Using rotational smoothing causes snapping and camera roll

[VZout]

Very high rotational smoothing values in combination with high sensitivity causes the camera to snap (to the closest quat rotation) and roll which is mildly dizzying 😅

[erwanvivien]

Having the same issue.

@h3r2tic Do you have an idea where in the code the bug is ? I would like to have a look, thanks

[erwanvivien]

I think it's there: https://github.com/h3r2tic/dolly/blob/main/src/util.rs#L33-L55 ?

[h3r2tic]

Don't know how the issue looks, so it would be hard for me to guess 😅 I don't currently have much time to investigate, sorry. I've seen issues with the look-at transform still having some numerical trouble; I've also seen cases where exp with large numbers can give NaNs, and also cases of quats "overflowing", and then (s)lerp suddenly switching to the other direction.

[erwanvivien]

I'll take a look 👌

[h3r2tic]

That's awesome, thanks :D

[erwanvivien]

I've dig into it a bit:

Some possible tracks:

• interpolate should not use delta time based for it's Lerp¹ ?
• Maybe use a smooth damp instead, because it needs start and end position, which we can compute (basically rotation without lerp / smooth damp)

1: https://gamedevbeginner.com/the-right-way-to-lerp-in-unity-with-examples#lerp_slowing_down

---

In my tests, moving the mouse real fast, produced an interp_t of around 0.06f32 which is not much, and could cause the lerp to be wrong? I don't know much

[h3r2tic]

The whole point of ExpSmoothed however is to create exponential smoothing just as the article shows, where the slowing down is a feature 😅

[erwanvivien]

Ohhh alright 🤔 Then maybe the problem is more around interp_t?

[erwanvivien]

I think there is less problem using the ExpSmoothed like you did, because in the article, it is using delta time, so it's always a small number, thus always on the slow part of the Lerp

Here the function is greatly helping the smoothing, I think ?

[h3r2tic]

The main difference is the framerate independence via the interp_t thing.