[01HNNWZ0MV43FF]

That code looks weird to me either way. You end up with a strange float. I represent framerates as rational numbers, e.g. 60/1. It works great for ffmpeg.

[niborgen]

The Kalman filter code does need floats, though. So here we did indeed need a conversion.

[01HNNWZ0MV43FF]

Sure. I haven't looked at the code so I could easily be off base but, even with floats I want to delay that divide for as long as possible