Chamfer Not Obeying Tolerances

The manufacturing process may be accomplished by series of line segments that are used to approximate an arc. How close you can approximate an arc depends on how many line segments you use.

You may get the dimension call out you want if you duplicate the edge of the chamfer and run simplifyCrv on the curve… You can then dimension the true arc.

why not? read the thread.
Brenda is measuring using the radius tool (or similar… like dimRadius).
others are measuring with crvDeviation.

chuck said the curve is off by 18x the width of an aluminum atom… brenda says no, it’s off by 0.00141 inches…
they’re measuring different things.

(edit- that said… i’m not quite sure how big 18x the width of an atom is in decimal inches… maybe it is .0014 inches and they’re actually measuring the same thing?)

I think that this is the most important conclusion in this post: Mcneel has added this issue to the RH6 development process. This means that hopefully in the future we will have exact solutions for these operations…yehaaa!

That’s 1.77e-07 inch or 0.000000177 inch

-Willem

1 Like

The reason it is inaccurate because you used a large arc span.
If you reduce the arc spans to 1 degree instead of 45 degrees the deviation woyld be less than 1/1000 of an inch. You could use 9 inch arcs to approximate a one inch circle accurately as long as you use small enough arc spans.

As I said the real world manufacturing process often uses straight lines to approximate arcs. One can approximate any arc as accurately as you want with straight lines as long as you are willing to make the line segments small enough.

hmm… yeah, i guess i don’t know how to explain what i wrote in that post…

i’m just pointing out that different people are using different measurements in the thread and it’s leading to confusion…

in my arc example… i didn’t say either one is right or wrong or accurate or inaccurate… so i’m not sure how what i said is inaccurate…

what i said was “Brenda is measuring radius – Others are measuring Deviation”
are you saying that ^ statement is wrong? or that i’m wrong about something else?

i hoped to clear up some confusion but it appears i’ve added to it :meh:

Presenting an inaccurate example does nothing but add to the confusion

what’s innacurate about the example?

(edit-- oh nevermind… over it)

1 Like

I am sorry, but I’ve had a loss happen in my life so I can’t be attentive to this for another day, but wanted to share this:

In upper example was done with the chamfer tool. If you use the dubious diameter tool to read its absolute dimension, you get what you get.

In the lower example, I rotated a subtraction tool, capped it, and subtracted it. Yes, of course the result was better. We all new it would be, but, if you again use the dubious dimension tool, it actually reads pretty good–even though the surface was subtracted after the woodruff key slot was added.

Chamfer 2.3dm (920.0 KB)

bummer

_Diameter or _Radius or _DimRadius aren’t dubious though…

it’s been said a few times in a few different ways already but you use those tools to measure arcs.

the shape you’re measuring using those tools is not a circle… those tools work on circles-- not the shape you’re measuring… in exaggerated form, the shape you’re measuring is something like :

the centerpoints aren’t common so while the dimension given by _Radius isn’t entirely useless in this scenario, it is for the way you’re trying to use it… in this example using your analysis technique, this shape shows a 350unit variance and is sorely to the nth degree inaccurate… (well, your exact technique is to use diameter so all this stuff doubles and we’re now at a 700unit discrepancy)

then add the circle which the shape is supposed to be:

the circle itself is only 96 wide but you’re saying the blue shape is 7x the width of the entire circle out of whack? it can’t be- the shapes are basically intertwined and we can deduce from the rest of the drawing/numbers that the shape is only a couple of units out compared to what it should be… definitely not 700. (1.9 to be exact_er)

if the centerpoints were always shared then you could make a comparison using the Diameter tool… going back to the other arc comparison except now sharing a centerpoint, these two arcs are 1’ away from each other.

if the top edge of a chamfer gave you the green curve when expecting the pink then yes, accuracy is out by a foot.


the chamfer tool isn’t giving correct results… you should be able to put your diameter tool anywhere along the top edge of the chamfer and get proper feedback… the reason you should be able to is because the top edge of a chamfer on a cylinder is a circle and circles are great shapes for using the diameter tool on : )

but it’s not a circle… my point in this post though is to show that the results of the current tool are much closer to a true circle than your analysis suggests… it’s inaccurate but not that inaccurate

2 Likes

pretty much, if the chamfer tool must be used in this scenario, you’d have to do it manually… measure the radius of the cylinder… subtract the amount of inset of the chamfer… edit a DimRadius using the number you calculated…

but if you’re sending the file to a robot to produce… it’s not too concerned with having a radius value to work from and instead will read the model in which case, the results will be much finer than suggested by the shape’s varying radius… more likely than not, the file, even with its inaccuracy, is finer than the machine can produce.

