Fusion 360 help

Hello there. I just got my Revopoint Pop 3D scanner. I have created a couple of scans, and I am trying to pull them into Fusion 360 in both OBJ and STL format. I would like to be able to sketch on the scans, and then export to G code to use on my CNC plasma cutter. When I load the scan into Fusion I set the units to inches, but when I start to sketch on the scan the units are wayyyyyyy off. Any suggestions or help on what I might be doing wrong would be greatly appreciated.

Thanks!

2 Likes

I am guessing but you will probably have to set the units to Millimeters rather than inches. Pretty much everything in the 3d printing/scanning/cutting/etc world uses the metric system.

Like i said, it is a guess as to a solution but it is easy for you to check if it works :slight_smile:

1 Like

@KnightRid thanks for the input. I tried loading the scan in both mm and inches. When I sketch in mm and then export to g code the design is way to small. When I sketch in inches the design is wayyyy to big.

I scanned in both Feature and Marker mode. I have seen where someone on the forums recommend possibly trying to scan in body mode. I might give that a try.

1 Like

I don’t particularly use Fusion, but ReCap, Inventor, Navisworks, and AutoCAD have all had consistent behavior: due to the density of the points, they can’t imagine that you’re working in a unit smaller than meters. (This works out for me- reverse-engineering stuff with less than a couple million points is difficult when your eyesight is as bad as mine, so I import a massive number of points, let ReCap create a point cloud in meters, then open it with Inventor scaled to .001 and I have nice, dense geometry that I can actually see when I slice thin cross-sections out of it.)

Set your scale factor to .001 when you import, and you should be fine.

1 Like

Setting different modes alters the operating distance from scanner to object (in addition to adjusting exposure presets). Body mode is the furthest from the object being scanned and will provide the least detail.

However, whichever scan mode you choose, the output file will have no dimensional units (mm, inches or otherwise), because that’s not how Structured Light 3D scanners work.

To get dimensional accuracy, you will have to measure a feature on your scanned object (with calipers, for best accuracy) and scale the scan you’ve got to the model you need. From there, you can start simplifying, smoothing, etc.

That’s odd: all models I scanned so far have exactly the right dimensions when I import the stl-files into PrusaSlicer for preparation to 3D-printing. I didn’t set any scaling factor anywhere.

Not entirely accurate- whether or not it’s explicitly assigning dimensions as meters or millimeters in the file, the output is metric. When you import it somewhere, your software should automagically assign millimeters (though b/c of point / face density some might call it meters instead). The dimensions you pull from the object in CAD should be pretty accurate at that point.

I think I have it figured out now. I basically did what @JeffLindstrom suggested. I measured an easily definable part of the actual object I scanned, and then used that measurement to determine my scale. For example when I measure the same object in my model it would show 83". In reality the object was only 3". I then divided 3/83 which provided the proper scale factor. Once that was set things seem to measure correctly.

I was under the impression that the scanner would have some form of spacial relation to be able to determine distance and size, but that doesn’t seem to be the case.

Thanks guys!

Uhmm… no. The POP hardware does not measure in absolute dimension (mm, inches, meters, etc.). Scan any object on a turntable and export the result. Set the POP to a different distance from the object, scan it again and export that result. Load the two objects into CAD software side by side and see that they are different “sizes”.

The reason you are seeing millimeters in your CAD software is because it can deal with absolute measurements and that is the default for your system (not the case for everyone). My CAD software’s default can be set as I choose (I usually use either inches or millimeters, but I can choose among feet, meters, yards, kilometers, miles, etc.).

To confirm, load either of the object files into MeshLab, invoke the Measure function and see that the result has no scale, only a count.

Lot of assumptions here.

Yes, AutoCAD is essentially unit-agnostic, but if you create a drawing and you’re thinking in inches, then you’re going to have to scale by 2.54 / 25.4 / 254 / 2540 / 25400 or whatever to get the appropriate metric dimensions, and by 1/2.54 or w/e to get back to inches.

However, the world has moved on since then. I’m using Inventor and ReCap, which absolutely do care about units. My default units are set to inches for both.

Let’s look at a workflow case study.

This housing is a part I needed to reproduce, because it’s broken:

Removing the shaft and two gears, I scanned it, and got this:

(note: it’s already been into Cloud Compare and had the RGB value per point changed to a scalar field based on Z height, and it’s been into Meshlab to get aligned on planes. The red circle shows where my scanned part was broken.)

This point cloud gets imported into Autodesk ReCap so that it can be saved into the format that Inventor wants to see. As I stated, my default units in ReCap are set to inches. ReCap says “oh, damn, this is not inches,” and automatically sets units to meters instead. (It assumes meters rather than millimeters due to the density of the point cloud, which is pretty insane right now.)

Obviously it’s not meters- this part is 2-3 inches long.

