3D printers don’t slow down for smooth curves they try to run at constant speed. Of course a “curve” is made from line segments and they meet at corners. Typically with 3D printers they take shallow corners at full speed if the instantaneous change in speed of each motor is less than the maximum “jerk” setting. They only slow down as they approach a sharp corner to keep within jerk limit.
CNC mills (which are generally heavier) can be instructed to slightly round off corners to avoid instantaneous speed changes.
It isn’t the CAM stage that does this. It produces G code with the desired speed setting. It is the motion controller that works out the acceleration and cornering speeds, etc.
Motion planning for standard trapezoidal acceleration is not particularly difficult. You have to do two passes, one forwards and the other backwards because the speed you can run on any particular segment depends potentially on all the previous segments since you last stopped and all the following segments until you next stop.
I am playing with a laser diode at the moment and the spot is more than twice as long as it is wide. Motion planning for that will be fun as for constant power density it will need to run at a speed depending on the angle.
Jerk in applied maths is rate of change of acceleration but in the 3D printer world it is the maximum instantaneous change in speed. So when it starts from stationary it will instantly jump to the jerk speed and then accelerate to the target speed linearly. When it takes a slight corner where the speed of each motor changes by less than the jerk limit it won’t slow down.
Obviously the head obeys the laws of physics so doesn’t change speed instantaneously.
It is expressed as a single constant. It might depend on speed but that is ignored.
So interesting in a device such as a laser cutter that’s going to be a constant, but clearly in a CNC mill it changes with the weight of the end-mill (a 1/64" ball mill vs. a giant 4" flat cutter clearly have huge differences in mass) so is that taken into account? (especially in a unit with a tool changer)
I assume on something like my Taz6 where I can swap heads (dual extruder, single, flex, dual flex) does one have to change the jerk settings (off some manual list I presume)
I think there is enough margin in the normal jerk setting that people set them and forget them.
In my own firmware I don’t have a jerk setting other than with a stepper motor the time between the first step and the second step defines the instantaneous start speed. I.e. you can’t ramp up from zero with a discrete step.
I’m far from an expert, but I do have an X-Carve and have spent a fair amount of time making things with it. In my experience, the weight of the mill isn’t a meaningful factor in the general case. I’m sure there are outlying cases where there are unusually heavy tools, but typically the mill is a very tiny component when you’re looking at the overall mass of the Y axis.
For instance, the Y axis on the X-Carve looks something like this (off the top of my head):
2 x End Plate Assemblies (End plate, pulleys, gears, stepper motor)
Y axis extrusion
Head carriage [ie. Z axis] (Stepper motor, lead screw, gear, belt, limit switch)
Router
Collet
Endmill
The aggregate mass of the Y gantry plus the cutting resistance (driven by type of material, depth of cut, speed, mill size/shape, etc) is more of a factor, I’d expect.
Oh yeah, didn’t mean the X-carve (I have one too, but yeah, the tools are all pretty much the same), I meant a real mill like a big tormach or something with a rotary tool changer. Some of those big flat cutter bits weigh many pounds…
Ah, gotcha. Yeah the constraints there are different. I wonder if the relative increase in rigidity and mass of the machine still makes it moot. Good question, now I’m curious too!
My guess is that the machines with the big honking cutters aren’t moving all that fast with them. And when you have a really big cutter you usually have a really heavy motor/etc to move it.
But if I might break in for a moment with a related question: has anyone else noticed cutting speeds being size-dependent? I made a bunch of tiny little rectangles to see how my cheapjack baltic birch would cut (about 1cm on a side), and got a through cut at about 100/80. But when I tried to make something bigger (a circle 6 cm diameter and a rectangle about 7x10, I had to drop down to 100/60 to get through.
It would certainly make sense in terms of acceleration and maximum speed and so forth, but it could complicate some designs.
My guess is the acceleration limit is such it can’t actually hit 80ipm over 1cm and stop again. I.e. it would need to go 0-80 over 5mm. In that case the motion planner would have to limit the speed. In a large circle that isn’t the case, so it will run full speed.
When they fix the over burn in corners, i.e. ramp the power during acceleration, then the small squares would also not cut through until you reduced the speed.
An interesting consideration here is the non linear nature of the lasers power supply and the effect of the beam on the material. I am sure the complexity of ramping down without extinguishing and ramping up again smoothly make for a testing nightmare.
Having thought about this, there is also another way low power mitigates over burn. With the power turned down you can score more slowly to achieve a given depth. When you run slowly the acceleration and deceleration distances get much less (by the square of the speed) so the length of the over burn will be much shorter.