S7-300 MPI Network: Master Device Limits and Segment Design

David Krause13 min read
Industrial NetworkingSiemensTechnical Reference
Licensed PE Working through this on a live machine? A Maine-licensed engineer can take it from here — included with IMD hardware, by the hour for everything else. Book an engineer

S7-300 MPI Network: Master Device Limits and Segment Design

Multi-Point Interface (MPI) is the legacy Siemens SIMATIC fieldbus that integrates S7-300 and S7-400 CPUs, programming devices, operator panels, and third-party nodes over a single shielded RS-485 trunk. When a project places one S7-300 CPU in front of several HMI panels, the engineering question is not whether the protocol can carry the traffic, but how many master-capable devices can share one segment before token rotation, address space, and cable physics force a redesign.

This reference consolidates the segment topology, addressing rules, master/slave semantics, and repeater rules that govern an MPI subnet. It is written for the engineer who must size a multi-HMI panel network, validate the bill of materials, and produce a commissioning checklist before powering up the rack.

Scope. This document covers the Siemens MPI (the SIMATIC proprietary RS-485 token-bus used by STEP 7 / TIA Portal). It is unrelated to the high-performance-computing Message Passing Interface standard. Do not cross-reference HPC MPI material when designing S7-300 networks.

1. MPI Physical Layer and Topology

MPI is implemented on an electrically isolated or non-isolated RS-485 differential pair. The CPU's MPI port is a 9-pin sub-D female connector carrying the two RS-485 lines (pin 3 = B, pin 8 = A), 24 V supply on pin 2 (only on later CPUs with PG power), and shield bonding on pin 5.

Table 1. MPI default electrical and timing parameters
Parameter Value Source / Note
Physical layer RS-485, 2-wire, half-duplex CPU port, non-isolated on most S7-300 CPUs
Default baud rate 187.5 kbps STEP 7 default MPI net
Maximum baud rate 12 Mbps (MPI/PROFIBUS shared port) Only on CPUs with combined MPI/DP interface set to PROFIBUS
Minimum baud rate 19.2 kbps Used for diagnostic repeater segments
Maximum segment length 50 m at 187.5 kbps Per RS-485 loading rules
Maximum nodes per segment 32 (address 0..31) Repeater counts as a node
Cable type 6XV1 830-0AH10 (purple) or 6XV1 830-0EH10 (burial) Siemens PROFIBUS FC cable; rated to 12 Mbps
Bus terminators 220 Ω at both segment ends, pulled to 5 V via 390 Ω Built into MPI connectors or external 6ES7972-0DA00-0AA0

The default 187.5 kbps selection is the only one that guarantees 50 m with up to 32 nodes, and it is the rate STEP 7 assigns when an MPI subnet is created without manual override. Pushing the baud rate up shortens the maximum segment: at 1.5 Mbps the segment collapses to 200 m, and at 12 Mbps to 100 m, and the maximum number of nodes is still capped at 32 per segment.

2. Address Space and the 126-Partner Ceiling

Siemens documentation states that an MPI subnet can address up to 126 partners when repeaters are used to extend the bus. This 126-partner limit is a PROFIBUS/MPI protocol number derived from the address byte length, not a physical-segment number.

Table 2. MPI addressing geometry
Scope Range Count Limiting factor
Single RS-485 segment 0..31 32 nodes RS-485 driver load and reflection budget
Subnetwork with repeaters 0..125 126 nodes MPI address byte (7 bits used, 1 reserved)
Reserved addresses 0, 126, 127 3 0 = master/class 2; 126 = default spare; 127 = broadcast
Default S7-300 CPU address 2 1 STEP 7 default for the first CPU
Default PG address 0 1 Programming device default
Default OP/TP address 1 1 First panel default

Per TIA Portal V20 MPI network configuration, an MPI subnet for S7-300, S7-400, and PC stations supports up to 126 addressable devices, with up to 8 devices at the slow 19.2 kbps rate. The 8-device limit at 19.2 kbps is a practical token-rotation ceiling, not an electrical one — it exists so the worst-case token hold time fits within the watchdog of passive stations.

