A notch filter to detect a smoke alarm
Here’s another class project from this past spring semester, this time from Introduction to Circuit Analysis class.
Background: There are already devices that detect a smoke detector beeping and do something with it. These range from inexpensive (Life+Gear Fire Safety Night Light, $8) to capitalize-on-the-disabled expensive (SafeAwake Fire Alarm Aid with Bed Shaker, $250), and all rely on a similar premise: Identify the temporal pattern of a smoke detector beeping, and do something useful with it.
Needing to design any kind of filter for ENGR 240 class, why not select a smoke detector’s ~3200Hz?
A further concept was a low-cost wireless interconnect, where a smoke alarm uses this circuit to detect a nearby smoke alarm beeping, and repeating the siren (at a slightly different frequency). Many people live and rent older houses, where wiring an interconnect system is impossible. RF interconnect alarms cost at least 6x as much, which makes them too costly for most people to consider.
But lets face it, smoke detectors save lives. And, every second counts in a fire, which is why cheaper interconnects must be developed, to bring more safe alarm systems to more people.
The basic parallel-series notch RLC filter was thus chosen:
The value of the inductor was chosen based on lab availability of parts. A 100mH inductor was then paired with a 1Mohm resistor to keep the filter highly selective, and the capacitor chosen based on our target frequency – determined to be 3353Hz for the Kidde smoke detectors found in my house.
1/C = (2*pi*3353Hz)^2*0.1H … C = 22.531nF
Due to insertion loss, the filter will have to be amplified. At the time RadioShack was going out of business and stores were selling parts at a great reduction in price, and of the available inventory, an LM386 audio amplifier was chosen.
The reason for an LM386 is its easy configuration, ready availability, and amplification using only ground and +V as sources (No -V needed).
Instead of overwhelming myself on this project, I elected to use a pre-amplified electret microphone from SparkFun.
The output of the amplifier is craftily connected to a diode in series with a parallel capacitor/resistor pair. The result of this is, one voltage level signifies no audio at 3353Hz, and another voltage signifies yes to 3353Hz. The real-world test produced a delta of 2 volts between yes and no signals, with a supply of 9V going to the amplifier (which is exactly what a smoke detector uses).
The output of this device would need to be piped into a microcontroller to detect the distinct pulse pattern of a smoke alarm, and then do something useful. For that, I’ll wait for a rainy day.
Powerpoint presentation slides can be found here: ENGR240 presentation.