A barrier gate that passes every commissioning test still comes down on a car about once a month at some sites.
The service log is unhelpful. The loop saw the vehicle. The beam was not blocked. The drive logged no torque trip. Three devices reporting normal, and a dented bonnet at the exit lane. What follows is usually a replacement of whichever sensor «must have failed», and the same incident again in six weeks.
The devices were working. They were answering the wrong question.
| Inductive loop | Photo beam | Servo torque / encoder | |
|---|---|---|---|
| Question it answers | Is conductive mass over this area? | Is this line broken now? | Did torque exceed threshold while moving? |
| Sees people | No | Yes, if mounted at the right heights | Only after contact |
| Sees a carbon-frame bicycle | Poorly or not at all | Yes | Only after contact |
| Acts before contact | If the vehicle is over the loop | If the beam is crossed | No |
| Legal status (EU) | Vehicle detector, not a safety device | Safety device if EN 12978 rated and monitored | Force limitation, EN 12453 category C |
| Typical failure | Broken turn, water ingress, crosstalk, splashover | Lens fouling, low sun, misalignment, wrong height | Threshold raised after false trips; latency from contact to reverse |
| Cost shape | Cheap device, expensive civil works | Cheap device, power and a mount on both sides | Included in the drive |
Read the table by column. Each column is a different instrument, and instruments do not add up to a safety function unless somebody designed the function around them.
The Door and Hardware Federation states the distinction plainly. Vehicle detectors, ground loops and radar devices included, are not safety devices under the EU Machinery Directive or the UK Supply of Machinery Regulations. Safety components for people, meaning safety edges, light grids and laser scanners, carry a Declaration of Conformity and meet EN 12978. A loop that opens the gate for a car is doing vehicle detection. No controller logic changes what a wire in the road can physically sense.
EN 12453 splits protection into categories that sales literature conflates. Category C is force limitation. Category D is a non-contact device supplementing it. Category E is non-contact protection good enough to work without force limitation, which on a barrier arm means a light grid or scanner over the swept volume, not one beam across the lane.
The force numbers set the design. Where a person can be crushed the limit is 400 N, decaying to 150 N within 0.75 s and 25 N within 5 s. Where a person can be struck but not crushed it rises to 1400 N, with the same decay. Measured at the edge, in the last 50, 300 and 500 mm of travel.
UL 325 approaches it from the other side. For a vehicular barrier arm it permits types A, B1, B2, C and D, and requires two independent means of entrapment protection in each direction where an entrapment zone exists. The two means must be different types, which rules out a pair of photo eyes at two heights. Since 2018 the operator also refuses to run when a required monitored device is missing or faulted.
An inductive loop is an oscillator. A vehicle body shifts its inductance, the detector card sees the shift and closes a relay. Healthy commercial loops read between roughly 80 and 300 µH at 30 to 130 kHz. Those two numbers are the best diagnostic on site, and almost nobody records them at handover.
What the loop misses follows from physics. Detection needs eddy currents in conductive material, so a carbon-frame bicycle produces nearly nothing, a motorcycle very little, a pedestrian none. A truck with high chassis and short wheelbase can straddle the loop whole. A vehicle towing a trailer presents a gap between axles, and a presence output that times out in that gap drops the call.
Then the faults misread as gate faults. A broken turn from a failed sealant line drops inductance below 40 µH, and the loop keeps catching trucks while missing small cars. The lead-in fails more often than the loop itself, usually to rodent damage or freeze-thaw at the pavement edge. Two loops closer than about 2 m report each other’s traffic, which reads on site as a gate opening for nobody. The cure for crosstalk is a different frequency on the neighbouring channel, not more sensitivity.
Most cards also apply a finite presence hold, often a few minutes, then drop the call while a vehicle is still there. That setting alone explains many «the gate came down on a stationary car» reports.
Sizing, saw-cut depth and double-loop logic live on the loop detector integration page. The point here is narrower. A loop answers a question about metal inside a rectangle of asphalt, and single-channel against double-channel detectors differ in how much of that ambiguity they resolve.
A through-beam photocell is the cheapest way to detect a person before contact, and the device most often installed in a way that cannot do the job.
