Noticed some offset on the left half box filter chop from the expected value. See video for example. Does not affect box, gaussian, or left half gaussian filters. Not sure what’s up. https://youtu.be/daupz566J_M
Hi @kendrick90,
having a hard time reproducing this in 2025.31550, could you see if attached file has the same behavior for you, and if not if there might be some differences between your setup and the test?
cheers
Markus
base_filterIssue.tox (2.0 KB)
Here is a pared down sample that matches your network. It looks like it has to do with the time slice and filter per sample options. When filter per sample is checked on the time slice option can’t be changed and gets stuck in the wrong option. Disabling filter per sample then turing time slice off and reenabling filter per sample fixes the issue. Also noticed that a new filter chop will come in with time slice set to on even though it thinks the default is set to off. https://youtu.be/gKv6cZKZzkk
FilterIssueExample.5.toe (5.0 KB)
Hi @kendrick90,
great - I can reproduce this.
Theoretically it should not be necessary to turn off timeslicing as the Filter per Sample option will set the CHOP to be not timesliced. So I guess the workaround is to leave the Timeslice parameter toggled on and enable the Filter per Sample parameter.
I logged the bug for us to take a look.
cheers
Markus