Selection guide
Running CiA 402 CANopen drives from an S7-1500 through a gateway
Short answer
Keep the CANopen drives for axes whose trajectory the drive itself generates — CiA 402 modes 1, 3 and 6 — with the segment's bus load held under about 40% at 250 kbit/s. Replace with PROFINET drives and Technology Objects for any gearing, camming or coordinated axis, because a conformance class B gateway gives cyclic data with no isochronous coupling and no Technology Object can close behind it. This rewinder's axes all generate their profiles drive-side, so the keep path wins.
On this page
- Why no Technology Object closes behind a class B gateway
- Sort every axis by where its trajectory is generated
- Bus load check: two Lexium 32 servos, a CMMS-ST and an STB at 250 kbit/s
- The dancer loop decides the reel axis
- What the hand-written CiA 402 layer must cover
- What the shaft does when the bus dies
- Choosing the gateway on what actually differs
- The decision
- Questions asked
Keep the CANopen drives and write the CiA 402 layer yourself — that is the right call for this rewinder, and it is not close once you sort the axes by where their trajectories are generated. An S7-1500 Technology Object will never run behind a conformance class B CANopen gateway: it needs a PROFIdrive telegram and clock-synchronous coupling, and the gateway gives cyclic data with neither. So the split is by motion class. Axes that index, home or hold a speed from a drive-side profile (CiA 402 modes 1, 3 and 6) stay on the CAN segment with bus load under about 40%. Any gearing, camming or coordinated axis goes to PROFINET drives with Technology Objects. Your segment — two Lexium 32 servos, a Festo CMMS-ST and an Advantys STB at 250 kbit/s — budgets out at a 20 ms cycle, so nothing on this machine forces the replace path.
Why no Technology Object closes behind a class B gateway #
A Technology Object axis is a fixed contract, not a general-purpose motion interface: a PROFIdrive-capable drive, a cyclic telegram carrying setpoint and actual values, and isochronous coupling so the servo cycle in the CPU and the drive's control cycle run in lockstep. The CANopen master gateways that fit here — the Siemens PN/CAN LINK and the IXXAT/HMS CM CANopen — are conformance class B: cyclic data exchange, no isochronous mode. Frames leave on the gateway's internal schedule, CAN arbitration reorders them by identifier priority, and the PROFINET side samples whatever the buffer holds. The transport delay is bounded but not constant, and a position loop closed across a delay that moves has a gain that moves with it.
The documented route for non-PROFIdrive axes — the MC-PreServo and MC-PostServo organization blocks in Siemens Industry Online Support entry 109741575, written for hydraulic axes fed from synchronous analog input cards — does not transfer to a gateway. Those blocks execute inside the servo interrupt and expect a fresh actual value every cycle. A gateway image that is a cycle or two stale, with jitter on top, is exactly the input they are not built for.
Sort every axis by where its trajectory is generated #
CiA 402 — the profile still widely called DS402 — standardizes the drive-side behavior for servo drives, inverters and steppers: the state machine, the modes of operation, the objects that carry them. That standardization is what makes the classification below portable across your axes.
| Motion the axis must make | Trajectory generated in | CiA 402 mode | Effect of gateway jitter |
|---|---|---|---|
| Point-to-point index, setup move, homing | The drive | 1 (profile position), 6 (homing) | The profile is untouched; the start time shifts by the transport delay |
| Speed held or trimmed by a PLC loop | The drive's ramp; the PLC writes the target | 3 (profile velocity) | Harmless while the loop's crossover stays far below the update rate |
| Electronic gear, cam, coordinated path | The PLC, every cycle | 7 (interpolated position), 8 (cyclic synchronous position) | Fatal: these modes are defined around a synchronized cycle the gateway cannot deliver |
On this machine the spreader indexes to setup positions — class one. The reel runs against the dancer on a speed trim — class two. Homing is mode 6 on every axis. Nothing is class three, and the moment something is (a flying splice cam, a true gear to a line shaft), the decision is made for you: those axes get PROFINET drives — an S210 or V90 PN, for example — running Technology Objects, with the mechanics covered in controlling a PROFINET servo drive with cyclic telegram data. One wrinkle the table exposes: the CMMS-ST speaks FHPP, a manufacturer profile rather than CiA 402, so the layer you write covers two dialects.
