Gantry XY orthogonality

I have built a gantry machine which does my PCB routing.

I have opted for tooling posts mid way along X axis at full Y axis span.( two pins as far apart as possible).

I will then drill tooling holes in the PCB blank laminate and proceed to drill and mill.

Tooling pins and associated data processing software allow me to operate the equipment in a kind of WYSIWYG mode.

Except I have a slight misalignment between the two layers.It seems X and Y axis of the machine are not perpendicluar to each other.

I have tried packing Y axis and have minimised the the misalignment however it is still there.

I was wondering if there were any tutorials or blogs pertaining to orthogonality issues.

Real live comments are also sought.

I have a program that will skew the X-Y coordinates in the gcode and output a new file. WarpDriver can be downloaded from PaulRowntree.weebly.com. Most of the code is to warp Z, but the skewing is described as well.

Have you measured the actual angle?

Cheers!

How does the skewing algorithm work?

About which ( x,y) point does the data get rotated?

Does it take into consideration the effects of spindle offset from the gantry face?

Do You have a general analytic solution for gantry skewness interms of spindle offset, XY angular displacement and tooling refference frame?

Errr.. apologies for this barage.

Quote:

---

Originally Posted by Zig

How does the skewing algorithm work?

About which ( x,y) point does the data get rotated?

Does it take into consideration the effects of spindle offset from the gantry face?

Do You have a general analytic solution for gantry skewness interms of spindle offset, XY angular displacement and tooling refference frame?

Errr.. apologies for this barage.

---

The program assumes that the X,Y coordinates in the file are good, that (0,0) in the file corresponds to (0,0) on the machine, and that it can correct the skew by shifting X for each specific value of Y on the real machine. The farther Y is from the Y=0 axis, the larger the skew correction applied to X. You provide the gantry skew angle, and all X are shifted to remove this error. It is not a rotation. Since this is working on toolpath gcode, it does not need to worry about reference frames etc. All control codes that are not motion related get put directly to the output file without changes. The spindle offset is not part of the problem, nor the solution, as long as you measured the gantry skew with the given geometry of the machine.
One issue is that it must expand g2/G3 arcs into a discrete set of steps (skewed circles are no longer circles) so the output gcode file is often larger than the source.
For the type of skews that you are probably working with (<<1 degree I hope) the Y calibration is probably not affected.

It would be much better if this type of correction was built into the controller software (eg Mach3) ...
Cheers!

The co-location of data file origin and machine frame origin is always guaranteed by the virtue of the fact that the tooling pins are part of soft tooling approach.

This is to say I use two 3.175mm ( 1/8") tooling pins located in holes drilled in delrin ( engineering plastic) plugs in the machine base at predefined coordinates which are referenced to the gerber processing software such that a direct one to one relationship exists between the machine and data processing space exists.

The tooling pins are located along a mid X axis value ( x=290mm).
Data is processed and any layer inversion is effected by flipping data about X=290mm.

The advantages of this approach are obviousRegistration top to bottom layer is ensured provided ( in main) the gantry X Y axes are orthogonal.

They are not.

I drill a set of holes in the corners of a sheet of laminate ( notionally 580mm by 470mm ).
Inscribe circles in top layer to observe they are centered on holes.

Same circles on the bottom layer in data are then flipped about X=290mm. The laminate is also flipped on tooling pins about X=290mm. The circles are then milled on the bottom side to observe lack of alignment on the bottom layer.

It is the difference between the orthogonal axes in data and less than orthogonal XY gantry axes.

I can make a measurement of the shift between the holes and the circles on the bottom layer ( error due to machine skewness).

I need to estimate the thickness of packing shim required to bring the gantry X axis into a truer alignment.

This would then be followed by replacing the delrin ( screw in ) plugs, drilling new tooling holesin to new plugs. A new sheet of laminate would then be drilled milled topside, flipped over and milled on bottom side using flipped data.
So newly made test plate would then be analysed and the packing process refined.

X axis structure is approximately 710mm long( distance between points where packing material could be inserted to correct for skewness).

Since each itteration is involved in terms of calling for new delrin plugs and possibly new sheets of laminate since the actual distance between tooling pin holes in laminate changes with each attempt to correct the skewnes I had hoped for some insight into an analytic solution for correction to machine geometry in order to minimise the number of itterations.

EDIT
As it stands machine output gets rotated about a point on X=290 line within the machine frame. The angle of rotation and amount of translation is dependent on the degree of lack of orthogonality. The first order approximation of origins being co located does not hold true since the gantry support points are outside the work space of the machine,

Thanks Zig for the very detailed reply. Ideally, as you say, you would be better to shim the gantry into squareness. For a long time I could not do this, and so I did the correction in software. In both cases, the fix is only as good as your measurement of the skew.

Do you have any idea of what the skew angle is? How far out of alignment are the holes after the flip?

On my site there is another program called SkewCalc that helps to determine the skew angle.

After initial efforts to minimise errors due to orthogonality issues and bearing in mind the that the errors are exhibit translation and dilation effects the errors within +/-100 from center of worktable are acceptable( approximately 0.05mm from through hole) which is approaching acceptable registration.

At extreme corners of the work table the errors are 0.15mm in various directions ( both into and away from the centre of the panel ( translation and dilation brought on by flipping ).

As intimated an analytic solution woudl have possibly been able to provide a handle on how to tackle shimming in order to minimise the waste of delrin plugsand laminate material.

Come to think of it... drilling sample holes along X-290 at farthet extremeties would provide a much clearer and less ambiguous outcome. A result which is directly and least ambiguiosly related to orthogonality. :idea:

Sorry about the chewed up ear

Zig, I still don't have a feeling for the skew angle, since I don't know the size in Y.

If you cut (or mark) a rectangle (as big as you can accurately measure), what are two diagonal measurements?

Paul,

Test holes were located nominally on a rectangle x=516mm y=413mm

Regards

If the differences of the diagonals are ~0.15 mm, I estimate that the skew is 0.0133 degrees.

Does this make sense?

paul ,

I have tried shimming the gantry with some successand noticed that 0.005" thick shim contributed significantly, disproportionately in fact to the shim thickness and shift in misalignment top to bottom.

So I guess the estimate is probably quite right.

I am going to start hunting for thinner shim material in the hope of at least halving the current error.

I appreciate Your time and effort / opinion in discussing this problem.

If you don't have uniform orthogonality, it is most likely caused by a non planar XY and or XZ and or YZ plane.

Attachment 301148

This tool is relatively cheap and you'll know right away what kind of problem you have but it's very likely you'll have to modify your machine to correct the problem in a controlled fashion.

I have an IronMan that I paid a fortune for and the orthogonality issues are so bad and what's worst, there is absolutely no practical mean to fix the problem.

Yves