Should CAD software anticipate your needs or do what you tell it to do?
In a comment to my recent Awkward Modeling Situations post, Roberto Ciarloni (CTO of Think3) said:
I think that , aside from bugs , there is
an important issue emerging from this post: should CAD software evolve to become
more and more intelligent , that is anticipating users’ needs , or should rather stay as a tool that works well under the user directives ? I am in favor
of this second approach but I see the marketing appeal of the first.
Let me first say that I have a serious problem with the word “should”. When you use the word “should”, you are implying that you are some sort of an authority on an issue. You are saying that you know better than the people who “should” do what you are telling them to do. Usually the parties on opposite sides of the word “should” have different goals, and the different goals usually turn out to be the source of the difference rather than what whoever should or should not be doing. When you use the word “should”, you “should” place it in some form of a context, such as to achieve a certain goal, you should do X.
So when I put the word “should” in quotes, it means that I’m trying not to take my question too seriously. It means that I recognize that there are different possible goals for creating a CAD system, and that there may be different valid answers depending on the goals of the person answering the question.
Anyway, Roberto is using the word “should” in the form of a question. He is asking users what we would like to see, in the context of what we need our CAD tools to do.
This is a theme I keep returning to on this blog. How automated “should” automated CAD tools become? Automation almost always comes at the price of control. How much control are we willing to give up? Is it ok for CAD tools like SolidWorks to continue on this downward spiral, where they are continually aiming at a less and less sophisticated audience just so they can sell more software? At what point “should” you just split the product into two, with one half catering to “SolidWorks for Dummies”, and the other half becoming a power tool that requires expertise to use?
Some automation is welcome. Whatever it is that SolidWorks (or Parasolid) does for fillets, for example, is very welcome. Manually creating fillets would be a very tedious task of offsetting edges, trimming faces, and creating new fillet faces through one of several techniques. I guarantee I would invent a lot of reasons why models didn’t need fillets if I had to create them manually.
On the other hand the new Lip/groove feature automates things also, but is fairly limiting in what it can do. This tool is meant to reduce work, but it only does it on the most basic parts. It doesn’t work when a rib intersects the wall of the plastic part at the face that is to be broken by the Lip/Groove. So the automation takes you part way, and then drops you off.
Or look at the Hole Wizard automation. It is so highly automated that a lot of changes simply can’t be done easily. It doesn’t follow the normal rules for sketches and planes, which makes it difficult to maniplulate the sketch to angle the hole, or change the geometry of the hole.
Some of the sketch relation automation in SolidWorks is welcome, but at a certain point it sometimes starts overdefining sketches automatically. This is where automation becomes counter productive.
The new “Convert to Sheet Metal ” sheet metal functionality? Yes, it automates things, but sometimes it makes a sloppy mess of your part as well. Sometimes the manual functionality is what is needed in order to complete the job.
Mold Tools are a great example of automation gone wrong. When using the mold tools, you are frequently forced into a certain type of design. Most of the SW mold examples show mold interlocks. I haven’t seen that many molds with interlocks, but the limitations in the SW Mold Tools seem to require them. So the automated tools force you down a certain design path. To some people this is acceptable, but not to me. Design “should” not be dictated by your tools.
I’m not suggesting that we throw away all of the automation, because much of it really is wonderfully useful. What I’m saying is that automation is not the goal that must be achieved. Completing your work is the goal that must be achieved. Sometimes you cannot have both. Sometimes you need to just turn off the fancy stuff, and get down and do some good old fashioned hard work.
Surfacing work is mainly manual, although some of the functions just take over and don’t allow you enough control to do what you want to do. Tangency and c2 edge conditions for example are often a mystery. Users don’t really understand how it all works, mainly because SolidWorks does not document this stuff. How does the Fill feature determine the UV grid direction? Why can you not manipulate this directly? Why does the Fill not find the simplest result with the fewest inflection points, rather than creating a crazy egg carton surface for a simple patch? What happens when you assign c2 all the way around a patch, but the boundary is not c2 to itself?
