The CAD industry is split by history
I understand confusion when I see it. Different segments of the CAD industry are moving in different directions, and it is by no means a given which direction is the correct one. In the end, I believe we are going to see more direct editing (non-history) based tools, but I don’t think FrankenCAD is the obvious choice.
In one corner we have the noisy, but still rather stuffy Siemens/UGS with its Synchronous Technology pom-poms and cheerleaders, claiming that history based modelers are going to give way to non-history based direct modeling, or at least mixed modelers which employ non-history based modeling within a history-based framework. This is an ugly FrankenCAD mix, and as details leak out, I don’t think it gets any prettier.
And then in the other corner we are seeing the budding romance of a non-history based modeler (Spaceclaim) with another non-history based modeler (Rhino), but then one of them goes and adds some history based functionality. Rhino with implicit and explicit history shows that the direct editing crowd understands parametrics has value that they can’t deliver. This is FrankenCAD from the other direction. How does adding random parametric history dependent features to a non-history based model work? This sounds rather like CoCreate software, which is another also-ran in the direct editing miasma. This story of Spaceclaim and Rhino teaming up and Rhino adding history isn’t getting as much press as the other Synchronous story, but I think it is the more important of the two.
Here’s why. Oreos. Yes, oreo cookies.
Spaceclaim on its own is limited. It has cool editing capabilities, but they are limited to analytical geometry – prismatic stuff with lines and arcs, single curvature, developable surfaces. It has some limited capability with general case NURBS shapes, but it is really limited. You can make simple machined parts and sheet metal parts, but more complex castings and plastic parts aren’t really part of the Spaceclaim equation. What Spaceclaim is missing is exactly what Rhino is best at – swoops. And frankly, Rhino, in all of its complex shape surfacing glory, is no match for a solid modeler when it comes to prismatic work. Plastic part engineering in Rhino would be tedious. What Rhino is most missing in its software is exactly what Spaceclaim does. Oh, and they are both direct modelers.
Everybody has been waiting for Dassault or Autodesk or PTC or UGS to swoop down and snag Spaceclaim, but it keeps not happening. What is happening is that these two small, rogue, direct editing CAD companies are starting to work together. It’s like an Oreo cookie. The wafer is good, but I wouldn’t eat it by itself. The filling is also yummy, but by itself isn’t satisfying. Put them together and you’ve got something that stands the test of time. I would like to see these two companies do more than just share files.
We’ve heard a lot of explanations about what is going on with Synchronous Technology, but the message is getting more rather than less confusing. At first it was just direct editing tacked on to parametrics, Spaceclaim bolted onto SolidWorks. Then it became something far more than that that mere mortals could not possibly understand, and self-proclaimed brilliant people could not explain. Now it appears to have many aspects to it, including everything previously mentioned, procedural features, form features, and feature recognition. That’s an awful lot of “features” for a direct editing (non-feature-based) modeling scheme.
Secretly, I’m hoping that it might be condensed to a statement something like “parametric relationships brought down to the final model faces, rather than intermediate features”. So you make geometry somehow, anyhow, and then put parametric relationships directly to the faces. You get the benefits of paremetrics without the overhead of history. This is something I wondered about in an earlier post. I’ve been writing a lot about the direct editing conundrum for about a year, mainly due to the appearance of Spaceclaim on the scene.
The alternative to a simple statement of the software seems to be something that is too complex to trust or use. Siemens is using the term “feature recognition”. Feature recognition is something I’m familiar with from two sources – FeatureWorks, which is one of those 40% solutions, and FeatureCAM, which also recognizes features for machining. In both cases, a lot of manual intervention is required to do real work, and in both cases, it works best on very simple geometry.
Another warning flag on the Synchronous stuff is that when I have mixed modes in SolidWorks (history based parts with direct editing features), you get a lot of confusion about how to change things at the end of the tree. Geometry gets “double jeopardy” because you could change it using either history based or direct editing features. Especially if multiple people do the work, this will add to the confusion. The mixed modeling that Synchronous Technology represents is an ugly frankenCAD, in my opinion. Cool idea, yes, but I think in practice this is going to be a bad idea. People will prefer a clean break.
In all of the times I have heard top SolidWorks users talk about the direct editing tools in SolidWorks, they have done it from the point of view that these tools are cool, and can make some tasks easier or simply possible, but they are also a best practice nightmare in terms of maintaining editability, and that its a sloppy way to work.
To me this suggests that if you’re going to move from history to non-history, it is best done either sparingly within a single CAD file or by moving from one CAD file to another, even moving to a different application. Once you start doing direct modeling on a part, switching back and forth causes a lot of confusion.
History based modeling is not going to disappear, let’s be clear about that. There are too many proven benefits. I’m sorry if sometimes history based modeling is intellectually difficult. I’m sorry about that. How else are you going to get the history based benefits of things like the shell feature, or fillets? History is not history.
What’s happening here is that the CAD industry is again appealing to the CAD bottom feeders, a trend that seems to be gaining momentum in an effort to expand the CAD market into new areas it hasn’t previously occupied. People who can’t hang with the intellectual/training demands of history based modeling are going to go for direct modeling. But I don’t think it’s a given that they are going to flock to the unnatural FrankenCAD combination of history and non-history modeling.
It boils down to this: Machinists and people doing simple concepting work will use direct editing because its a no-brainer (think Sketchup) and less expensive. Engineers responsible for production models will still use history based modeling because it represents more complete control. To the extent that you have both types of people in the same organization, it may make sense for that organization to get one tool that does both, but honestly, direct editing comes at the end and the beginning of the project. At the beginning, their models are used as reference, but the production model is made from scratch. At the end of the project, it doesn’t matter because edits are for mfg, for example adding stock, and won’t be pushed back up to the engineering model.
As an aside, all the references to Synchronous Technology so far have been with respect to solid modeling, not necessarily to surface modeling. Add to that the seeming lack of capabilities around complex shapes, and I for one will not be able to make use of this concept at all. I will be much more likely to be able to use whatever comes of the marriage of Spaceclaim and Rhino.
Spaceclaim + Rhino = solids, surfaces, prismatics and general NURBS, mainly direct editing with some history
Synchronous Technology = solids, prismatics, mainly history with some feature recognition and a lot of unknown details
To me, the Spaceclaim/Rhino FrankenCAD – SpaceRhino – seems like the more attractive combination. A big part of the reason for that is that you can see what that combination is, and what it does. This Siemens thing is still too much of a pig in a poke.
Matt,
Interesting summary of the current trend towards mixed modeling techniques. I think you are dead on regarding the problems it will create.
I’ll be curious to see how the Spaceclaim/Rhino marriage progesses.
Have a great Memorial Day weekend!
I think your comment “People who can’t hang with the intellectual/training demands of history based modeling are going to go for direct modeling.” is a bit harsh. I use both SolidWorks and CoCreate OneSpace. I see the benefits of each and use the one best suited for the job.
For conceptual work, the direct editing approaches win hands down. The downside is that none of the design intent is captured in the model. A BIG problem in the direct editing approach is that the base feature used early in the creation of the model is lost and it is almost impossible to tweak that base feature later on in the design after all the direct editing actions have been performed. Being able to work with that base feature later on in the design process is what makes the history based approaches shine.
I agree though, combining both approaches just adds to the complexity. I think the best approach is more of a feature based modeler less dependent on a “history tree”, retaining as much as possible the design intent.
****
I debated with myself the use of the word “intellectual”. In the end I was too lazy to come up with a better word.
I used HP SolidModeler for a brief time a few years back, and I remember it being a struggle to get some shapes correct. Does the product have an aspect that is in some way history or parametric driven?
Thank you very much for the FUD!
I am amazed how you can comment on smth which you have never seen and have never worked with, except for a few videos. I am sorry I can’t support this with arguments at this time because of the agreement I signed.
