Posts made by roiki11

@cjm

In a cnc program you want to adjust your toolpath based on measurements. So in one point you might see that the features are slightly too small. So you add cutter comp to cut a tiny bit smaller. You can also use it to walk your holes to a specific tolerance(for example). Add cutter comp to undersize the hole. Then walk back the offset and repeatedly run that program segment.

Cncs are not accurate like you imagine 3d printers to be. The machining process often requires fine tuning to get proper results and continuously regenerating the program is very slow and unnecessary. Just better to change a single value on the console and get on with it

@deckingman

3dprintingnerd has a video on YouTube that shows the bed a bit better. It's basically individual pcb heaters with their own thermistors hooked up to a central "hub" that's under the bed. It also allowed things like segmented heating, hotspot measurement and compensation and heating only specific areas of the bed.

It's a pretty neat concept, albeit a complex one.

Lacks a 3 post leveling. A pretty basic feature.

Worth it is a very subjective term. One basic thing those people trying to make their printers very fast is that the easiest way to remove vibrations is to make your frame heavier. Just by adding enough mass to the frame you can make a print head of several kg move at 1000mm/sec.

Just ask essentium. Or pnp machines.

@visionary

i didn’t mean actual stewart platform, just the kinematics of it since what you’re trying to do is mathematically very similar to it with the two platform rotation axes.

Anyway, a 3 dof parallel platform is a simpler stewart platform. In this diagram, You can count the x1 and y1 rotation axes and you get the height of the point B from the I actuators and trigonometry.

If you look at this Instructables link, i’d think you get what you want when the T vector is the height and is constrained to the base Z vector.

At least thats my understading of the math.

@joergs5

Have you though about looking at Stewart platform kinematics? It's a pretty similar in concept, it just has more movable axes.

@o_lampe

Not true. In CNC world the 5 axis programs are supplied as tool vector coordinates (X, Y, Z for the tip, I, J, K for the vectot point) and it's the machine itself that applies the appropriate transformations to run the machine. The CAM software doesn't need any machine specific information to create the code, though specific limitations of individual machines are not taken into account(there's Sim software for that).

The same approach would have to be done on th slicer in order to not lock it for a specific machine configuration that it's designed for. You're right about that.

@joergs5

Linuxcnc is a machine controller, like duet and RRF. The G-code is largely the same, aside from some codes that might be specific to machining or 3D printing. If you had a slicer that would output a 5 axis G-code with the standard tool vector co-ordinates, you could build a 5 axis printer with linuxcnc. Though it’s missing the more advanced 3D printing focused features.

@joergs5

a non-planar printing is simply anything that uses the 3rd axis while printing to create non-planar layers. techically mesh bed compensation falls in to that. kind of. though its not printing non-planar layers.
there is already a fork of slic3r that does non-planar printing. some guy did it as a masters thesis in germany a few years back. People have made machines for 5 axis printing, or a tlting bed printers. Mostly for university work. They just aren’t usable in the real work because there’s no slicer to run them. i don’t see it as a chicken-or-egg problem because the hardware has never been the problem. Making the 5 axis hardware is quite easy. Even the matrix transformations for a 5 axis 3d printer aren’t that complex if you use the BC trunnion method, which mostly comes down to two matrix transformations since no work offset coordinates are needed.

Linuxcnc actually has that, has had for years. It’s just not very usable in a machining context and there’s no slicer to try it with a 3d printer.

3D printing didn’t really kick off until a workable slicer emerged. Its the most complex problem to solve.

@joergs5

thats honestly the wrong way to go about it. The slicers would need to be developed first, then the hardware. A normal 3D printer can already do non-planar printing. There’s no modifications required(though you would need a special nozzle for optimal performance). for a true 5 axis printing only a simple trunnion with 2 motors is needed. which is trivial to make with a 3d printed parts .

I get it, im more of a hardware guy myself. But this really is putting the cart before the horse(or climbing a tree ass first as we say). We’d need slicer support before any hardware work is really usable. and i dont see any work being done on that front. If i remember the slic3r maintainers were quite hesitant to bring the non-planar slicing work into the mainline slic3r. which is a shame.

There's no slicers(except a few proprietary ones) that can do anything but basic 2.5d slicing.
Even non-planar slicing hasn't been brought over to slic3r despite it existing for years.

The biggest hurdle for any 3d printing in more than 2.5d is the slicer. It's a CS problem that is hard to solve.

@t3p3tony

But open loop motors have no benefits over feedback motors. Quite the opposite. With feedback you can get the maximum performance out of a motor.

@dc42

You could make a completely separate board with hardware quadrature counters or biss-c master ic(ic-haus makes them) then just use SPI to talk to them. The master ic supports SSI too.

@coseng

You might want to check the temp rating on the linear rails. Most of them are rated around 85c and the plastic parts are nylon which melts at around 120. It's possible the thermal expansion binds them beyond that.

You'll also want to use grease that works at high temperatures and doesn't melt away.

@neueklasse

If you get them used. New they are 300-500 1k and over.

Also you have no way to interface absolute encoders to duet. No protocol support. And using incremental doesn't work as you have no way to actually know if the system has moved or not. So you need to home it to be sure.

Or just don't shut down the machine so you don't need to home it.

What you'd need is a pid controller on the duet, implementing the outer position loop. This isn't currently possible.

Also if you need it to maintain position across power cycles you need absolute encoders. Which are quite expensive.

@xyzdims said in 5-axis CNC/FDM: Tool vs Machine Coordinates (Inverse Kinematics):

@roiki11 You might look at my blog post https://xyzdims.com/2021/04/19/3d-printing-sub-volume-segmenting-non-planar-slicing/ regarding 4- and 5-axis FDM and non-planar slicing.

I have seen it, It’s good work.

@droftarts said in 5-axis CNC/FDM: Tool vs Machine Coordinates (Inverse Kinematics):

@xyzdims impressive progress! From what I understand, 5 axis machining is mostly a slicing/tool path generation problem, and the firmware doesn’t resolve the input coordinates; it just goes where the Gcode tells it to. So it’s slicing where the heavy lifting happens. But I haven’t played around with it, and I’m not a programmer.

As for tools to generate 5-axis Gcode, Fusion 360 has it built in I think, but it’s only available for paid-for versions, and is limited to machining rather than 3D printing afaik. I’m sure there are other professional tools on the CNC side, but haven’t looked recently. 5 axis 3D printing seems very much in its infancy!

Ian

Its completely the opposite. Most industrial machines do the kinematics on the controller itself. The gcode is tool center point data and you use a specific g code to turn inverse kinematics on and off. They can also jog in these coordinate systems and have dynamic work offsets where the program origin can be anywhere on the part. This also means programs are machine agnostic. Mostly.

The biggest obstacle for 5 axis printing is Indeed the slicing side. There’s currently vety little information and research in 5 axis slicing algorithms. Its much more complex than 3 axis. We even havent made any progress on non-planar slicing.

@nraynaud

What that needs in the machine is inverse kinematics of the machine to Calculate the positions for a given tool center point position. No open source controller currently does this properly, though linuxcnc has it on experimental basis.

@theruttmeister

If you have a well insulated enclosure then it's much more efficient to heat it up once and keep it at set temperature. Instead of constantly blowing hot air and fighting against the significantly colder ambient. It's stupidly inefficient.

The longer the print time. The more inefficient it is.

PEEK is notoriously hard to print. Even on high end machines so I'll believe it when I see it. Especially on usable parts. Also PEEK parts should usually be removed hot and annealed straight away. Letting it cool first is much more likely to warp.

I'm a little skeptical about that being PEEK though.