Airbags deploy when sensors detect a sudden, severe deceleration consistent with a crash, typically equivalent to hitting a solid wall at 8 to 14 mph. The process from initial impact to the decision to fire takes as little as 2.5 milliseconds in some vehicles, making it one of the fastest automated safety responses in any consumer product. What seems instantaneous actually involves a rapid chain of sensing, electronic decision-making, and chemical inflation.
How Crash Sensors Detect an Impact
The core technology behind airbag activation is a tiny accelerometer, a microelectromechanical systems (MEMS) chip mounted on a circuit board and bolted to the vehicle’s frame. This chip continuously measures changes in acceleration. When a crash occurs, the rapid deceleration produces an electrical signal that reflects the severity and pattern of the impact. These acceleration signals are the dominant factor in determining whether airbags fire.
For side impacts, the detection system works differently. Pressure sensors installed inside the door cavity continuously measure atmospheric pressure and detect changes caused by the door deforming inward. The entire door essentially acts as one large sensory element. This approach, developed by suppliers like Bosch, makes it significantly easier to distinguish a real side collision from a harmless bump. Many vehicles combine these door pressure sensors with acceleration sensors on structural components, giving the system two independent physical measurements to cross-check before deploying side curtain airbags.
The Electronic Decision in Milliseconds
Raw sensor data alone doesn’t fire an airbag. The signals feed into an airbag control unit (ACU), a small computer that runs crash algorithms to evaluate whether the impact is severe enough to warrant deployment. This is where the system separates a real collision from hitting a pothole or slamming the brakes.
Modern ACUs use a dual-processor architecture specifically designed to prevent false deployments. One processor runs the main crash algorithm, while a completely independent second processor handles a “safing” check, a confirmation step that must agree before any airbag fires. These two processors cannot influence each other’s decisions. If the main processor malfunctions or experiences a software glitch, it physically cannot override the safing processor. Both must independently confirm the crash before the system sends an electrical signal to the airbag inflator.
The system also filters out electrical noise that could mimic a crash signal. The detection circuit uses triple sampling, meaning a signal must register as a crash three consecutive times before it counts. Filter circuits are tuned so that even measured levels of electrical interference entering the control unit won’t produce a false positive. The result is a system biased toward caution: it needs consistent, redundant confirmation before it acts, yet still makes the call in under 15 milliseconds in most crashes. In a study of GM vehicles, the median deployment time was 15 milliseconds from impact, with 75% of deployments occurring within 22.5 milliseconds.
What Happens Inside the Airbag Module
Once the control unit sends the fire signal, a small electric charge ignites a chemical propellant inside the airbag’s inflator. The most established propellant is sodium azide mixed with an oxidizer like potassium nitrate. Sodium azide decomposes rapidly when heated, producing large volumes of pure nitrogen gas. The reaction begins around 395°C and completes within a narrow temperature window, generating enough gas to fill the airbag in a fraction of a second. The bag itself inflates at speeds up to 200 mph.
Newer inflator designs are moving toward alternative chemical compositions. Some use mixtures of aluminum, copper oxide, and copper-based compounds that generate gases like nitrogen and oxygen through a high-energy thermite reaction. These formulations allow for smaller, lighter inflator modules while maintaining the same gas output. The shift is partly driven by the fact that sodium azide is highly toxic in its unreacted form, and incomplete combustion can leave residue behind.
Why Airbags Don’t Deploy for Every Passenger
Your vehicle’s airbag system doesn’t just decide whether to fire. It also decides who to protect and how aggressively. Seat weight sensors classify whoever is sitting in the passenger seat, and this classification directly controls whether the front passenger airbag activates at all.
Different manufacturers use different sensing technologies. Some use strain gauge sensors that measure tiny deformations in the seat frame under load. Others use capacitive sensors that detect changes in an electrical field when a body is present. A third approach uses a fluid-filled bladder inside the seat cushion that registers pressure changes. Regardless of the method, the system categorizes the occupant as an adult, a child, or an empty seat, using standardized calibration weights (typically 10, 30, and 50 kilograms) as reference points. If the system detects a small child or a rear-facing car seat, it suppresses the airbag entirely, because a deploying airbag can injure or kill a small occupant positioned close to the dashboard.
Side Impacts Require Faster Decisions
In a frontal crash, there’s a relatively large crumple zone between the bumper and the occupant. In a side impact, the door is the only barrier. This means the control unit has less than five milliseconds to decide whether to deploy side curtain airbags, roughly a third of the time available for a frontal deployment decision.
The door-mounted pressure sensors help meet this deadline. They detect a side impact faster than a central acceleration sensor alone could, because pressure changes inside the door cavity register the moment the door begins to deform, before the force even reaches the vehicle’s main structure. Combining pressure and acceleration data from multiple sensor locations allows the system to identify both the type and severity of the crash quickly enough to get side curtain airbags inflated before the occupant’s head reaches the window.
When Airbags Deploy Without a Crash
Unintended deployments are rare but not unheard of. The most common non-collision trigger is a hard impact with a curb, especially at an angle that mimics the force signature of a side collision. Hitting a curb at speed on a wet road, for example, can generate enough sudden lateral force to trip the side impact sensors and deploy curtain airbags. Severe undercarriage impacts from deep potholes or road debris can occasionally produce similar results, though the triple-sampling and dual-processor safeguards make this uncommon in newer vehicles.
Electrical faults, water damage to the control unit, or deteriorating wiring in older vehicles can also cause false deployments, though these are increasingly rare as redundancy in the system has improved. If your airbag warning light is on, it typically means the self-diagnostic system has found a fault, and the airbags may either fail to deploy or deploy unexpectedly until the issue is resolved.
Low-Speed Crashes and Non-Deployment
Airbags are calibrated to stay packed in minor fender benders. Below the 8 to 14 mph equivalent threshold, the system determines that the seatbelt alone provides adequate protection, and deploying an airbag would add unnecessary risk and cost. This threshold is measured as equivalent force against a rigid barrier, so real-world speeds may be higher when the crash involves a glancing angle, a softer object, or another vehicle that absorbs some energy. Two cars colliding head-on at 30 mph each may trigger deployment, while a 30 mph rear-end collision into a car moving 20 mph in the same direction may not, because the relative speed difference and deceleration force are much lower.
On the rare occasion airbags deploy at very low speeds (under 5 mph equivalent), or fail to deploy at high speeds (over 20 mph equivalent), it usually points to sensor placement, algorithm calibration, or an unusual crash geometry that the system misread. These edge cases are tracked by safety researchers and fed back into algorithm improvements for future vehicle models.