Removing noise from the speed sensor of a motorcycle

Question

I am trying to make a digital dashboard for a motorcycle using a Raspberry Pi, a display and some sensors.

I am using this speed sensor.

When I put the sensor on the axis of the motor wheel and run the engine, it shows random values even if the wheel is fixed and not turning.

The magnetic field generated by the engine and the alternator is affecting GPIO value, so I am trying to move all that noise from the signal. I added a low pass filter.

The maximum of speed is 120km/h so the frequency should be at maximum fmax= S / ( R x 2 x Pi), S is the speed , R is the tire radius.

In my case fc = 13.71 Hz and (fc = 1/(2 x Pi x R x C))

The capacitor used is a 22uF 40V and the resistance R= 530 ohm.

When running the engine the GPIO pin is still detecting noise and is reading like 10 pulses per second. Would it be better if I add an LC low pass filter?

This is the electrical diagram of the sensor and the opotocoupler 4n25.

Can someone please tell me how to do the conditioning for this type of sensor?

The idea of adding a low pass filter comes from this video tutorial at min 5:41.

He made his own dashboard and it is on sale so the low pass filter should do the work and filter all that noise.

If nothing is wired (the Raspberry is supplied by the power supply so the Raspberry is totally isolated from the sensor and the engine's battery) and I run the engine and the GPIO is configured as pull up and an interrupt so it is conected to an internal pull up, it detects nothing but if I just add any type of wire (just simply to add an external pull up resistor without even connecting the sensor's output to the GPIO) the wire acts like an antennae and detects all the noise coming from the engine and the alternator. I tried also with the coaxial cable and got the same thing. The problem comes from the GPIO.

Answer 1

I've had similar problems with similar speed sensors on trucks. The problem might be the cable itself acting as antenna and picking up radiated noise. Check the signal with a scope, are you picking up stray spikes etc along with the pulse train?

The first thing I would try is to replace the cable with a shielded version and then obviously also try to keep it away from noisy things like the alternator. Also there's the classic last EMC resort: ferrite beads.

Also please note that adding custom electronics which is powered while the vehicle is running requires type approval and mandatory 3rd party EMC testing. If you put on some Rasp Pi there and the bike is involved in an accident, you'll get in some serious trouble with the law.

Answer 2

That sensor has an NPN output. That means, an NPN transistor pulls the output down if the sensor is triggered, otherwise the output is disconnected.

This is your current circuit.

^{simulate this circuit – Schematic created using CircuitLab}

The lowpass filter is between the internal 100k pullup in the Raspberry Pi, and your capacitor. Your resistor actually diminishes the filter effect as it limits the current the NPN transistor can draw from the capacitor. Remove it. If you want to increase the filter effect, add another 1kΩ pullup resistor to +3.3V in parallel to the internal 100kΩ resistor instead. That way the capacitor is recharged quickly after it was discharged by the NPN transistor and spikes cannot trigger the Raspberry input any more.

Answer 3

Here's information I found on the sensor:

• some hookup information: http://pliki.aksotronik.pl/NJK-5002c.pdf

• a demo using an Arduino: https://www.youtube.com/watch?v=Vc0LKeh_0FQ

• previous SE question: How to replace a switch with a NJK-5002c hall sensor?

The interesting tidbit in the SE question is:

The NJK-5002c doesn't have a lot of documentation, but I've found that if I connect the brown wire to V+ (12V), and blue wire to V- then the black wire will toggle between V- and V+ when a magnet comes close to the sensor.

But the fan demo at the 2:35 mark in the Arduino demo suggests an open-collector type output.

Without being able to inspect or experiment with the sensor itself, I would suggest the opto-coupler circuit below:

^{simulate this circuit – Schematic created using CircuitLab}

Note that V+ is the power to the sensor and has to be at least 5V.

The circuit uses an optocoupler so that the Raspberry Pi power supply can be isolated from the sensor supply.

This configuration should give you clean square waves on the GPIO pin.

Answer 4

The NJK-5002C can sink up to 150 mA, so the 100 Ω resistor provides a strong pull-up to 12 V. The sensor transistor provides a strong pull-down to ground sinking 120 mA when the magnet passes by the sensor. That should provide a high degree of noise immunity.

Next, the voltage is stepped down to about 3.2 V using a voltage divider and then low pass filtered for a cut-off frequency of 14 Hz. It can work with a duty cycle of about 5% to 100%.

Finally, a Schmitt trigger reshapes the signal to produce clean edges for the GPIO interrupt pin. The low threshold is 1.1 V and the high threshold is 2.2 V Either inverting or non-inverting can be used.

You may have to adjust the component values and Schmitt trigger threshold levels if you want to work with a duty cycle of less than 5%.

Click to simulate.