PROFIBUS Position Sensor: GSD File and Module Selection

The 195 series GSD runs to only 187 lines, far shorter than an EtherCAT or PROFINET description. Short does not mean simple: it contains hard limits on speed, on how many modules fit, and on how the bytes are arranged. None of them show up during configuration, and all of them cost time once the machine is on site.

What the file actually declares

The header carries the identity, and the controller uses these entries to recognise the station and the model.

Item	Value in the file	What it does
GSD revision	2	Older masters may not read it fully
Ident number	0x04C3	How the master files this model; must be unique on a mixed bus
Model name	GJ 195 series	The name shown in the catalogue
Order number	GJ_195	Matches the label on the unit
Station type	DP slave	Not DP-V1, so diagnostics follow the standard slave path
Minimum slave interval	1 base unit (100 us)	Floor for the poll cycle when sharing a bus
Input ceiling	60 bytes	Most data one station can send
Output ceiling	8 bytes	Most the master can write
Diagnostic length	33 bytes	Length to read when pulling diagnostics

One hard limit in the speed list

The file declares each supported rate individually: 9.6k, 19.2k, 45.45k, 93.75k, 187.5k, 500k, 1.5M, 3M and 6M, with automatic baud detection enabled.

There is no 12 Mbit/s. That matters because 12M is a common default on production lines: the master and the other slaves may all support it, and one station that does not declare it forces the whole segment down to 6M. Where cycle time is tight, that slows the master scan.

Two ways out. Confirm nothing on the bus needs 12M and set the network to 6M or below; or split the segment so this station sits on its own 6M run, with a repeater or gateway back to the main network. The second costs more and only earns its keep when 12M is a genuine requirement.

The file also declares support for the master writing the slave address remotely. Replacing a unit on site does not require opening the cover and setting the switches, which saves a cabinet entry.

One module slot, not several

The file allows a maximum of one module. The project takes a single module in a single slot; you cannot stack them the way some modular slaves allow.

Module	Input	Output	Fits
1 magnet	4 bytes	1 byte	Single position point
2 magnets	8 bytes	1 byte	Two positions on one axis
3 magnets	12 bytes	1 byte	Three points, long stroke or several stations
Generic	16 bytes	16 bytes	Where data has to go both ways

Choosing three magnets gives 12 bytes in and 1 byte out, with no spare room for anything else. Work out how many position points are needed before ordering; adding a second function later means a different module and a fresh configuration.

Watch the output ceiling of 8 bytes. The generic module declares 16 bytes of output, which is above the per-station limit, so verify against what the master actually accepts rather than against the module definition alone.

Four bytes per magnet, four possible arrangements

Each magnet takes 4 bytes: three of position and one of status. The file offers four arrangements, presented as a drop-down in the configuration tool.

The four differ in exactly two things: which byte holds the status, and whether the position bytes run low-first or high-first. It looks like a formatting detail; on the PLC side it is the code that extracts the value.

Pick format 2 while the code still reads format 0 and the result is not a slightly wrong number: the status byte becomes part of the position, so the reading jumps to a wild value. It only jumps when the status changes, which makes it intermittent and therefore among the hardest faults to trace.

There is a quieter version of the same mistake. Change the format, download the parameters, and leave the extraction logic untouched. The sensor does exactly what it was told and the master still reads it wrong. Changing format means changing parsing code, not changing what is displayed.

With several magnets, each 4-byte group follows the previous one and the numbering runs from the electronics head outwards. Three magnets make 12 bytes, and the chosen arrangement applies to all three groups at once; you cannot mix formats within one station.

Three parameters that change the data structure

The user parameter block totals 10 bytes and six items are adjustable in the tool. Three of them do more than change a number.

Data format — the four arrangements above. Changing it changes how the bytes are read.

Measuring direction — Normal or Reverse. This is separate from how the rod is mounted. Turn the rod around and also set Reverse, and the two reversals cancel: the number looks perfectly normal while disagreeing with the datum on the drawing. The cheaper habit is to leave the mechanics alone and change only this setting.

Error handling — four options covering what the output does when a magnet is missing or a reading is invalid. You can zero a single position or all of them, and you can either zero or hold the old value. This affects safety logic more than it looks: zeroing presents a legitimate zero to the supervisory system, while holding the old value makes it look as though nothing moved. Either can be mistaken for a healthy signal, so decide how the supervisor detects faults before choosing.

The other three are numeric: resolution in six steps (5, 50, 100, 200, 500 and 1000 um), synchronisation (Unsynchron or Synchron), and measuring cycle (range 300 to 10000, default 2000).

Resolution follows stroke length. A few hundred millimetres at the 1000 um step makes one count equal a whole millimetre, and fine adjustment then feels like an inaccurate sensor when the accuracy was discarded during configuration. The opposite error is just as real: a ten-metre stroke on the 5 um step runs out of count range.

The file declares support for both synchronise and freeze modes. Use Synchron where several stations must sample the same instant; a single station is better left on Unsynchron, because synchronised operation expects the master to broadcast sync telegrams, and half-configured sync makes the refresh less stable, not more.

Traps seen on site

Symptom	Likely cause	What to do
Station will not join a 12M bus	12 Mbit/s is not declared	Drop the segment to 6M, or split it
Cannot add a second module	The limit is one module	Choose the module that fits and reconfigure
Reading occasionally jumps to a huge number	Format and extraction code disagree	Check the format and change the code too
Format changed, master still reads it wrong	Parameters downloaded, logic untouched	Change parameters and parsing together
Supervisor sees the carriage parked after a magnet is lost	Error handling set to hold the old value	Reselect based on how the supervisor detects faults
Refresh uneven with several stations	Synchron selected but no sync telegrams	Use Unsynchron for a single station, or complete the sync setup

That last row deserves a note. Synchronised mode is not automatically better. It depends on the master broadcasting sync telegrams on time, and one slow slave on the bus is enough to break the cadence, which shows up as uneven refreshing. With a single position point to read, Unsynchron is both simpler and steadier.

In one line

A PROFIBUS GSD is short, but every declaration in it is a hard boundary: speed tops out at 6M, one module slot, four byte arrangements. Confirm those three at selection time and commissioning comes down to aligning parameters. Discover them on site instead, and the cost runs well past that of a different module.

This piece stays with the file itself. Whether an existing line should keep PROFIBUS is covered in Series 195 on an existing PROFIBUS line; importing the GSD and attaching the slave in GSD import and slave attachment; dropped stations in termination and segment isolation. Still comparing buses, the PROFINET version on Series 199 makes a useful contrast.