Welcome to the Onshape forum! Ask questions and join in the discussions about everything Onshape.
First time visiting? Here are some places to start:- Looking for a certain topic? Check out the categories filter or use Search (upper right).
- Need support? Ask a question to our Community Support category.
- Please submit support tickets for bugs but you can request improvements in the Product Feedback category.
- Be respectful, on topic and if you see a problem, Flag it.
If you would like to contact our Community Manager personally, feel free to send a private message or an email.
Can this component be created using OnShape?
me_again
Member Posts: 18 ✭
I've uploaded a relatively simply component https://cad.onshape.com/documents/b35681d2df1a1c777f2d53b9/w/9ed7662fa935730066b0247a/e/fc5ef8edfc6d008377db3785, and I'd like to know how (preferably watch a video?) of this being re-created here in OnShape.
Anyone up for the challenge?
Anyone up for the challenge?
0
Answers
Sketch the main body profile and use revolve
create a helix offset from the top face
use the "Use" which is the U button on your keyboard and select the outside edge
Use that sketch to create a helix as shown
create a plane that starts on the helix point
sketch on that plane a rectangle for the spiral cutout
now sweep/remove the rectangle using the helix as the path
I have to say the accuracy with which you recreated it is surprising (0.749 not 0.75
However, I think it kind of reinforced a conclusion that I've rapidly been approaching for a while now;namely that I'm never going to find this place useful for the creation of components, because the whole sketcher philosophy of drawing something that roughly approximates what you want to draw, and then "constraining" it into submission by adding dimensions that tell you the sizes that you have no interest in, just so that you can change them to be the sizes you wanted in the first place, is just so laboriously tedious.
Especially as the only 'benefit' of this philosophy seems to be the avoidance of using a sketch plane grid which would allow you to quickly and accurately draw what you need without the secondary and tertiary steps.
I just re-created the component in my CAD program and it took 55 clicks and 37 keystrokes from beginning to completion.
I attempted (several times) to count the clicks/keystrokes in your video -- whilst discounting those that appeared to be there only for the purpose of demonstrating the progress (ie. engaging section mode and dragging it around a few times) -- but (given the vagaries of video quality) found it impossible to come up with an accurate count.
My best informed guestimate is some thing like 10x as many clicks and close to 20x as many keystrokes.
My best guess is that the philosophy arises from the desire to support fat-finger painters on touch devices to construct back-of-an-envelope rough approximations using (inevitably) broad brushstrokes, and then constrain them to some semblance of engineering using stepwise numerical refinement.
But would it kill OnShape to provide -- those of us that are using an input mechanism that allows for a degree of positional accuracy -- to opt to have a grid displayed on the selected sketch plane, so that we could avoid the need to define and then redefine our requirements?
Please do not read this as any kind of critique of you personally; I am truly grateful for you having confirmed my rising belief that it is not just my lack of familiarity with this site's mechanisms that prevents me from producing components in an efficient manner here.
I'd love to be convinced that there is some down-stream benefits to be had from the laborious. finger-paint-an-approximation-and-then-constrain-that-to-your-requirements-by-stepwise-numerical-refinements; but I'm just not seeing it.
Cheers, Buk.
Onshape works exactly like every other parametric modelling system - the idea being that during concept design, you have a good idea of the shape you want, but no idea what the final sizes will be (concept is 0.749 but for detailing purposes you would round it up). Adding dimensions and constraints enables you to define design intent and manufacturing detail and more importantly, change it quickly if you need to as the design gets more refined. A grid and coordinate input assumes you know the end goal upfront, everything is in round numbers and nothing will ever change.
If you want to change this behaviour, please create an improvement request in the category on the right, detailing your preferred method of working and if it gets enough votes we will add it to the product.
Cheers, Neil.
There are some fundamental differences here between these two methods of sketching.
You appear to come from a 2D autocad style world. Where every line is just a collection of points in a 2D plane.
Sure it is easy to just draw a line that is .749mm long by just clicking a point, move the mouse to the right then type .749<enter>
but that is what we call dumb geometry.
if you want to change that line to .750mm you will need to find the stretch command, select the line (or group of lines) and move the mouse to the right type .001mm
when we draw in a parametric sketch, we utilize constraints and dimensions to do this automatically by just double clicking one of our dimensions and enter the final value.
It's not about the journey it's about the destination.
This style of sketching is for making faster and more accurate modifications later on. Or to have other geometry in a totally different sketch drive the result of another sketch with potentially no user input required.
When you count the clicks for the first draft, sure you may come out slightly ahead if you know your final product like you're having god himself whisper the length of each line into your ear as you go. But we live in reality and know every line has the potential to be modified. With this system, we can spend more time up front to create an intelligent sketch that can me modified with little to no effort. And time is more valuable at the end of the project, not the beginning.
Lets give an example:
Say you have a rectangle with a circle in the middle. You know the circle should always be 1" from the top of the box to the center of the circle. and the circle should always be centered in the box.
So you draw your box and circle in position. Now someone asks you to make your box 1" taller and 2" wider.
you have to use stretch and manually change the box, then manually position the circle by half that distance. Or change each side wall by half the distance.
Now lets to the same thing with a parametric model:
notice how many more clicks you need to edit the above version versus this version?
in fact you can remove the dimensions and just grab a corner and see how things move instantly.
So in the end the parametric way is leaps and bounds faster and more efficient in the long run.
try and edit the 2D as fast as i did in the second animation.
- I have no experience of autocad; 2D or 3D, nor any other 2D CAD.
- FreeCAD (in all its hideousness) is a full "parametric" modeller; and it's Sketcher provides an (optional) sketch plane grid with snapping and auto-constraints. I can easily and quickly draw things to the right size directly -- without the need for OnShapes secondary and tertiary definition steps --; and it can infer a whole bunch of constraints -- which lines are parallel; which corners right-angles; which
circles are concentric etc. etc. -- automatically. And everything can be easily changed later if required.
- Your hole example.
I realise that was just a quick example in an attempt to make a point, but the whole idea on needing to constrain a hole relative to one edge of a rectangle is a complete anathema to me; moving the hole whether when sketching it, or after extruding the component is so simple: 2 clicks and drag. Hell, filling it in (2 clicks) and creating a new one at the required position (7 clicks and a drag) is so quick and simple that I fail to feel the need to try and constrain its position at some early stage in the designs evolution. But this thread wasn't about a Direct vs Parametric modelling discussion; just the absence of a grid when sketching. The former discussion will rage and rage, and depending upon which you encounter first, or which fits best with what you do, or your thought processes, people will come down on one side of that or the other; and both are equally valid ways of achieving the final result. The absence of/refusal to provide a snap grid when sketching seems to serve only to require the user to explicitly define dimensions and constraints that could easily and accurately be inferred by the software. And in my book, software is meant to save work not create it.My point here is that the provision of a grid on the sketch plane does not in anyway compromise the 'parametric' nature of the modelling; it simply obviates the need to construct something random before you can start applying your constraints; you simply draw the thing you actually want and the constraints fall out of that process automatically.
In my preferred CAD, adjusting the size of the rectangle takes 4 clicks and 3 drag operations. (I wish I had a screen video set up here, but I don't)
Click to select Move tool.
Click bottom line and drag it down by 1".
Click the left line and drag by 1"
Click the right line and drag by 1"
Oh. I suppose I should add hitting esc to exit the move tool.
More importantly, once the rectangle with hole has been pulled into 3D, changing the size of the rectangle without moving the hole's relative position to the top edge requires the exact same number of operations, using the exact same tool.
Click to select Move tool.
Click bottom face and drag it down by 1".
Click the left face and drag by 1"
Click the right face and drag by 1"
Hit esc.
I don't use (or like) freeCAD; it was just an example of another parametric modeller that does allow you to use a sketch plane grid.
Again... the vast majority of my dimensions could not be easily and accurately inferred by the software. Constraints, yes. Dimensions, no.
Additionally, if I am creating a parametric model, I'll need to manually replace the inferred dimension with my variable reference or equation anyway.
I'm sure there are use cases where the opposite it true, but count me in favor of the current setup.
In my preferred CAD (SpaceClaim), the grid spacing adapts to the zoom level. Working in close, its 1mm spacing, zoom out a bit and it becomes 2mm, a bit more and 5mm; etc.