Address 0 is reserved for programming devices / class-2 masters. A CPU or HMI configured with MPI address 0 will conflict with the PG and will not enter data exchange cleanly. Use addresses starting at 2 for CPUs and 1, 3, 4 ... for HMIs.

3. Master and Slave Roles on MPI

MPI is a token-bus protocol: only the station holding the token may initiate a request. A master is any device that can hold the token; a slave can only respond to a master's read/write request. This distinction is the source of most multi-HMI confusion.

Table 3. Typical MPI device classes
Device class Examples Token capable? Default address Notes
Class 1 master (active master) S7-300 CPU, S7-400 CPU Yes 2 (first CPU) Initiates read/write to slaves, responds to class-2
Class 2 master (PG/ES) Field PG, TIA Portal runtime Yes 0 Diagnostic, programming; participates in token ring
Active HMI (HMI master) TP/OP/MP panels with area pointers, Comfort Panel MPI Yes 1, 3, 4 ... Each panel issues read/write requests; counts as a master
Passive slave ET 200S on MPI, SINAMICS with slave profile No 3..31 Waits for master's acyclic or cyclic frame
Repeater 6ES7972-0AA01-0XA0 No (transparent) — Counts as one node in the segment count

The practical rule: every HMI panel that polls tags over MPI is a master. A project with one S7-300 CPU and four Comfort Panels therefore places five token-holding stations on the bus. The CPU is the only mandatory active station, but the four HMIs are also masters and they share the token rotation time with the CPU.

4. Token Rotation Time and Bus Loading

The MPI token rotates in a logical ring. The time for one complete rotation, TTR, is the sum of each master's token-hold time plus the propagation delay of the physical segment. The watchdog, TWD, is the maximum time a passive station will wait before declaring a master failure; on MPI it is fixed at approximately 9.6 seconds at 187.5 kbps.

The governing inequality is:

T_TR = Σ (T_hold_i + T_frame_i) + T_propagation < T_WD

Where:

  • T_hold_i = time master i holds the token before passing it (depends on number of pending requests).
  • T_frame_i = on-the-wire transmission time of the frames master i has queued.
  • T_propagation = ~5 bit-times per segment, dominated by cable length and repeater count.

For a project with one CPU and four HMI panels, each requesting tag updates of, say, 200 bytes every 500 ms, the per-master frame budget is roughly 32 bytes per transaction (one read job). The rotation time stays well under the 9.6 s watchdog as long as:

  1. Total polled data per master < 200 bytes per poll cycle.
  2. Poll interval ≥ 1 s for non-critical areas.
  3. Number of masters ≤ 8 per segment.
Field rule. Once an MPI segment has 8+ active masters, drop the baud rate or split the masters across repeater segments. Otherwise the combination of long token-hold time and slow polling can stall the CPU's OB1 cycle indirectly, because SFC calls (X_PUT, X_GET, BSEND) wait for token ownership before issuing.

5. Segment Limits: 50 m, 32 Nodes, 187.5 kbps

Per the STEP 7 / TIA Portal MPI configuration documentation, the recommended single-segment design is:

  • Length: 50 m maximum at 187.5 kbps.
  • Nodes: 32 maximum, including any repeaters.
  • Speed: 187.5 kbps (do not raise unless you have measured the segment).
  • Topology: linear bus with terminator at both ends; no stubs > 1 m if avoidable.

This is not an aspirational target; it is the only configuration in which the bus will behave deterministically regardless of vendor mix (Siemens CPU, third-party HMI, SINAMICS drive with MPI card). A project that intends to deploy, for example, eight HMI panels and one CPU should still land inside this envelope by default.

