Chamfer Not Obeying Tolerances

Wim, doesn’t seem you you are highjacking this thread. I don’t want to have to revolve every single object that might need a chamfer, either. Yet, I pride myself in not making crap, so if I cannot quickly and reliably measure the result, I will have to avoid the chamfer tool all together.

I am not after the relative tolerance of two curves, if we cannot even trust the first one because it’s been split.

Is it impossible to indicate the resulting diameter of an edge of chamfer?
Or to specify an accurate distance relative from it?

It sounds like you might be better served using a solid modeler that is intended for mechanical modeling, instead of an organic surface modeler that is intended for industrial design and product styling.

Harley Davidson uses a parametric, feature based, solid modeler for the mechanical components of their motorcycles like the frame, engine, transmission, primary drive, etc.
They use Rhino to model the fairings, fuel tank, side covers, fenders, saddle bags, and tour pak.

Have not we traveled afar because I wanted good chamfers?

It depends on how you measure “good:”. Because of the organic shapes in industrial design work, Rhino uses tolerance based fitting to find surface intersections, for Splitting/Trimming, and for Joining. If tools designed and intended for that use don’t produce what you need, then why would you use them?

What are you building? What is the manufacturing process? What tolerances can that build process hold?
Clearly machining parts for nuclear missile or submarines is different than modeling sun glasses, consumer electronics, or soap bottles.

Hmmm… (as in ‘sigh’)
Personally, I would say that Rhino has added way too many drafting features for that statement to hold water. Take out the layouts and you’ll start making a better argument. Also, there is a very fine line between ‘mechanical’ and making a fuel tank. Finally, I would venture to say that you would have to turn away a lot of current Rhino users (@dan comes to mind as only one example) if Rhino policy would be ‘No mechanical design allowed’.

As a side-note to @brenda, I personally wouldn’t dimension a chamfer by setting a radius or diameter on the end of the part, but rather by either stating the size and angle or 2 sizes. Not? [I still want Chamfer to make a simple surface as well…].

I feel that one of Rhino’s strength is that it’s a tool can create objects that can be brought into the real-world.

It’s your company. You can do anything with it you want, but perhaps you might take another look at what your customer base is doing with your product. A surprising number of machinists use Rhino, as reported on threads on Cnczone.com. One of the people I introduced Rhino to, used it for designing bike frames. A friend of mine had looked at for designing high-end speaker enclosures, and now that it’s available for the Mac, it might be irresistible.

I guess I am taking another side of things. When I see things made with Grasshopper, I think, “That’s amazing,” but then I think, “How are they going to build that.”

If you want the answer to be with Rhino, then some of the more mundane aspects of design, the persnickety CAD side of things could also use some attention, in comparison to that which garners brain-share.

One of the things I am making with Rhino for a friend is the roughly the equivalent of riding lawn mower. The tolerances in this machine are about the same as you would find in a car. I feel that I have learned from using Rhino in this capacity, and I wanted to share my experience. Yes, I’ve asked for some changes in Rhino, such as the block manager, this thread, the Drag-Line, and Name Call-Out, one bug, and one UI regression.

What I do know is: while you four professionals are answering threads, I am using your product for a real-world application. It’s up to you to decide whether or not the issue would be better changed, or the mail-carrier questioned for delivering the message.

I used “their product” for 8 years as an independent model/prototype maker (actually what got me to be a reseller), CNC machining all kinds of stuff in engineering plastics and metals. I call that a real-world application.

In that time, I never had any issues with Rhino’s accuracy. Machine accuracy, my own accuracy, yes, but not Rhino’s.

–Mitch

1 Like

Mitch, yes Rhino’s geometry is usually pretty darn good, but I have a problem whereas, I can’t either create or specify/indicate the diameter of an edge of chamfer within .001" even though my tolerances are set stupid-fine.

Writing down dimensions so indicated sizes can be called out seems a little impractical.

The resulting chamfer has 148 control points : \

[*One thousandth of an inch is the thickness of a cellophane wrapper, or half of the diameter of a hair from your head, or likely all that holds the bearings from spinning in your alternator or water pump on a car.]

Stupid-fine is probably a good description for it, there’s absolutely no reason for this:

My absolute and relative tolerances are both set at an overkill setting of: .0000001. My angular tolerances are .001 Degree.

A ten-thousandth of an inch is a tolerance that only a extremely precise surface grinder can achieve, you are asking for 1000x smaller than that…

Your chamfer will have a LOT fewer control points if you set your tolerance to something more reasonable like .0001.

http://wiki.mcneel.com/rhino/faqtolerances

–Mitch

Yes, I understand my tolerances were quite fine, but perhaps the edges of that cylinder shouldn’t have issued hundreds of splits in the first place. I was worried that something would happen–like happened: It have a part that has the appearance of being out of tolerance.

