen object for each machine and follow its tag, then that tag's Connection, to the CPU it ends at. Check the acquisition mode while you are there — with cyclic acquisition, a tag updates only while a screen that uses it is open, so test each machine on a screen that is actually displayed. A connection that never establishes surfaces on screen as question marks instead of values; the timeout behavior behind that pattern is described in why HMI values show question marks.

Decision: sorted by the Connection column, the tag tables must form four groups, one per CPU, with no machine 2 to 4 tag referencing machine 1's connection or machine 1's PLC tags; rebind them machine by machine.

What if a CPU refuses the connection or its program does not match? #

A CPU rejects the runtime when its protection level requires an HMI access password the connection does not carry, and it fails the symbolic requests when its downloaded program does not match the project the HMI was compiled against, so check both on all four CPUs, not only the first.

The password is configured in each CPU's protection settings and is entered on that CPU's own connection row; carry it exactly as that CPU demands.

Rebuilding the project into a four-PLC layout changes data block layouts even when the tag names survive, so a CPU still holding an older program answers requests for symbols it no longer contains. Download all four CPUs from the four-PLC project, not only the first. Take the machine out of production before any CPU download; a download can change outputs while the line runs.

Open the system message view of the runtime with logged messages, not only active ones: a connection that fails at runtime leaves entries there even while the screens stay calm. Three silent CPUs with an empty log point at connections that were configured but never established — back to the partner fields of the connection table, not to an error code. To catch the next drop instead of finding it on a screen, raise a SCADA alarm when the link goes down; the setup is described in creating a SCADA alarm for PLC link loss.

What is the order of checks that fixes SCADA reaching only one PLC? #

Run the checks in this order and stop at the first failure: address uniqueness first, the S7ONLINE binding second, one connection per CPU third, tag bindings fourth, password and program match last — in the one-updates-three-do-not pattern, the fault almost always sits in steps three or four.

  1. Addressing: ping each CPU once, run arp -a, require four distinct MAC addresses matching the CPU labels, and require TcpTestSucceeded : True on port 102 for all four; a failure here stops everything until the addresses are unique.
  2. Access point: S7ONLINE set to the connected card's .TCP entry, with Accessible devices listing all four CPUs; a failure here stops everything until the binding is corrected.
  3. Connections: four connection rows naming four different partner CPUs, each partner address matching the MAC check from step 1; duplicated copies get repointed or recreated.
  4. Tags: tables sorted by Connection, every machine 2 to 4 tag rebound to its own connection and PLC tag path.
  5. Password and program: each CPU's HMI access password carried on its own connection row, and all four CPUs downloaded from the four-PLC project.

The PC hardware, the network card and the runtime were already proven by the single-PLC project, so nothing in this list touches them. In the pattern where one machine updates and three never move, steps 3 and 4 hold the fault; a duplicated address caught in step 1 explains the rarer case where green pings and dead screens coincide. When all five pass, all four machines change status on screen and the cell is done.

Questions asked

Can several S7-1500 CPUs share one IP address if they use different ports?

No. Every S7 CPU listens for HMI and SCADA sessions on TCP port 102, so the port cannot tell two CPUs apart; the IP address is the only discriminator. Two CPUs with one address also corrupt the ARP table of the PC station, which makes both connections drop intermittently even when they first appear to work.

Can the SCADA PC reach the four CPUs through a router on different subnets?

Yes, if TCP port 102 passes end to end. Each connection still carries that CPU's real address, and ping across a router proves the ICMP path, not the S7 path. Test each CPU with the port check from the PC station itself, and remember one access point binding serves one network card, so routing must be true IP routing, not a second card with a second subnet.

Why does the project compile with no errors while three CPUs never update?

Because nothing the compiler checks is wrong. Every tag resolves to a tag that exists somewhere in the project, every connection has a partner, and the download to the PC station succeeds. The compiler never compares a tag's Connection property with the machine the tag name suggests, so bindings left on the first CPU pass compilation and only show up as silent screens at runtime.

Does one CPU's protection setting affect the other three connections?

No. Protection levels and the HMI access password are set per CPU, so a cell configured over time can easily end up with one CPU open and three demanding a password. That asymmetry produces exactly the one-connects-three-do-not pattern: enter each CPU's own password on that CPU's connection row rather than one password for all four.

How do I prove the fix works before putting the screens back in service?

Open a screen per machine and watch one tag from each CPU change with the process, with cyclic acquisition limited to open screens. Then force one clean failure: disconnect the network lead of CPU 2 only and confirm the runtime logs the drop and CPU 2's objects show a failed connection while CPUs 1, 3 and 4 keep updating. Restore the lead and confirm CPU 2 reconnects.

Keep reading

  • Troubleshooting

    Why a SCADA write to a PLC real address fails silently

    The link is usually fine. A SCADA write to a PLC REAL address fails over configuration: the ladder compares the double word as a DINT instead of REAL, the SCADA tag type does not match, raw-to-engineering scaling is not 1:1, or the key sequence is shorter than the entry format. Run the four checks in order and re-enter the value.

    SCADA and data logging

  • Troubleshooting

    Commissioning CP1L Host Link serial reads to SCADA

    CX-Programmer going online proves only the programming path, not the SCADA read. Set the CP1L serial port to Host Link, enter its unit number (0 to 31) as the driver's device address, match baud, data bits, parity and stop bits, prove the cable end to end, and free the computer port of other software. Read one known word before adding tags: a timeout means no valid reply arrived; a PLC protocol error proves the link works.

    SCADA and data logging

  • Troubleshooting

    Connecting STARTER to a PROFIBUS drive through an S7-300

    Being online with the S7-300 does not put STARTER online with the drive. Confirm the drives are on PROFIBUS, make the drive's station address match the one the project expects, select the routed path through the S7-300 in STARTER, and save the new parameters to EEPROM with P0971 = 1.

    Industrial networks