Pocket accuracy

[Tony Pagett]

I have been running my CNC with V-carve for some weeks now but mainly tinkering and using MDF with no need for accurate dimensions. The results have been very pleasing.
But, I tried to make an accurate piece out of Corian and my pockets are all under size by approx 1%
For example a pocket 70mm x 40mm came out as 69.3mm x 39.6mm and a hole of 7mm diameter was 6.9mm.
I have a laser attachment on my CNC and by using this I know that the CNC is very accurate. For example a 500mm x 500mm square came out as 500mm x 500mm.

Is there some tolerance setting that I am missing? or could it be to do with the measurement of the tool bit?

[Adrian]

It can be a number of things. The bit dimensions could be off, it could be bit deflection or deflection in the machine itself.

Their are no forces involved when using a laser whereas when you're routing there is a lot of force on the bit and on the machine itself.

Try shallower passes and/or reducing the feed rate.

[sharkcutup]

Adding to Adrian's suggestion in shallower passes, feed/speed rate

Try ramping into the pockets --- I have found using ramping into pockets tends to eliminate most machine flexing. It adds a little bit more time to your toolpath but the end results is better.

I would also recommend checking the actual bit size too that can sometimes make all the difference in the accuracy of your toolpath.

Sharkcutup

[Tony Pagett]

I always use ramps and a very slow feed rate ( I am in no rush to damage my bits or machine )

I think I can rule out bit deflection or machine deflection because the pockets are extremely accurate and square (albeit too small) Surely if there were any deflection the edges of the cuts would be "wobbly"

[Adrian]

I always use ramps and a very slow feed rate ( I am in no rush to damage my bits or machine )

I think I can rule out bit deflection or machine deflection because the pockets are extremely accurate and square (albeit too small) Surely if there were any deflection the edges of the cuts would be "wobbly"

Not necessarily as the bit deflection would be constant over the pass.

Do a simple profile cut square with the same bit, depths and feed rates. Measure that. Now, with the same vectors, change the toolpath to 0,5mm depth or so so it just barely cuts the surface. If there is no bit deflection or machine deflection then the results will be identical.

[Tony Pagett]

Do a simple profile cut square with the same bit, depths and feed rates. Measure that. Now, with the same vectors, change the toolpath to 0,5mm depth or so so it just barely cuts the surface. If there is no bit deflection or machine deflection then the results will be identical.

I have tried this and the results are identical

I have just tried "fooling" the machine by editing the tool diameter. My 70 x 40 hole is now 70 x 40 but this only works with a single tool pocket. If I use a two tool (clearance and final) then the second, smaller bit does not remove any material apart from a very small piece in the corners

[Adrian]

What type of pocket toolpath are you doing? If it's a raster have you got a final pass option selected?

[Tony Pagett]

What type of pocket toolpath are you doing? If it's a raster have you got a final pass option selected?

I have tried just about every combination (offset, raster, climb, conventional etc etc)

I have noticed that the profile tool path is also out by approx 1% making objects oversized (130mm is cut to 131mm). "Fooling" the program again also corrects this.

[Rcnewcomb]

It is easy enough to validate if the G-code is correct. Draw a square. Do a toolpath ON the vector. The values in the G-code should match up with the square.

[LittleGreyMan]

Try tracing a very shallow (a few tenth of mm) square with a sharp new Vbit in MDF.

If the size is correct, this is definitely a flexion issue. If not, double check the size of your laser cut.

Coud you post a picture of the machine and the spindle?

[ger21]

I think I can rule out bit deflection or machine deflection

Any OpenBuilds machine will have a lot of deflection. They're not that rigid.
Everyone has seen deflection. Even $100,000 routers will have deflection.

As Randall said, check the g-code. The coordinates will be off from the actual position by 1/2 the bit diameter. You'll almost certainly find that there are no errors in the g-code.

[LittleGreyMan]

Any OpenBuilds machine will have a lot of deflection. They're not that rigid.

Gerry, you ruined my attempt to suggest that in a much softer way

[wb9tpg]

Using a Shapeoko 3XL to make mating box tops and bottoms they did not fit using the same vector to cut both top and bottom (profile on bottom one and pocket on top). I measured them with the micrometer and the top was smaller than the bottom I assume due to deflection. I tried a number of cutting options with little improvement. I ended up adding an -0.015 allowance on both the pocket and the profile cut tool paths and I have a nice fitting box (sightly loose to leave room to apply a finish to final fit).

[ger21]

Something as simple as changing the direction of cut can make a big difference.

Climb cutting will tend to remove less material, while conventional cutting will remove more.
The reason is that climb cutting pushes the tool away from the work, while conventional cutting pulls it into the work.