6. Extending the Network with Repeaters

When a project exceeds 32 nodes or 50 m, a Siemens RS-485 repeater (6ES7972-0AA01-0XA0) is inserted. The repeater:

  • Regenerates the electrical signal, allowing another 50 m of segment downstream.
  • Is electrically isolated between the two segments (it has its own 24 V supply).
  • Counts as one node in each segment it joins. Adding a repeater therefore consumes one of the 32 node slots in both the upstream and the downstream segment.
  • Does not consume an MPI address — its address is not part of the 0..125 logical ring; only the bus-load count.
Table 4. Repeater rules of thumb
Condition Action
Adding 1..30 more nodes < 50 m away Add one repeater; treat as new segment
Extending cable beyond 50 m Add a repeater every 50 m of new cable
Number of repeaters in series Max 9 in series between any two nodes (MPI limit, not electrical)
Maximum logical nodes in subnetwork 126 (address 0..125)

7. Configuring an MPI Multi-HMI Project in STEP 7 / TIA Portal

7.1 Prerequisites

  • STEP 7 V5.5 SP4+ or TIA Portal V15 or later (V20 referenced in current Siemens documentation).
  • S7-300 CPU with MPI port (e.g., 6ES7313-5BF03-0AB0, 6ES7315-2EH13-0AB0) or S7-400 (e.g., 6ES7414-2XK05-0AB0).
  • MPI connectors with PG port (6ES7972-0BB12-0XA0) for diagnostics, or without PG port (6ES7972-0BA12-0XA0) for endpoint nodes.
  • PROFIBUS FC cable (6XV1830-0AH10) cut to length, with shields bonded at the cabinet entry, not at the device.
  • Panel addresses assigned in advance (e.g., 1, 3, 4, 5 for four Comfort Panels, CPU = 2).

7.2 Address plan

Table 5. Sample address plan for 1 CPU + 4 HMI
Station Role MPI address Highest MPI address in net
Field PG Class 2 master 0 31
TP/Comfort Panel 1 Active HMI master 1 31
S7-300 CPU Class 1 master 2 31
TP/Comfort Panel 2 Active HMI master 3 31
TP/Comfort Panel 3 Active HMI master 4 31
TP/Comfort Panel 4 Active HMI master 5 31

STEP 7 must be told the Highest MPI Address on the net (default = 31). It is the upper bound of the token ring scan; setting it too low causes long startup, setting it too high extends the token-passing initialization. For a 6-master segment, leave it at 31.

7.3 Panel-side configuration (WinCC / TIA HMI)

For each Comfort Panel, the HMI connection in TIA Portal must be set to "MPI" and the panel's MPI address must match the address plan. The connection resource on the CPU side is the S7 connection list in NetPro / Devices & Networks:

  1. Open the project, switch to "Devices & Networks".
  2. Drag the CPU and each panel onto the same MPI subnet.
  3. For each panel, double-click the connection line and confirm: type = S7 connection, partner = CPU, MPI address = the address plan value.
  4. On the panel, configure the connection > PLC address = CPU's MPI address (2 in this example).
  5. Compile, download the hardware configuration to the CPU, then download the HMI project to each panel.

7.4 Verification

After download, the following checks confirm the bus is healthy:

  • CPU diagnostic buffer: no entry of SF (bus fault) or BF (bus fault on MPI interface).
  • Online & Diagnostics > MPI/DP interface > "Operating mode" = "MPI active", and "Bus participants" list shows every assigned station with the expected address.
  • On each panel, the connection status icon should be solid green (no flashing yellow), and the project's status force list should update within one poll cycle.
  • On the PG, run "Accessible nodes" — every node, including the PG itself, must appear with its assigned address.

8. Common Multi-Master Pitfalls

Table 6. Multi-HMI MPI pitfalls and remedies
Symptom Likely cause Remedy
Panel loses connection randomly Token rotation too long, watchdog expires Reduce polled tags, raise poll interval, split masters with a repeater
PG cannot go online Two stations with address 0 Change the HMI/CPU from 0 to a free address
CPU goes to SF, BF3 flash Cable break or missing terminator Verify terminators at both segment ends; inspect connector pins for bent contact
One panel sees CPU, others do not Duplicate MPI address Run "Accessible nodes" from the PG; addresses must be unique
Token rotation lag grows with each new panel Too many active masters on one segment Insert RS-485 repeater; isolate panels onto downstream segment
Communication works at 187.5 kbps, fails at 1.5 Mbps Cable length or stubs exceed 1.5 Mbps budget Lower baud rate or shorten cable / remove stubs