Geometry first. A beam at 700 to 800 mm crosses most adult torsos and passes above the head of a four-year-old. A beam at 300 mm catches feet and ankles, plus litter, snow banks and any bumper parked 200 mm too far forward. EN 12453 tests force at 50, 300 and 500 mm for that reason. On a barrier arm the practical answer is two beams, one low and one at tip height, both monitored. Under UL 325 that pairing is still one protection type, so it has to be combined with a contact edge, a force-limited drive, or different technology.
Optics second. A beam that passes a noon commissioning test can fail at 07:40 when low sun enters the receiver on an east-west lane. Wet concrete and polished vehicle panels reflect it at angles nobody tested. A spider web across the lens attenuates the signal without breaking it, which is the worst failure available: the device still reports clear.
Alignment third. A wing mirror clipping the receiver housing moves it a few millimetres, enough to halve the received signal. Nothing announces this. The gate keeps cycling, the beam stays «clear», and the protection is gone until somebody walks the lane with a meter.
Monitoring separates a safety device from a decoration. A monitored beam reports faults to the controller, and under UL 325 the operator then refuses automatic operation. An unmonitored beam fails silently in the unsafe direction, which is the one failure you cannot afford.
Servo drives are better machines. Closed-loop control over torque, angle and speed gives smooth acceleration, programmable force, and recovery to commanded position after a disturbance. On a 0.3 s gate that control is what makes the speed possible, which is why fast arms run on DC servo operators rather than AC gearmotors.
None of it changes the sequence at contact. Torque rises, the drive measures it, the controller compares it to a threshold, then commands a reverse. Every step takes time, and the arm is moving through all of them.
The arithmetic is unforgiving. A 2.5 m arm of 3.5 kg opening in 0.3 s sweeps 3.93 m of arc, so the tip averages 13.1 m/s. Treated as a uniform rod, its rotational inertia is 7.3 kg·m², which puts kinetic energy near 100 J. A 4.5 m arm of 8 kg at 0.9 s gives 82 J. A 6 m arm of 12 kg at 3 s gives 20 J.
| Arm | Mass | Cycle time | Average tip speed | Rotational energy |
|---|---|---|---|---|
| 2.5 m | 3.5 kg | 0.3 s | 13.1 m/s | 100 J |
| 4.5 m | 8 kg | 0.9 s | 7.9 m/s | 82 J |
| 6.0 m | 12 kg | 3.0 s | 3.1 m/s | 20 J |
The fast short arm carries more energy than the slower long one. Speed does not buy margin here, it spends it. SUNFRE servo operators run at 0.3, 0.6 or 0.9 s, adjustable up to 6 s.
At contact, if the arm decelerates over 100 mm of compressible obstruction the average force is about 1000 N. Over 50 mm it is about 2000 N. The 400 N crush limit assumes that energy was never there, or that the arm was already slowing before it arrived. A torque threshold cannot remove kinetic energy stored in a moving arm.
Latency compounds it. At 13 m/s, a 50 ms delay between contact and commanded reverse puts the tip 0.65 m past the contact point. At 100 ms it is 1.3 m. On a vehicle panel that is a dent. On a person it is the difference between a bruise and a fracture.
One more behaviour matters. A position-controlled servo holds commanded position, so a person leaning on the arm may feel the drive push back rather than yield. Correct for mechanics, wrong for people, and the reason force limits have to be configured rather than assumed, and why the position-hold behaviour of the drive is worth confirming with the supplier at specification stage.
Servo drives still earn their place. Programmable torque, self-recovery after a vehicle clips the arm, and no limit switches to wear out are real advantages at high cycle counts. The servo drive is rated at 5,000,000 cycles MTBF, with a machine design life beyond 10,000,000 cycles. The planetary reduction gearbox used across the high-speed barrier gate range is part of why an arm can be stopped and reversed repeatedly without the gear train loosening. Torque monitoring is the last layer, not the first.
| Site profile | People in the lane | Sensible stack |
|---|---|---|
| Staff depot, 4 to 6 s arm | Rarely | Loop for open and presence, one monitored beam at 300 to 500 mm, force limiting set by test. A scanner is money spent on false trips. |
| Public car park, pedestrians cut through | Daily | Loop, two monitored beams at different heights, EN 12978 safety edge or safety radar, force limiting as backup. Signage and a separate pedestrian gate, because the cheapest protection is not having people in the lane. |
| High-speed lane, 0.3 to 0.9 s | Sometimes | Non-contact area sensor over the swept volume, loop for presence, force limiting tested to the standard. Torque alone is not a plan at 13 m/s. |
Environmental ratings belong in the same decision. A beam housing that fogs internally at 90% humidity fails exactly as silently as a dirty lens. SUNFRE housings are rated IP54 across an operating range of −20 °C to +75 °C, and the choice between a safety beam and a radar sensor depends on the site profile.