I DO THINK that you are too much into SW; you are jealosu and I bet that if SW had come up with ST you would be the most passionate supporter.
Hopefully, the time will tell.
OT: I liked the three “fish” posts.
****
I re-read the article, and I think it’s a fair assessment. Maybe not eloquent, but the ideas are in there.
FUD? “Pig in a poke” is a fair description because it is not available for public scrutiny. I think that’s a very justifiable comment.
Yeah, right now we don’t have much to go on other than what we have heard. I simply can’t take what the talking heads say about it at face value, so I’m trying to read between the lines. Granted, it is only slightly informed speculation.
Its funny that you think I’m too into SW. SW people think I’m insanely critical of SW. I guess I must be doing the right thing if I’m pissing off both sides. I don’t think the questions raised are completely new, because they come up while using SW.
Anyway, yes, eventually time will tell. This is all about opinion, anyway.
Matt-
Thanks for “letting the cat out of the bag”.
Nutter Butters are better than Oreos, way!
SpaceClaim (I prefer SpaceClam), hmmm… their website and blog are dead in the water. Their last postings on both are from Jan. 2008! Are the lights still on in the building? Anyone from Concord, MA walked by and knocked on the door? Are there old flyers and newpapers out front?
FrankenCAD, that’s rich!
What, me worry?
Devon
I am new to MCAD, just start using SolidWorks & Inventor, this could be stupid question,
I want to know; what is the difference between the new UGS history free CAD and Catia Imagine and Shape technology?
****
To be honest, I have no idea about the Catia applications.
Ricky Jordan did a recent blog post on 3dvia Shape. http://www.rickyjordan.com/2008/05/3dvia-shape.html
It’s news. I don’t think it’s not big news, though maybe a subtle landmark in CAD history and evolution. Seems there’s only one under-bridge dweller who’s truly excited about this.
Like any engineer, I’m more interested in how it doesn’t work than how it should work. Curious how it would work on a part like this: (scroll down to picture, download & view .AVI).
Matt – sorry I missed this until today. I took the 4 day weekend option so am just catching up on things this morning. Anyway, there were a few things said at the Siemens PLM Software analyst event that you might want to take a look at:
First, a few video interviews, one I did with Dan Staples, the development lead for Solid Edge: http://siemens.pmhclients.com/index.php/video-interview-dan-staples-on-synchronous-technology/ and another I did with a Solid Edge beta customer that has actually used synchronous Technology: http://siemens.pmhclients.com/index.php/video-interview-cory-goulden-national-steel-car/
Second, Chuck Grindstaff (head of development for all of Siemens PLM Sofwtare) got asked during the wrap up Q&A to very succinctly define synchronous technology and he said: “The key thing we are doing is taking direct editing, feature recognition and adding state of the art constraint management to be the only solution that does all three simultaneously.”
Agree with your ‘pig in a poke’ description somewhat. It is a technology that not many people outside of Siemens have used yet. But with Solid Edge and NX6 shipping in the next few month the poke will be empty soon and I think you’ll see its not such a pig after all.
****
Chris,
Thanks for trying to shed some light on this. The Grindstaff comments are probably the most enlightening, but they leave out some things I expected to hear. I’m assuming that this is all on top of the existing parametrics in Solid Edge/UG? Also, isn’t “constraint management” just a euphamism for something simpler? Could you replace that phrase with “parametrics”? Parametrics on “dumb” geometry would be about the best thing I think could possibly come out of this arrangement. It would have all of the best functions of direct editing tools with the best of parametrics. It would be parametrics without history. Is that essentially what we are looking at with “constraint management”?
According to Chris Kelley’s comment, Siemens developers didn’t mention “parametric” in the list of technologies used (confirming my opinion about hybrid parametric/direct-modeling being an oxymoron). They mentioned “constraints manager”.
Indeed, [ Feature-recognition + constraints-manager = good direct-modeling/editing ] is the equation and it makes perfectly sense. Feature-recognition and constraints-managers are the key ingredients of SpaceClaim, validating my suspect that SpaceClaim’s technology and Synchronous Technology are not that different, despite all Siemens marketing effort to convince us of the opposite.
