Seems like Input Shaping plugin implementation is not usable

gnydick

I see it this way...

Implementation

The Input Shaping plugin measures straight line movements, one at a time.
There is also an option to not record from the very beginning of the move, I assume because that would show a large impulse at the beginning.

Someone please correct me if I'm wrong, but isn't Input Shaping for smoothing out changes in velocity?

Problems with Implementation

There is nearly no change in velocity for a single, straight line segment that is both fast and high acceleration, shouldn't it be recording explicit velocity changes only? The current way just measures vibration and not the effect of corner changes or speed changes.
Why would you not record the beginning of the movement? There are many different circumstances where the tool head will be moving from a dead stop.

alankilian

@gnydick I'm confused why you are complaining about an OPTION of not recording the beginning of a move.

If you want to record that data, don't choose the option.
If someone else wants to ignore that part of the move for a reason YOU don't understand, they can choose that option.

How about explaining to us how you tried to use the input shaping plugin and post the results you had from using the information it provides? That might help the person who provided it to you learn about your specific situation and might be able to help you get what you want out of this deal.

gnydick

@alankilian I'm not complaining about anything. I'm asking a question based on my understandings and hoping someone can either corroborate or correct me.

Try actually reading the post instead of being offended by the title.

jens55

@gnydick sorry but the way I read it, you are complaining ... and the title is part of your post. If your intent was to ask a question then that intent was poorly executed. Most of us are guilty of that on occasion but when it is pointed out that something should have been phrased differently, digging in one's heels is not advised.

achrn

@gnydick said in Seems like Input Shaping plugin implementation is not usable:

There is also an option to not record from the very beginning of the move, I assume because that would show a large impulse at the beginning.

If you don't like the option don't use it.

FWIW, I assumed that it was because "Since a sampling move can introduce low-frequency responses (< 10Hz), you can choose if you want to record data either while the move is being performed (default) or after the move has finished." (https://docs.duet3d.com/User_manual/Tuning/Input_shaping_plugin).

I tried both and get a cleaner spike (though lower) with "capture data during the whole length of the move" (which is the option I think you are referring to) off (i.e. unticked).

dc42

@gnydick said in Seems like Input Shaping plugin implementation is not usable:

There is nearly no change in velocity for a single, straight line segment that is both fast and high acceleration, shouldn't it be recording explicit velocity changes only?

The velocity changes that excite ringing are the acceleration and deceleration segments - more specifically, the changes in acceleration/deceleration. The ringing occurs during the whole of the move after the initial acceleration, and after the move completes. The frequency resolution is half the reciprocal of the duration of the data collection, for example collecting data for 1 second give a resolution of 2Hz, 2 seconds give a resolution of 1Hz, u.s.w. So the way it works is that the acceleration and deceleration segments of the move excite ringing, and the accelerometer measures that ringing.

@gnydick said in Seems like Input Shaping plugin implementation is not usable:

Why would you not record the beginning of the movement

If you record the whole move then you not only record the acceleration caused by ringing, you also record the acceleration and deceleration segments. These mostly but not completely manifest themselves as acceleration close to zero frequency. If you take measurements starting immediately after the move stops, then you just measure ringing. On some machines (e.g. my E3D tool changer with a Hemera tool) this works really well. On other machines it doesn't pick up the ringing as well as measuring the whole move does. That's why we give you the choice.

gnydick

@jens55 yeah, you're right. I've been having neurological issues lately; everything is really, REALLY hard to parse, grok, and reply. I'm amazed I was even able to post the post.

P.S.

It's very frustrating, and I don't know what's causing it and neither do my dr's. Not asking for pity, just venting.

gnydick

@dc42 so, by that logic, if I had a resonance of 40Hz, it doesn't matter what the impulse was that caused it? That is very counterintuitive.

Since the ringing is a physical phenomenon, not something that manifests in the motors, how does this work?

How can the algorithm universally adjust no matter what the initial impulse was?

gloomyandy

@gnydick You may find some of the discussion in this thread of interest. In particular the comments about the sampling methods provided by "pixelpieper" who can clearly remember more of his signal processing classes than I can.

Edit: Helps if you remember to include the link:
https://forum.duet3d.com/topic/25975/comparing-klipper-and-rrf-input-shaper-data-collection

engikeneer

@gnydick said in Seems like Input Shaping plugin implementation is not usable:

it doesn't matter what the impulse was that caused it? That is very counterintuitive.

Since the ringing is a physical phenomenon, not something that manifests in the motors, how does this work?

Ringing is caused by the motion of the printer exciting some oscillation/mode shape. At its simplest form, think of a mass on a spring. That will always want to bounce/oscillate at a frequency of sqrt(k/m). k is stiffness and m is mass. You can hit it as hard or as gently as you like, but it will always want to respond at that frequency. Twang a ruler on the edge of a table and it will make a certain sound. The note it makes will be the same frequency, no matter how soft/hard you twang it, as long as you don't change how far it overhangs (I.e. the mass/stiffness)

Essentially your print head is a mass moving around on a spring (flexing belts, gantries etc), so will always want to respond at a certain frequency (40Hz in your case). The accelerometer is a way of measuring that response and find what that frequency is.

It actually gets a hit more complex as there are lots of masses and lots of little springs in a real system, so you will get many frequencies the system may want to respond at, just some will always be more dominant. Again, using the plug in you can look at these and pick a shaper profile to cover the worser ones.

Theoretically an impulse is an excitation that includes every frequency there is, hence the move you make should excite all the possible modes and so you can find them.

Once you apply the shaper, it essentially removes those frequencies from the acceleration part of moves when you are printing. If you then aren't exciting those frequencies, it is harder for the system (printer) to respond at those frequencies (ringing). Same idea as purposely waggling the ruler at a different frequency to the sound it wants to make - it'll never be as loud!

gnydick

thanks, @gloomyandy