Small variations in timing of spikes produce several clusters

[guilhemibos]

Hi

I'm new to Kilosort. I'm using KS3, recording from 32 channels linear stereotrodes (50um between contacts), with open pehys. I have a hard time calibrating KS in order to have decent results which do not need very long manual curation.

Often, I observe some sorting like this: A large spike is separated arbitrarily into 2 or more units because their is a small temporal drift in the waveform, messing up with correlagrams of neurons. Is there a reason why this happens and how can I get rid of this issue? Thanks