Hi, The new firmware is working great in general regarding the new acceleration algorithms. Speeds up to 160mm/s are possible. Of course in real life limited to lower values because of the print quality and maximum flow of the extruder with a specific filament.

Only: I had problems with detailed and curved objects: when printing the loops at 100mm/s (The Draw speed) the printer was making odd sounds (rrrrrrrr sound). A perimeter of a curved object at 100mm/s was very bad, while the perimeter of a straight object was still good.

I thought it was of to small segments, and by that the planner chocking, but tweaking the "Path smoothing" to higher values wasn't improving much.

As the earlier firmware versions seemed to improve with an acceleration tweak I tried it as well with this version.
The following values added to the start Gcode seemed to improve those detailed curvy prints when printing at high speed:

M1203 A3000 D3000 F1200

And to spice up the hold torque of the stepper motors I increased the current of the motors by adding these lines:

M906 X50
M906 E50

So with a Draw speed of 100mm/s all seems te be printed well. Of course setting the perimeter and top fills to 30% will result in a nicer skin.
The speed of the infill to 120mm/s seems ok too.

I think the support structure is standard printed at 150% of the Draw speed: so 150mm/s in this case. I asked CU to add a variable to influence that speed too.

It would be great if others would try this settings as well, to see if it's improving with other prints too or not.
Bart

January 31, 2016

Bartaar Bartaar
Service partner
1677 posts

21 replies


I would need to translate these settings into CURA 15.02.X (old cura) since as much as I would love to use craftware on OSX its still functionally broken beyond usable for me.

Also is this with a stock Craftbot? or CB+ .. or CB/CB+ with your new modified hotend replacement?

January 31, 2016

Bixit219 Bixit219
Superstar
1451 posts

Hi Brixit, what do you mean with translate to Cura? I use that startGcode settings for all the slicers. So I added that lines to Kisslicer too. (KISSlicer is for me the Mac version of Craftware).
Or do you mean to change all the speed settings in Cura to 100mm/s etc?

It's only working when you have an up to date Firmware version in the CraftBot. So 8249 or higher (I think it's now 8263).
I have put that firmware 8263 on my original CraftBot as well the PLUS CraftBots. Because of testing: all my CraftBots are printing with the alternative hotend, so it's working with that too.
No reason it won't work with the original hotend. In fact: I think the original hotend is capable to extrude more filament per second (the max flow) with PLA. With other filaments it's different.




January 31, 2016

Bartaar Bartaar
Service partner
1677 posts

It kinda looks worse to me, I bumped up my speed to 70mm/s (from 50) and turned off slow down on perimeters.
The left is before the m1203 change right is after, same Gcode.
Both are far from perfect but before the change I at least had some spots that looked smooth. In the photo the left cylinder has the smoother spot showing, other sides are almost as bad as the right but the right cylinder has no smooth "side".
M1203 A3000 D3000 F1200
XY stepper currents left at defaults.

Acceleration tweak 8249+ firmware for curved objects

For comparative purposes here is the STL file used. (named "tube 11id 14od")

January 31, 2016

Matthew L Matthew L
Exalted
703 posts

That's not improving at all! Just to be sure: which firmware did you use?
I just tried it with a much larger round object, so I'm gonna try with yours now.
Thanks!

January 31, 2016

Bartaar Bartaar
Service partner
1677 posts

HMI: 1.1.7864
pr3dator: 1.1.8263

Also just realized cooling was enabled so it limited the max speed, the objects above really printed at around (target speed) 50mm/s.

January 31, 2016

Matthew L Matthew L
Exalted
703 posts

Would you mind to try another (bigger) curved / detailled object too? Maybe there is influence because it's a small object, the layer below is still soft because of the speed. And with the firmware tweak it's printing faster and by that less change the layer below is cooled?
Just hoping and thinking, now trying myself ;-)

January 31, 2016

Bartaar Bartaar
Service partner
1677 posts

Sure, I will make a object for another test unless you have a good model to test it with.

I just tested this with A/D at 3000 and f left at default, it looked almost the same but the smooth area was slightly smaller.

Perhaps this issue is not a acceleration issue but a over loaded/busy controller causing a delay and micro pauses in the tool path? The none functioning pause button made me think of that, no matter how much I tried I could not get this object to pause.

January 31, 2016

Matthew L Matthew L
Exalted
703 posts

Uploaded a new object to use. Same spot, named "curvy test1"

I have scaled it to 150% it's showing your rrrrrrrrr issue well:)
Printing with no top/bottom or infill.

February 1, 2016

Matthew L Matthew L
Exalted
703 posts

I forgot I made one myself too:
http://www.thingiverse.com/thing:1179244
The curves of it all have a different 'resolution" The separate 'spline" has the most detailled surface.
I try yours as well!

February 1, 2016

Bartaar Bartaar
Service partner
1677 posts

You are correct, looks like the small object was too small.
This printed at 70mm/s no perimeter slow down.
Lower object is before the M1203 A3000 D3000 F1200 command.
Not all of the bumpy surface is solved but it is definitely better.
Acceleration tweak 8249+ firmware for curved objects
Acceleration tweak 8249+ firmware for curved objects
Acceleration tweak 8249+ firmware for curved objects
Acceleration tweak 8249+ firmware for curved objects
Acceleration tweak 8249+ firmware for curved objects

If you print it be sure to scale to 150% to keep the test consistent.

February 1, 2016

Matthew L Matthew L
Exalted
703 posts

The acceleration and deceleration values I understand but do you know what the minfeed value is doing?

Just did another test with your changes but at a slow 35mm/s all looks good, no strange effects.

