A light curtain does not stop a machine. It starts a stop, and the machine keeps moving while that stop runs. The distance between the beams and the hazard is the whole of the protection, and it comes out of a calculation that takes ten minutes and is skipped more often than any other step in a machine build.
On a cell we were asked to review last year, the light curtain in front of a servo press was mounted 300 mm from the tool. It was a Type 4 curtain, correctly wired to a dual-channel safety relay, tested at handover, and every document in the file said the cell was safe. The curtain worked exactly as specified: it detected a hand, opened both channels and commanded a stop within 20 ms. The press then took a further 190 ms to come to rest. In those 190 ms a hand travelling at ordinary speed covers close to 400 mm, and the tool was 300 mm away. Everything in that cell was right except the one number nobody had calculated.
This is the most common serious finding we make on machines that arrive for retrofit or for a second-source build, and it is worth writing down plainly. A protective device does not stop a machine. It starts a stop, and the machine keeps moving while that stop runs. The safety distance is what buys the time. If the distance is wrong, every other part of the safety system is decoration.
The formula, and what each term really means
ISO 13855 gives the calculation for the minimum distance S in millimetres: S = K x T + C. It looks trivial, and the danger is in the detail of the three terms.
K is the assumed approach speed of the body part, in millimetres per second. For a hand or an arm approaching perpendicular to a vertical detection plane, the standard uses 2000 mm/s. If the result of the calculation comes out above 500 mm, the calculation may be repeated with 1600 mm/s, but the answer may then never be less than 500 mm. This is not a guess about how fast a careful operator moves. It is a conservative figure that covers a person reaching quickly for a part that has jammed, which is precisely when hands go where they should not.
T is the total time between the moment the beam is broken and the moment the hazardous movement has stopped. It is the sum of at least two parts: the response time of the protective device and its logic, which the datasheets state, and the stopping performance of the machine itself, which they do not. The machine part includes the reaction of the safety relay or safety controller, the drop-out time of the contactor or the reaction of the drive, the response of a valve, and the physical deceleration of the moving mass. On an electromechanical press or a hydraulic cylinder, the last term dominates and it is never in a catalogue.
C is the intrusion distance: how far a body part can reach through the detection plane before it is detected at all. For a curtain with a detection capability d of 40 mm or less, C = 8 x (d - 14) mm, and it is never negative. A finger-protection curtain with d = 14 mm therefore adds nothing. A curtain with d = 30 mm adds 128 mm. A curtain with d = 40 mm adds 208 mm. This term alone is the reason a coarse curtain saves less money than it appears to: what is not spent on the device is paid for in floor space, and often in cycle time when the operator has to walk further.
A worked example
Take the press from the opening paragraph and do the arithmetic properly. The curtain has a detection capability of 30 mm, so C = 8 x (30 - 14) = 128 mm. The response time of the curtain and its relay is 20 ms. The measured stopping time of the press, at its worst case of maximum load and maximum speed, is 180 ms. T is therefore 0.2 s.
At 2000 mm/s, S = 2000 x 0.2 + 128 = 528 mm. That exceeds 500 mm, so the calculation may be repeated at 1600 mm/s: S = 1600 x 0.2 + 128 = 448 mm, which is below the floor of 500 mm, so the answer is 500 mm. The curtain at 300 mm was not marginal. It was 200 mm short, and the gap could not be closed by better wiring, a faster PLC or more training. Either the curtain moves back, or the machine stops faster, or the tool is guarded another way.
That third option matters, because the calculation also shows where money should go. Fitting a drive with a safe stop function that brakes the servo axis in 60 ms rather than letting a contactor drop out and the mass coast, brings T down to 0.08 s and S to 288 mm. A shorter stopping time buys back floor area and reach, which on a manual load station is worth real money every cycle. The safety calculation and the productivity calculation are the same calculation.
Horizontal and angled approaches are a different sum
A curtain laid horizontally over an access area, or an area scanner covering the floor in front of a robot, is not covered by the same constants. Here the approach speed is taken as 1600 mm/s and the intrusion distance depends on the height H of the detection plane above the floor: C = 1200 - 0.4 x H, and it may never fall below 850 mm. A scanner field 300 mm above the floor therefore carries an intrusion allowance of 1080 mm before the stopping time is even considered. Detection planes must also sit within the height limits the standard allows, because a plane mounted too high can be stepped under and one mounted too low can be stepped over.
Angled installations, common where a curtain has to cover both the opening and the approach along a conveyor, are treated as the more demanding of the two cases. If the geometry is not clearly vertical or clearly horizontal, we calculate both and take the larger distance. It is cheaper than arguing about it during a factory acceptance test.
The number that has to be measured
Everything above depends on the stopping time of the machine, and the stopping time is the one input that degrades. Brake linings wear, hydraulic valves slow as they age, a clutch on a mechanical press changes with temperature, and a machine that stopped in 180 ms when it was new may take 260 ms after two years of three-shift work. That is why the stopping performance is measured with a stop-time measuring device rather than estimated, measured at the worst case rather than at a convenient setting, and measured again at a defined interval. We write the measurement interval into the maintenance plan, with the measured value and the distance it justifies, so the next engineer can see at a glance whether the machine has drifted into its own safety margin.
Two related traps are worth naming. The first is the retrofit that adds mass: a bigger tool, a heavier gripper, a longer stroke. The mass is in the mechanical calculation and nobody returns to the safety calculation, although the stopping time has changed. The second is the spare part fitted in a hurry, where a contactor or a valve with a different response time goes in because it was on the shelf. Both are found by re-measuring, and by nothing else.
What a light curtain cannot do
A distance calculation answers one question: how far away must the detection plane be. It says nothing about whether a light curtain is the right device at all. If a person can pass through the plane and stand between the curtain and the hazard, the machine will happily restart with somebody inside it. That is a whole-body access situation, and it needs either presence sensing inside the cell, a physically restricted opening, or a fence with an interlocked door and a trapped-key or restart discipline that makes it impossible to be inside when the cell is enabled. A curtain also does nothing about parts thrown from a tool, about hot or sharp surfaces, or about the reach-over and reach-around distances that ISO 13857 covers. Reflective surfaces deserve a mention too: a polished guard or a shiny fixture close to the optical axis can reflect a beam around an obstruction, which is why the minimum distance to reflective surfaces from the device manual is part of the layout, not an afterthought at installation.
Muting deserves its own paragraph, because it is where good installations most often go wrong. Muting exists so that a pallet can pass through a curtain while a person cannot. It works only when the muting sensors are arranged so that no combination of sensor states can be produced by a person walking through, when the muting time is limited, and when the muting lamp is visible. A muting arrangement designed by cutting and trying at commissioning tends to end up as a curtain that is switched off for a quarter of every cycle.
How we handle this on a build
On our own machines the safety distance is a dimension on the layout drawing before the steelwork is ordered, not a note discovered at installation. The risk assessment names the required performance level for each function, the device selection follows from it, and the calculation appears in the file with its inputs: detection capability, response times from the datasheets, measured stopping time, the resulting distance and the distance actually built. At the factory acceptance test we measure the stopping time again on the assembled machine and check the built distance against the fresh number in front of the customer, because a number in a file is worth very little if nobody re-measured it after the machine was finished.
None of this is complicated. It is one formula, three inputs and a tape measure, and it takes ten minutes. It is also the difference between a guard that protects a person and a guard that only looks like it does, and it is the first thing we check when a machine arrives from somewhere else.
— GANI Engineering engineering team