9. When to Migrate from MPI to PROFIBUS DP or PROFINET

Once the panel count exceeds 4–6 active HMI masters, or once the total tag volume exceeds ~2 kB/s of cyclic traffic, the engineering-cost benefit shifts to PROFIBUS DP at 1.5 Mbps. PROFINET is the modern alternative and is mandatory for any new S7-1500 deployment. The migration rules:

  • S7-300 CPUs with combined MPI/DP port (e.g., 6ES7315-2EH13) can be reconfigured to PROFIBUS DP master, freeing the segment for higher speed and addressing.
  • HMI panels with PROFINET port migrate directly; older MPI-only panels (e.g., OP77B) must be replaced with Comfort or Unified panels.
  • Repeater rules of PROFIBUS DP follow the same RS-485 physics; the segment speed budget is the same (50 m at 1.5 Mbps, 100 m at 12 Mbps).

10. Diagnostic Counters Worth Watching

On the S7-300 CPU's MPI/DP interface, the following online diagnostic counters expose token-rotation and frame-error health:

Table 7. CPU MPI diagnostic counters (online view)
Counter Healthy value Indicates
Frame retries Steady, no growth Excessive collisions, bad cabling
Token losses Zero Token-ring stability
Bus errors (CRC, parity, overrun) Zero at steady state EMC, ground loop, termination
Active connections Equals the number of configured HMI connections + PG Missing partners when value is lower than configured
Token rotation time (ms) < 100 ms typical for 1 CPU + 4 HMI Worst-case round trip for token
Field tip. A token rotation time > 500 ms in a multi-HMI setup is the early warning that the segment is at the edge of its deterministic envelope. Before adding a fifth or sixth panel, re-evaluate the polled area pointer count and reduce non-essential data.

11. Commissioning Checklist

  1. Confirm the cable is Siemens PROFIBUS FC (purple), not generic RS-485, with shield bonded only at the cabinet entry.
  2. Install a terminator at both physical ends of the bus, including any segment created by a repeater.
  3. Verify unique MPI addresses for every device; reserve 0 for the PG.
  4. Set Highest MPI Address to 31 unless the largest real address is < 15, in which case the CPU can be told to scan only the real range (faster startup).
  5. Connect the PG last, after the CPU and HMIs are online, so a duplicate address on the bus does not black-screen the HMI commissioning.
  6. Run "Accessible nodes" and save the screenshot as the as-built record.
  7. Cycle power and confirm every panel recovers within one watchdog (9.6 s).

12. Frequently Asked Questions

What is the maximum number of master devices on a single MPI segment?

The practical limit is 32 nodes per RS-485 segment, of which the CPU plus each HMI panel counts as one master. A single segment with one CPU and up to four active HMI panels runs deterministically; beyond that, token rotation time and watchdog pressure force the addition of an RS-485 repeater or a move to PROFIBUS DP/PROFINET.

Is the 126-partner number a real segment limit?

No. The 126-partner figure is the MPI/PROFIBUS logical-addressing ceiling (addresses 0–125), not an electrical-segment ceiling. It is reached only after repeaters extend the bus; each RS-485 segment between repeaters is still capped at 32 nodes and 50 m at 187.5 kbps.

Why does the documentation say "up to 8 devices at 19.2 kbps"?

At 19.2 kbps the on-the-wire time of each frame grows by an order of magnitude. To keep the token rotation time under the ~9.6 s MPI watchdog, the protocol limits active participants to 8. This is a token-time constraint, not a physical-layer constraint.

Do I need to set the Highest MPI Address above the largest real address?

For token ring initialization, the CPU scans addresses 0 through the configured "Highest MPI Address". Leaving it at the default 31 is safe; lowering it to the highest real address shortens initialization but means you must update it every time you add a station.

Can a single S7-300 CPU serve more than four HMI panels on MPI?

Yes, but at 187.5 kbps the deterministic margin shrinks. With six or more active HMI masters, switch the bus to PROFIBUS DP at 1.5 Mbps (if the CPU has a combined MPI/DP port) or migrate the panels to PROFINET. Keeping the panels on MPI past that point invites intermittent watchdog disconnects.

Back to blog