Thank you for all who replied. Most of it is being supportive to my friend, though what I witnessed was terribly sad.

What I uploaded was a actual part of a machine, truncated. Regardless of what the dimension tool reads, it still reads better for the rotated VS the chamfered edge. I realize while it’s perfect because it is can’t be totally round, we still need tools to call out these type of dimensions to communicate our ideas for manufacture. We also need tools to make sure what we make is in good shape, or do we create something that can no longer measured?

I am disappointed in the chamfer tool, and perhaps also the dimensioning tool.

[quote=“Brenda, post:53, topic:21742, full:true”]
What I uploaded was a actual part of a machine, truncated. Regardless of what the dimension tool reads, it still reads better for the rotated VS the chamfered edge.[/quote]

that’s because the rotated edge is an arc… do this:

• ungroup then DupEdge the two curves you’re measuring… move the edges along with the centerpoints to the side for analysis

• run the _What command on the curves… notice the one made by revolving is an arc and the one made with the chamfer tool is a nurbs curve…

i don’t think you’re making the connection as to why it’s not advisable to use the DimRadius tool on the nurbs curve so for now, pretend there is no radial dimension tool in rhino and you have to dimension it in a more manual way… a Radius is “a straight line from the center to the circumference of a circle”… so:

•draw a couple lines from the centerpoint to the curve:

• now dimension those lines (_DimAligned):

…you’ll see the bottom curve, the Arc, is perfect in all places… it’s an arc " a round plane figure whose boundary (the circumference) consists of points equidistant from a fixed point (the center)." …all points along the curve are exactly the same distance from the centerpoint and therefore it’s appropriate to use DimRadius on this curve…

the top curve, the nurbs curve, is an approximation to an arc… but when measuring the ‘radius’ manually as shown above, you’ll be able to see how close of an approximation it is… it’s incredibly close and it’s within your tolerance… in some places, the line is two ten-millionths inch longer and in other places, it’s eight ten-millionths shorter than true radius…

idk, if something in the above doesn’t click with you then i give up :smile:

the funny thing, to me at least, is that rhino is so accurate that it allows you to measure it’s inaccuracies down to near atomic sizes.
i get it that you keep thinking this part is out by the width of a piece of hair but it’s not… it’s way more precise than that… i also get it that you think people are responding here in a sort of manner to defend their almighty rhino (or whatever ; ) ) while ignoring its faults but that’s not the case either… you really are analyzing the situation inaccurately and that’s what people are attempting to point out to you.

V6 will use surfaces of revolution for fillets and chamfers wherever it decides it can. While this change seems simple, and is for the cylinder-plane case that started this thread, it really is a fairly major coding change for the general case. For that reason, I cannot put it in a v5 service release. Here is the result, using the current v6 code, for the original chamfer.Chamfer56.3dm (506.8 KB)

4 Likes

hey chuck, while you’re in there messing around with chamfers and fillets, don’t shy away from sweeping also :wink:

being able to sweep this instead of revolving each radius individually would be a sweet timesaver

sweepVSrevolve.3dm (131.8 KB)


(sweep1 on left… revolve on right)

1 Like

Hi Jeff,

Interesting idea. I’ll pass it on to @lowell , the head sweeper.

Thanks.

2 Likes

@Chuck, Thank you. Excellent work!

I exploded the result, which doesn’t have any more nodes than if it were hand-revolved.

[I understand that it must be tricky, as your code probably has to make a determination that the edge, or the section of the edge is circular within tolerances, such as checking to see if the edge can be described as a 2nd degree curve. In the case of a circular edge, the chamfer edge distance describes: either an external or internal conical section that can either be used to split the hull, which in turn can be used to split the cone. Cull the scraps, flip the surface normals to whichever way the right way is(?), and copy the materials onto the remaining conical edge surface. Ask the mesher to make a mesh. Log it in the undo stack… Phew! ]

Thank you.

~

@jeff-hammond, LOL, you fight me tooth-and-nail on this, and now that things look good–you suggest an accessory request? Tsk-tsk.

not really ‘now that things look good, you try to ride my shirttail…’ (or whatever)…

the first post-- the first sentence of mine in this thread was about sweeping…

Well, I hope we are in agreement that Chuck’s effort produces good results : )

1 Like