Mouse sensors are usually ~16x16 CMOS cameras that just return signed int X/Y movements. Shutter speed is goverened by both luminosity and polling rates. There's no raw pixel output, and even in sensors that have it, it's for focus adjustment only and not useful in realtime sensing. They're also crazy cheap.
Therefore the cheapest way to do ball twist is to just put two sensors and use the error as the twist amount. You could use a singular high resolution highspeed camera pointed at the ball instead, but that won't be the cheapest way to do it.
With only one you can't detect rotation around the axis that goes through the sensor, and for rotation close to that axis the resolution will be very poor.
Comments
Why does it require two sensors?
Mouse sensors are usually ~16x16 CMOS cameras that just return signed int X/Y movements. Shutter speed is goverened by both luminosity and polling rates. There's no raw pixel output, and even in sensors that have it, it's for focus adjustment only and not useful in realtime sensing. They're also crazy cheap.
Therefore the cheapest way to do ball twist is to just put two sensors and use the error as the twist amount. You could use a singular high resolution highspeed camera pointed at the ball instead, but that won't be the cheapest way to do it.
With only one you can't detect rotation around the axis that goes through the sensor, and for rotation close to that axis the resolution will be very poor.
Okay, sure, but that's a very narrow and specific angle though? Not sure I could deliberately rotate the ball in that direction even of I wanted to.
Especially given the limited use kensington uses rotation for (at least to my knowledge)