Mini-challenge
-
-
Unfortunately - the lines are not parallel:
-
@thomthom said:
Unfortunately - the lines are not parallel:
[attachment=0:1xhfujx7]<!-- ia0 -->Accuracy.png<!-- ia0 -->[/attachment:1xhfujx7]bah!
I was hoping gilles nailed it -
@thomthom said:
Unfortunately - the lines are not parallel:
What's missing? I redid it from scratch. I would suggest drawing a guide-line inferenced perpendicular to the new diagonal lines. (I group the newly drawn diagonals so I can rotate exactly to the perpendicular.)
-
My solution - many posts back - using my True-Tangents script's 'True Intersections' - finds the correct point to be used...
It also uses "trigonometry" to calculate the values for the point - like the DC...
It's as accurate as you can get.My subsequent "plugin-free" approach does it by say three rotate/scale combos until the accuracy is so tiny it could be ignored...
-
I'm also still adamant that the previous approach I showed will work fine -- the only sticky thing about this approach is you may have to manually enter the rotation as the rotate tool will not allow for fine enough control of placing the construction point. The width of the beam stays absolutely perfect -- and the only precision issue is the height of the beam before rotation (as set by the construction point).
Bonus here -- no maths
Best,
Jason. -
@tig said:
My solution - many posts back - using my True-Tangents script's 'True Intersections' - finds the correct point to be used...
It also uses "trigonometry" to calculate the values for the point - like the DC...
It's as accurate as you can get.Nice, yes, that method gets the intersection point correctly as not relying on the segmented SU cricle for the intersection. Thanks for pointing out your tangents tool. I still will use ACAD when it comes to using arcs and circles.
-
@jason_maranto said:
I'm also still adamant that the previous approach I showed will work fine --
but fine is not exact... that's the whole point isn't it.
-
Yes, but I am getting down into 1000ths of a degree on that rotation -- I'm not sure how much more precise it can be without being fully automated... we humans tend to fudge it, but computers can go much more precise in practice -- regardless, you will never build something with wood that will be more precise than what we are doing here. So I think any of these are decent enough for most users... after all this isn't meant to be an engineering program.
Let me put it to you another way, when I do a solid operation (unite/outer shell) there is no gap or extra faces -- it's perfect -- and I think that is about as good of a precision as you are going to get from SketchUp.
Here's the file check for yourself: challenge_altfinal.skp
How I got there:
Best,
Jason. -
@jason_maranto said:
Yes, but I am getting down into 1000ths of a degree on that rotation --
that's fine and as i said earlier, there are plenty of ways to get it 'close enough' for construction purposes..
the challenge though, at least for this thread, is to get it 100% perfect..
and if you give up at 'close enough' then you'll quit exploring ways to make it precise
so while a 100% solution to this challenge isn't totally necessary for me to get a dimension for a cut-list, a 100% solution is necessary to relax the
in some people..
-
Jason may have a point about the precision. The idea is that in doing work that requires this precision (such as survey), where close can give results further out that are way off-- would you EVER be using SU anyway?
Jason you mention entering the rotation manually. Where do you get the angle from? I guess, on the other hand, I think a failsafe technique was sought rather than a function of someone's eyeball ability and time available (which we seem to have ample between us...
)
-
Problem is that tiny fractions of inaccuracy can very easily cascade into the rest of the model where you start getting issues. Ever so often when I import DWG files I end up with lots of issues where things are nearly square, nearly perpendicular etc. Initially SU treats them as "good enough" bug eventually this inaccuracy creeps over to the point where SU doesn't treat them as square and parallel. In worst scenarios I get models where splitting faces creates duplicate overlapping faces. That's why I don't like settling for "good enough"...
-
Jason, I think the issue is that you are manually placing an inference where it doesn't exist in SU. Like thomthom says, that can lead to inaccuracy down the road.
-
I got it as close the the guideline as I can (with the tolerance of the rotate tool maxed out) and then adjust using the VCB to fine tune (if needed)... I'm not very experienced with advanced math (calc and trig) so I try to keep it simple.
I'm sure there is some math that could explain why this works and how to setup the construction point absolutely perfectly without rotating it into place -- but I'm far to math illiterate to figure that out.
But like I say, all the dimensions in the file I just posted are completely accurate as far as SketchUp can measure -- there's no flaw that SketchUp can find (it doesn't mean one doesn't exist).
Also for clarity, this is just my version of the same technique somebody else posted here earlier -- It's not my solution, just one that seems to work really well.
Best,
Jason. -
Well, I meant was that there are lots of ways to approximate the result, but only a few ways to use internal SU inferencing to get there. Oh well, no use arguing over a thousandths of a degree is there...
edit: sure there is!
-
Absolutely -- I don't disagree, I'm waiting for somebody to give the math to make it predictable/easy... also really, open the file, there's no error you can find in SketchUp.
We are talking rhetorical "issues" at this point because I've posted a file with no errors.
Best,
Jason. -
I'd just negative invert copy and connect the top and bottom middles... but I'm sure there is a better way to do it.
I like puzzles
Best,
Jason. -
@jason_maranto said:
Absolutely -- I don't disagree, I'm waiting for somebody to give the math to make it predictable/easy... also really, open the file, there's no error you can find in SketchUp.
We are talking rhetorical "issues" at this point because I've posted a file with no errors.
Best,
Jason.close.. no cigar
re: the math.. i've shown one approach in the DC.. there are about 50 ways i can think of doing the math but they all involve trig to find the angle of the hypotenuse then subtracting the angle of the board's hypotenuse to arrive at the final rotation angle..
-
I get 10 even -- I think you may have mis-measured.
I'm not talking about the angle -- you don't need any math for that at all, what you need to know is how long the board should be.
Best,
Jason. -
@unknownuser said:
@jason_maranto said:
Yes, but I am getting down into 1000ths of a degree on that rotation --
a 100% solution is necessary to relax the
in some people..
an example of a much easier challenge would be this:
there's a pure geometrical approach to finding the solution.. and finding these can actually be geeky fun instead of 'i need to build this' (or whatever)..
so with the challenge i've presented, sketchup has a speed bump in that we need to find this without using the rotate tool in the way we think we should be able to.. (a similar situation would occur with the example in this post if we didn't have midpoint snaps⦠we'd have to find another way and in that particular scenario, there is another way to do it without using a mid snap)
Advertisement