[FR] M592 From RepRapFirmware (nonlinear extrusion)

[ArtemKuchin]

Just discovered the problem with extrusion error vs extrusion speed on my own experience.
See discussion
http://forums.reprap.org/read.php?262,802277
Don't have any option other than to print at low speeds at least for external perimeters to get better quality.
RepRap firmware guys did M592 for nonlinear extrusion compensation.

Hope such thing will be implemented in Marlin.

---

http://reprap.org/wiki/G-code

“

Most extruder drives use toothed shafts to grip the filament and drive it through the hot end. As the extrusion speed increases, so does the back pressure from the hot end, and the increased back pressure causes the amount of filament extruded per step taken by the extruder stepper motor to reduce. This may be because at high back pressures, each tooth compresses and skates over the surface of the filament for longer before it manages to bite. See forum post http://forums.reprap.org/read.php?262,802277 and the graph at http://forums.reprap.org/file.php?262,file=100851,filename=graph.JPG for an example.

Nonlinear extrusion is an experimental feature in RepRapFirmware to compensate for this effect. The amount of extrusion requested is multiplied by (1 + MIN(L, Av + Bv^2)) where v is the requested extrusion speed (calculated from the actual speed at which the move will take place) in mm/sec.

Nonlinear extrusion is not applied to extruder-only movements such as retractions and filament loading.

”

[thinkyhead]

Marlin's LIN_ADVANCE feature (recently updated to v1.5) is recommended to deal with back-pressure and regulation of line-widths. http://marlinfw.org/docs/features/lin_advance.html

[ArtemKuchin]

Yes i know, but does it really compensate for speed extrusion error when extruder just does not extruder the needed amount of filament?
I though LIN_ADVANCE is more like coasting and assumes that extruder physically extruder the given amount of filament. But it does not, see graph in the first message in the forum link i gave.

[ArtemKuchin]

For clarification: Here is my test
http://forums.reprap.org/read.php?1,812325,812339#msg-812339
It is just a constant extrusion, not speed ups or speed downs. So, as i understand LIN ADVANCE will not help here at all.

[Sebastianv650]

@ArtemKuchin for me it's not clear what you try to compensate. Is it for slipping hobbed bolt? You table in the link is also confusing for me, at 360mm/min you got 1.13g, at 60mm/min you got only 0.69g? Doesn't make much sense to me, if something is slipping it should be the opposite way round.

If it's slipping, that's something that has to be solved by hardware in my opinion. I never had a problem with slipping drive gears, at every volumetric speed I ever tried the limit was Marlin calculation speed in most cases and only if I realy push it to extreme I'm hitting the melt capabilities of my hotend. In both cases, adjusting esteps/mm will not help.

[AnHardt]

In short.
Some observe under extrusion at high extrusion speeds. The try to calibrate steps/mm for different speeds and get different results even if there are no, visible signs for a grinding bobbed bolt, or audible lost steps. What is causing the effect is not entirely clear and how to properly test even more.
An attempt to compensate for that was made by dc42. He calculates (somehow) steps/mm from e-speed and some constants (http://reprap.org/wiki/G-code#M592:_Configure_nonlinear_extrusion).

Linear advance is another pair of shoes. It deals with a delayed reaction of outflow when extrusion speed changes - not the absolute amount.

[thinkyhead]

Presumably the issue has to do with the expansion of the filament, where at high speeds it might not expand as much. It sounds like something that would be sort-of akin to a long Bowden tube adding more flexibility to the drive.

[AnHardt]

The different steps/mm seem to be measurable at the input side of the extruder.

[thinkyhead]

From what I've read online it is attributed to what amounts to "slipping" (or "missing") the teeth on an extruder that has a typical MK8-style drive wheel. I'm sure it doesn't affect a Bondtech Extruder, where there is constant grip. But if there are gaps between the points where grip can occur, I can see how the filament might "spring back" between those points, especially as pressure builds up.

[ArtemKuchin]

@Sebastianv650 The described effect is observable on both direct drive, geared and belted extruders with hobbed bolts and with toothed wheels, direct or Bowden. No matter how tight the extruder is. All people who actually did the test see then actual extruded length depends on extrusion speed. The physics behind this is not 100% clear. You can read the forum post i provided and links from there. People did the graphs to see the correlation.

About 1.13g - plase, look carefully. 1.13 is mm - it is diameter, weight is 0.54-0.55g
0.69 is mm too, weight is 0.58g. You mixed the columns.

@AnHardt Not some, but everyone i know who did the tests. Maybe others just don't have the tools, time or will. or don't care.
From what i read myself (did not test it personally) the effect is not present when nozzle is removed. And error is measurable on input side as filament length OR on the output side as filament (model) weight.