****
I personally hope “parametrics” is implied by “constraint management”. Feature recognition is exactly the part of this that scares me. To me, it seems that feature recognition means that the software is trying to assign design intent. In the other software I have used that employs feature recognition, it is the kind of intelligence that even if it worked right, it would never read your mind, and so it will be limited to certain types of parts such as simple machined or sheet metal parts. I think that feature recognition will be the weak link here.
The main difference with ST is that it is on top of a traditional parametric model. This is where the FrankenCAD bit comes it. Either technology on its own is wonderful, and mixing even a small amount of emergency direct editing with the parametrics makes for something sloppy, but combining them both full-bore I think will create a mess.
History modeling is very process based. Direct modeling disposes with the process. I don’t think these can exist in equal parts for production modeling.
Franco – the key difference with ST is that the solves for all of the elements you mention (contraints, features and direct editing) happen at the same time. STREET solves them simultaneously rather than in sequence like other systems. When things are solved synchronously construction order doesn’t matter and therefore there is one less thing to think about while designing and one less thing to have to figure out when reusing someone else’s model.
****
Isn’t a part of the part’s tree history based? My idea of it was that there is a part of the tree that is traditional history, and then you can add direct editing features to the part. This is something I would like to understand better because its not clear. I’m thinking that ugs is handling this like SW handles direct editing features within the history framework. Can you explain this a little?
Simultaneous equations can have multiple solutions. This is what gets SolidWorks in trouble with D-cubed, flipping mates and sketch relations.
OK. Seems like some pretty intelligent people out there with some pretty intelligent questions, so I am going to jump in with some answers (facts actually). Bear in mind that my answers are solely for the Solid Edge implementation – someone from NX would have to speak to the NX POV.
One person said –> The main difference with ST is that it is on top of a traditional parametric model. This is where the FrankenCAD bit comes in.
And another said –> Isn’t a part of the part’s tree history based? My idea of it was that there is a part of the tree that is traditional history, and then you can add direct editing features to the part. This is something I would like to understand better because its not clear. I’m thinking that ugs is handling this like SW handles direct editing features within the history framework. Can you explain this a little?
========
Answer –> OK. So this is where a base misconception lies. With the Synchronous Technology implementation in Solid Edge there is no traditional linear history tree. There is a feature COLLECTION where all the features are gathered, but they are all peers – they are not in a linear dependency graph. So there is no regeneration when you edit something you created earlier. It will only compute its local surroundings and things it is officially related to. No blind regen from top of tree to bottom.
Another person said: Parametrics on “dumb” geometry would be about the best thing I think could possibly come out of this arrangement. It would have all of the best functions of direct editing tools with the best of parametrics. It would be parametrics without history. Is that essentially what we are looking at with “constraint management”?
Answer –> Not totally off target. To consider this, let’s talk about what you mean by the terms “parametrics” and “dumb geometry”. From my POV, Parametrics equates strongly to “dimension driven” and from that perspective you are right on track. Solid Edge with Synchronous Technology allows you to put dimensions on the 3D model at any point in the process and between any points of the model – not just within the sketches. And then you can drive the size by changing those dimension values. However, it’s totally localized in that it will move only the minimum required – it won’t do a big honking regeneration like you are used to in traditional systems.
The second point on “dumb body” I think is very interesting. We are used to thinking of bodies without a history tree as “dumb” because they cannot be replayed. But they are certainly not dumb when brought into Solid Edge with Synchronous Technology. Think about it – a body has lots of smarts – some faces are tangent to one another, some are coplanar, some are concentric. A concept called “Live Rules” finds and maintains these relationships in real time, while allowing the user to override what the system found and decide whether he wants to maintain it for a given edit or not. So IMO the word “dumb” and “body” don’t go together anymore since the announcement of Synchronous Technology.
Good dialog guys. Keep it coming.