Bus load check: two Lexium 32 servos, a CMMS-ST and an STB at 250 kbit/s #
Inputs, each with its source:
- Segment bit rate: 250 kbit/s — the segment as the machine is wired today.
- Bit time: 4 µs, from 1 s ÷ 250,000 bit/s.
- Frame size: an 8-data-byte frame with an 11-bit identifier occupies about 112 bit times before bit stuffing, and worst-case stuffing adds roughly 24 — budget 136 bit times per frame.
- Cyclic mapping for the port: each Lexium 32 sends two TPDOs and consumes two RPDOs; the CMMS-ST carries one each way on FHPP; the STB carries one each way while neither direction exceeds 8 bytes; one SYNC per cycle.
The rule in words: bus load is the time the cycle's frames occupy the segment, divided by the cycle time. Hold it at or below about 40% so the worst-case arrival time of a frame stays something you can bound on paper.
The arithmetic:
- 136 bit times × 4 µs per bit = 544 µs per frame.
- Frames per cycle: (2 RPDO + 2 TPDO) × 2 servos = 8, plus 2 for the CMMS-ST, 2 for the STB and 1 SYNC = 13.
- 13 × 544 µs = 7072 µs, about 7.1 ms of segment occupancy per cycle.
- At a 10 ms cycle: 7.072 ÷ 10 = 71% load — rejected.
- Minimum cycle at the 40% line: 7.072 ÷ 0.40 = 17.7 ms.
- At a 20 ms cycle: 7.072 ÷ 20 = 35.4%.
Decision: run the cyclic exchange at 20 ms. That lands at 35%, below the 40% threshold with margin for the odd heartbeat or emergency frame, and 10 ms is simply not available on this segment. If a fifth node or a wider STB rack later pushes the number over, trim the mapping before anything else — statusword, modes-of-operation-display, position actual and velocity actual are what the PLC consumes.
The dancer loop decides the reel axis #
The dancer loop is the outer loop: dancer arm position is the process variable, and the manipulated variable is a speed trim added to the reel's line-speed reference. The dancer is a mechanical integrator with compliance — absorbing the difference between web feed and take-up is its job — so a rewinder dancer with usable stroke crosses over in single-digit hertz.
Read that crossover from the controller being retired (PI gains and sample time), or step the arm and measure. Below about 5 Hz, run the trim over the gateway and stop worrying: at the 20 ms cycle the loop gets ten updates per period of its slowest motion (1 ÷ 5 Hz = 200 ms = 10 × 20 ms). If the existing loop is tuned at 15 to 20 Hz or beyond, it is fighting the mechanics rather than riding the dancer — fix the mechanics before blaming the transport. And plan to re-tune rather than port the gains: loop stiffness is specific to the axis, the same effect behind why Kv factor varies when axis loop tuning differs.
What the hand-written CiA 402 layer must cover #
- The controlword and statusword state machine: shutdown through switch-on to operation enabled, plus the fault-reset path.
- The setpoint handshake in modes 1 and 3, so a target is never accepted twice.
- Homing in mode 6, with method and limits per axis.
- Runtime parameter reads — the drive's error code after a fault, homing parameters at commissioning — which are SDO traffic, so the gateway must expose acyclic access to the PLC and not only to its configuration tool.
- The FHPP dialect on the CMMS-ST, tested separately.
Write it as a documented, tested block set. The honest risk on this path is not technical; it is that one person understands the driver and takes holidays. The gateway choice below is what makes that survivable.
What the shaft does when the bus dies #
Modes 1, 3 and 6 put the profile in the drive — which is why jitter cannot distort it, and why a dropped network leaves the stored profile executing. Configure the reaction explicitly: consumer heartbeat time on each node, producer heartbeat on the gateway, and the abort-connection option code and drive error-behavior setting set so a timeout drives a quick stop rather than a coast. Then verify it the only way that counts: pull the CAN connector on a running axis and watch the shaft. A reel in profile velocity that holds its last target after a bus fault unwinds the roll onto the floor.