Why can users not control rebuilds? All of this is taken care of automatically behind the scenes. If users could control rebuilds, we could probably get an honest 40% rebuild speed bonus immediately. How about all of that extra data that SW stores inside part and assembly files? Do we really need that? If users could say, yes, I need this, but I don’t need that, we could save time and disk space.
But no, CAD has to be as “intuitive as a light switch”, according to Jeff Ray. Well, I’ll tell you, a light switch only has two options, and the kind of work I do requires more control than that. If SolidWorks is so hell-bent on converting AutoCAD users that they want to make the software as “intuitive as a light switch” with only two options, I’m using the wrong software. I don’t need dumbed-down CAD. I need CAD that empowers me.
So what “should” SolidWorks become? It is headed a direction that I cannot use. It is being formed into a mass market 80/20 market maximizer to reach the stated goal of over 1 million users, not to act as a tool for me or you to create what we can envision. It isn’t getting more powerful, it’s just getting more automatic. It isn’t enabling me with more capabilities, it’s removing options and homogenizing the results.
So, I think SolidWorks “should” split into two tools, or have a set of advanced options. In one mode it enables tools like extrude and revolve, and in the other, it gives the user more control. You really can’t please everyone with a single piece of software. It’s time they realized that and stopped trying.
I wouldn’t mind so much if I could just shut off the trash that gets in the way of what I really need to do. In fact, I still do much of my modeling in a way very similar to what worked in 1997–partly because that’s the only way I could learn to do it back then and partly because it’s proven always to be the most reliable method of getting what I need to do done.
Several of the things you mentioned above–mold tools, sketch relation automation and even Toolbox are things I cannot possibly make use of most of the time. I only started using the Hole Wizard about three years ago because I couldn’t count on it to properly call out my holes, or switch from default of minor pitch to major pitch. Ever had that one bite you? You get parts/prototypes back and all the holes are so cored out that the fasteners have no metal/plastic to bite? How does this help me?
When it comes down to the tools I rely upon, they’re simple and (for the most part) tested and fail-safe. I frequently get tooling cut worth more than my house–how could I justify trashed tooling because my CAD tool failed me when I wasn’t looking? It’s too much to ask.
Maybe that’s just my personality. I charge clients a decent amount of money to get the job done, and I cannot fathom failing that task (or worse, costing them money). It doesn’t matter if it’s a glitch, I’m the guy doing the design and they must ultimately trust me.
I know I mentioned it before, but it’s like back in my physical model-making days. Simple set of tools. Almost everything I needed (apart from a Bridgeport, of course) fit in a wooden cigar box. Not many tools. X-Acto knife, oodles of spare blades, custom-made sanding block, small files, radius gauges, and plastic-working solvents, adhesives, etc. along with my trusty vernier calipers, 1-2-3 blocks, and squares. With those tools, I could create models to within +/- 0.005″ fairly quickly. I learned a lot then, too. Master a few tools well, and you can do anything. Clutter your repertoire with complex, specialized tools, and you face a diminished return on that investment, as well as your own productivity.
CAD is a tool. Is it cluttered with slow functions that might fail when I’m not looking and cost many thousands of dollars? Then I don’t need it. I’ll go “old-school” and create what I need with more reliable tool sets. Sure, sometimes it takes longer, but it won’t fail me and it often takes even LESS time.
I think this is a terribly relevant discussion, as I’ve also lamented some of the poor choices in flaky tools in recent releases of SolidWorks. And worse, many of the flaky tools haven’t been redressed in future releases, but left to languish in permanent uselessness. I don’t leave tools of that sort around my office, and I guarantee tools of that sort will never make it into the trusted cigar box.
I prefer a tool that is organized and used according to principles. The alternative is one that has little structure and a cookbook or menu of loosely related functions. When I define a sculpted shape I prefer to use a few well placed splines, curves and sections. The alternative is one with many small features glued together, it will never be perfectly fair, but will closely approximate the shape.
