13 — SERCOS III

High-precision motion bus over real-time Ethernet.

Estimated 22 minutes.

A bus built for motion

Sercos III is the third generation of a bus that has driven CNC machines, printing presses and packaging lines since 1985. The defining requirement: many servo axes, all synchronized to the same instant, with cycles measured in microseconds.

From fiber to Ethernet

Sercos I and II ran on plastic optical fiber in a ring. Sercos III, released in 2003, kept the ring topology but moved to standard 100 Mbit Ethernet hardware. Same wires as your office network — completely different timing.

Why a ring?

Why does Sercos III physically wire its devices in a ring (or a line that closes back)?

  • It looks nice on diagrams
  • So one telegram visits every device once and returns to the master
  • Required by Ethernet hardware

See the ring

Each slave has two Ethernet ports. The frame enters port A, is processed in hardware, and leaves port B in less than a microsecond.

Free redundancy

Because the topology is a ring, the master can send the same telegram in both directions at once. If a single cable is cut, the master detects the break and the bus keeps running — no lost cycles. This is built-in cable redundancy.

Two telegrams per cycle

Every Sercos III cycle has two essential frames: the MDT (Master Data Telegram) carries setpoints to the slaves, and the AT (Acknowledge Telegram) carries actual values back. They look like normal Ethernet frames — but their timing is sacred.

MDT vs AT

Who fills the MDT?

Inside one MDT, who writes the data into each slave's slot?

  • The master writes the entire MDT before sending it
  • Each slave writes its own slot as the frame passes

Processing on the fly

Sercos III slaves do not "receive then forward". A FPGA inside each slave reads its bytes and writes its bytes WHILE the frame is still streaming through the port. Forwarding delay per slave: roughly 1 microsecond.

One frame, all slaves

Because the frame passes everyone in turn, a SINGLE MDT serves 50, 100, even 200 axes. You don't send one packet per drive — you send one packet that scoops up everyone's data on the way around.

Real-time channel + UC channel

Each Sercos III cycle has two parts. The real-time channel (MDT + AT) has guaranteed slots for motion. The UC (Unified Communication) channel uses the leftover time for normal TCP/IP — web pages, file transfers, video — without disturbing the servos.

See the cycle slice

Sercos III cycles range from 31.25 µs (extreme servo loops) to a few milliseconds. Pick a cycle and watch how much room the UC channel has.

Trade-off check

You shorten the Sercos cycle from 1 ms to 62.5 µs. What happens to TCP/IP throughput on the same wires?

  • Increases — faster cycle means faster everything
  • Stays the same — Ethernet bandwidth is constant
  • Drops — the UC channel gets a much smaller slice of time

Synchronization, the real prize

Cycle time is impressive, but the real reason to choose Sercos III is synchronization. Every drive applies its setpoint at the same instant — typically within ±20 nanoseconds across the whole bus. That is what enables true multi-axis interpolation.

Jitter, before and after

Why nanoseconds matter

A 5-axis CNC mills a tight curve at 60 m/min. If two axes fire 200 ns apart instead of together, what happens at the tool?

  • Nothing — humans can't see nanoseconds
  • The path develops a microscopic stair-step that ruins surface finish
  • The CNC stops with an alarm

Communication phases (CP0 → CP4)

Sercos III boots through five phases: CP0 (find devices), CP1 (assign addresses), CP2 (configure telegrams), CP3 (download parameters), CP4 (operational). Motion only happens in CP4. If the bus refuses to leave CP2, configuration is wrong.

Step through the phases

IDNs — every parameter has a number

Sercos uses IDNs (Identification Numbers) to address parameters inside a drive. S-0-0100 is "velocity loop proportional gain". S-0-0040 is "actual velocity". Standard IDNs work on any vendor; P-IDNs are vendor-specific. It is the original "object dictionary" idea.

Cabling — copper or fiber

Standard Cat5e copper handles 100 m between devices. For longer runs, EMC-heavy environments, or galvanic isolation, the same protocol runs over fiber-optic links — same telegrams, same cycle, just a different physical layer.

CIP Safety / Sercos Safety

Sercos defines a safety profile (CIP Safety on Sercos) that piggybacks safe data inside the same MDT/AT telegrams. One cable carries motion AND functional safety up to SIL 3 — no second safety bus needed.

Sercos III vs EtherCAT

Both Sercos III and EtherCAT use processing-on-the-fly Ethernet for motion. What is the most distinctive Sercos III trait?

  • It uses 100 Mbit Ethernet — EtherCAT does not
  • Native physical ring with built-in cable redundancy and a guaranteed UC channel for normal TCP/IP
  • It can talk to any Ethernet device

Common failures

Bus stuck in CP2 → telegram lengths or IDNs misconfigured. Random axis trips → poor cable, wrong shielding, or a slave with broken DC sync. UC channel slow → cycle too aggressive. Ring won't close → port B unplugged on the last device.

Where you meet Sercos III

Bosch Rexroth machine tools, Schneider Lexium drives, Indramat / Rexroth servo presses, large CNC, printing presses, packaging lines with dozens of synchronized axes. If a machine has more than ten servos cooperating tightly, Sercos III is a strong candidate.

What you learned

  • Sercos III runs over standard 100 Mbit Ethernet, but in a physical ring with built-in redundancy.
  • Each cycle has an MDT (orders out) and an AT (status back). Slaves write their slot in the AT on the fly.
  • Cycles range from 31.25 µs to a few ms, with a parallel UC channel for normal TCP/IP.
  • Synchronization across all axes lands within ±20 ns — the real prize for precision motion.
  • Boot phases CP0 → CP4 lead from topology scan to operational motion. IDNs address every parameter.

Final challenge

A 12-axis press uses Sercos III at 250 µs cycles. The HMI loads pages slowly, but the press runs perfectly. What should you change first?

  • Switch to a different protocol
  • Increase the cycle time (e.g. 500 µs or 1 ms) to give the UC channel more bandwidth
  • Add a second Ethernet network just for the HMI

Sercos III — done.

  • You understand ring topology and built-in redundancy.
  • You can describe MDT vs AT and processing-on-the-fly.
  • You can reason about cycle time vs UC bandwidth trade-offs.
  • You know why nanosecond synchronization defines Sercos III.