GJ 197 EtherCAT Sensor: ESI Walkthrough and Setup

On EtherCAT, a 197 series magnetostrictive displacement sensor is not just a sensor: it is a slave described by an ESI file, with three 32-bit inputs, no output process data, and parameters that travel through the mailbox. The file fixes all of that, so reading it removes the guesswork from commissioning.

Every index and value below comes from the manufacturer file GJ 197 EtherCAT (1 magnet).xml, not from general practice.

The file in three layers

The header declares EtherCATInfo Version 1.6. Below it sit Vendor and Descriptions; the vendor block carries the manufacturer identity, the descriptions block carries the device itself.

Item	Value in the ESI	Where it matters
Vendor Id	#x000005FC	how the master identifies the vendor
Device Type	ProductCode #x00000197	how the master matches the ESI file
RevisionNo	#x1	description revision, checked against firmware
Physics	YY	physical layer is EtherCAT
GroupType	GJ EtherCAT Sensor	grouping in the master device tree

RevisionNo is the item most often skipped. Fit new firmware without the matching file and the master reports a mismatch; replace the file or switch that check off, but do not run past it.

Three inputs, nothing going down

The file holds a single TxPDO: #x1a00, named Sensor Inputs, mapped to SM3 and flagged Mandatory and Fixed, so entries cannot be removed or reordered.

Index	Name	Data type	Meaning
#x6000:1	Position 1	DINT, 32 bit	position
#x6000:2	Velocity 1	DINT, 32 bit	velocity
#x6000:3	Status 1	DINT, 32 bit	status

Three entries of 32 bits make 96 bits, 12 bytes, all upward. The file contains zero RxPDOs: there is no process data the master can write, no control word, no setpoint. Parameters can only travel through the mailbox, over CoE SDOs.

One detail raises eyebrows: SM3 declares a default size of 6 bytes while the three entries add up to 12. Most masters recompute the size from the PDO mapping and run fine; if yours reports a sync size mismatch, follow the 12 bytes of the PDO rather than trimming the mapping to fit 6.

Only #x2000 is really yours to set

The dictionary holds 17 objects. Fourteen are generic EtherCAT entries (device type, error register, identity, sync parameters, PDO assignment) that nobody touches on site. Three relate to the measurement: #x2000 parameters, #x6000 inputs and #x8000 debug.

Sub-index	Name	Role
1	Stroke	measuring range
2	Resolution	resolution
3	Direction	increasing or decreasing reading
4	Max no of magnet	number of magnets fitted
5	No of missing magnet	missing magnet count
6 / 7	Reserved1 / Reserved2	reserved
8	Save	commit to non-volatile memory

Order matters: the first four entries, then Save. Without Save the values stay in RAM and revert at power-off, and with stroke and resolution both at 0000 the position cannot be converted to millimetres, which reads like a communication fault.

The ESI flags every one of these objects ro, which looks read-only. That flag is not the last word: whether a write is accepted depends on the firmware and the manual. The quick test is to write with an SDO and read straight back; if the original value comes back, the write went through.

Free Run or DC Sync0

The file offers two operating modes: Free Run (AssignActivate #x0000) and DistributedClk (#x0300, Sync0 event). Commission on Free Run to prove the link, then switch DC.

Once DC runs, watch #x1c32 and #x1c33, twelve sub-items each. The ones used on site are Synchronization Type, Cycle Time and Sync0 Cycle Time, plus three error counters: SM-Event Missed, Cycle Time Too Small and Sync Error. When those start counting, stop tuning the controller and reconcile the cycle time with master load.

Leave headroom as well: DC settles when sampling and when the master collects it, and does nothing for the sensor's own measurement delay. Add that separately when budgeting overall loop response.

#x8000 is the debug door

The debug object has eight sub-items: Out 1 to Out 6, Reserved and Command. Nothing like it exists in a generic EtherCAT device; it is vendor-specific, and the ESI gives names without semantics, so each Out needs the manual.

It still earns its place: when the device looks suspect, reading this group tells you roughly whether the problem sits in measurement or in communication, which saves pulling the sensor off the machine.

Six steps to running

1. Import the ESI and check identity. ProductCode #x00000197 and Revision #x1 both have to match.
2. Choose the sync mode. Use DC Sync0 when available; prove the link on Free Run first.
3. Check the PDO. Three 32-bit entries, 12 bytes, position before velocity before status.
4. Write #x2000. Stroke from the real travel, resolution from the model, direction from how the cylinder faces, magnet count from what is fitted.
5. Write #x2000:8 Save. Read back immediately and confirm nothing rolled over.
6. Verify. Read #x6000:1, convert to millimetres with stroke and resolution, and check both ends against a steel rule.

Three traps

Trap one: the wrong ESI file. Both the file name and the device name carry (1 magnet); one series ships different ESIs for different magnet counts. Swap the configuration without swapping the file and communication may still come up, but the parameters tied to magnet count no longer line up: readings arrive, wrong ones.

Trap two: polling position over SDO. The mailbox is request and reply, so a handful of axes saturates the bus and the update rate drops. Position belongs in the PDO: #x1c13 (TxPDO assign) already defaults to #x011a, pointing at #x1a00. Do not delete that assignment.

Trap three: mistaking the SM default size for the data length. The 6-byte figure above is the example. Data length always follows the PDO mapping; the SM default is only a hint.

The ESI is the manufacturer's and stays as it is. What you control is how the master reads and writes. Get the three inputs right, set #x2000 and Save, and what remains is scaling and mechanical zeroing.

This is about the file. Wiring and bring-up order sit in ESI files and IN/OUT wiring direction, the sync and state-transition trade-off in DC configuration and the slave state machine. For whole-line selection see choosing between four buses, and the PROFINET sibling is GSDML setup on Series 199.