[ev3] Gyroscope angular resolution

[jaguilar]

Hi there, I'm working on writing a program to run the gyroboy robot from scratch. I'm using a PID controller to control the bot. Things seem to be working okay-ish, but I'm wondering if the gyroscope has any more resolution than what is available through the angle() method. It seems like it might help with the p component of the controller in particular if I could see sub-degree signals.

[jaguilar]

For anyone finding this in the future: it appears that the gyro speed variable has a lot more resolution than the angle variable, because angle is an int.

Think about a scenario where you're sampling the data at a rate of once per 0.015s. Suppose your robot is rotating at a rate of 4 degrees/s. How long will it take you to know that with the two methods?

• With gyro.speed(), you know immediately, modulo any smoothing you need to do. My program smoothed over a window of 4 observations with an average sample rate of around 100hz, which is enough to filter out most of the noise without excessively increasing latency. That means that we would know the speed within about 45ms of the movement starting (with the first sample being taken at 0ms). And we get the first inkling of the signal immediately.
• With gyro.angle(), the first you know you're moving at all is when you pass the 1 degree mark, after 250ms. But you won't know that your rate is 4 degrees per second until you get another datapoint 250ms later. So you don't know your velocity until about 500ms after you start looking, 10x as slow.

It seems that gyro.speed() can be used to compute a decent approximation of gyro.angle() by summing it and multiplying by the change of time. This works decently well especially if you have some mechanism for calculating a "true" angle by observing some other variable.