@thinkyhead I did not read about Bondtech, pretty rare beast and don;t know how it works, but the problem is very much present on E3D titan and on belted extruders. Also, if that was slipping then tightening the extruder should at least do some change, but no, in my test tightening the extruded results exactly in the same numbers. So, there was no slipping. I have direct extrusion extruder (not Bowden) and see the effect very much. BTW: Where did you read that online?

Also, note than the problem is present for at least PLA and ABS filament types. I tested only PLA. Have SBS (softer, more like stiff rubber, 220C printe T) also, will test and see if the graph form is the same.

[thinkyhead]

Could it be that the material extruded at higher speeds is simply denser and less spongy at the microscopic level? And if so, how does this effect manifest in different materials?

Perhaps we can do tests and actually weigh the filament that is extruded down to the milligram. If the weight-per-length is different that would fit the hypothesis.

[ArtemKuchin]

@thinkyhead i will test PLA vs SBS today on non-print extrusion and weight to 10s of milligram. My scales have 0.01 accuracy and resolution.
For PLA i did test for weight here http://forums.reprap.org/read.php?1,812325,812339#msg-812339
already
So, i'll do SBS.

[ArtemKuchin]

Okay, did synthetic tests with SBS and real prints.

L - extrusion length, also, how many times and pause time
T - temperature
F - feed rate
D - extruded diameter
W - weight of extruded filament

L can be just number (400) - which is mm
Or like this 4x100x2 - means 4 times 100mm with 2 second pause

L       | T   | F   | D    | W    | NOTE
    400 | 220 | 60  | 0,66 | 0,93 |  
    400 | 220 | 360 | 1,07 | 0,93 |  
    400 | 220 | 480 | 1,33 | 0,88 |  
    400 | 220 | 540 | 1,43 | 0,87 |  
    400 | 220 | 540 | 1,45 | 0,85 | 3 skips
        |     |     |      |      |  
4x100x2 | 220 | 480 | 1,36 | 0,91 |  
8x50x2  | 220 | 480 | 1,31 | 0,92 |  

SBS is very flexible filament. Has very low friction and good flow.
But as you can see it still have underextrusion problem when speed grows.

But look, if extrude the same amount in 4 to 8 segments with pauses the problem goes away. So, it seems like pressure build up.

Then i though. Well, how does it behave in real print. And i printed a test cube 20x20x10 with 80% infill

layer | F  | Accelartion  | w    | print | 
   -- | -- | --jerk--     | --   | time  | 
  0.2 | 40 | 500/10       | 3,45 | 9:34  | 1484.4mm (3.6cm3)
  0.4 | 40 | 500/10       | 3,45 | 7:49  | 1484.4mm (3.6cm3)
  0.4 | 80 | 500/10       | 3,45 | 4:36  | 1484.4mm (3.6cm3)
  0.4 | 80 | 2000/20      | 3,44 | 4:36  | 1484.4mm (3.6cm3)
  0.4 | 80 | 2000/20      | 3,45 | 4:36  | 1484.4mm (3.6cm3)

As i thought, since there is accelatation/decelation happens pretty often the pressure does not build up.
Then i thought what is i have model with longs walls.
And i printed 100x20x5 mm test with 80% infill too. 0.4mm layer,accel/jerk 2000/20
The test at 40mm/sec weighted 9,06g
At 80mm/sec it became 8.92

I have to notice here that i do not know the correct weight because at 40mm/sec there is a possibility of overextrusion. It it just a relative comparison point.

This is a test for SBS. For PLA, i am sure, the result will be mode dramatic as it does not flow very well and pressure builds up faster and stronger.

The filament feed rate in test prints.
Layer 0.4, width 0.7 -> s=0,28 mm2
Speed 40: 0.28*40=11 mm3
filaments s=2.4
feed rate: 16/2.4=4,6 mm/sec

For speed 80mm/sec: 9.3 mm/sec

PLA at 200C starts to skip at 6mm/sec. SBS at 220C does not skip until 9mm/sec contstant extrusion, but does not skip at real print (too short extrusion). ABS should be somewhere in between.

So, it seems like for models with many corners and not so long lines the problem virtually not present. However, for models with longer walls it does pose a problem.

[thinkyhead]

Hmm, so that's a small difference, but still significant. A 2% reduction in mass at double the speed. I find it mildly amusing that the solution is to feed more into the extruder, thus increasing the pressure even more. But, it is perfectly sensible to mitigate against this effect nevertheless.

I'll see about adding M592 and borrow the formula given on reprap wiki:

(1 + min(L, A * v + B * sq(v))) where v is the requested extrusion speed (calculated from the actual speed at which the move will take place) in mm/sec.