Tachometer design and constructionAccurate engine (N1) and rotor speed (N2) are necessary for the realtime system to be able to keep the engine running efficiently. When the aircraft is climbing the blades will be at a higher angle of attack than in a hover, creating more lift and more drag. The engine must open the throttle to maintain the same engine RPM and a smooth climb. Rather than using elaborate throttle / collective curves to deal with this, we can just have a PID control loop maintain a constant RPM for us. Additionally, detecting an engine failure is important for initiating an autorotation to the ground. The source code for tracking these pulse counted inputs is available.
There are three parts to each tach -- a wiring harness to connect to the realtime system, the Hall effect sensor and magnets on the gears or engine fan. As the fan rotates the magnets pass by the Hall effect sensor. The sensor detects the magnetic field and sends a signal to the realtime board that has an interrupt configured to count the pulses.
The engine fan needs to be drilled to accept the magnets. This is the scariest part of the installation. Be careful and try to ensure that the fan remains balanced -- it does spin as fast as the engine. The sensor is only sensitive to one magnetic pole, so install each set of magnets in opposite orientations. Secure them with glue. I used thick CA and had to sand it smooth afterwards.
Next the sensor needs to be wired to the harness. I used a two lengths of twisted pair spun together to make the harness, but if you have three conductor wire you can make it cleaner. The pinout for the Hall effect device is Vcc, Gnd, Data when viewed from the top. The top is the beveled side -- it took me a while to determine that from the fuzzy images in the DigiKey catalog.
Then you have to install it in the aircraft. I zip tied the sensor the the mainframe rails, just in front of the fan duct. It gets an accurate pulse count from the magnets, so I'm ready to work on the PID control loop for engine speed governing.
We've tested the sensors on the bench and in the field and have seen a stable, accurate count. At high idle we receive 50 interrupts per 8 ms time slice, easily within the AVR's capability.