February 1, 2016

Matthew L Matthew L
Exalted
703 posts

That's looking better indeed! I didn't tweak that settings, so things should improve when trying some variations of it. The last time I tried it with the old firmware the F setting had the most influence.
I have to get my laptop and try with an USB kabel... to much walking with USB sticks / forgetting it / taking the wrong USB stick / agitated returning / tripping over the cat / shouting at the CraftBot....

February 1, 2016

Bartaar Bartaar
Service partner
1677 posts

I'm not sure, but I think that the F value seems to be the threshold were de/acceleration is applied. So under that value no acceleration? F1200 is 20mm/s that can be done unaccelerated. (at least in the old days.)

February 1, 2016

Bartaar Bartaar
Service partner
1677 posts

You get the bumps from uneven speed. The flow of the filament at the nozzle is about constant, but the head speed varies, so you get bumps at the slowdown points, because the head deposits the same amount of material to a shorter distance. If you really want to avoid bumps, you have to do the perimeter slow, about 20 mm/s IMHO.

The exact formula is this: slowdown/speedup occurs if

F > sqrt(0.5/(1-cos(direction_change))*F_in_M1203

So if you do it faster, you get bigger blobs. If you increase F in M1203, then you get smaller blobs, but bigger noise/resonances, it is a tradeoff. Also, if you increase A/D, then the slowdowns are shorter, you get smaller blobs, bigger noise.

February 1, 2016

psanyi psanyi
Exalted
958 posts


Bart ter Haar wrote:
So with a Draw speed of 100mm/s all seems te be printed well. Of course setting the perimeter and top fills to 30% will result in a nicer skin.
The speed of the infill to 120mm/s seems ok too.

I think the support structure is standard printed at 150% of the Draw speed: so 150mm/s in this case.


It was these settings I was referring to . .. I added your gcode to my gcode.start .. ill see how it goes (im on the most recent firmware)

February 1, 2016

Bixit219 Bixit219
Superstar
1451 posts

Still playing around with A/D F settings, I have found i get more ringing with the 3000 1200 values, as would be expected with faster direction changes. It hard to find the happy balance:D

Default:
Acceleration tweak 8249+ firmware for curved objects

M1203 A3000 D3000 F1200:
Acceleration tweak 8249+ firmware for curved objects

Default:
Acceleration tweak 8249+ firmware for curved objects

M1203 A3000 D3000 F1200:
Acceleration tweak 8249+ firmware for curved objects

February 2, 2016

Matthew L Matthew L
Exalted
703 posts

Matthew L wrote:The acceleration and deceleration values I understand but do you know what the minfeed value is doing?

The minfeed value is the allowed instantaneous speed change. At corners the head slows, then abruptly changes direction, and speeds up. If this direction change is done at a speed, then the speed vector changes in an instant and the acceleration is infinite in theory, but in practice no nasty things happen, if you do it at slow speed. This is what you set with the minfeed.

I've rearranged the slowdown formula to show the maximum angle change, that can be done without slowdown.

angle = acos(1-0.5*(minF/F)*(minF/F))

Some numbers: with the default F240 you have to slice a full circle to 200 slices. With Bart's F1200 40 slices is enough.

February 2, 2016

psanyi psanyi
Exalted
958 posts

Thanks Psanyi I think i get it.
So the higher the F value you have the less speed change (Acceleration/Deceleration) you will be need to do before you can take a corner? Ie it will not slow as much before it takes a corner?

Or to put in another way the F value is the speed the bot will be moving at the end of deceleration, then takes the corner and accelerates again?

Edit:
That may be a bit over simplified, I guess the the angle of the corner is also taken in to account as your formula would imply.

February 2, 2016

Matthew L Matthew L
Exalted
703 posts

This image show several corners. The black line is the speed vector before the corner, the blue is after and the red is this speed difference, that is limited by the M1203 F
Acceleration tweak 8249+ firmware for curved objects
The speed difference is the same in all cases, but the slowdown depends heavily on the angle. For turns over 60 degrees it slows down to a lower speed, than the minfeed value. This is why it is important to split the curves into many straight segments.

February 2, 2016

psanyi psanyi
Exalted
958 posts

Nice drawing, I can visualize that better then the formula:)

If I understand correctly:
Only velocity change magnitudes (red line above) that are greater then the m1203 F value have acceleration applied to it.
Angles over 60 degrees have a extra modifier applied.
Any other direction changes are not slowed down before hand.


By "Enough" you mean before acceleration will need to be applied?
"Some numbers: with the default F240 you have to slice a full circle to 200 slices. With Bart's F1200 40 slices is enough. "

Thanks for being patient with me, I feel like I made this more complex then it needed to be:)


February 3, 2016

Matthew L Matthew L
Exalted
703 posts

Matthew L wrote:Only velocity change magnitudes (red line above) that are greater then the m1203 F value have acceleration applied to it.
Yes.

Matthew L wrote:Angles over 60 degrees have a extra modifier applied.
Any other direction changes are not slowed down before hand.
No. The the three vectors happen to be equal size at 60 degrees, that is all. The formula is applied to any direction change and if the result warrants it, then slowdown as well.

Matthew L wrote:By "Enough" you mean before acceleration will need to be applied?
"Some numbers: with the default F240 you have to slice a full circle to 200 slices. With Bart's F1200 40 slices is enough. "
Yes. If you make a circle with 200 segments at F7200, then it will be done at 120 mm/s without slowdowns. If you use Bart's parameters, then even a 40 segment circle will be done at 120 mm/s.

Matthew L wrote:Thanks for being patient with me, I feel like I made this more complex then it needed to be:)
No problem, these vectors are tricky :-)

February 3, 2016

psanyi psanyi
Exalted
958 posts
To start a discussion or reply to a post please Login or Create an account.