PS> Back to the original blog, I will point out a significant flaw in the SpaceClaim – Rhino marriage theory. Who would do the drawing creation? Not a strong point for either system…
****
Dan,
Can you state who you are and how you know this info? Thanks a lot for contributing.
Ok, I understand that features are all solved simultaneously, and traditional “history” is gone. This leads to more questions, though. In traditional history modelers, there are some advantages of history, not just disadvantages. How does ST deal with the advantages? Meaning, how would you make history based edits to features like shells and fillets?
I think I get the “Live Rules” too. This sounds like the feature recognition/constraint management. This is the part that scares me most. If the software is making assumptions about things, I want control over that.
The bit about “body parametrics” is great news. I can see a lot of use for that. I think solving all the features at once is a concept that needs to be proved in actual use. Feature recognition is something I could live without, unless it is an order of magnitude better than anything else that exists currently.
Spaceclaim does drawings. Drawings are not a major concern for me personally, and I think they are a trivial commodity with respect to CAD in general – you can go to a number of places if what you really need is drawings.
about Dan Staples – http://siemens.pmhclients.com/index.php/site/video-interview-dan-staples-on-synchronous-technology/
As a user I can say:
You can control the rules in “live rules”.
You design the way you a want, but there still a feature tree. Meaning that all the steps are still recorded and you can go and make edits. It is just organized and functions in a different and more powerful way.
It is based on “Only if I could…” Well, now you can!
There is a lot more but let’s wait until shipping starts.
****
You understand why I put very little credibility in people who just say “it’s way better, beleive me”? If you can’t answer details, why bother posting “pig in a poke”?
I want someone to describe to me how you deal with the types of issues you can handle with history and shells and fillets. You know the rock-paper-scissors game I often describe with shell, draft and fillets? How do you control stuff in ST?
Please, no more “trust me” answers.
Matt says –> Can you state who you are and how you know this info? Thanks a lot for contributing.
Dan says –> I’m Product Development Director for Solid Edge. So I would say I know the details rather intimately.
I am really trying to avoid being too biased and “salesy” here – not the place for that nor my big thing. But I thought I would try to answer your questions since you asked. Let me know if I wear out my welcome, but in the meantime, I appreciate your interest and will do my best to answer things in a (relatively) unbiased manner.
Matt Says –> Ok, I understand that features are all solved simultaneously, and traditional “history” is gone. This leads to more questions, though. In traditional history modelers, there are some advantages of history, not just disadvantages. How does ST deal with the advantages? Meaning, how would you make history based edits to features like shells and fillets?
Dan says –> Very good question. And you are right, existing approaches have brought some really key things to the table – heck, hopefully the industry has been doing at least some things right for the last 20 years. But let’s not confuse history with features. These two have always been synonymous – but they are no longer. The goal with Synchronous Technology is have features and dimensions, but not pay the ordered rebuild price that comes with existing history-based implementations.
As you note, designs are full of shells and rounds (fillets), so those are an integral part of the software design. How do they work? Let’s start with shells. In Synchronous Technology a shell is not a feature somewhere upstream in the tree that gets executed at a specific point in time. Rather, each face of a shell feature understands its partner. So if you move or rotate one of these faces, its partner knows to do likewise to maintain “shellness”. Doesn’t matter whether you move the originating face or the generated face – you get the same result. This is obviously unlike history-based where you have to move the originating (parent) face. Also, as you can imagine in this scheme, there is no rebuild penalty – just a change to the two shell faces only. It’s very cool, but definitely different — we probably need to post a YouTube video on this aspect.
For fillets/rounds, these are special features that know the faces that created them and update automatically when those faces change. Don’t need a history tree for that. So again, it’s a very localized thing – drag a rounded boss and all the fillets drag with and readjust to the geometry they encounter. See the end of this video to see this when the cone feature is dragged (worth watching the whole thing of course): http://www.youtube.com/watch?v=E9LLpG4oSXQ
Matt says –> I think I get the “Live Rules” too. This sounds like the feature recognition/constraint management. This is the part that scares me most. If the software is making assumptions about things, I want control over that.
