PROFIBUS DP Multi-Master Systems: Design and Tradeoffs

Overview: PROFIBUS DP Mono-Master vs Multi-Master

PROFIBUS DP (Decentralized Periphery) is specified in IEC 61158 and IEC 61784-1 (CPF 3) as a token-passing master-slave fieldbus optimized for high-speed cyclic I/O. A mono-master PROFIBUS DP segment has a single Class 1 DP master (DPM1) that cyclically polls all assigned DP slaves. A multi-master segment has two or more Class 1 DP masters sharing the same physical RS-485 trunk; bus access is arbitrated by token passing within the Logical Token Ring (LTR). Multi-master is a native PROFIBUS capability, not a workaround or a vendor extension. The trade-off is structured: more flexible, partitioned control and isolated failure domains, paid for with token rotation overhead, inter-master data exchange, and tighter bus-parameter discipline.

The user question — "when is multi-master useful in PROFIBUS DP if it increases response time compared to a mono-master?" — is a real engineering decision. The answer is architectural, not performance: multi-master is justified when the application itself is partitioned (multiple cells, multiple vendors, multiple owners, geographic spread, fault isolation, redundancy), not when an engineer wants more bus bandwidth. If a single CPU can host all slaves and meet the cycle time, mono-master is the correct answer.

Protocol Foundation: Token Passing and the Logical Token Ring

PROFIBUS DP runs on the FDL (Fieldbus Data Link) layer defined in IEC 61158. The MAC mechanism is a hybrid of two well-defined behaviours:

1. Token passing between active stations (Class 1 DP masters, Class 2 DP masters).
2. Master-slave polling between the master that currently holds the token and its assigned DP slaves.

Every Class 1 master and every Class 2 master (engineering PG, SCADA, diagnostics) is an active station. Active stations are organized in the Logical Token Ring (LTR), ordered by PROFIBUS address. The token is a short, fixed-length frame; the holder has exclusive bus access until it releases the token to the next higher-addressed active station. With N active stations, a complete token rotation visits N stations; the period of one full rotation is the actual Token Rotation Time T_TR.

Key bus parameters (set in STEP 7 / TIA Portal, derived from the PROFIBUS standard; see Siemens PROFIBUS bus parameter manual and IEC 61158-6):

1 t_bit = 1 bit time. At 1.5 Mbps, 1 t_bit = 666.7 ns; at 12 Mbps, 1 t_bit = 83.3 ns. The token telegram itself is approximately 33 bits (SD1, DA, SA, FC) plus t_min per handover. With three masters the token-passing cost per rotation is roughly 3 × (33 + 11) = 132 t_bit, or about 11 µs at 12 Mbps — usually small compared to slave poll cost, but it is a deterministic floor.

Multi-Master Operation Mechanism

In a multi-master DP network the bus behaves as follows:

1. Each Class 1 master is configured with a defined subset of DP slaves. The assignment is made in HW Config (STEP 7) or the device view (TIA Portal).
2. Bus access rotates: master A holds the token, polls its assigned slaves (one or more message cycles), then passes the token to the next active station (master B, master C, or any Class 2 master on the ring).
3. While a master is not the token holder, it cannot initiate frames on the bus. Its slave diagnostics, however, are still serviced when a Class 2 master (PG) or a DX subscriber queries the bus.
4. Within its own token hold, the master may execute several slave poll cycles. The number is bounded by T_TH (token hold time) derived from the gap update factor G and the configured T_TR.
5. When the master releases the token, it executes the gap maintenance routine, then waits for the next token arrival. It is in the GAP list of every other master.

The token rotation time budget for N active stations (mostly DP masters) with K DP slaves is approximated by:

T_TR ≈ Σ T_slave_i + N_masters × T_token_passing + N_repeaters × T_repeater_delay + T_class2_acyclic

where T_slave_i is the per-slave poll time (I/O data length × 11 bits / baud rate + protocol overhead), T_token_passing ≈ 44 t_bit for each master-to-master handover, T_repeater_delay ≈ 2 t_bit per RS-485 repeater, and T_class2_acyclic is the bandwidth used by engineering or SCADA traffic. The cycle time observed at any single slave is a fraction of T_TR proportional to the I/O throughput of its own master.

