What Have You Done To Your Tdm Today?

[fixitsan]

2005 & 2006 TDM900A (ABS) still have the same speed sensor on the front sprocket along with the front & rear wheel speed sensors used by the ABS ECU.

Latter ABS models only have wheel speed sensors, no sprocket speed sensor.

 

 

 

Some time ago I worked on embeded GPS & accelerometers sensors.

The problem with the accelerometers is the noise & error building. Noise may be filtered (with a Kalman filter or just a low-pass filter) but the error increase remains a problem when trying to derive speed from acceleration.

The idea was to overcome the GPS speed data output - data is outputed once per second on most embeded GPS - to give a more frequent speed data using an accelerometer, and fixing increasing errors every second with the GPS data.

 

One thing to know with a GPS is that the speed data is not a computed data using the distance between coordinates divided by time, which would be very inaccurate (up to 20% error with few satellites in sight). They actually use the doppler frequency shift on received signals to compute a much more accurate speed.

Maybe using only the GPS speed data could allow to compute acceleration data without the noise & error issue with an accelerometer, but that would be only once per second. No idea if this is enough for a brake warning signal...

Just a though on your very interesting project... :good:

 

 

I once built a gps based nixie tube speedometer and fitted it to my old xj600. The fragile neon filled tubes didn't last long....the glass envelope was fine but the fine metal cathodes shook loose !

 

I was looking at using a Kalman filter, but then a Kalman filter isn't actually filter but more of a predictor of state-space variables, such as future direction or speed.....since the first order output of the accelerometer is also deceleration (deceleration is just negative acceleration) then I already have the output from the accelerometer telling me everything I need

 

Like you said, it's VERY noisy..... so I'm using an 11-pole median filter, at 500Hz  (median filter because they reproduce edges of data changes very well)...and feeding that to a 150 point shifting buffer. Applying another median filter to the last 80 buffer elements gives me a recent and very smooth representation of the input signal which is slow to respond to rapid changes (such as acceleration or deceleration).  The accelerometer output is signed 16-bit, so that's converted to unsigned offset data around a midway point to keep the maths easy. Comparing the latest accelerometer reading to the long term average, plus a reference calibration point, tells me if this is an acceleration, deceleration due to braking, or a reduction in a previous acceleration period

 

 

At the moment I have too much sensitivity  - easier to get rid of than try to improve sensitivity !, and I need to add some hysteresis.

 

 

 

 

 

Looks like yesterday's wasted day of effort was actually a much needed refresher and learning session !

 

 

This video shows what I have at the moment. The red light is being switched by relays (ready to drop into the motorcycle wiring) and it will be clear that the device doesn't interact with the bike brake light yet.....but the mechanical brakeswitch and this device will blend together seamlessly

 

I just need to add a calibration routine and fix the two little points above and it's ready for active duty !

 

 

 

https://www.youtube.com/watch?v=oVFiWU9VY-w

This post was last modified: 23-04-2016, 06:32 PM by fixitsan.