First a background.
Since we are dealing with corexy, where a pair of motors operate the toolhead, It is really important to get the belt tension correct AND equal on both motors.
Input shaping uses clever algorithms to alter motion in real time, thus negate unwanted frequencies and reducing ringing/ghosting.
In this link you can read more on how to equal belt tension with the help of automated scripts.
Measuring Resonances
The graphs I embedded are a function of frequencies during input shaping and their respective amplitudes. In other words “how much the toolhead shakes at each frequency”
There are multiple input shaper algorithms such as ZV, MZV, EI…
Each use different frequency and amplitude range to try and negate vibrations.
You can say, that the more aggressive the shaper, the more smoothing you get at edges, decreasing dimensional accuracy.
You can see in the graphs, that before I had two separate frequency spikes, thus input shaping algorithm recommended 2HUMP_EI at 97Hz negating both spikes at @60Hz and @130Hz at the cost of higher smoothing and lower accelerations.
Once I equalled the belt tensions the spike at @130Hz hugely dropped in amplitude and shaper recommends MZV @69Hz with 0 vibrations, higher accelerations and reduced smoothing
In conclusion. Klipper is awesome 