When Multi-Master Is Justified

Multi-master is the correct answer when the system requirements themselves are partitioned. The seven engineering scenarios that justify it are:

1. Logical / Process Separation

Independent machine cells, each with its own controller. A PROFIBUS DP trunk is the lowest-cost way to share process data between cells (recipe transfer, line balance, hand-shake bits) without forcing a single, large CPU. Practical example: a packaging line with three form-fill-seal machines and a palletizer. Each machine has a CPU 315-2DP; they all share one DP trunk and exchange status via DX or S7 connections.

2. Subsystem Autonomy and Fault Isolation

If one machine trips, only its segment of the bus drops out. With a mono-master, a CPU failure halts every slave. Multi-master is therefore common in high-availability lines where each cell must remain locally operable even if the line controller is down. Fault domain isolation is the single strongest argument.

3. Performance Isolation

Heavy I/O traffic on cell A does not extend the worst-case I/O cycle time of cell B. Each master's cycle time is independent of the load on the others, capped only by the token rotation. In a multi-master with three DPM1s, a cell running 8 slaves will see roughly T_TR/3 cycle time, not T_TR.

4. Distributed Ownership / Mixed Vendors

Two different control systems (e.g., a Siemens S7-300 line and a third-party machine tool with its own controller) must exchange status and a small set of commands. PROFIBUS DP is the lingua franca of industrial fieldbuses. A multi-master trunk with both systems acting as DPM1 is the cleanest integration path, avoiding the need to wire their slaves to a third "bridge" CPU.

5. Geographic Segmentation

Plant sections separated by hundreds of metres. Each section has its own master and an OLM (Optical Link Module) fibre link joins the segments. Multi-master is the only practical way to keep the bus segment count sane, and avoids a single CPU at one end of a long fibre chain.

6. Redundancy

Hot-standby master pairs (e.g., SIMATIC S7-400H) where one redundant controller pair assumes the DP master role, but the bus still contains a second DPM1 from another subsystem. PROFIBUS DP supports two redundant masters on a single trunk in DP-V2; see Siemens SIMATIC S7-400H manual for H-system redundancy configuration.

7. Migration and Capacity Ceiling

When the existing master CPU has hit its max-DP-slaves limit or its cycle time budget is exhausted, adding a second master splits the load. Note that an S7-300 CPU 315-2DP historically supports up to 32 DP slaves per master interface (DP-V0) and up to 125 slaves on newer firmware with DP-V1; the trade-off between firmware version, I/O count, and timing must be checked against the specific Siemens S7-300 CPU 315-2DP manual.

Multi-Master System Limits

The 3 – 5 masters practical limit is not a hard standard: PROFIBUS allows up to 126 active stations in one network. The limit comes from token rotation overhead. Each master hands over the token, and if a master is in its gap update phase the rotation is forced to continue. With a 1.5 Mbps bus, 32 slaves per master, and 5 masters, the worst-case T_TR can reach 50–100 ms. For higher speed, fewer masters or fewer slaves per master keep T_TR tight.

Hardware: Siemens S7-300 DP Master Interfaces

Common CPU 315-2DP order numbers and firmware relevant to multi-master networks (always verify with the current Siemens S7-300 CPU 315-2DP manual):

All variants expose a 9-pin D-sub RS-485 PROFIBUS DP master interface. The interface runs DP-V0 by default; DP-V1 (acyclic services, alarms) is enabled on newer firmware. DP-V2 (publisher/subscriber, isochronous mode, redundancy) is supported on selected CPUs with matching firmware and is required for H-system redundancy scenarios.

ET 200M DP Slave Interface Modules

ET 200M is the typical DP slave under each CPU 315-2DP. Relevant IM 153 modules (see Siemens ET 200M manual):

Segment Infrastructure

• RS-485 repeater 6ES7 972-0AA02-0XA0: extends a segment beyond 32 nodes or 100 m; up to 9 repeaters in series (10 segments).
• Active RS-485 terminating resistor 6ES7 972-0DA00-0AA0: powered terminator for clean signal on long backbones.
• Diagnostic repeater 6ES7 972-0AB01-0XA0: per-segment diagnostics for cable and slave health; supports slave location and cable fault detection.
• OLM / ELM (Optical Link Module / Ethernet Link Module): converts PROFIBUS DP copper to fibre for long-distance (up to 15 km) or noise-immune runs between masters.
• PROFIBUS DP cable 6XV1 830-0EH10: violet, shielded twisted-pair; characteristic impedance 150 Ω, loop resistance 110 Ω/km, capacitance 30 nF/km.

