Fig 1. Fires follow a progression from start to full involvement, with both temperature and smoke content rising rapidly once full combustion is underway.
Fig 2. Traditional optical smoke detectors depend on an isolation chamber to prevent ambient light from reaching the optical sensor and triggering false alarms.
Fig 3. Using proximity-detection optical sensors can eliminate the need for a special chamber to hold the sensor.
Fig 4. Measurements in a calibrated smoke tunnel show the proximity-detection sensor operates correctly even without an isolation chamber.
Smoke detector systems are a vital part of building automation and occupant safety, providing early fire alerts to give inhabitants more time to escape. While some smoke detector designs use an ionization chamber containing an Americium 241 radioactive source, most recent designs use an optical detection approach that provides a simpler and more cost-effective solution. It is possible to lower costs further while preventing false alarms by adapting proximity detection techniques to smoke detection.
To understand the basic requirements of smoke detector design, it is helpful to examine some key characteristics of fires. A fire evolves over time, generally progressing in three phases (Fig. 1). The first phase occurs when the fire starts. There are no flames and no smoke and the temperature in the room remains unchanged. At this point, the fire generates only gases. This phase may last for seconds to hours depending on what sparked the fire and the combustible characteristics of the materials in which the fire started.
The second phase of a fire is the most critical in terms of smoke detection because it is the first to generate smoke due to combustion of inflammable material. The pace of the fire’s evolution in this phase varies widely depending on the burning material’s characteristics. The smoke generated in this phase is invisible to the eye because particle sizes are in the micrometer range. Temperature elevation in the room is not yet measurable, and no flames are present.
The third and final phase of fire progression begins when flames start to appear. At this point temperature and smoke particle size both increase very rapidly.
It is important to note that some materials, such as natural gas, may burn without generating any smoke, i.e., generating only heat and gases. This type of fire can only be sensed by heat or gas detectors, especially if there are no combustible materials around the fire to generate smoke.
The earlier the fire is detected, the easier it is to contain and stop—and the more time occupants have to evacuate the premises. To give occupants as much time as possible, it is important to detect low levels of smoke as soon as possible and before the smoke reaches the visible stage.
In addition, it is important to avoid false alarms. Vapor clouds generated by boiling water should not trigger a smoke detector, for example, nor should airborne dust particles. Frequent false alarms may lead some end users to remove the batteries from their smoke detectors, which is a dangerous practice.
Traditional Optical Smoke Detection
The optical sensor approach used in most smoke detector designs involves measuring the light from an infrared (IR) light-emitting diode (LED) that the smoke fog scatters to an IR sensor. The two components are mounted at an angle to one another so the detector cannot see the LED’s normal beam.
The optical elements are housed in a detection chamber (Fig. 2) that limits the IR sensor’s exposure to ambient light that might interfere with the smoke measurement. As smoke particles enter the detector chamber they scatter some of the light from the LED’s normal beam so the sensor can detect the light. When the measured light level goes above a threshold value, the system triggers a smoke alarm.
The smoke detector does not use a single measurement or simple threshold, but it must be “intelligent” enough to discriminate against false alarms. This discrimination is based on the detection speed.
If the signal’s rise rate is too sudden it is likely to be a false alarm caused, for instance, by an insect entering the detection chamber. Water fog coming from boiling water in the kitchen or steam from a shower could also trigger a false alarm. These types of events evolve quite rapidly, however, and result in large-size particles, both causing the detection’s signal level to change quickly.
While such events should not trigger a smoke alarm, they have the potential to “blind” the smoke detector to a real fire. Smoke detectors may therefore also provide a maintenance alarm that sounds if the sensor signal level remains high after a reasonable delay. Government regulations typically define the levels that should trigger the alarms.
Proximity Sensors for Smoke Detection
The optical principle used in traditional smoke detectors is similar to the one used for proximity detection except the scattering material is different and proximity detection typically depends on a surface that reflects back the light emitted from the LED. Still, optical sensors such as Silicon Labs’ Si114x that target proximity-detection applications can serve in a smoke detector design and bring with them several important benefits.
One benefit sensors like the Si114x family bring is that they contain both IR and visible light photodiodes. The ability to detect ambient light enables the design of a smoke detector that can reject unwanted ambient light and work with only the scattered IR, eliminating the need for a bulky detection chamber (Fig. 3).
Eliminating the chamber significantly reduces system cost because the chamber in traditional designs must be built to very precise specs and the manufacturing process must be tightly controlled. Any light leakage or plastic reflectance changes—even the moisture level in the plastics is critical—can completely alter the effectiveness of traditional chambered systems.
To prove the effectiveness of ambient light detection and removal in eliminating the need for a chamber, we made measurements in a calibrated smoke tunnel (Fig. 4). The smoke comes from paraffin oil and has particle sizes ranging from 0.5 µm to 1 µm.