One safety note, here where it belongs: safe torque off stays hardwired to each drive's safety inputs and is validated against that drive's safety manual. The heartbeat quick-stop is an operational reaction, never the safety function.
Choosing the gateway on what actually differs #
The Siemens PN/CAN LINK and the IXXAT/HMS CM CANopen both sit at conformance class B, so neither buys motion capability the other lacks. Equivalent CANopen-master gateways from other vendors document the same master functions — network management, SYNC telegrams, node guarding. Choose on maintainability, and check four items in the datasheet before ordering:
- Runtime SDO access: whether the PLC can issue SDO reads and writes while running, or only the configuration tool can. Without it, every parameter change is a laptop visit.
- Cyclic image size: total mapped PDO bytes in each direction and the per-node limit — the STB rack grows.
- Configuration tooling: whether EDS import and PDO mapping happen inside the PLC engineering suite or in a separate tool whose project file must be archived beside it. The second case hurts at year six, when that tool version no longer installs.
- Diagnostics: NMT state per node, EMCY objects and bus-off or error-passive counters visible in the PLC image. Without them, intermittent CAN faults are invisible at the panel.
The decision #
| What the machine asks for | Commit to |
|---|---|
| Class-one and class-two axes only, segment at or under 40% at the chosen cycle (this machine: 35% at 20 ms) | Keep the drives: CANopen master gateway on the S7-1500, hand-written CiA 402 and FHPP layer, STB on the same segment |
| Any class-three axis, now or planned | PROFINET drives with Technology Objects for those axes; CANopen remains for the rest and the I/O |
| Dancer trim measured above about 5 Hz and mechanics you will not touch | A mechanics job first; if the loop genuinely needs that bandwidth afterwards, that one axis moves to PROFINET |
For this rewinder, commit to the keep path: both Lexium 32 servos, the CMMS-ST and the STB stay on the CAN segment behind a CANopen master gateway, the cyclic exchange runs at 20 ms, the dancer trim is capped at a 5 Hz crossover, and the heartbeat quick-stop is proven by pulling the connector. Your own hours estimate supports it — porting roughly 40 program blocks, the comms and the rebuilt operator station costs the same either way — so spend the difference only when a class-three axis appears. The operator station itself is where you choose an industrial panel PC to run the visualization client at the machine, whichever way the drives go.
Questions asked
Can an S7-1500 Technology Object position a servo behind a CANopen gateway?
No. The Technology Object expects a PROFIdrive telegram and clock-synchronous coupling with the servo cycle, and a conformance class B gateway provides cyclic data with neither. The MC-PostServo route documented for hydraulic axes assumes a fresh actual value from a synchronous input card every servo interrupt; a gateway image arrives a cycle or two late with jitter, and a loop whose delay varies is a loop whose gain varies.
What actually goes wrong above 40% CAN bus load?
Worst-case latency stops being calculable by hand. Above roughly 40%, low-priority COB-IDs are delayed behind higher-priority traffic by amounts that grow with load, so you can no longer bound when a node's PDO will arrive. The fix is a smaller mapping or a slower cycle, not a faster bit rate dropped in blindly.
Can the load problem be fixed by raising the bit rate to 500 kbit/s?
Not without re-validating the physical layer. The 250 kbit/s rate was chosen for a segment spanning drive cabinets at both ends of the machine; a higher bit rate shortens the maximum segment length and tightens the sampling-point requirements. Re-check cable length and both terminating resistors first, and trim the PDO mapping before touching speed.
Does the Festo CMMS-ST use CiA 402?
No. It runs FHPP, a manufacturer motion profile on the same CANopen layer, with its own object layout and handshake. Your driver therefore covers two profile dialects, and each needs its own state-machine and fault tests, so budget it as extra commissioning work rather than a reuse of the servo code.
Is the heartbeat quick-stop a safety function?
No. A heartbeat timeout driving a quick stop is an operational reaction and belongs in the fault matrix, not on a safety drawing. Safe torque off stays hardwired to each drive's safety inputs and is validated against that drive's safety manual; nothing routed over CANopen or through the gateway counts toward the safety function.