wdm0006

Anyone interested in following the progress on the Electronic Terrain Predictive Suspension project can do so here: http://www.wandertechnologies.com I've managed to get the system working properly on the chip, and should have the servos and actuators in later this week to start progress on actual fork parameter modification. Early testing results have been very promising thought.

Turns out innovation is kinda hard.

There were some decent ideas in the top 20, but the well drawn ones have been those to advance.

Yeah I could see that being helpful. For the overshoot part, yeah a feedback loop for position of the servo may make some sense, but using good servos or stepper motors would probably work out well enough. Hard to say without building one and trying to break it.

Yeah, that surprised me too. I would have guessed that they way they seem to be doing it, they would have to sample much faster.

I read that they are sampling every 2ms, and the accelerometer is an Analog Devices ADXL78, but I am not sure about the processor.

Thanks man, I'm glad someone brought up Cannondale, because I really like what they did with Simon. Their implementation seems to be an acceleration hard wired to the damper unit, which 'opens' it up in real time when certain loads are applied, which is limited by requiring ridiculously fast latency. My design is a touch different by analyzing a recent window of data using different methods, which reduces the need for low latency hardware and software, which would greatly lower cost.

sorry, didn't mean to do this twice.

A rotary or linear potentiometer can find the position of the servo or actuator which is physically changing the damping very easily and without much noise at all, if that was something you wanted to do. Because it is all done in a %open or %closed fashion in the damper, true values aren't really nessesary for the functioning of the open loop system. That's the cool thing, the system never 'thinks' the damping is anything, its not relevant to the algorithm. That's the main difference between the open and closed loops. There is no value fed back from the system to the controller, it goes straight from voltage from accellerometer input to voltage sent to the actuator. If you were measuring the speed of fork deflection for estimating damping, differentiating a linear pot would be much much better than integrating 2 accellerometers. Brake dive could be very very difficult to model because the pitch of the bike changes soooo much on a trail, and the weight distribution changes constantly (the rider moves around). Certainly possible, but that would require many many more sensors.

I completely skipped one, the second is just a plot of the GPS location, with the color being the roughness value. All of the others are bumped down one due to the ommision.