I don't understand this argument. As is, you draw something to a bunch of random dimensions and then manually resize them to the required dimensions. With a grid, if your initial dimensions weren't exactly grid sized, then you'd need to manually tweak them. No change.
But your initial sketch would (could easily) be much closer to your final requirements; and any dimensions that did coincide with the grid spacing -- and in my work most do; I'll generally choose 10mm over 10.123456789mm -- then you've saved the need to redefine those dimensions.
In any case, if I were advocating change here (and I wasn't, just noting why I prefer to produce my components elsewhere), then I would make it optional. It would not affect those that chose not to use it in any way at all.
When I posted the challenge it was because I was looking to find the 'killer application' that would cause me to bite the bullet and my overcome my considerable reservations and force myself to get used to the OnShape way of working. So far, I'm just not seeing it.
Personally I can't turn this function off quick enough.
The grid invariably snaps to something that is not what I want and have to go back and disconnect the references.
I can think of a couple of ways to generate some grid functionality with in OS if I felt the need but.....
But the only reliable way I have found to get (some) of my components into SimScale is via OS -- OS's ACIS conversion is better than SimScale's.
I also intend to use OS to assemble my components before transferring them to SimScale to run FEA.
And I am much enamoured with OS Assembly facilities.
Given all the parts of OS that I really do like, I'd really like to like its (obviously powerful) sketcher, but every time I try to use it, I give up before I get started; because the impedance match between it an me, makes converting what's in my head into a model so laborious.
(For me) it feels very similar to the difference between being used to writing software in a good high-level language and then being forced to go back to using assembly language.
PhD, Mechanical Engineering, Stanford University
For those that chose not to use it, it would have zero effect at all.
For those that chose to use it, it would mean that at the 30second point in Cris' movie above, instead of having this:
He could have constructed this:
And then skipped the next 105 seconds and gone straight to the revolve, like this: http://vimple.co/da89e3688e004daba120f2ce4f06a90d
(Sorry that's not embedded, but I couldn't work out how to do that here.)
Providing a grid in the sketcher doesn't convert OnShape to a direct modeller at all, it just makes it more efficient to use.
HWM-Water Ltd
With that in mind, here's a link to video of me creating that part (with a small error I couldn't be bothered to correct and re-record) in SC:
http://vimple.co/24985586e24d4edda83402a71d3f0948
It also has several unnecessary intermediate steps where I drag the component so that you can see what's happening at each stage.
It could come in handy for electrical drawings where you just want your connecting lines to be straight/in-line with each other. but the actual dimension of the line is a non issue.
I would vote it up if you make it. But arguing with everyone that our way takes a couple more clicks is a non issue. It's all muscle memory in the end. I show extra steps in videos so people can "hopefully" understand the process. but in reality it use a lot more keyboard shortcuts and window selection tricks that I bet I could draw it twice as fast as your grid method.
I was being careful to make the dimensions and location relative to the origin exactly as your original model.
with a couple of mistakes here and there.
A grid wouldn't help much anyways, because you have a lot of decimal values like .002, .749, .5 So really the grid would hurt more than help in this case.
When the bottoms of the radial slots coincided exactly with the edges of the helical slot, the upload translation resulted in a badly formed solid here. (The same problem occurred at SimScale and CadExchanger and couple of other PC-based CAD programs.) I've found through experience that moving one of the coincident edges by a tiny amount seems to prevent the problem most of the time, but not always. This is one (of many) reasons I would like to be able to construct components directly within OS, but as is that is a non starter for me.
If the grid was available, you could ignore it, I could use it and we could both be happy. As is, for me it like trying to use a camera without a viewfinder, clumsy and inept; but I see from the preponderance of negative reaction to the mention of the idea that I am on a hiding to nothing.
Another of the reasons I'd like to get comfortable with OS, is that I can see the potential of the parametric model. For example, I have to redesign this very component. Running FEA on it shows me that the 3 concentric cantilever beams are too flexible; by an order of magnitude under the expected loads. To correct this, I need to either shorten the beams, increase their cross-sectional area, or a combination of the two.
As this is such a simple component, it wasn't hard to re-draw it from scratch, using a wider pitch and less rotations for the helical slot, and come up with this:
But, that might prove to be too stiff, or still too flexible, and there are a near infinite number of variations of those two parameters I could try. In OS, I assume that I could just tweak those two parameters in the helix and the rest would take care of itself.
