Sketch is failing to recalculate correctly when anchored to origin

[chris.midgley]

Hi all - I have been struggling with a problem for a few days now, and have it simplified to just one sketch with several constraints, driven by some parameters.  If the sketch is floating (not anchored to anything), it recalculates just fine.  If I anchor the bottom right corner of the sketch to the world origin, changing a height parameter results in sketch errors (on tangents and height).  If I simple drag the bottom line (which is constrained, so it doesn't move) the sketch jumps around and recalculates just fine.  It's 100% reproducible - see the screencast and attached model for details.

This feels like a bug, but perhaps there is a workaround or something I'm doing wrong?  

Thanks!

Chris

[chris.midgley]

I forgot to click "Insert" on the screencast attachment - sorry about that!  See below...

[davebYYPCU]

What you say is just fine in the screen cast is not fine, you should not be able to drag a fully constrained line to make it recalculate.  I have not worked out all your formulas, but the air gap on the vertical to the peg, is flipping.

I think you can make it recalculate by dragging - weird - because under error conditions it can't be fully constrained.

Might help....

[chris.midgley]

Thanks for looking at it.  Sorry, I really misspoke.  I didn't mean the entire sketch was constrained, but that line and the other nearly lines were constrained such that they don't move when dragged - as best I know, only parameter changes (or direct editing of some distance) allows any change in position.  Did you find otherwise?

I also didn't understand your comment

...the air gap on the vertical to the peg, is flipping.

What do you mean by "flipping", and under what condition?    For the line, do you mean the vertical construction line that goes through the center of the concentric circles?  

I do have other occasional problems, where I find a few lines in sketches "flipping" (on either vertical or horizontal axis, when parameters change).  I've never been able to figure out how to stop that - perhaps these are interrelated and you have some hint or trick to understand / fix???

Thanks so much for your time.

Chris

[davebYYPCU]

I rotate the screen 90 to get better detail into the photo.

Ok, In this case, Fusion does not know that the short line has to be above the long line.  So 50 - 50 it gets it wrong.

Dimensions from the origin will stop the flipping

Short line will be (HingeGap + RotationalEdgeGap) from the origin. 

(I thought I saw this flip in the early part of the movie, not sure now.)

Might help....

[chris.midgley · Accepted solution]

Wow - I can't believe how much effort you are putting into this.  Thank you so much.

As for that long line along the bottom (in your rotated image), the second line from the bottom isn't used in anything interesting and can be deleted without changing the problem.  Also those lines vary in height from the bottom substantially based on the parameters, so even if you could constrain them (which I can't seem to do as they would become over-constrained) it wouldn't work for my model.  Perhaps I'm not understanding what constraint you are recommending?  If you found a constraint that works, if it's not asking too much perhaps you could do a quick screen grab?

Having said that, I may have found a solution for now.  I started studying more on the constraints, and noticed that two of my lines that are constrained in angle via "PrintableOverhangAngle" (45 degrees in your version) are receiving their constraints (tangent, coincidents and angle) all from mirroring another line.  In your picture, it is the two top-most angled construction lines flying off to the left and right.  I removed the mirror, and instead directly added the tangent and angle (in theory, exactly the same as the mirror) and now the error has gone away (at least for now).

This feels like a bug in Fusion to me, and not a "flipping" problem per-se.  But you know far more than I...

Lastly - if you do have any hints on how to detect flipping, and hints (or links to videos/pages) on how to fix it, I'm all ears!  I constantly struggle with that, as I often make large / complex highly parameter driven models to dynamically scale and flipping is my constant nightmare!

I really appreciate your help!

Chris

[davebYYPCU]

I am first to admit I have not cottoned onto all your formula outcomes, 

what I was speaking to is this top dimension.

Not helping is that different behaviours happen when constrained to the origin.  Based from the Origin should be paramount, because driving an unconstrained sketchs leads to this sort of flipping and unpredictable results.

when changing chain height to and above 30mm your outer circle flips over the construction line, and I was thinking the 0.15mm dimension was preventing an even tighter movement.  Is 30mm even a legit test?

Flipping happens when 2 or more legitimate outcomes are available.  Fusion then takes a random decision, the trick is to test the sketching for this type of problem, as you build the sketch.  Most stable is to set the variables from a stable position like the Origin.

Might help...