Set it as overlay in Inventor–again, in a native-inch file–and Inventor recognizes that this is still a metric file, and we scale by .001 to get mm vice meters.

Next we open up the model of the part I’d previously made based on calipers and micrometer and compare the two. (Note- I also machined this part previously, and it’s pretty nearly a perfect drop-in fit, though a few areas that I’d simplified from the original part needed to be modified slightly.)

Looks pretty close… let’s overlay them directly.

If the original file was not created by Handyscan such that the location distance between points was meant to be understood using SI units, then those holes would not line up.

1 Like

Also, re: meshlab’s measure, that large hole in the part above where the planes intersect? That’s about 12.82.

The actual hole measures .507" (was .500 +/- .002, but it’s zinc alloy and has a steel shaft spinning in it).

Convert that to millimeters: 12.87. And remembering that we shrank it by 1000 in Inventor, that’s pretty close to the 12.82 that Meshlab gave us.

(edit- the hole in my part is larger, obviously. It was intended to be milled to the size shown, then bored and reamed, then have a bushing pressed in, which would then itself be reamed to size. The hole shown was for the initial machining operation, not for the final size where the shaft would sit.)

This is all great information. Thank you guys for the feed back, and detailed explanations

I would very much like to get a quick tutorial on how that works. The documentation for MeshLab is all over the map and not very well organized, so if you could point me in the right direction, I’d appreciate it.

Good point, but remember that the POP had to be designed in the first place to output a scan, so using the metric scale is reasonable for initial calculations and the range of depths for a Feature scan is on the small side, so it is not unreasonable that the resulting scan would be fairly closely aligned to millimeters, but it’s not guaranteed. I figure you just got lucky that the alignment was so close… this time.

Did you run the experiment I suggested about scanning the same object on a turntable with the POP set to two different distances?

That was scanned on the turntable, with multiple heights / angles / distances, and the part flipped 3 or 4 different ways to capture all the faces possible, and the resulting scans merged with Handy Studio.

Re: alignment to a plane in Meshlab- it’s pretty easy to do.

Start by identifying a planar feature in your part and selecting some polys or points, depending on what you’ve got:

The average of these faces will become the ZX plane in a moment.

You don’t need clean selections or to get every single face- you can see what my selection set here looks like.

(Incidentally- using the workflow detailed a few posts above, this scan comes straight into Inventor with the dimpled panel in the middle measuring 30.5" across, which is at least within the margin of error of the tape measure that I’ve previously measured at. Now, granted, the scan says that it’s curved when the actual part has a nice straight line and most everything is square or at 45 degrees, but the raw measurements are very close.)

Go through the menus as follows:

I’ve picked the ZX plane here, but you can pick whichever you like.

For the first alignment, you want to have rotation on any axis.

Click Apply.

image

With the axes turned on, you can see that we’re now at least reasonably parallel to the ZX plane, though not exactly laying on it. (The face is, as mentioned, curved where it shouldn’t be.)

You can also see that we’re rotated wildly about that plane. I’ve selected some faces on a panel which is (in the real world at least, but not in this scan) square to the first one.

This time I’ve selected the YZ plane, and I’ve told it that I want to only rotate on the Y axis. This should have us square (more or less) to two origin planes.

And it does:

From here, just use the transform: translate tool to handle any remaining alignment requirements by moving in the + or - direction along any of the three axes. Make sure you turn preview on, though, so that you can see what you’re doing, and once you click the apply button ALWAYS close the tool and re-open it before you do an additional translate, because at some point you will forget to re-set something back to zero and you’ll wind up moving it a second time by accident.

1 Like

Thanks for the tutorial. I can easily adjust the orientation so that any one flat facet is placed on a given plane, but my CAD software doesn’t average out the ‘fuzz’ that comes with scans of flat surfaces.

No problem!

Meshlab can do it pretty well. Cloud Compare is another useful one, but unfortunately while it’s translate/rotate tools are far more usable than Meshlab, it doesn’t give you the visual reference of the axes, so you it’s harder to get the alignment perfect. (It’s better for some other tasks, though.)

Convert .obj or .stl into quadmesh using netfab then covert into solid object. Everything will be easier in fusion after that.

1 Like

@Codejockey thanks for the suggestion. Is there something similar to the program you are suggesting but doesn’t cost $4k a year?

Releases · 3DprintFIT/netfabb-basic-download · GitHub - Netfabb used to have a free version but Autodesk (like all companies) has gotten greedy and took it away in favor of fusion360 (which is getting harder and harder to use for free it seems too). The manual can be downloaded from here - netfabb_basic7_manual_en.pdf - Google Drive

All this information came from Download the Old Netfabb Basic (Free) – Jules Gilson - which I had bookmarked years ago :slight_smile:

I am not sure what features are eliminated from the older netfabb versions or if it will do what you want but at least you can give it a try.