Marlin: lround function used to calculate position might create wooble

[Nutz95]

Hello,
I have Z wooble issues and i believe that this beahvior comes from the firmware.

I'm using RC6 and here is my config:
lead screw: 1.5mm lead
stepper: 16 microstep (4988)
layer height : 0.21 or 0.12
Zsteps/mm: 2133.33

I have wooble on my printer when using this leadscrew and i have not when using standard M8 leadscrew.
I've been tweaking my printer to make sure my z axis was moving perfectly straight by placing a laser on my z axis and monitoring the projection of the laser spot on my bed when moving up.
The conclusion is that when i move up the laser spot stay on the same location. => no oscillation or similar behavior.

This said, i can confirm using prusa calculator that when i use a 1.5mm leadscrew i need to use 2133.33mm as Zsteps/mm.
And when i double check the compatible layer heights i find 0.12mm and 0.21mm being compatible, meaning that they need an integer amounts of steps to perform this layer height.

But when i look at the code we are always using axis_steps_per_unit[Z_AXIS] inside a lround() function.

for example inside planner.cpp : target[Z_AXIS] = lround(z * axis_steps_per_unit[Z_AXIS]); (around line 582)

the lround function is not well documented and from the wooble i can notice i understand that it's doing a lower rounding on the value inside the ().

and to me if z is 0.21 and axis_steps_per_unit[Z_AXIS] = 2133.33 i get
0.21 * 2133.33 = 447.9993
when rounding down i have 447.

Again from calculation (we can confirm using prusa calculator) the proper amount of steps i need to perform is 448 not 447.

200 steps/turn * 16microsteps * 0.21mm(layer height) / 1.5mm (screw lead) = 448

And i believe the function used in planner.cpp needs to be rounded to the closest integer not to the lower one.

Does this makes sens to you people? or Am i wrong somewhere ?

[brainscan]

The only way I've found to eliminate the z wobble is to use layer heights that divide nicely into the lead of my z screws. For 1.5mm lead I'd use 0.3, 0.15 or 0.075 or for M8 0.25, 0.125, 0.0625 otherwise I always have wobble. Sorry if this is of no help to you.

Sent from my iPhone

On 2 Sep 2016, at 12:18, Nutz95 <[email protected][e-mail]> wrote:

Hello,
I have Z wooble issues and i believe that this beahvior comes from the firmware.

I'm using RC6 and here is my config:
lead screw: 1.5mm lead
stepper: 16 microstep (4988)
layer height : 0.21 or 0.12
Zsteps/mm: 2133.33

I have wooble on my printer when using this leadscrew and i have not when using standard M8 leadscrew.
I've been tweaking my printer to make sure my z axis was moving perfectly straight by placing a laser on my z axis and monitoring the projection of the laser spot on my bed when moving up.
The conclusion is that when i move up the laser spot stay on the same location. => no oscillation or similar behavior.

This said, i can confirm using prusa calculator that when i use a 1.5mm leadscrew i need to use 2133.33mm as Zsteps/mm.
And when i double check the compatible layer heights i find 0.12mm and 0.21mm being compatible, meaning that they need an integer amounts of steps to perform this layer height.

But when i look at the code we are always using axis_steps_per_unit[Z_AXIS] inside a lround() function.

for example inside planner.cpp : target[Z_AXIS] = lround(z * axis_steps_per_unit[Z_AXIS]); (around line 582)

the lround function is not well documented and from the wooble i can notice i understand that it's doing a lower rounding on the value inside the ().

and to me if z is 0.21 and axis_steps_per_unit[Z_AXIS] = 2133.33 i get
0.21 * 2133.33 = 447.9993
when rounding down i have 447.

Again from calculation (we can confirm using prusa calculator) the proper amount of steps i need to perform is 448 not 447.

200 steps/turn * 16microsteps * 0.21mm(layer height) / 1.5mm (screw lead) = 448

And i believe the function used in planner.cpp needs to be rounded to the closest integer not to the lower one.

Does this makes sens to you people? or Am i wrong somewhere ?

You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHubhttps://github.com/MarlinFirmware/Marlin/issues/4753, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ACHemBFOxe1nMTW-aen43FqS1wgn0m7Hks5qmAXmgaJpZM4Jzm5e.

