Relative acceleration between bed and gantry

[olhev]

As of now the input shaper measurement method measures the absolute vibrations of the printing head over the inertial reference of the Earth, but when printing defects result from not from the absolute errors, rather from relative errors in the motion printing head vs bed.

I think this is obvious, just take the extreme case of a printer hanging from the ceiling and with a super heavy tungsten head, while the frame is made of plastic: the adxl would measure about zero vibrations while the rest of the print moves like spaghetti with terrible prints.

Likewise, installing vibration absorption feet would decouple the frame from the table, resulting in the ratio of frame weight over printing head weight being significantly lower, and as consequence the adxl less able to measure the actual relative motion head vs bed.

I searched in the forum and I found only Interpreting Input Shaping results from 2 different printer mount methods - #13 by Sineos which I don’t agree with. The reasoning is correct only when the vibrations are caused externally and you want to avoid transmitting then to the printer, but for dissipation of the vibrations from the printer itself, bolting the printer would increase its mass to huge amounts and the assumptions of the input shaper calibration more realistic (the input shaper correction during print is not the topic of my post).

Were tests performed to measure the actual relative positioning error head Vs bed? Is the sampling rate (or communication speed) of the adxl even suitable for a concurrent measurement of two adxl?
Would two sequential runs give valuable information about the vibrations of the bed caused by the accelerations of the printing head?

I’m assuming corexy here.

@dmbutyugin since you worked on it.

[TheFuzzyGiggler]

I’m not sure what you’re getting at here.

measures the absolute vibrations of the printing head over the inertial reference of the Earth

An accelerometer measures the variation in acceleration due to gravity, which, yes, is Earth’s gravity because it’s the dominant gravitational source. Unless you want to take your printer ~198,000 miles away and measure the resonances near the moon I don’t think you’re going to get around that.

Were tests performed to measure the actual relative positioning error head Vs bed?

You’re measuring resonances in the frame due to movement. It’s generally assumed the bed is attached to the frame of the printer.

In general, It’s finding the frequencies where movement causes the frame to resonate and avoid those. All machines have resonance no matter how big or heavy they are or what it’s coupled to. 30 ton milling machines have resonances.

Granted the frequency of that resonance is far different than a 3d printer make of bolted together aluminum extrusion.

Also the ADXL is a 3 axis accelerometer and generally you’re only worried about the resonance at the print head since that’s where the filament comes out of. You’re trying to avoing ringing artifacts caused by constructive resonant frequencies feeding back on themselves and causing “wobble”.

You can bolt the printer down to concrete pillars that are buried to a depth of 100 ft, but I promise your gantry it still going to resonate. That’s just what things do when they move.

I don’t see how your bed would be moving unless your printer is shaking itself apart. I mean sure, It probably moves a little but the general assumption is that it’s rigidly attached to the frame of the printer so the entire printer moves and it nullifies the movement. Meanwhile the gantry can move independantly.

Especially on CoreXY flying gantries like the Voron, It has linear rails but it’s just attached by 4 machine screws on a 3d printed part that flexes. Otherwise it’s held up by the belts which obviously aren’t ridgid at all. That’s going to be 99.99999999% of your vibration.

[olhev]

@TheFuzzyGiggler

Unless you want to take your printer ~198,000 miles away and measure the resonances near the moon I don’t think you’re going to get around that

That’s not my point at all.

Due to acceleration, two things move: the print head, and the frame. If you hang the printer from the ceiling, the relative mass of print head vs frame will cause a motion of the frame equal to about (mass head) / (mass frame+bed). Not much, but it’s there. For example, a cube of 50 cm is 2 kg and a lightweight bed assembly another 2 kg, the print head all inclusive 0.5 kg, so if the printer is hanging, with one accelerometer, you measure only the motion of the print head which is 1 - 0.5 / (2+2) = 87%. You can put different values, the concepts stays. And if the printer is standing on the floor, the frame gets stiffer.

The defects on the print result from the relative motion head vs bed, so if the frame is not bolted to the ground and perfectly stiff, an error results from omitting one measurement.

My questions is simply: has a combined measurement been performed, and if so how big is the approximation due to the omission of the bed accelerometer?

I guess I’ll have to do it myself once my printer is ready. I can put an extra adxl on the carriage, one on the bed, and read them while klipper does his thing, and compare the magnitude of bed vs head.

[dmbutyugin]

I generally agree that in some instances it may be beneficial to measure the relative motion of the toolhead against the bed, but it depends. The resonance test moves the toolhead by some fraction of a mm during the test (it depends on the frequency, for example at 40 Hz it moves it by ~0.1mm), so you cannot truly expect that these vibration would really reach the bed (even if the actual higher-amplitude vibrations may). So I don’t think it is particularly necessary for the current resonance test. For the print test, it is already accounted for.

In my opinion, the printers where the measurements one the bed could really be beneficial is when the bed itself is weakly mounted (e.g. from one side) and thus tends to vibrate itself; this you could do by mounting an accelerometer on the bed and just taking the measurements from it alone to find the resonances one should avoid.

Then, to show an example of the measurement of cross-resonances between the toolhead and the bed:

We can generally see that the registered vibrations on the bed are some order of magnitude smaller than the ones on the toolhead, so it does not make sense to account for them at least in this instance.

[olhev]

Now I’m happy, thanks!