Exponential easing/interpolation might be incorrect #50011

New Issue

Robert Forsman · 2016-11-12T21:01:50+01:00

Robert Forsman commented

2016-11-12 21:01:50 +01:00

blender 2.78a
Linux bubastis 4.0.5-gentoo #1 SMP Mon Nov 16 11:48:34 EST 2015 x86_64 Intel(R) Core(TM) i7-4800MQ CPU @ 2.70GHz GenuineIntel GNU/Linux

I created an fcurve with two keyframes: (1,1) and (100,1024) When I zoom in on the curve it appears to jump from (1,1) to (2,2) before resuming an exponential track. I would have assumed it didn't reach a y of 2 until keyframe 10 or 11.

In any case, this discontinuity in the first frame after the keyframe seems incorrect.

blender 2.78a Linux bubastis 4.0.5-gentoo #1 SMP Mon Nov 16 11:48:34 EST 2015 x86_64 Intel(R) Core(TM) i7-4800MQ CPU @ 2.70GHz GenuineIntel GNU/Linux I created an fcurve with two keyframes: (1,1) and (100,1024) When I zoom in on the curve it appears to jump from (1,1) to (2,2) before resuming an exponential track. I would have assumed it didn't reach a y of 2 until keyframe 10 or 11. In any case, this discontinuity in the first frame after the keyframe seems incorrect.

Robert Forsman commented

2016-11-12 21:01:50 +01:00

Changed status to: 'Open'

Robert Forsman commented

2016-11-12 21:01:50 +01:00

Added subscriber: @mutantbob

Brecht Van Lommel commented

2016-11-13 02:15:15 +01:00

Added subscribers: @JoshuaLeung, @brecht

Brecht Van Lommel commented

2016-11-13 02:15:15 +01:00

Confirmed: exponential_discontinuity.blend.

I guess this is one for @aligorith.

Confirmed: [exponential_discontinuity.blend](https://archive.blender.org/developer/F398204/exponential_discontinuity.blend). I guess this is one for @aligorith.

Joshua Leung commented

2017-03-06 13:27:43 +01:00

This is very strange...

A while back, I spent some time investigating this issue. I went and double-checked whether the implementations of the easing equations were were using were correct - I went as far as copying the code out of one of the reference implementations, making the relevant adjustments, and trying again.

While there are some slight errors I picked up this way (affecting other cases), ultimately they didn't seem to be related to what we're seeing here. The code used to calculate these curves is practically the same as is used everywhere.

The main point of different I suppose is that we stick to the 0-1 range (on both the x and y axes), and then try to scale up the y-range to accomodate the amount of change between the keyframes. My current best guess is that this is a numeric precision error/discontinuity in the exp function that might not show up if we weren't trying to scale things up by using only the 0-1 range. If anyone has any better suggestions about what we could try here, patches are welcome :)

This is very strange... A while back, I spent some time investigating this issue. I went and double-checked whether the implementations of the easing equations were were using were correct - I went as far as copying the code out of one of the reference implementations, making the relevant adjustments, and trying again. While there are some slight errors I picked up this way (affecting other cases), ultimately they didn't seem to be related to what we're seeing here. The code used to calculate these curves is practically the same as is used everywhere. The main point of different I suppose is that we stick to the 0-1 range (on both the x and y axes), and then try to scale up the y-range to accomodate the amount of change between the keyframes. My current best guess is that this is a numeric precision error/discontinuity in the exp function that might not show up if we weren't trying to scale things up by using only the 0-1 range. If anyone has any better suggestions about what we could try here, patches are welcome :)

Sybren A. Stüvel commented

2019-12-18 10:50:31 +01:00

Added subscriber: @dr.sybren

Sybren A. Stüvel commented

2019-12-18 10:50:31 +01:00

Changed status from 'Open' to: 'Resolved'

Sybren A. Stüvel closed this issue

2019-12-18 10:50:31 +01:00

Sybren A. Stüvel self-assigned this 2019-12-18 10:50:31 +01:00

Sybren A. Stüvel commented

2019-12-18 10:50:31 +01:00

Closing, as this doesn't seem to be an issue any more.

Closing, as this doesn't seem to be an issue any more. ![image.png](https://archive.blender.org/developer/F8231732/image.png)

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Exponential easing/interpolation might be incorrect #50011