I get the benefits of parametric, I just can't get past the sketcher.
I'll shut up now as its obvious it ain't gonna fly here (I wonder how many other potential customers, used to using a grid, have taken one look and walked away?)
Thank you for taking the time to produce the videos.
It also seems to make more sense now why he's in the stance that's the initial sketch is the most important.
I can respect that.
Because that is the way it is done in some systems like AutoCAD for example. It works really well for those who are used to it. but for others who have gone to the parametric side, we'll never go back.
I'm sorry @me_again the fact you consider a re-draw of a part from scratch was enough to make a modest designer cringe. But in the end whatever works for you, is the correct way.
I still will vote for your grid, Onshape is all about making a great system. There are good reasons for it. But that doesn't mean everyone else is wrong
Until then, you can create a grid by drawing a single point on the orign, then making a 2D pattern in a configiration.
From ther you can import that sketch and set a pre-defined grid for a part studio. Then you can create a new sketch that uses the grid sketch as a plane.
That will effectively be the same thing. It won't be a one click thing you turn on... but it will be fast enough to get you in familiar territory.
If you look at my video (assuming you're not blocked from it) you'll see that I created it in a quite different way. I tend to model components in (roughly) the way and order that I envisage them being machined.
For this component, that would mean starting with a cylinder, pulling a hole through it for the bore, reducing the diameter for the thin end, cutting the helical slot first (because cutting it after the radial slots were cut might result in distortion due to flexing, or even the cutting tool binding), and finally cutting the radial slots front and back.
However, this component is never intended to be manufactured as it is a gross simplification of a two piece design that will end up looking much more like this early attempt which was too ridged and had a tendency to twist rather than translate under load. :
The simplification (reducing it to a single part without the threads) is because it vastly reduces the size of FEA mesh required, and thus the run times (costs) in order to solve.
The 3 layer nested beam design is intended to make the flexure component more flexible (which it has in spades, but too much) and also to allow the axle (the narrow bit sticking out the front of the blue component above to translate (offset radially whilst remaining axially parallel) which it does nicely.
However, the need for the helical slot denies the possibility of using a thread on the outer circumference of the flexure to join the two parts, so my current thinking looks more like this (don't look too closely at this as I knocked it up just now and I need to loose at least 2 and preferably 3 mm on the axial length of the assembly.):
By moving the flexure into the axle mount, and separating the bearing surfaces of the axle and integral thrust washer into a different component, it facilitates applying different heat treatment strategies to the two. Or even using different materials.
The purpose of my OP was to see if my perception of the OS sketcher as clumsy and laborious was down to the way I was using it; by seeing how someone familiar with this place would do it. Having watched both the original and the later 'fast' version, it still seems that way comparatively.
John. The fact that you see 3 minutes work (in OS; more like 45 seconds in SC) as cringe worthy makes me cringe. This harks back to my software development analogy from above; it suggests that parametric designers see hardware development in the same way as practitioners of the waterfall method of software design saw software development. Which I guess makes direct modelling more analogous to the modern rapid application development practices like
XP Agile & Scrum.
This component is a throw-away, whose only purpose is to allow to run FEA analysis. It is one of at least 8 possibilities that might allow the required flexibility in some dimensions, and stiffness in others, to be achieved within the required physical space. None of those 8 designs could be parametrically derived from any of the others. (I thought about posting a image of the 8 basic designs to prove that; but I suspect it would be seen as taking this thread even further "awry".)
However, each of those 8 designs could be varied parametrically -- beam/web thickness; slot width etc. -- which is why I feel drawn to parametric for the purpose of my FEA analysis of this component.
Thank you for your patronage; but I have never suggested anyone was "wrong"; only that there are multiple ways of achieving any goal and (for me, and I suspect many others who've been and gone), the addition of an optional grid would permits alternate ways of working for us, without compromising the working practices of those that chose not to use it.
I suspect that any such ad-hoc solution to the provision of a grid would be less than usable. Ie. would it adapt the grid spacing to zoom level? Would it provide snap? If so, would that snap be defeatable on click-by-click basis? Would snap to local features override the grid snap? etc. etc.
I guess that is more than enough from me; so -- barring further questions I feel obliged to answer -- I'll go away.