Hello All,
I have been scanning the forum and not yet found any detail on how to interpret input shaping graphs.
I get the general gist but I am looking for more.
The intention is to have an open discussion where those with factual knowledge on the subject can share.
A couple of items to get the discussion going:
-
What are the dashed vs full fill lines?
-
Understanding that each printer is different, are there some general guidelines on the magnitude of spikes? What could it tell us if the magnitude is greater than this?
-
Should we be shooting for 1 spike and then a smooth downward slope or is it ok / expected to have secondary (or even tertiary) bumps?
-
Are there some guidelines on what frequency range we should expect to be seeing the spike? What can we take away from a higher or lower frequency?
I hope I haven’t missed all this explained elsewhere, if I have I apologise, if not then I hope this can form a place for people to gain a better understanding of what the input shaper output is telling us.
The solid lines (as shown in the graph at Measuring Resonances - Klipper documentation) are the measured frequencies picked up by the accelerometer. The dashed lines represent the theoretical vibration reduction for a given input shaper configuration. The scale of the dashed lines is shown on the right side of the graph. In general, you want to see the dashed lines at a low point at the frequencies where the solid lines have a high point - this indicates that the input shaper should (at least in theory) cancel the found vibrations.
As a general “rule of thumb”, a lower magnitude is preferable. When considering magnitudes it’s very important to look at the scale on the upper left side of the graph (eg, “1e4” or “1e3” in the graphs shown in the document linked above). It’s not clear to me that the magnitude has a direct correlation to a tangible physical quantity, but lower is better. The magnitude can be useful for comparison purposes - for example adjusting the belts may result in a change in magnitude and/or frequency.
As a general “rule of thumb” single clear spikes are preferred because input shaper is better at cancelling them. However, when comparing graphs it’s important to take into consideration magnitude (again, be sure to check the scale on the top left of the graph). A single spike is desirable, but not at the cost of a spike that has significantly higher magnitude. Due to the way the graphs are drawn, a high magnitude spike can visually look “more clean”, but this visual appearance can be misleading. For example, a wide “lump” showing active frequencies between 40-80hz at “4 * 1e3” magnitude is better than a single spike at 60hz with a magnitude of “4 * 1e5” - in the latter case the wide range of frequencies are still there, they’re just not drawn as prominently because a worse problem is highlighted.
A wide range of spikes can also be an indicator of a failure during the resonance test (for example, a fan left on during the test which results in a spurious spike). I don’t know of anyway to describe which graphs look suspect and which look normal, but I have seen graphs which just “don’t look right” at first glance.
As a general “rule of thumb” higher frequencies are preferable because they are easier to cancel with input shaper and because they are less likely to result in print artefacts in general. However, the measured frequencies are often related to the physical hardware, and it’s often not simple to “change them”.
Cheers,
-Kevin