Data Exchange Between Masters

Once you have two DPM1s on one bus, they must exchange data. Three mechanisms are available, all defined in IEC 61158 and implemented in Siemens S7 communication libraries:

1. Direct Data Exchange (DX) — Publisher/Subscriber

A slave can be configured as a publisher (DX-sender); one or more other masters can subscribe to its inputs (DX-receiver). The publishing master polls the slave as usual; the subscribing master reads the same input image via the slave's DX-slot. No additional cyclic bus traffic is generated. Configuration in STEP 7: open slave properties → Configuration tab → tick the slot as "DX"; in the master project that should receive, add a DP slave with the same PROFIBUS address and slot, marked as "DX subscriber" with the source master address. Example: ET 200M under master 1 publishes 16 bits of line status; master 2 subscribes and reads them as input without polling the slave itself.

2. Master-Master Acyclic Communication (S7 PUT/GET)

The masters establish S7 connections over PROFIBUS FDL and exchange data via PUT/GET blocks. This is a one-to-one master-master channel; bus traffic is acyclic, latency is not strictly bounded by T_TR, and the typical use is configuration, recipe, parameter set, and slow diagnostics. Blocks (STEP 7): PUT, GET, USEND, URCV, BSEND, BRCV. PUT/GET is unidirectional; USEND/URCV and BSEND/BRCV are bidirectional with handshake. Connection is established in NetPro (STEP 7) or in the Connections editor (TIA Portal).

3. Master-Master via Shared DP Slave (DP-V1 Shared Device)

A single DP-V1 slave may be polled by two masters in turn; each master reads the same input image. This is the "shared device" pattern when two masters need overlapping I/O, for example a cell controller and a line controller both reading an ET 200M. Bus traffic roughly doubles per cycle, so reserve this for small I/O counts and isochronous-free slaves.

PROFIBUS DP Versions and What They Unlock

Multi-master networks exist at all three versions, but DP-V2 unlocks the cleanest publisher/subscriber model and the deterministic cycle time required for high-performance motion. Older CPU 315-2DP firmware (V2.x) is DP-V0/V1 only; for DP-V2 features such as isochronous mode, choose a CPU 315-2DP with FW V3.x and matching IM 153 firmware.

Configuration in STEP 7 / TIA Portal

Multi-master configuration follows the same steps as mono-master, plus inter-master connections. Sequence:

1. Insert each master CPU in HW Config (STEP 7) or device view (TIA Portal). Drag a CPU 315-2DP from the catalog into the project.
2. Assign distinct PROFIBUS addresses to each master (1, 2, 3, …). Address 0 is reserved for PG; 126 is the default initial address; 127 is broadcast. Edit the DP interface properties → PROFIBUS address.
3. Set the bus parameters (baud rate, profile). TIA Portal's "Network view" or STEP 7's NetPro will compute T_TR, T_SL, and related values from the topology. Stick to a single baud rate network-wide.
4. Insert DP slaves under the correct master, ensuring each slave's PROFIBUS address is unique network-wide. The tool checks for duplicates within a master but not always across masters; manual cross-check is required.
5. Configure inter-master S7 connections in NetPro (STEP 7) or Connections (TIA Portal). Define an "S7 connection" from master A to master B with a unique connection ID.
6. Configure DX links for cyclic data sharing. In the slave properties on the publishing master, mark the relevant I/O slot as "Direct data exchange"; in the subscribing master, add the same PROFIBUS address as a DX subscriber with the same slot mapping.
7. Compile and download. TIA Portal validates slave addresses for collision; STEP 7 may require manual verification via the "Bus nodes" report.
8. Online & Diagnostics to verify token rotation, slave states, and bus timing counters in the diagnostic buffer of each master CPU.

