How to Model Based On Complex Existing Geometry in Fusion 360 (Or Go Mad Trying)

For years now, I’ve had an Android tablet mounted (semi)permanently in my car. It’s been falling apart for a while and last week, I finally repaired it to death, so now I’m faced with the task of making a new mount for a new tablet. It’s been a long time coming anyway, the old mount was total crap and JUST functional to leave in place since 2014. Temporary solutions usually last longer than you expected, even if they’re made out of bent sheet steel, PCB, romex, a home depot yardstick, and some sugru. True story.

The general task is to turn this:

dashboard without tablet

into this:

Dashboard with tablet held in place by hand

But, like, without my arm holding it in place. And yes, I realize the tablet is enormous. After much mockup and test fitting and sitting, the top edge is out of the field of view in my normal driving position. The massive screen will mean less time focusing on a tiny Google map and more time focusing on the road.

The problem, of course, is making something that neatly fits that dashboard shape. My initial thought was to design something similar to this mount from Kuda:

Kuda brand E90 dashboard mount

The problem was figuring out the specific shape of the dash in that area, to model off of.

My solution to that problem was photogrammetry. I’ll leave the details out, mostly because I actually did all this work back in 2017 and REALLY don’t feel like going back to put together a whole post about it. The gist is, cover all reflective surfaces with matte blue painter’s tape, take a whole ton of pictures from all kinds of angles but with super flat diffuse light, and use something like Autodesk ReMake to reconstruct a 3D model from the data. Unfortunately, ReMake is no longer a product and ReCap, its all-things-3D-capture successor, isn’t free. Like I said, out of scope. Sorry. Anyway, this is what I ended up with, after lots of pain and suffering:

Looks good, right?!

It’s pretty tricky to work with meshes while modeling in F360, though, so I had to strategize. My thought was to build a block around the mid section of the dash there, right where the Kuda mount goes. I’d build a solid, then use the mesh to split it, leaving a surface on the inside/underside of it in the shape of the dash. You can’t cut solids with meshes in F360, though. After trying a whole bunch of things, I landed on the best solution being to loft a bunch of mesh section curve fits. Here’s what that looks like:

Pardon the haphazard process and the texts. I’d already done it once and didn’t want to do a complete re-do, so I just kind of turn stuff off and demonstrate the concept. This is basically a complete rehash of this Fusion 360 Quick Tip, which I strongly encourage you to follow instead if you need to recreate this technique.

I ended up with just the block I wanted.

Block cut to shape of dash
Just the shape I was going for.

To check the fit without printing a big time consuming structure like that, I extruded a short little .4mm section off of each side, which could be printed much more quickly.

Coupons to test the profile fit
Print these two out and see that the left and right profiles are good.

Unfortunately, checking up in the car didn’t go quite so hot.

One coupon against the dash with the model obviously scaled wrong.
Obviously, the model was scaled incorrectly here.
Coupon against dash - scale is right, but the fit is poor.
The scale is just about spot on, but the accuracy just isn’t there.

I ran into two problems here:

First, the scale is tricky to get spot on. Because the photogrammetric reconstruction is missing the finest details, there aren’t any really good fiducials to key off of to get the scale right. I ended up using the corners of the radio faceplate, since they’re sharp, recessed corners, but with pretty open angles so they show up in detail on the 3D model. That and the distance is pretty significant, so error as a percentage of that should be relatively lower. After scaling the model based on that dimension measured in the car vs the model, the coupon was about the right size.

Second, the fit just isn’t that great. Unfortunately, even with the scale right, the photogrammetric reconstruction just isn’t quite accurate enough, I don’t think, to produce a useable output for a task like this. A better strategy would be to re-model the dash itself from measurements, but I don’t know the best strategy for that. Even after trimming the coupon a bit, I could never get a satisfactory fit with the dash contour.

Next Steps (Recalibrating After Failure)

Unfortunately, I won’t be knocking out a 3D model I can get on the printer tonight, ready to toss on the car tomorrow. Luckily, there are a couple other options.

1. Remix another design

Luckily, I’m not the first to have this problem, and it looks like the other guy is a better modeler than I am. This mount on Thingiverse is for a Nexus 7 (my old tablet, in fact), but the dash interface can probably be repurposed for my needs and interfaced with a cradle for my much larger tablet. The biggest disadvantage of this route is that the author sticks it to the dash with tape, and hell if I can get anything to stick to mine. https://www.thingiverse.com/thing:2787046

2. Mix 3D printing and metalwork

Another solution is one I actually used back in the very first cardboard-and-duct-tape prototype iteration of this project, in like 2012. Thin material can be jammed both under the vent fascia in the middle of the dash (where the Kuda mount secures itself), and behind the wood veneer trim in the notch. If I simply cut a 56mm wide strip of aluminum and bend it in two spots, it should jam nicely between the notch and the vent fascia. Hopefully, that’s secure enough.

That’s all for now, thanks for reading!

2 Comments

  1. Leigh Whittemore
    January 10, 2019

    My first thought, after seeing your extrusion concept, was to wonder why you just didn’t use a simple contour guage (just do a Google image search for “contour tool”). Cheap. Simple. A WHOLE lot faster than photogrammetry. Or, only a LITTLE more work, how about a simple clay-molded curve. Are you perhaps over-thinking this and not using the KISS methodology?

  2. January 10, 2019

    A contour gauge (or yes, even better, a clay model) would definitely get the detail. But the problem is still digitizing that into a reproducible, printable form. In an ideal world, I’d make a 3D model that’s printable, and when I decided I needed to adjust it or change it or whatever, then all I have to do is a minor digital tweak, and I can completely re-make a new one with minimal marginal effort. Doing something completely by hand locks me in to whatever I do the first time (and means I can’t share it!). Plus, I know I’m going to 3D print something for the tablet-side of the interface, so it might as well be one monolithic structure instead of having to design and fabricate two separate things.

    But anyway, you’re right in that there’s a simpler solution – I’ll post about it next time 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *