I'm hoping someone can provide a little more detail on how to use the

I’m hoping someone can provide a little more detail on how to use the “Junction_Deviation” setting. I understand it’s essentially the jerk setting on Marlin, however, I’m not sure how the numbers translate. For example, I’m running 10-12% jerk of my max speed on my Cartesian and that setting works nicely. However I feel like the Junction_deviation setting might need to be adjusted on my Kossel pro, i’m just not sure how or what sort of calculations need to be done in order to get the result I’m looking for?

Thanks in advance,

A good default is 0.05. Multiply it by 10, and test. Divide it by 10 and test. Then choose whichever direction you like most, and adjust from there.

@Arthur_Wolf Perfect - Thank you, I’ll give that a shot. :slight_smile:

@Arthur_Wolf weird! it sounds like a “pragmatic” hack (not to say dirty), doesn’t it?

@Jeremie_Francois Well that’s about how people determine acceleration too. There are so many parameters here you can’t be fully deterministic …

@Arthur_Wolf yep, I never saw any good model, there are always a few semi-arbitrary constants. I would bet that the majors also rely on secret and heavily guarded black magic :wink:

@Jeremie_Francois If you had some way to feed Smoothie a “print quality” signal, that’d tell it if a print is more bad, or more good, it’d be able to auto-tune all that. But baring that, it has to be done by a human.

just to understand it, what exactly does the parameter do? I prefer to understand things before changing them.

@Hakan_Evirgen junction_deviation essentially defines how much to slow down when turning, proportional to the angle. Higher values means you’ll take corners faster, lower values means you’ll slow down for corners more.

so that means if my prints are not exact at the round parts, I should lower this value?

@Hakan_Evirgen Yes, also if your machine shakes too much, lower either acceleration or junction_deviation

@Arthur_Wolf I very much appreciate the time you take to explain this to us. Not everyone who develops and markets something, takes the time with their followers/product owners. It’s very refreshing, I for one am thankful.

@Chris_Wilson \o/

100% ack with Chris Wilson

ditto, thanks @Arthur_Wolf for explaining it!Understanding what goes on is the first step to improvement! And 100% agreed with some machine learning there. But we need more closed loops to do so indeed. Sensing/detecting the results properly is something yet to be done. I fear that separating the impact of every and each of the sensor may never be possible.
This is also why my next printer will be ARM32 (ie. why I consider the Arduino as obsolete). We need room to program real time feedback, analysis and eventually control.