6 Axis linear Delta

tkln

@dc42
So the kinematics calculation is completed and tested with Matlab, so far everything works. Now I want to implement the refraction in the firmware. For this I would like to cut out everything unnecessary from the LinearDeltaKinematics.cpp. So that I can enter x,y,z and the three rotation angles u,w,v and the machine moves. Is it correct that I can delete everything in LinearDeltaKinematics.cpp from line 245 or does everything have to remain in place for it to work?

dc42

@tkln you could write stub functions to replace the ones after line 245 in LinearDeltaKinematics.cpp.

tkln

@dc42
So is it correct that from line 245 everything else can be commented out? Or is any of this required for the Reprap firmware to work if I just want to test motion without any calibration?

tkln

@dc42

I have several questions about the LinearDeltaKinematics.h file.

Is it enough to change these things in this file UsualNumTowers and "Axis names used internally" + "Derived values" for my kinematic calculation?

static constexpr size_t UsualNumTowers = 3; 
-> static constexpr size_t UsualNumTowers = 6;```

	// Axis names used internally
	static constexpr size_t DELTA_A_AXIS = 0;
	static constexpr size_t DELTA_B_AXIS = 1;
	static constexpr size_t DELTA_C_AXIS = 2;
	static constexpr size_t DELTA_D_AXIS = 3;
	static constexpr size_t DELTA_E_AXIS = 4;
	static constexpr size_t DELTA_F_AXIS = 5;

Can i use SUPPORT_LINEAR_DELTA for my Six Axis Delta or do I have to change something there?

tkln

@dc42
Is it correct that machinePos[X_AXIS], machinePos[Y_AXIS], machinePos[Z_AXIS] calculates the Z coordinate of the linear slides?

// Calculate the Cartesian coordinates from the motor coordinates
void LinearDeltaKinematics::ForwardTransform(float Ha, float Hb, float Hc, float machinePos[XYZ_AXES]) const noexcept
{
	// Calculate RSUT such that x = (Uz + S)/Q, y = -(Rz + T)/Q
	const float R = ((Xbc * Ha) + (Xca * Hb) + (Xab * Hc)) * 2;
	const float U = ((Ybc * Ha) + (Yca * Hb) + (Yab * Hc)) * 2;

	// Note, Ka + Kb + Kc = 0 so we could calculate just two of them
	const float Ka = coreKa + (fsquare(Hc) - fsquare(Hb)),
				Kb = coreKb + (fsquare(Ha) - fsquare(Hc)),
				Kc = coreKc + (fsquare(Hb) - fsquare(Ha));

	const float S = Ka * towerY[DELTA_A_AXIS] + Kb * towerY[DELTA_B_AXIS] + Kc * towerY[DELTA_C_AXIS];
	const float T = Ka * towerX[DELTA_A_AXIS] + Kb * towerX[DELTA_B_AXIS] + Kc * towerX[DELTA_C_AXIS];

	const float A = fsquare(U) + fsquare(R) + Q2;
	const float minusHalfB =  Q2 * Ha
							+ Q * (U * towerX[DELTA_A_AXIS] - R * towerY[DELTA_A_AXIS])
							- (R * T + U * S);
	const float C = fsquare(towerX[DELTA_A_AXIS] * Q - S) + fsquare(towerY[DELTA_A_AXIS] * Q + T) + (fsquare(Ha) - D2[DELTA_A_AXIS]) * Q2;

	const float z = (minusHalfB - fastSqrtf(fsquare(minusHalfB) - A * C)) / A;
	machinePos[X_AXIS] = (U * z + S) / Q;
	machinePos[Y_AXIS] = -(R * z + T) / Q;
	machinePos[Z_AXIS] = z - ((machinePos[X_AXIS] * xTilt) + (machinePos[Y_AXIS] * yTilt));

dc42

@tkln, I suggest you create a new kinematics file and use the new type number that I allocated for you. Unless your kinematics uses the same inverse kinematics as for a linear delta, you will need to enable segmentation. This is done by supplying default segmentation parameters when calling the base class constructor. See the RotaryDeltaKinematics files for example.

tkln

@dc42
I understand that would be the best option for clean and orderly documentation.
I would like to test my kinematics first and for that I would like to have only the most necessary things in the kinematics file and keep everything as short as possible. For that I would like to not add something like autocalibration etc. in the first place.

tkln

@dc42
Ah the tip with the Rotary Delta helped me, thanks. Now I can compare a bit.

tkln

@dc42
Okay, last question for today, it's getting late in Germany
Sorry for all the questions but I'm so excited to finally see the machines in action.
If I just want to test the 6-axis kinematics first, is it possible to create 6 axes with "Creating additional axes" and feed them with motor positions? I would then convert a Gcode with my 6-axis kinematics and then give the values to the Duet.
Is there a command that I do not control the ABCDEF axes using Cartesian coordinates (X10 Y10 Z5) but directly using steps (A5 B10 C10 D10 E100 F10)?
Is there anything against it?

o_lampe

@tkln said in 6 Axis linear Delta:

Is there a command that I do not control the ABCDEF axes using Cartesian coordinates (X10 Y10 Z5) but directly using steps (A5 B10 C10 D10 E100 F10)?

I guess by defining them as rotatry axes and setting a propper steps/degree value, you can mimic steps . (eg. with 16 microsteps it would be 'M92 A16')

tkln

@tkln
Solution: https://forum.duet3d.com/post/315912