The bit about “body parametrics” is great news. I can see a lot of use for that. I think solving all the features at once is a concept that needs to be proved in actual use. Feature recognition is something I could live without, unless it is an order of magnitude better than anything else that exists currently.
Dan says –>As you say, you want control over this process and you have it. This I do think you really need to see, rather than explain. There are logical defaults (maintain tangency, symmetry, concentricity and coplanar), but the user can control exactly what he wants. See the youtube video http://www.youtube.com/watch?v=Prt1Jod7BJo for an introduction. The UI is a bit hard to see in this video, but about mid way through you’ll see the user interact with a UI panel on the left (titled Live Rules) where you control this. You can turn on and off what you want recognized, as well as it highlights in bold those items that are currently being found by the system for the edit you are in the middle of doing. Another pretty good YouTube for this is http://www.youtube.com/watch?v=ACneq2GcFLM that uses an imported rather than native Solid Edge part. It shows the UI for a pretty long time during the first minute or so, so you can get more of a feel.
Matt says –> Spaceclaim does drawings. Drawings are not a major concern for me personally, and I think they are a trivial commodity with respect to CAD in general – you can go to a number of places if what you really need is drawings.
Dan says –> OK. Don’t get me started here. All I will say is “Do not take production drawing creation for granted. Make sure any system that you like for modeling can create YOUR production drawings.”
******
Ok, lets worry about drawings later.
Thanks for the explanations, but you are skipping over some detail and some critical stuff I’m digging for.
Back to the shell. You’re assuming that the overall face count stays the same. What happens if you add a new “feature” to your part and you want the new feature shelled out too? This is different from just moving a face.
Also, back to fillets, the order of fillets produces different results. How can you produce all of the same range of results if there is no such thing as history?
This is sounding more like CoCreate than Spaceclaim.
Actually, we are not like CoCreate or SpaceClaim. The only similarity is the purposeful lack of ordered rebuild. But I think those systems stopped far short of what is needed. (See Live Rules, Feature-based, Dimension-driven etc.) We felt neither history-based nor history-free had solved the problems completely and each had it own set of warts. Our goal was to take the best of both these. And I think a lot of people feel we succeeded. I think if you look closer, this will eventually be your POV as well.
On the questions of shell and round. Again, these are problems we felt were poorly addressed in the current generation of history-free systems. But to understand the solution, you have to think differently. OK. Let’s take the example of the shell with faces needing to be added. In history-based it is logical to roll back and add them and then pay the price to roll forward. However, this is not very logical — just how we are all used to doing it. Instead, why not choose the shell command and just identify new faces to shell. The system is smart enough to know the model is already shelled and to add these faces into the shell, adjusting only the local topology.
Same thing with rounds. Reordering is just one way to get a result — and not even that natural a way even. If you have a patch corner and wanted the torus type corner instead, rather than the user figuring out how to order that to get that result, why not have him say “give me the other result” and let the system manage *locally* reordering the blends in that area to achieve the result?
I think in general, as people come to learn about Synchronous Technology, they will start to see that “tree” and “reorder” are not the only way to achieve a goal — just the way we all (me included) have been doing it for so long. We have to rethink these things and as you say in your recent article be the devils advocate and say “why” — why have we been doing this this way for 20 years — is it the best way or is there a better way? Many ideas in Solid Edge with Synchronous Technology started out as “well you can’t do it that way” and people took the contrary position of “what if you could” and then invested to make “what if you could” be the way that it is.
****
Ok, identifying faces to shell at least explains how it works. In some cases I can see that being a benefit and in others a real pain in the butt. I definitely don’t see this as “breakthrough” sort of material. Toggling through the fillet options is also in SolidWorks, and truthfully, I don’t use it. Thanks for the explanations. I think this will all be rather anticlimactic. Because it can’t do anything with general NURBS faces, I don’t have any use for it personally.
This may indeed be a “breakthrough” when seen from the development side, but from the users side, I think it’s just more of the same.