Bus parameters (T_TR, T_SL, etc.) are auto-calculated by the engineering tool from baud rate, number of masters, and number of slaves. Manually override only if cycle-time measurements demand it. Reference: Siemens PROFIBUS bus parameters manual.

Sample bus profile settings in TIA Portal (Properties → PROFIBUS interface → Bus parameters):

Commissioning and Diagnostics

• Online diagnostics: right-click the master CPU in STEP 7 / TIA Portal → "Online & Diagnostics" → PROFIBUS DP shows slave status, retry counters, and bus errors. The diagnostic buffer stores the last 100 bus events with timestamps.
• Bus fault LED (BF): solid red on the master CPU means configuration mismatch, missing slave, or bus short. Yellow flashing means station failure on at least one slave (slave missing or diagnostic alarm).
• Diagnostic repeater 6ES7 972-0AB01-0XA0: gives per-segment error counters (CRC errors, retries, slave dropouts, cable reflection quality) on the diagnostic buffer of the master CPU. The diagnostic repeater is the fastest way to localize cable faults in a multi-segment trunk.
• Token watchdog: if a master crashes, the GAP list update will eventually remove it from the LTR; the remaining masters continue. Typical recovery is one token rotation. A second watchdog (FCNT retry counter) catches scenarios where a master is alive but does not release the token — a much rarer failure mode.
• Class 2 master (PG): always assign a Class 2 master address (default: 0) to the engineering station. It co-exists with Class 1 masters on the bus without token conflicts. In TIA Portal, the PG address is set in "PG/PC interface" properties.
• Live list (BT200, AMPROBE, Softing PROFIBUS tester): a bus monitor displays the live station list, baud rate detected, and the number of token rotations per second. It is the most reliable first check at commissioning — if a station is missing, the bus monitor reveals it before the master raises an alarm.

Decision Matrix: Mono-Master vs Multi-Master

Calculating Token Rotation Time for an S7-300 Multi-Master Network

Use this estimator for a sanity check before configuring. Example: 3 × CPU 315-2DP, 24 slaves total (8 per master), 32 bytes I/O per slave, 1.5 Mbps baud, average distance, no fibre.

1. Baud period: 1 / 1,500,000 = 666.7 ns = 1 t_bit.
2. Per-slave poll: 32 bytes × 11 bits/byte (start + 8 data + parity + stop) = 2816 t_bit. Add 100 t_bit protocol overhead → 2,916 t_bit ≈ 1.95 ms per slave.
3. Per-master poll: ~8 slaves × 1.95 ms ≈ 15.6 ms per master.
4. Token handover: 3 × (33 + 11) t_bit ≈ 0.09 ms.
5. Total T_TR: 3 × 15.6 ms + 0.09 ms ≈ 46.9 ms worst case.
6. Typical cycle for any one slave: T_TR / 3 ≈ 15.6 ms — same as if the same 8 slaves were under a mono-master, because each master only polls its own 8 slaves.

Compare to mono-master: 24 slaves × 1.95 ms = 46.8 ms. The total T_TR is similar, but the multi-master network has isolated the failure domains and the per-cell cycle time is roughly one third of the total. The user's intuition that multi-master "increases overall response time" is partially correct: it increases bus engineering complexity and may slightly inflate worst-case T_TR due to handover overhead, but the per-master cycle is unchanged and the failure isolation is the actual value.

Migration from Mono-Master to Multi-Master

When a single CPU is at its I/O or cycle-time ceiling:

1. Inventory all DP slaves; group by physical location and process function.
2. Select the new CPU (e.g., add a second CPU 315-2DP 6ES7 315-2FJ14-0AB0 with FW V3.3).
3. Repartition the slaves. Each master ends up with roughly the same I/O count to keep T_TR balanced.
4. Re-issue GSD files for all slaves in the engineering tool to validate; check the GSD revision matches the slave's hardware revision.
5. Configure the inter-master connection (PUT/GET for acyclic, DX for cyclic) and verify the data mapping in the symbol table.
6. Re-validate cycle times using the online diagnostic buffer's PROFIBUS statistics and a bus monitor.
7. Update functional safety review if any slave is safety-related; SIL slaves must remain under the certified master with their PROFIsafe slot configuration untouched.