This stack is not free, and on some sites it is the wrong purchase.
A safety radar or scanner must be aimed, powered and kept clean. Snow, heavy rain, foliage and passing birds all produce inputs the controller has to interpret. A false trip on one lane running 5,000 cycles a day at a 1% rate means 50 unnecessary stops daily, each one a driver who now distrusts the gate and an operator who eventually asks for the sensor to be disabled. That outcome is worse than the risk it was bought to reduce.
There is a regulatory floor too. On a vehicles-only lane where EN 12453 applies, force limitation plus a low non-contact device plus a vehicle detector is the compliant arrangement. Going beyond it is a risk decision rather than a legal one, and it should be justified by pedestrian exposure, not by catalogue positioning.
On a 6 s arm carrying a light aluminium boom, the energy in that table is 20 J. A well-set force limit, a monitored beam and a loop do the job. A scanner there is over-specification with a maintenance bill attached.
Record the baseline. Loop inductance and frequency at handover, written on the cabinet door. A reading that has drifted 30% is a loop about to fail, visible six months before the first complaint.
Audit thresholds rather than hardware. Every torque limit raised to stop nuisance trips and every beam realigned with a screwdriver is a decision somebody made under pressure. Those decisions are the failure mode, and they leave no fault code. SUNFRE units are commissioned at the factory with remote-guided on-site commissioning support, and a supercapacitor auto-lift with manual release is available for power-loss situations.
Why did the gate close on a car when the loop is working? Three common causes. The presence hold expired while the car was still over the loop; the car stopped short of the loop or straddled it with a trailer gap; or a single channel was used on a lane that needs two, so the controller never confirmed the boom had cleared.
Is an inductive loop a safety device? No. It is a vehicle detector. It responds to conductive mass and does not detect people, carbon-frame bicycles, or a pedestrian’s foot. It belongs in lane logic, not in the people-protection argument.
Can a servo motor prevent crushing by itself? It limits force after contact and reverses quickly, which reduces injury. It cannot prevent contact, because detection begins when torque rises. At 0.3 s the arm is near 13 m/s and carries roughly 100 J, and no threshold setting removes that energy before arrival.
Can I use two photo beams instead of a beam and a loop? Not as two means of protection. UL 325 requires two different types, so a beam pair counts once. The EU question is different again: whether the device is EN 12978 rated, monitored, and covers the swept volume rather than a line.
At what height should the safety beam be mounted? Low enough to see legs and feet, high enough to survive litter and snow. In practice 300 to 500 mm for the lower beam plus a second at tip height. Force testing at 50, 300 and 500 mm exists because the body presents differently at each height.
Why does the beam trip at sunrise but not at noon? Low sun entering the receiver, combined with marginal alignment and a signal margin nobody checked. Through-beam units tolerate this better than retroreflective ones, and a 2x range margin at installation is the usual fix.
Do I need people protection on a vehicles-only lane? The standard applies to anyone who can reach the moving arm, including people who should not be there. A delivery driver cutting through a closed lane is foreseeable use, not misuse.
Send four numbers and the rest follows: cycles per day on the busiest lane, arm length, required open and close time, and whether pedestrians share the lane. Add the jurisdiction, because EN 12453 and UL 325 ask for different evidence, and a supplier who does not know which applies will default to whichever datasheet is shorter.
SUNFRE builds AC, BLDC and DC servo barrier gates on the same platform, exported to over 50 countries and regions with an 18-month warranty from factory shipment (extended to 20 months where long sea freight applies), which means the anti-crush argument can be settled on the site rather than on the motor. The T10 barrier gate and the K10 high-speed BLDC servo unit cover the two ends of that range. If you are comparing motors before choosing a detection stack, start with servo versus AC versus BLDC and the specification sheet guide. Broader selection logic sits in the barrier gate guide.
Send the four numbers to our engineering team and we will return a detection stack, a force-limit setting, and a written statement of what the design does not cover. That last page is usually the most useful one in the document.
