I have a heavily modified YOOCNC6040 with a 4th axis mounted on the table.
I’m thinking of adding a 1:100 or 1:50 harmonic drive between the Z stage on the Y gantry and the spindle to get a B axis.
Now harmonic drives don’t have a perfectly linear relation between input angle and output angle but instead a very slight sinodial function superimposed on that.
What do you think about my plan, given this issue?
You could add a position feedback sensor to the output and conpletely bypass the nonlinearity of the drive. Or maybe you could take advantage of the ballscrew error compensation available on the machine control software.
I’m not familiar with MACH3 screw mapping feature, but you could machine an adapter to couple the drive to a rotary encoder instead of the spindle just for the experiment, and use something like an Arduino to provide the step signal and count the encoder pulses from the origin. You get the number of encodoer pulses per step, and that should allow you to construct the look up table for the axis. You may even add some extra weight to the axis to account for the missing spindle, and get a more realistic (?) measurement.
@Citrus_CNC That would work. Good idea! Weight should not influence this. The harmonic drive has a 100:1 gear ratio and zero play (half of all teeth engage at any one time). So holding torque should be through the roof.
The 4th axis still has a 1:8 belt drive (I just replaced the NEMA23 motor with a very beefy one to get decent holding torque). If this one works out well, I may replace it with a harmonic drive too.
BTW…it will get an automatic tool changer from USOVO too. I finally found a silent (60dB(A)) compressor that does an 8-10 bar hysteresis to keep an 8 bar minimum for the ATC at all times.
As far as g-code goes I plan to do 3 Axis code in my good old Deskproto, then use cnc-toolkit with the surface normals to determine A and B angles. I got cnc-toolkit running well but not having any kind of tool-size-compensation is a problem. Thus the combination.
No it doesn’t. At least not freeform. It has 2 strategies that follow contours but for organic shapes without contours desiged in Fursion 360 in the first place or detected on import these two are useless.
Also it needs very elaborate machine configurations in a very complex language.
I’m not 100% sure yet it works well at all for machines that have no TCP correction in theirs control software.
I’m using it to machine masters and molds for casting that I currently do using 3D printing (Ultimaker II extended with a ton of added, custom sensors and remote operation) and weeks of plain manual labour to get a perfect surface finish.