Token Ring Reconfiguration After Master Failure

PROFIBUS DP's GAP maintenance mechanism handles master dropouts gracefully:

1. Each active station maintains a GAP list of all known active stations, sorted by address.
2. When a master expects the token from a station that does not respond within a configured time (FAS = "Failed Active Station"), it removes that station from its GAP list and continues the token rotation.
3. All other masters learn the new ring within one token rotation.
4. When the failed master returns, it executes a beacon process to rejoin the LTR.

Default FAS timeout is configurable in TIA Portal → PROFIBUS properties → "Bus parameters" → "Time-out for active stations". Typical value 100 t_bit at 1.5 Mbps = 66.7 µs is far shorter than a token rotation, so the GAP update is fast.

Class 2 Master (PG/SCADA) on a Multi-Master Trunk

Class 2 masters do not poll DP slaves. They participate in the token ring for acyclic traffic only: reading diagnostics, downloading configuration, writing parameters. Default address is 0. Common Class 2 masters:

• Siemens PG with STEP 7 / TIA Portal (programming device).
• Siemens HMI panels (e.g., TP, MP, Comfort series) on the same bus.
• SCADA gateway (e.g., SIMATIC WinCC) via CP card.

Do not assign the Class 2 master an address that collides with a Class 1 master. A Class 2 master with an aggressive acyclic poll rate can starve Class 1 cyclic traffic on a slow bus; default STEP 7 acyclic traffic is conservative and acceptable.

Field-Proven Caveats

• Address conflicts are the #1 commissioning fault. STEP 7 will not always catch them across multiple masters; manual cross-check is mandatory. A bus monitor is the fastest verification.
• Termination: every copper segment must have exactly two terminating resistors, one at each physical end. Active termination (6ES7 972-0DA00-0AA0) is preferred on long trunks because it survives unplug events.
• Baud rate mismatch: all stations on a PROFIBUS segment must operate at the same baud rate. Multi-master networks do not auto-negotiate per-master speed. A single Class 2 master at the wrong baud rate can hang the bus.
• DP-V0 vs DP-V1 mixing: mixing V0 and V1 slaves under one master is fine, but DP-V1 acyclic services require V1 firmware on the master (CPU 315-2DP FW V3.x or later). Older masters will silently drop the acyclic traffic.
• Token jitter under heavy diagnostics: Class 2 masters (PG, SCADA) consume token time. Do not let a busy HMI/SCADA poll the bus every 100 ms; it can dominate the rotation and increase the worst-case T_TR noticeably.
• EMC: RS-485 is sensitive to ground loops. Use fibre (OLM/ELM) for segments running alongside VFD power cables or in high-EMI areas. A single fibre OLM in the middle of a copper trunk is a common multi-master deployment pattern.
• Shielding and grounding: the PROFIBUS cable shield must be bonded to ground at both ends (360° bond to connector backshell). Floating shields or single-end bonding degrade noise immunity significantly.
• Repeater orientation: RS-485 repeaters (6ES7 972-0AA02-0XA0) must be powered before any master on the segment. A cold repeater can cause transient bus errors that other masters interpret as CRC failures.

PROFIBUS vs PROFINET Migration Considerations

For very large multi-master installations or new builds, consider PROFINET over Industrial Ethernet. PROFINET offers:

• Full-duplex, switched Ethernet — no token ring, no shared medium.
• 100 Mbps / 1 Gbps / 10 Gbps line rates.
• Native multi-controller (multi-master equivalent) via shared I/O device and shared device with up to four controllers per device.
• Same application model (cyclic I/O, acyclic parameter, alarms) but a different transport.

Migration path: install an IE/PB Link (6ES7 141-1BF12-0XB0) to bridge a PROFIBUS DP trunk to a PROFINET controller. The CPU 315-2DP can serve as a gateway; the PROFINET line is the new backbone and PROFIBUS segments remain in service. This is a common way to extend the life of an installed multi-master PROFIBUS base while upgrading to PROFINET for new cells. See Siemens PROFINET documentation for further details on PROFINET and PROFIBUS integration.

FAQ