[thinkyhead]

lround() is actually "round to nearest and cast to long integer."
So it actually rounds up if the decimal part is over 0.5.

Where'd you get "lower round"? Usually we call that floor().

[thinkyhead]

use layer heights that divide nicely into the lead

Also make sure the coupler connecting the stepper motor is of the flexible variety, and don't constrain the lead-screw. Let it hang as loose as possible, otherwise it will divert the axis.

[emartinez167]

Wild thought occurred (out of ignorance). Could this rounding/Z wobbling be
inadvertently causing issues with bed levelling?
On Fri, 2 Sep 2016 at 20:21, Scott Lahteine [email protected]
wrote:

use layer heights that divide nicely into the lead

Also make sure the coupler is of the flexible variety and don't constrain
the lead-screw. Let it hang as loose as possible, otherwise it will divert
the axis.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
https://github.com/MarlinFirmware/Marlin/issues/4753#issuecomment-244359743,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AGcPbyQjbyRCMCnJZoaDp8ny-H6SmK_sks5qmBTOgaJpZM4Jzm5e
.

[Nutz95]

Thanks you both,

• I'll try what you say Brainscan, i've never really tried theses layer heights with my 50mm tower.
• Thanks thinkyhead for clarifyng the lround!
  if it's really doing this then i guess it's fine on this part.

On the Coupler and coupler side i've tried everything, and i could'nt notice any difference.
i've even tried

• to let everything free
• to contrain the bottom
• to contrain the top
• to contrain bottom and top
• to push the lead screw all the way down against the motor shaft inside the coupler
• to let some space from the motor shaft and the leadscrew inside the coupler
• to let the Z nut free on one direction

I spent a few weeks testing everything, mounting/unmounting my printer... making only 1 change at a time.
And everytime i get wooble

About bed leveling i haven't tried to disable it to check the difference.
But i guess it could amplify the issue as the Z axis is moving both upward and downward during the print.

[thinkyhead]

inadvertently causing issues with bed levelling

No. The cause of the current bed leveling bug (errant Z position following G29) is well known. We will be meeting to work on a comprehensive solution soon.

[thinkyhead]

everytime i get wooble

Please post some pictures of the banding on your prints. This will give some clue as to the nature of your wobble or irregular layer heights.

[Nutz95]

sure i'll post some this coming week end

[Blue-Marlin]

2133.33mm as Zsteps/mm

The max error is one step -> 1/2133.33mm -> forget it.

[Blue-Marlin]

If you want to use DISABLE_Z it is advisable to use a laeyer height with a 'even' full step.

[Nutz95]

The max error is one step -> 1/2133.33mm -> forget it.

I don't say the total height of my plastic part is not correct.
I says that each of my layers don't have the same height, some are smaller, and some are higher, but the overall height is good.

A 50mm tall rectangular tower is 50mm high

[thinkyhead]

All kinds of things can affect this. For example, a direct-drive extruder can get pulled upward by the tightly-curled filament near the end of a spool. Periodic thermal expansion of the heated bed can produce uneven layer heights. And as mentioned, if you use layer heights that don't comport with your whole steps, and/or if you allow the Z steppers to sleep between layers…. Many possibilities. Is this a "Prusa i3" type model with moving bed?

[Nutz95]

Yes it's a prusa i3 like printer.
I print on a cold bed so no expansion issues on this part
I've tried with 0.15 and 0.3 so far and i still get wooble.
But i'm doing a more complete and documented set of tests i hope i can bring you some nice pictures by the end of the week end.

I also want to take some time to thanks all of you for

• your support
• the great job you are doing, guys you are all amazing.

[Nutz95]

I couldn't make the photo yet but i could solve my wooble issues.

I changed 1 thing at a time and i could see that what Brainscan proposed had some improvement on the wooble.
It was not the root cause but it definitely helped.

The root cause was a slight mechanical oscillation of my nut holder on my Z axis.
So i used 0.3 or 0.15mm layer height + increased the rigidity on this holder => no wooble anymore.

I could also test the impact of using 2133.33 and also 2134 as ZStep/mm and the lrounding seems to work properly as it had no impact on the final print.

I'll close this git issue and thanks again for your great advices.

[thinkyhead]

On Prusa style printers, one end of the z rods must be let free. Constraining the rods only transfers wobble in the rods to the carriage.