Extruder Acceleration - Holy Crap!

General discussion topics
User avatar
zemlin
Posts: 453
Joined: Thu Feb 04, 2016 9:32 pm
Location: Indianapolis, Indiana

Re: Extruder Acceleration - Holy Crap!

Post by zemlin » Wed Feb 22, 2017 3:11 pm

Jerk in this case is not the derivative of accel. The value represents the velocity change required before acceleration is implemented. Smaller Jerk values result in a smoother velocity profile. I have not done much testing with these values either to see what impact that has ... setting XY jerk to 1, for example.

User avatar
jimc
Posts: 2888
Joined: Wed Apr 09, 2014 11:30 pm
Location: mullica, nj
Contact:

Re: Extruder Acceleration - Holy Crap!

Post by jimc » Thu Feb 23, 2017 2:44 am

the acceleration values being lowered is good to a point. what I have to add here is that its a balancing act. the lower you make the acceleration the less ghosting you'll get but it also increases the time it takes to do a travel move so now your fighting increased ooze. you need to retune your ooze control settings and make those more aggressive to counteract the longer "hang time" you have on a jump. the e value is just your extruder acceleration. that is going to affect your retraction speeds etc. the xy acceleration are the key ones to look at. jin, myself and a ton of others did a lot of messing with this some years ago. the default in marlin is 3000 and I think most of us settled in around 1000-1500. some tried as low as 800 but then the oozing becomes too hard to control. on short distance jumps no big deal but longer ones become a problem. things just get too sloppy. of course printing a test cube or a single small item will never show any of this. I have been at 1500 for a couple years now. seems to be the happy middle. ive never really noticed a difference when I changed jerk settings. I just set mine to 15 and left it there.

hybridprinter
Posts: 141
Joined: Sun Dec 21, 2014 8:04 pm

Re: Extruder Acceleration - Holy Crap!

Post by hybridprinter » Thu Feb 23, 2017 9:46 pm

what does the 'jerk' setting do? what is the default setting in the firmware on the Rev. E machines?

jsc
Posts: 1863
Joined: Thu Apr 10, 2014 4:00 am

Re: Extruder Acceleration - Holy Crap!

Post by jsc » Sat Feb 25, 2017 7:38 pm

zemlin explained jerk above, but check out this description from the Repetier manual:

https://www.repetier.com/documentation/ ... tallation/

Search down for "Jerk control".

If the change in velocity is below MAX_JERK, the planner just assumes it can make that jump instantaneously, otherwise it will adjust the speeds until the transition is exactly MAX_JERK.
You want high jerk values, because

printing time is reduced.
print shows less blobs.

You want low jerk values, because

it causes less mechanical stress to your printer.
moves are smoother.
filament has better adhesion at directional changes.
reduces printer noise.
you loose steps with higher values.

User avatar
zemlin
Posts: 453
Joined: Thu Feb 04, 2016 9:32 pm
Location: Indianapolis, Indiana

Re: Extruder Acceleration - Holy Crap!

Post by zemlin » Thu Mar 02, 2017 1:20 pm

I've been messing with stuff again.
I'm thinking my early results were mainly the result of very slow overall speeds which was not building up any pressure in the nozzle so resulting ooze was low. Build times, however, were not what any of us would be willing to live with.

My spare time over the last few days has been spent studying ringing and how to minimize it without killing build times. I have been relying on slower speeds for print quality, but I'm hoping I can push my speeds up and still maintain the quality I've been getting - or maybe improve on it.

I'm still working on it, but I believe much of what we see as ringing may be the result of interference patterns between the XY and the Extruder motion profile - like intersecting wave patterns on a wave table (remember high school physics?) One test part had X Accel at 200 and Y Accel at 2000. This part had no ringing on the X axis, but did ring on the Y axis.
If the ringing was due to dynamic instability of the X Axis, it would show on the Y axis, not the X axis.
I was able to get fairly minimal ringing with Y Accel at 950 and X Accel at 1250. I've seen start scripts from other users with similar numbers. With these accelerations I'm getting a bit of corner bulge, probably from ooze as JimC mentions above

I also tried taking X & Y Jerk values down to 1. I didn't really see a difference in print quality, but build times went up.
Looking at Extruder Jerk settings, the extruder speed is so slow using numbers similar to the X and Y axes doesn't seem to make sense - so perhaps extruder acceleration doesn't matter if it's supposed to be tracking with the gantry movements, but what impact does extruder Jerk have on that?

I'm printing a larger test part to get more real-world results - build time is about 4 hours at this point - with my current speed settings S3D says it should print in 2:31.
At an extruder jerk setting of .01 I did not get a good build. Printing a large arc had a bulge at each tessellation vertex and under extrusion between those points - i aborted that print.
Extruder jerk of 0.1 resulted in a pretty good print with a bit of corner bulge and very little ringing - but a bit of roughness on the perimeters. I'm currently printing at 5200 mm/min and a 50% underspeed for outer perimeter and full infills.
Extruder jerk of of 0.5 shows slight ringing, but better surface finish, slight ringing, and less corner bulge.
The print currently running has an extruder jerk of 0.25 and I've increased by X and Y accel settings to 2500 and 1900 to see what impact has. I'm thinking the low extruder jerk will keep the ringing under control and the higher XY accelerations will reduce corner ooze/bulge.

User avatar
ednisley
Posts: 1020
Joined: Fri Apr 11, 2014 5:34 pm
Location: Halfway up the Hudson
Contact:

Re: Extruder Acceleration - Holy Crap!

