Fusion 360 CAM doesn't use G03 where it should

[redbull00]

I realized that it's very hard to "force" Fusion 360 CAM to generate arcs instead of tiny linear motions. The underlying logic behind this is very unclear to me. For example in the project attached, I'd expect that the simplest circular cut operation should result in a very lightweight GCODE that uses G03 to express the contour with arc. Instead, it generates a huge file that breaks down the smooth contour into hundreds of small linear movements, which overwhelms the motion controller of my machine, that tries to shake the poor thing out of its existence. I first thought something is wrong with the machine I bought, but then I saw the huge GCODE and the reason behind that shaking became clear.

However, I realized that if I increase the precision of the contour operation to 0.01mm and use smoothing and set it to 0.02mm, it uses G03 correctly, only this really doesn't make sense to me, because this setting doesn't work consistently accross my projects. E.g. in another project where I used Create->Revolve operation to create a 3d shape, the same settings no longer result in G03 arcs - it's back to the same horrid heavy GCODE that uses thousands of tiny linear motions. How do I use this correctly?

[BrandonTBFBF]

In this particular case you're using a 3D toolpath to machine a 2D feature. Using a 2D Contour strategy rather than 3D contour yields a small file size even with smoothing turned off.

With CAM software, the better the inputs, the better the outputs. 

[seth.madore]

What @BrandonTBFBF says is very true; use the best tool for the job.

That said; with the 3D Contour (not the right tool for the job), it can be made to be much smaller code, simply by turning on Smoothing.

Before:

 

After:

 

(all those dots represent separate lines of code)

[HughesTooling · Accepted solution]

In  your example file you're using a 3d toolpath and they all generate linear movements unless you use smoothing to fit arcs (a lot of the 3d ops turn the model into mesh to calculate the toolpaths). You usually get best results if the tolerance is 2 - 5 times smaller than the smoothing tolerance. For this simple example the 2d ops will create the simplest toolpath and for arc no need to use smoothing at all.

One problem you can run into with the 2d ops is machining convex splines. When Fusion offsets the curve it does corner rounding but on a spline this means lots of very short lines with tiny arcs between, the arcs are the same as the cutter radius. Sometimes you can get smoothing to work if you set the op's tolerance very fine like 0.001mm and set smoothing to 0.01mm, might even need a finer tolerance. This is only a problem with in computer compensation, if you use in control smoothing works with more sensible tolerances.

The 3d ops actually do a better job on convex curves.

Mark

[redbull00]

Thank you, this really offers a valuable insight into the process. Also replying other comments - I used 3D contour to reproduce what I did with another example which is a bit more complex and it requires that operation.

[redbull00]

I used the 3D operation to represent what I did in a bit more complex example that I have (it's a half circle profile that's turned). However, what you wrote clarifies the issue for me. Thank you! Also, how did you manage to get the individual segments displayed in the screenshot above?

[redbull00]

I noticed one more issue - the smoothing now works, but it seems to affect only the cutting toolpath. Ramps and transitions don't seem to get smoothed. Any way to fix that too?

[seth.madore]

---

@redbull00 wrote:
how did you manage to get the individual segments displayed in the screenshot above?

---

You can toggle that display here:

---

@redbull00 wrote:
I noticed one more issue - the smoothing now works, but it seems to affect only the cutting toolpath. Ramps and transitions don't seem to get smoothed. Any way to fix that too?

---

 Smoothing only effects the cutting moves, it does nothing for the transition or ramp moves.

[redbull00]

Thank you for your answer! Is there some way to also smooth transition or ramp moves? The issue is that it overwhelms my motion controller and the machine slows down significantly. I wonder if there's any way to avoid this. Having cutting moves smoothed is already a huge improvement!

[seth.madore]

No, but if your controller and machine struggle with this sort of motion, I would suggest doing what you can to minimize them. For starters, I'd suggest turning off vertical lead-in/out radii. Plunging into a pre-drilled hole or off the part is also a great way to keep code condensed.

[redbull00]

Thanks for answering so quickly. I noticed that vertical plunge and turning off lead in/out help a lot too! Glad to hear from an expert that my observations seem valid. Is there any plan to have a feature that smooths transitions or ramp moves as well? Is such the feature perhaps available in Inventor?

[seth.madore]

No, I'm not aware of any plans to improve ramp and transition moves (in regards to smoothing) and no, I don't believe behavior in Inventor is any different.

What is your machine and controller, what sort of work do you do?

[redbull00]

I am a hobbyist working wood and ocassionally metal. I hardly ever need anything more precise than 0.1mm tollerance.

The machine uses RnR Motion board (a Chinese USB motion controller) with Mach 3. I think Mach 3 is mainly at fault here, because its motion planning is pretty horrid. I just did a benchmark where it has to make a circular motion with 55mm diameter (interpolated with 4 equal arc sections) and unless I set the tolerance to 1 mm for the constant velocity mode (G64), the machine still makes 4 distinct stops and starts (accelerating and deccelerating) even though it should not do that at all, because there's literally nothing between the 4 arcs GCODE. When the tolerance is set to 0.1, it really noticeably shakes making those 4 distinct pauses, which makes no sense, because if it can execute each arc section within acceleration limits (which it does perfectly), it should as well be able to continue this motion to a full circle with no change of velocity (which it doesn't, unless G64 tollerance set to 1mm).

This test rules out any USB bandwidth or too rich GCODE issues, because now it's nothing more than just 4 arc sections and there's still this horrible shaking present, unless the tolerance is set to an abysmal value.

I think the control software (Mach 3) is simply really bad and outdated. I should probably consider an upgrade. I am curious what your opinion is on this subject.

[seth.madore]

I think it sounds like you need to spend some time digging into Mach3 and it's many areas of motor tuning. It's been over 20 years since I built my first CNC (and used Mach3), but I recall going thru much the same frustration. 

I think @HughesTooling might still use Mach3 on one of his machines, he may be able to make a suggestion or two

[redbull00]

Thank you. I was digging for the last couple of days in those settings (CV and motor tuning), but it really seems that no matter what I try, this ridiculous problem persists. The machine weighs around 120kg and the accelerations are at 800mm/s^2 for X and Y axis and I wouldn't be comfortable going past that value, because the whole frame the machine stands on shakes violently from anything past that!

If nobody will be able to help me with this, I think I will really have to consider getting a better software and controller. It almost hurts me to watch this poor machine suffer so much from jittering - at this rate I will have to start tightening the screws every couple of hours of use lol. In any case, your answers were hugely helpful, because now I understand smoothing in Fusion 360 much better and it already offers a great improvement in performance - even with this annoying motion control software.

[seth.madore]

I would encourage you to ask these same questions on forums that are more dedicated to the Mach3 platform; cnczone.com, reddit, official Mach3 forums (if they even still exist)

[redbull00]

Will do! Now thanks to what I learned here and some experiments I did, it's clear what is at fault 🙂 I can imagine more people struggle with those issues, because I saw similar posts here. If I learn any way to improve this situation with Mach 3, I will also try to remember to mention it here.

[HughesTooling]

Not used Mach3 for a while but from memory make sure you have constant velocity enable and also try increasing the look ahead.