I want a few good reliable tools. I do not want the software interrupting with useless dialog boxes. (My current gripe is the X to close a sketch that has not been edited in any way that brings up the “Do you want to save changes?” dialog) The automatic relations that break the sketch are also most unwelcome. The wizard functions are awful. Automatically generated mistakes are the worst.
you are right absolutely.
“You really can’t please everyone with a single piece of software.”
I think you’re contradicting yourself when you say just before that, “In one mode it enables tools like extrude and revolve, and in the other, it gives the user more control.”
And there in lies the solution. The best comparison I can think of is photography and cameras. Many of the newest cameras, compact and even dSLRs, have BOTH manual and automatic modes and controls. For those who don’t know, need help, or are just lazy, you can use the automatic controls, and get decent, good, well exposed pictures. BUT, if you KNOW what you’re doing, and want more than just a good picture, but a great picture, a stunning picture, and have the creativity, then you can use the manual controls to get the look, feel, color, etc. that you want to create your masterpiece.
So I think what you and others (myself included) are really asking for, is MORE manual controls. I have absolutely no problem if they want to add automated functions, but give us more manual controls for the “CAD artisans” who want to create CAD masterpieces we envision in our heads.
****
Yes, the camera analogy is a good one. Is there a single camera that is best at both automatic and manual photography? Probably not.
I think that Bruce is right in using the
analogy of the cameras that have both
automatic and manual UI.
Also the last sentence is perfect : design
happens in designer’s mind not in the
software.
I want to point out also that I am very
skeptical of the possibility to write
really intelligent software , it seems to
me that intelligence is deeply linked to
context knowledge and the design context is
(like most) incredibly complex.
So trusting a supposed-intelligent CAD
program will usually end-up in decisions
that the designer will not like and maybe
he/she will discover it too late…
Matt,
Have you thought about SW dumbing down in order to separate it from Dassault’s premium product CATIA? In other words, if experienced users cannot get the “functionality is what is needed in order to complete the job”, maybe we are using the wrong Dassault product? Maybe a sophisticated SW and the CATIA product were overlapping a little too much?
Phil Sluder
****
Phil,
I’ve often thought I’m using the wrong product. I don’t think I ever thought I was using the wrong DS product. Catia is probably not the correct tool for an independent doing product design, at least I haven’t seen anything to indicate that. There are plenty of other tools that accomplish what I need to do.
I think there are situations where the CAD system can help the user with automated “wizard” like functions. Take rendering for example. Up until 18 months ago rendering was a chore in any system that required skill, knowledge and sheer luck frankly. Now with Hypershot and PhotoView 360 I can get results ina fraction of the time that before would have taken me days. That kind of intelligence/functionality is good.
In modelling terms every user who has to create anything more complex than a extrude/fillet/shell combo can list numerous situations where having TRUE design intent would be useful. I used to use Think3 products and the GSM stuff in that was great for that 10-15% of times I struggled to get the shapes I needed via standard methods.
I think there are three stages in modelling.
1. Initial shape generation – defining the major surfaces and features
2. Detailed modelling – defining things like ribs, lips, cut outs etc.
3. Editing modelling – going back and redefining to meet either changed criteria or advice from the toolmaker.
I think some apps to 1 well (Alias/Rhino/Ashlar), some do 2 well (SolidWorks/Pro/E), hardly any do 3 well (maybe Think3 with GSM or CATIA with functional part – both $$$) unless you know what the changes are likely to be and can build this into the model from the start.
Of course this also sums up neatly exactly what most big companies (and some small ones) do during development.
Stage 1 – concept – IDers using above tools
Stage 2 – development IDers and engineers using SolidWorks
Stage 3 – tooling – mainly off loading the mods to the toolmaker using something like Delcam Powershape or Think3 or CATIA – or – a total rebuild in SolidWorks or whatever.
