Hacker News Reader

Comment by diafygi | Hacker News Reader

It's my understanding that higher quality plotters/3d printers/mills' accuracy and precision come from higher quality motors (e.g. stepper motors instead of servos). Will it be possible to someday just put cameras on the plotter which watch the current print and self-correct as needed? That's basically what our eyes do when we write/sculpt/carve.

Will cheap motors + cameras + software eventually meet or exceed the quality of expensive high precision motors/controllers?

View on HN

Rigidity is a major factor--if you inspect one you'll notice that high speed and high repeatability machines (e.g. a 6-axis +/- 0.020mm repeatability manipulator) are made from rigid (often high-mass) components with robust joints.

I see. I wonder if my hypothesis is still valid, even given lower quality rigidity. Human arms aren't made of big heavy metal rods, yet we can still draw beautifully. Couldn't cameras and software compensate for cheap unstable draw arms?

Replace large "cheaper servos" with a human and a self correcting small servos and you sot of have this:

https://www.shapertools.com

Its very cool...

That's in interesting thought but instead of a camera what if it used an optical mouse sensor? Even cheap ones in $5 mice can do hiDPI now and track on most surfaces. Maybe it could slide on the paper next to the pen with a fixed offset telling the software exactly how far the pen has moved.

Optical sensors will drift over short timescales so you will never have a repeatable "zero" point. it would probably work for small organic shapes but anything more geometric would not turn out well.

> (e.g. stepper motors instead of servos)

Servos are more accurate than steppers.

See: https://en.wikipedia.org/wiki/Servomotor#Servomotors_vs._ste...

https://en.wikipedia.org/wiki/Visual_servoing seems related.

I believe cheap servos are inherently very discrete due to noise in the "potentiometers" used: it will not be possible to do precise positioning no matter how exact your timing pulses are. But thats just a guess. If its a noisy potetiometer/wiper but monotonic it should be possible to correct in software. If its non-monotonic and has reverses gradient it will be much harder or impossible I guess.

precision and accuracy don't come from motors. Mostly it comes from rigidity and squareness of the assembly, the smoothness of the rods, the quality of the bearings. The steppers I work with are not really accurate or precise, mainly they just average out errors as long as you're doing roughly the same number of moves in each direction.

Interestingly, servos mean different things. The servos used here are cheap crap (but work well enough) but in the Real World, servos are actually more like steppers that have feedback based on absolute or relative positioning.

I am also intriguided by the "can cameras be used for realtime feedback" but I don't think they have enough resolution to determine accurate sub-mm movements.

There are similar techniques used to self-correct 3D printing. You can actually license a system from NASA.[0]

[0]https://technology.nasa.gov/patent/MFS-TOPS-70

I can see this helping with low-force motion such as a pen-plotter but I think with things like 3-d cutting there just isn't a replacement for rigidity and high-power motors that can maintain a consistent cutting feed rate for a quality finish.

You can improve the precision by using visual feedback, but there are movements that are impossible to do if you don't have precise enough motors.

Something like the "Shaper Origin" ? but it's still not cheap

Reply on news.ycombinator.com