[BUG] LIN_ADVANCE produces horrible E jerking on first layer (depending on object)

[AndKe]

Bug Description

Some files, produce violent E travel, causing it to skip steps, making awful sound, and not print correctly. (underextrude)
in this case: 0.5mm thin wall test file is affected badly.

Configuration Files

gcode:
0.5mm-thin-wall.zip
Configuration.zip

Steps to Reproduce

Print Gcode as-is ,
M900 K0.6 is set, so LIN_ADVANCE will be active, and produce horrific result on the first layer.
then change gcode to
M900 K0
observe normal print of all layers.

Additional Information

[swilkens]

There are some known issues with LA specifically in combination with some SKR boards.

Is the sound only occurring on the first layer, or on all layers?

In your Configuration_adv.h file, please look for //#define LA_DEBUG and enable it, report back here if you see any significant debug output over the serial interface.

Possibly related;
#10272

[AndKe]

I have only really observed it on the first layer. (I have no explanation for that)
Also: I am fairly sure I've heard this before on the Atmega2560 board I upgraded from. Last time I thought it was the hobbed gear that skipped on the filament. (I almost never see a print job start over than via octopi , this is detected because of the fact that I replaced the 2560 board with SKR1.3 and were nearby for all basic tests.)

[XDA-Bam]

@AndKe I see your DEFAULT_EJERK is set to 5.0. In my experience, this is way too low for LA to work correctly. Maybe try it with 15.0. You could also try using CLASSIC_JERK in stead of JD. The combination of LA and JD can cause some odd issues.

It is strange that it only happens on the first layer. But I would test these two changes first, before investigating any further.

[AndKe]

@XDA-Bam Thanks - I did not tweak DEFAULT_EJERK - maybe the default should be very different then?

[XDA-Bam]

As a follow-up: The DEFAULT_EJERK is fine, because LIN_ADVANCE is not on by default.

Did you try any of the suggested settings? Did they improve anything?

[AndKe]

No, have not tried a different DEFAULT_EJERK as it's unclear what would be a decent setting.
I did not try CLASSIC_JERK eighter.

From what I see in this thread, the default config files should have a clear warning against LIN_ADVANCE & DEFAULT_EJERK (if it is known to cause trouble) and DEFAULT_EJERK should have a good default vealue.

[XDA-Bam]

No, have not tried a different DEFAULT_EJERK as it's unclear what would be a decent setting.

See above: "Maybe try it with 15.0". I ran up to 20.0 on my bowden setup (standard Creality extruder), but started getting some skipped steps there, so I turned it down again.

I did not try CLASSIC_JERK eighter.

Well...

From what I see in this thread, the default config files should have a clear warning against LIN_ADVANCE & DEFAULT_EJERK (if it is known to cause trouble) and DEFAULT_EJERK should have a good default vealue.

I agree that some comment/warning in the LIN_ADVANCE section of the config would be appropriate. But as stated above, the defaults for DEFAULT_EJERK are fine, because the default for LIN_ADVANCE is, that it's deactivated.

[AndKe]

Please implement XDA-Bam 's conclusion.

[thisiskeithb]

We're accepting Pull Requests!

[Vertabreak]

would be happy to submit a comment change what exactly should it say.
something to the effect of linear advance may require adjustment of default_ejerk?
perhaps a if condition that switch's the value when linear advance is enabled?
come up with the most ideal way to handle it and ill put in the request should others be unable to them self.

[XDA-Bam]

I'd suggest a warning in the style of

#if ENABLED(LIN_ADVANCE) && DEFAULT_EJERK<10
    // Warn user about possible problems
#endif

[Vertabreak]

sounds like more of sanity check something like the following.

#if ENABLED(LIN_ADVANCE) && DEFAULT_EJERK<10
    
#error DEFAULT_EJERK less then 10 results in issues when LIN_ADVANCE is enabled.

#endif

is this the consensus and is it universally true,

[thisiskeithb]

is this the consensus and is it universally true,

It's not universally true, so I'd change it to a #WARNING instead of an #ERROR and inform users that they may experience issues.

Note: There will probably be configs from the Configurations repo with matching settings that will cause builds to fail with that #ERROR as well.

[Vertabreak]

so we are leaning more to something like.

#if ENABLED(LIN_ADVANCE) && DEFAULT_EJERK<10
#WARNING DEFAULT_EJERK less then 10 with LIN_ADVANCE enabled may cause issues.
#endif

are we thinking sanity check or should this be added to the lin_advance config section.
when we have a consensus on how it should be handled i can then submit it see what thinkyhead has to say about it.
thisistheithb would you be in support of this or no and if not why given you could have done a pull request yet you seem to have abstained. i do hope others would be candid in there words should i be over eager to be of assistance.