The yellow plot is the official calibration measurement in dB/m from the smoke tunnel’s optical obscuration meter. The blue plot is the Si11xx optical sensor’s output count averaged over 20 measurements. The optical sensor measurements clearly follow the reference measurement without the use of an isolation chamber to ensure darkness.
Smoke Detector Design
Most smoke detectors require an IR LED and optical sensor for smoke detection, a microcontroller (MCU), a test/mute button, a buzzer alarm, and a battery. Our proof of concept design used a Silicon Labs C8051F9xx MCU, which is an ultra-low-power device with fast wakeup-time performance.
This MCU choice allowed a significant reduction in the system’s overall power consumption in all modes of operation through efficient use of sleep modes. The MCU supports a variety of interfaces, including digital I/O to the test button, an LED, an alarm buzzer, an interrupt line, and an I2C bus connection to the optical sensor.
The MCU also supports additional interfaces that expand the design’s capability beyond smoke detection. A capacitive touch-sensor, for instance, can be added to provide humidity detection. With an external thermistor, the MCU can measure room temperature.
These additional measurements help the smoke detector discriminate against steam and can use room temperature changes to detect smokeless fires. A UART interface allows the MCU to send measurements in real time to a home networking system if desired.
To minimize and eliminate sources of false alarms, the MCU software does not simply use the optical sensor’s measurement. It defines minimum and maximum “trends” in the signal to look for the characteristic rise in smoke levels associated with fire progression. Every sensor signal variation in the signal level that is below a certain trend is considered to be system “noise,” and every variation in the signal level above a certain trend is also dismissed.
The process by which the MCU determines whether or not to trigger the alarm follows these steps:
- At startup the MCU initializes the optical sensor to perform measurements every second in autonomous mode and generate an interrupt to the MCU after each measurement. Then the MCU enters a sleep mode.
- On receiving the interrupt, the MCU wakes and reads the sensor’s measurement over the I2C bus.
- The MCU subtracts the last noise value from the raw measurement data and compares the modified measurement to the trend constants.
- If the modified measurement is below the minimum trend value, the measurement is considered to be a noise signal and the CPU exponentially averages the raw value with the previous noise value to get the new noise value.
- If the modified measurement is above the maximum trend value, the MCU discards the measurement as a false alarm. Such measurements, which have an unusually rapid evolution, are typically generated by an insect or other object, not smoke. If this is the first false alarm occurrence after any other type of measurement, the MCU triggers a counter that will generate a maintenance alarm if false alarms persist throughout the counter’s time duration.
- If the modified measurement value is in between the two trend values, the noise level remains unchanged and the MCU saves the raw measurement value. The MCU then sets a flag to signify that the next measurement should be compared to this saved value instead of to the noise level.
- Finally, the MCU clears the optical interrupt register and returns to sleep mode to await the next wakeup interrupt.
This process continues until either the measurement steadily rises to reach the alarm threshold or levels fall back to system noise and the MCU resets the measurement flag.
Additional Features
Since the MCU has additional interfaces available, our design can do more than smoke detection. By making a temperature measurement with a thermistor connected to the MCU’s on-chip analog-to-digital converter (ADC), we can provide a heat alarm as well as a freeze warning.
Also, a capacitive touch-sense button directly on the printed circuit board allows the MCU to measure humidity. This humidity measurement is an easy and economical means of detecting water vapor “smoke” and discarding it as a false alarm. The MCU simply averages the touch sensor measurements exponentially over time and compares it to a threshold. If the averaged value is greater than the threshold, this indicates the presence of humidity, and the MCU sets a flag that will disable the smoke alarm.
Each measurement cycle, the MCU looks at all alarm flags to generate the correct alarm in time. For a smoke alarm, the MCU triggers the buzzer and blinks a red LED. For a heat or freeze alarm, the MCU triggers the buzzer and blinks a yellow LED. A low-battery alarm sounds the buzzer and lights a red LED once every 10 seconds (to save power). Pressing the test button triggers the smoke alarm and also turns on a yellow LED.
Conclusion
The combination of an IR/ambient light optical sensor and an ultra-low-power MCU provides an ideal platform for a smoke detector design. The optical sensor’s ability to detect smoke in the presence of ambient light enables an economical chamberless smoke detector design, which is smaller, safer, and more cost-efficient than those based on radioactive materials as well as less expensive and more accurate than other optical design approaches.
The MCU’s ability to measure humidity using capacitive touch-sensing provides a cost-effective way to distinguish water fog and avoid false alarms from this source. The potential combination of smoke detection and heat detection is a useful feature that allows detection of fires generating little to no smoke as well as typical smoky fires. By using ultra-low-power MCUs, the smoke detector’s power consumption can remain low enough to enable operation from a single coin cell battery for several years.