Post by ednisley » Thu Mar 02, 2017 1:57 pm

zemlin wrote:early results were mainly the result of very slow overall speeds
AFAICT, everything depends on hot end pressure, which is nonlinearly affected by everything else. Absurdly low accelerations and jerks reduce the overall XY speed to the point where the pressure remains more-or-less constant, so all the flow-related problems Go Away, but you could get the same result by reducing the maximum speeds.

The very short straight paths in most real-world models have acceleration-limited maximum speeds: there's not enough distance in the straight sections to get very far up the leading edge of the trapezoidal speed profile. Test pieces with artificially long straight sections have velocity-limited maximum speeds that actually reach the top of the trapezoid profile.

Have you tried just reducing the XY printing speed, while keeping "normal" acceleration & jerk values?

Low maximum speeds should reduce the hot end pressure and eliminate most of the flow-related variations, without slowing non-printing travel motions (thus reducing drool). That will improve the appearance of test pieces with artificially long straight sections by limiting the max speed and (probably) won't dramatically increase the total print time for real-world models that rarely reach the max speed.

User avatar
zemlin
Posts: 453
Joined: Thu Feb 04, 2016 9:32 pm
Location: Indianapolis, Indiana

Re: Extruder Acceleration - Holy Crap!

Post by zemlin » Thu Mar 02, 2017 3:18 pm

ednisley wrote:
zemlin wrote:early results were mainly the result of very slow overall speeds
AFAICT, everything depends on hot end pressure, which is nonlinearly affected by everything else. Absurdly low accelerations and jerks reduce the overall XY speed to the point where the pressure remains more-or-less constant, so all the flow-related problems Go Away, but you could get the same result by reducing the maximum speeds.
What's been bugging me about ringing and the frequency of the ringing is that it isn't speed dependent. From what I've seen, finding it tied to the extruder velocity profile makes sense. If Extruder jerk is too high and the extruder speed is stair-stepping and not following the ramp of the trapezoid while the linear movements ARE ramping, that would create humps in the stream that would show as ringing.

Reducing the XY jerks to 1 didn't make a huge impact in print speed - only a few minutes increase on a small test part.

User avatar
ednisley
Posts: 1020
Joined: Fri Apr 11, 2014 5:34 pm
Location: Halfway up the Hudson
Contact:

Re: Extruder Acceleration - Holy Crap!

Post by ednisley » Thu Mar 02, 2017 5:59 pm

zemlin wrote:extruder speed is stair-stepping
Marlin dispenses step pulses to all the motors from the same IRQ handler, so I'd be completely unsurprised to see some interrupts produce XY motion without corresponding extruder motion.

On the other hand, the extruder's 5:1 gearbox should smooth out much of the intermittent motion, because moving 1 mm of filament requires 500 (micro)steps. Printing at XY = 50 mm/s runs the axes at 4500 step/s (and the extruder at 2 mm/s = 1000 step/s, so I don't think the velocity increments, even at the start of a long straight section, will be coarse enough to be visible.

On the gripping hand, direct-drive extruders, along the lines of the old-and-cranky Makerbot designs, definitely produced visible artifacts during low-speed extrusion.

It'd be interesting to watch the XYE step pulses on a logic analyzer triggered at corners of the model, which you might be able to gimmick using the camera-triggering M-Code command.

I made some measurements along those lines a while ago, but on a scope without enough channels:
https://softsolder.com/2013/06/04/marli ... pt-timing/

User avatar
zemlin
Posts: 453
Joined: Thu Feb 04, 2016 9:32 pm
Location: Indianapolis, Indiana

Re: Extruder Acceleration - Holy Crap!

Post by zemlin » Sun Mar 05, 2017 1:43 pm

I'm done playing for a while - need to get back to work, but a couple of additional observations.

I didn't find the speed increase I was looking for. The last 'official' ABS print I did had a print speed of 5000, 30% underspeed on outlines, and 60Z underspeed on solid infills. For now I'm going to stay there. I have found some improvement in print quality with the extruder jerk and acceleration settings.

i have settled on an Extruder Jerk setting of 0.33. I have seen a definite correlation between the Extruder Jerk setting and ringing. Impact on print time isn't major.

At one point I was thinking about the relative speeds of printing vs extruder velocities. Extruder will depend on nozzle diameter as far as mm extruded vs mm printed. Extruder acceleration high, 1000 I think. my print acceleration are currently X 1875 and Y 1425. i did a quick calculation and found the print/extruder ratio to be about 10:1 for the .5mm nozzle (20:1 for .35, 40;1 for .25). I tried setting the extruder accel to 1/10 of the print accel. I used 180.

With the high accel setting I was getting some ticks in the XY motion on long arcs that would result in a small bump on surface. With the extruder accel at 180 those ticks are gone, corner bulges are reduced, and print time was reduced - about 12-15 minutes reduction on a 3.5 hour build. The reduction in build time was surprising, but if it results in smoother motion ...

The effects i'm seeing between settings are pretty subtle which make it hard to photograph without turning it into a production, so for now I'll just share the numbers and words.

The last test part I ran was at the same speeds I had been using - 5000 with 60% underspeed on infills and 30% on perimeters. Extruder accel is 180 and extruder Jerk is 0.33. As stated above, X/Y accel are 1875/1425. XY jerk is at 20. With those settings, I think it's the cleanest ABS print I've ever run. Virtually no ringing or corner bulge. I still have a ridge at the start point, but not on the leading edge - only where it stops, so i'll be upping my coast (currently at .33mm).

With these settings i don't yet know where I stand on build times from before I started - I didn't time a 'control' print.

Jackenpacken
Posts: 1
Joined: Mon May 08, 2017 5:14 pm

Re: Extruder Acceleration - Holy Crap!

Post by Jackenpacken » Mon May 08, 2017 5:21 pm

There should be different values for acceleration and deceleration. Most problems comes when you try to slowdown in corners.

Post Reply