And there’s the thing. I don’t think you can get the best of everything in one system. Like in many things in life you get the best by using the best not by sticking to a single brand. So for the mainstream sector where does that leave the SolidWorks user? Well I think the core application is pretty good. I would like to see the focus on modelling retained and I would like to see more basic functions added (like conic curves, or the ability to delete part history without having to export and import it). I’d like to see Rhino or Ashlar like curve handling (something ALL the MCAD sector gets wrong IMHO).
The common thread here is if your absolute basic building blocks are sound and robust and very easy to use and edit you have a strong foundation to build from – or rebuild if that is the case.
My skepticism on intelligent CAD is
also what makes me suspicious on
Syncronous Technology. I did not yet
use it so I may be well wrong but
when they say that the software is
inferring features and constraints
I cannot avoid to think that the software could easily infer wrong things given a dumb model.
History approach is for sure more rigid
but here the designer is explicitely
telling the system what are the features
and what are the constraints.
****
Roberto,
Interesting to hear your point of view on ST. There is actually a way to see and disable/limit the constraints it makes as you edit. After seeing this aspect of ST, it is easier to trust.
Once again you have provided us with a thought provoking post, and good points in the comments. As an instructor I see the people coming into the solid modeling arena with a deer in the headlights look. There needs to be a balance in functionality. The software has to have ease of entry to produce longevity. Once you have longevity, five plus years say, then you have the luxury of a broader view. The functionality we need now is not the same as when we first started.
I see the the 80/20 rule in day-to-day tasks, 80% mind-numbingly tedious get it done stuff so I can do the 20% really cool stuff. Some of examples of automatation mentioned fall into the 80%, and some can leave me to my own devices. As long as they work when I need them I am not going to be too picky about how I get there. Back to the point: get the job done.
I believe Autodesk tried to be everything to everybody which diluted their products. And the lesson was do what you do well. Take risks, expirement, make mistakes but somewhere somehow figure out what you do well. Enhancements are great but the users ulitmately detrermine value. We all know the software will evolve. If the software is truly user driven, the users will keep it on track. Maybe they can increase the functionality of Cosmic Blobs a little and have the second tool?
Hi Matt,
This discussion and all point raised are very good and for each person in similar positions the solutions will be ‘similar’.
What I personally believe the software industry need to do is understand functionality and flexibility of software is of the greatest importance and ease of use and automation should be a distant second. For those following set processes to produce product variations, ease of use/automation plays a significantly important role in both time to market and cost containment/reduction.
But for the many of us who do jobbing/training work and work across a number of disciplines, automation is of lesser value; it is often a handicap.
Bruce Buck says “You really can’t please everyone with a single piece of software” and he is partly correct. However, I believe a piece of design software should, and can, have ALL the necessary tools to complete any design in any discipline and that means the choice of tools must range over both 2D and a choice of 3D methodologies simultaneously and those tools must be able to be used in any combination not the in segmented way we currently have – this is one reason why I like Synchronous Technology. I feel it makes Solid Edge more complex to implement and use but the flexible has a greater value.
As one firmly entrenched (supplying, training and supporting) in Autodesk’s world I also agree with beekerly, Autodesk have made a ‘dogs breakfast’ out of their software. Their pursuit of ‘verticalization’ and 3D at any cost was and is their greatest mistake.
However flexibility comes with a penalty, complexity; as a jobbing person heavily involved in training I believe that is more important. If CAD software is flexible and complex enough to satisfy ‘jobbers’ and trainers, and includes an appropriately suitable (customizable) interface those that would benefit from automation will also be accommodated, but the reverse is not true.
Autodesk’s software is the obvious example: AutoCAD, ‘not great’, very ‘complex’ and criticized by many but just look at what people have done with that ‘tool box of bits’ and their own ingenuity. On the other hand, the criticism(s) leveled at Inventor by many is its lack of flexibility; as one guy puts it, “it doesn’t work the way I do and slows me down”, and it is struggling to gain the same acceptance as its so called obsolete predecessor.