I should think that tolerances, at largest, should be set to 1/10th of the manufacturing process, which .0001 is not. My Mitutoyo mic is accurate to 1/2 of a tenth (-thousandth). Would it have made a world of difference if I made the tolerances coarser. Would it made my result accurate, as…

The problem remains: I still have a edge reading ±.002

Overall, I feel bad for the responses I have received on this thread.

If you know you need perfection for this part, just draw a pair of perfect circles and loft, or revolve the chamfer profile. And then done, and perfect, and move on. Chuck even said in the very first reply that perhaps Rhino should recognize chamfers of circular edges as a special case.

You need to know your tools, and how to evaluate the results, and chamfer as a tool in this case gave an imperfect, although still well within tolerance result. Again, dimensioning the diameter of a non-circle curve is giving you misleading information (you know it’s not an exact circle because the diameter is changing). You seem to be condemning Rhino for one very particular situation and even then the geometry is still fine, and it might very well be improved in the future. Not sure what more you could ask for…

There is another problem: I haven’t condemned Rhino, but people are sure responding like it.

I know it’s better to revolve everything–including in this case, a revolved cylinder with a chamfer to subtract from my object, still I wished the chamfer result or its indication was better. Also, in this case, the process would have been subtracting two objects would have had them in a co-planer situation.

Have you looked at the result of the file I uploaded?

Fair enough, just insistent that tolerances aren’t being observed. I actually think people have been helpful and patient in this thread.

Yes, I looked at the file at one point. And did a test to check the results of different methods. Again, I’m not sure what else you want to be done at the moment. Chuck said, in the very first reply, he would look into special casing this kind of situation.

Gregb, I am sorry, by my take was he didn’t even seem to be sure it would make it into V6 in the second reply. So, I am thinking maybe the others don’t even really believe there is a problem, but there is, unless for instance think that .001" is only “18 times the width of an aluminum atom” .

http://www.rhino3d.com/accuracy

If you reread this thread, and see how it has gone for me.

Okay Brenda, I will try one more time. In the attached example, there is a blue true circle with a diameter of 10 units and a highlighted irregular somewhat circular curve. Is the irregular curve 0.16 out of tolerance as CrvDeviation reports? Or is it 23 out of tolerance, the difference of the high and low diameter values? You have been claiming it would be 23.

The reported diameter of the non-circular curve does give not give any helpful information in determining tolerance.

http://mcneel.myjetbrains.com/youtrack/issue/RH-30821
At least it’s on the list now - you can add your comments, votes, … there.

For the time being, as with all other reported issues, we’ll live with it and apply work-arounds where necessary.

1 Like

another place where this shows up is when sweeping… if you sweep around a circle, you’ll see similar results as the chamfer.

pretty much any time i do something involving arcs, i’ll take the extra steps of revolving those particular sections even though a sweep would be faster.

the revolves are lighter but they’re also dead exactly stupid accurate… feedback will be 8.0000000"
and to me, that’s important… it tells you exactly what you’re expecting it to say and you’re actually expecting it to say something because arc geometry is an easy one for us to know…

if i’m sweeping something like this:

…and were to check the radius at the red dot, it wouldn’t be a big deal if it said 8.123456 or 8.123467… because, for one, i wouldn’t know what it’s supposed to be anyway and 2) it’s such an incredibly small number anyway that it doesn’t matter…

but if we’re talking circles, the “it’s such a small number” reasoning doesn’t really fly because doubt has already been injected into the user…

point being, i agree with what chuck said in his first post… “I do think that we should recognize when a chamfer is a simple shape (cone in this case) and give the exact answer.”

if it’s an arc or circle or line (or anything where the exact answer is obvious to a user), then an exact solution should be given.

1 Like

hey Brenda, i agree with the issue you’ve raised here and would also like these things to be perfect… (especially when we know rhino is capable of such perfection as witnessed with some of its other tools).

that aside, there does seem to be some confusion in the thread about who’s measuring what… using this drawing of two arcs:

…the way you’re saying it is that these two arcs are off by 1 foot… the way others are saying it is that they’re off by 7/8"

again, i don’t think they should be off by anything …but just mentioning this as a means to clear up some (possible) miscommunication in the thread.

Hi Jeff
I would say that the position is off by 7/8" and the curvature radius is off by 1 foot … no ? :wink:
( Then you may be after position or curvature of course … )

No that is not quite accurate. You chose to compare an 8 foot and a 9 foot arc that overlap for a span of 45 degrees. If the overlapping arc spans were reduced to 1 degree the difference between an 8 foot and a 9 foot arc would be less than .001 inch. If the spans overlapped by only .1 degree the deviation would be around a millionth of an inch.

When you use the radius command to take arc readings of a nurbs curve, you are
measuring arcs over a very very tiny spans so your readings will tell you nothing about how much the nurbs curve actually deviates from a true arc. If you want to know the actual deviation, you need to create a true arc and compare using CrvDeviation command.

The problem here could be solved if Rhino had SimplifySrf command like it has a simplifyCrv command. SimplyfySrf would convert nurbs geometry that is withing tolerance of analytic entities. So in this case the chamfer would be converted to a cone.

2 Likes