Do you build top down or bottom up?
That is correct. Building each component in its own file will avoid that problem.
The reason is that it prevents you from making all sorts of references between components in the same assembly.
Using edges from one body in one component to project onto a sketch in another component creates an interdependence between these two and that is not always intended. That's jus one example of such references.
But with a little practice and discipline it can be avoided or used to ones advantage.
Something that helps is the BORN modeling technique.
I am not aware of any comprehensive courses in Fusion 360, but have considered putting together a curriculum to support my private tutoring efforts.
The reason this isn't easy to find own youtube videos is that the vast majority are created by hobbyists for hobbyists that often done have the needed experience in any industry to explain this in some context and this does need some context.
In general Fusion 360's strength is more of a top down design workflow where you design parts in place in their final location. Terms such as To-Down and Bottom-Up really mostly serve to explain some basic approaches, but most more complex designs are a mixture of both.
This is a great topic - one of my favorites. I find the labels "top/down" and "bottom/up" to be a bit misleading in this discussion. A maybe more important related question is "local components" vs "external components". You can build in a "bottom/up" fashion using all local components (though it does take a bit of self-restraint to do so). It's harder to try to build "top/down" using all external components, maybe not even possible. So, the topics are certainly related.
In my admittedly biased view, the ability to have local components is one of Fusion's coolest features. Most of us came from Inventor, and it was a pain (a UI pain, a data management paint) to create a bunch of parts on disk just to get a quick concept design in place. So, I tend to like to work with local components when I can. Certainly when exploring design options, this is very handy. There is also the related benefit of "design in place". Again, this is much harder to do in a traditional parts/assembly modeler. It is very convenient to be able to build up a design with all the components built in the right place. And the As-Built Joint is the culmination of that workflow. If the components are in place, then it is very easy to create joints between them. Finally, the ability to do position-based modeling is also very powerful in this workflow - to be able to reference a component inside another component, at a certain position, I find very useful.
So, when should you use external components? Your mileage may vary, but I find that external components are very useful for components that don't change much (off the shelf stuff), because then you don't need to have all the features for building those components cluttering up your timeline. There is also a performance aspect to it. For small designs, having all the components in a single timeline is great. Once you get above 50 components or so, it becomes a nightmare to manage (more on that later), and can be slow to modify one of the early components in your design. So, if you know you are going to be building a large design, it's good to think up front how you might want to break it up. Maybe along subsystem boundaries makes sense. A lot of this just takes time and experience to find the way that works best for you.
And, the points raised here about it being very easy (too easy, maybe) to create a dependency nightmare using local components is a very good one. Be very aware of what you are referencing where. Sure, sketching on the face of another component, or auto-projecting an edge/face from another component into a sketch is convenient. But, every time you do that, you are creating a dependency in your design. It's easy to end up with spaghetti dependencies, and as mentioned here, editing becomes very fragile. So, turn off auto-projections. Use datum planes without dependencies for sketch planes when you can. Be very deliberate about projecting things into your sketch. When you do need to project, choose the most stable thing you can project (a face is more stable than an edge, a body is more stable than a face, and origin work geometry is the most stable of all).
Use R.U.L.E. #1 and #2 to keep your components self-contained. You will be much happier.
One last word about local components/top/down modeling. If you want to save yourself future nightmares when you come back to edit them months later, be very careful in organizing your timeline. Turn on "Component Color Swatches" to help you visualize your component organization in the timeline. If you make an edit to a component, roll back to its area of the timeline and insert features, rather than just sticking them at the end of the timeline. I take it one step further, and create a timeline group for each component, but that's just me. And, as rule #2 says, rename everything. It's so much easier to know which extrude created this notch if you have it named in the timeline. Name your joints, workplanes, sketches, everything. It is a pain at first, but you will thank yourself later.
Jeff Strater
Engineering Director
@kb9ydn wrote:
@jeff_strater wrote:And having to rollback the timeline to make edits to a component is not only tedious, it's counterproductive.
Here is an easier way to do this:
Activate your component.
Roll the timeline back one step
Roll it forward one step.
Now you are at the end of the timeline of that one component and any thing you add now is aded to the end of the timeline.
Definitely not a one-click solution, but Fusion 360 is not yet at the point of refinement of Solid Works for example. I generally shudder in disbelieve when people start comparing Solid Works with Fusion 360 and wonder what in the world these people do with Solid Works. The breadth and maturity of functionality Solid Works has today will take another 10-15 years to appear in Fusion 360.
I agree with many of your assessments. It is very easy to create relationships between components and the timeline is not particularly informative, neither does Fusion 360 offer any more verbose ways to visualize relationships, let alone fix them.
@TrippyLighting wrote:
@kb9ydn wrote:
And having to rollback the timeline to make edits to a component is not only tedious, it's counterproductive.
Here is an easier way to do this:
Activate your component.
Roll the timeline back one step
Roll it forward one step.
Now you are at the end of the timeline of that one component and any thing you add now is aded to the end of the timeline.
Definitely not a one-click solution, but Fusion 360 is not yet at the point of refinement of Solid Works for example. I generally shudder in disbelieve when people start comparing Solid Works with Fusion 360 and wonder what in the world these people do with Solid Works. The breadth and maturity of functionality Solid Works has today will take another 10-15 years to appear in Fusion 360.
I agree with many of your assessments. It is very easy to create relationships between components and the timeline is not particularly informative, neither does Fusion 360 offer any more verbose ways to visualize relationships, let alone fix them.
Ahh, that's quite clever! I'll have to remember that trick.
Re comparing Fusion and Solidworks; yes it probably is unfair and I'm fully guilty of making this comparison too often. 
Given it's relative newness it's understandable that there will be incomplete features and that the UI won't be very streamlined. But the entire idea of the timeline when working with assemblies is quite fundamental to the workings of Fusion, and it has severe (I would say negative) implications when it comes to working with large models. So it's hard for me to imagine that the developers didn't foresee this being an issue and build in a way to handle it from the beginning. I'm SO hoping there is a plan for making Fusion better at working with larger models, but to date I haven't seen any indication of that. And that is what worries me the most.
C|
@kb9ydn wrote:
In other words, allow assemblies to be built without a timeline.
You can do that already. Simply start a new empty design and turn off the timeline and start inserting componebts/subassemblies.
This is generally a good workflow as it is indeed much easier to re-organize assemblies.