"Z compensation" on 4x8 foot CNC router

[ryanjg11]

I would like to do some miter folding on my CNC router, but this requires leaving .005" of material at the bottom of the V-groove. While my machine is very accurate in the X and Y axis, I have never really been able to dial in my Z axis and get the table perfectly flat...

I am running Centroid Acorn on an Avid Pro 4896. This is a 4x8-foot gantry-style CNC router made of mostly 80/20 aluminum extrusion. I knew the aluminum would be a weak point, so I welded my own base using 1.5" diameter Schedule 90 pipe:

On top of this, I have an MDF vacuum table made from two 4x8 sheets of MDF. The bottom platen was coated in wood glue after this image to reduce leakage through the bottom:

I'm running an Airtech L160B rotary vane vacuum pump which is a bit undersized for this size of router, only pulling 112 cfm. While hold-down is good, the MDF top is actually too porous and kerf leakage also contributes to parts lifting up. The somewhat counter-intuitive solution I found was to add another sheet of 3/4" thick, 4x8 "normal density" MDF, not skinned, on top. For a total of three sheets of 3/4" MDF making up the entire vacuum table. This results in good part hold-down and minimal leakage.

I mention the vacuum table construction because I think it might be one contributing factor to the challenge I've had getting accurate Z-height across the entire table. Of course, I am flattening the table using a 2.5" SlabSlayer flattening bit routinely, and while the vacuum is turned on during this operation, there aren't any parts on top of the table to apply downward clamping force. I suspect that the spoilboard is being flattened correctly, but once a real sheet of plywood is put on the table to be cut, the vacuum table pulls the table harder toward the middle of the table, likely creating an valley in the middle of the table.

It could also be that the MDF is compressing slightly, and maybe inconsistently, forming Z height irregularities.

I was wondering if anyone has had this challenge before, and if there's been any creative solutions to solving for it? Is there a software solution that could incorporate Z-height measurements at different locations of the table, say every square foot, and use that to compensate for Z height irregularities across the table surface?

[ryanjg11]

One more idea I had to perhaps solve for this during flattening the spoilboard: use a 8x2 foot wide panel and place it adjacent to each cut pass (flattening is done across the long side of the table). I'd have to turn on/off the vacuum and pause the spoilboard flattening process in CNC12 constantly. Would take forever, and not particularly practical.

[Ken Rychlik]

I glue the top spoilboard down. That won't help with the machine construction, but it sounds like a lot of your issue is the loose spoilboard. I use wood glue and a paint roller to apply. I put it on the pads of the grid with a light coat, being careful not to get glue running into the grooves. Then when you place the board, you have to put it EXACTLY in place and not slide it around. Throw a sheet of plastic over it and turn on the pump while it dries. I walk it down also while the vac is on. Then when you surface, you don't need to run the pump and it works pretty well. You can either shave the top of the grid to put another piece on, or put a perimeter bead of glue with a few lines down and put one over the top of the old.

[ryanjg11]

I glue the top spoilboard down. That won't help with the machine construction, but it sounds like a lot of your issue is the loose spoilboard.

This is precisely how I attached the vacuum table top to the platen. The third panel isn’t glued (yet), just composite nailed around the perimeter. With the vacuum on, it’s held down pretty well across the top.

Feeling like this might be more of an issue of unequal clamping pressure between spoil board surfacing and putting an actual panel to be cut on it. I can probably confirm by measuring the flatness of the top just after flattening, and then take similar measurements with a thin plastic sheet over the entire surface, and using my z touch probe at different spots and documenting any variance.

[tblough]

Have a look at volumetric compensation: https://centroidcncforum.com/viewtopic.php?f=67&t=9266

[Ken Rychlik]

I work on cnc routers all the time and I will say on a mdf spoilboard, having .010 difference in a day is pretty normal in Texas. This is with heavier machines than what you have, so the light duty machine may be adding to the issues. With high humidity the spoilboard is thicker. then as the day goes on and you are pulling vacuum through the board, it dries out shrinking. Sometimes as the day goes on with a lot of sheets cut, I either adjust the tool measure switch height setting, or surface again, depending on my needs. Cutting through .01 on my cutout parts was normal.

Building cabinets, exact depths were not critical and 5-10 thou wasn't an issue for me. With you having 3 sheets of mdf to give issues, it might be worse than normal for what I have seen. You may just be expecting perfection out of a moving target. Aluminum or phoenlic for a vac plenum would help some, but the top spoilboard will still have movement, and the lighter machine will still cause some of it.

[ryanjg11]

Have a look at volumetric compensation: https://centroidcncforum.com/viewtopic.php?f=67&t=9266

Tom, this is super interesting and an almost perfect solution for me. Sadly, it looks like this feature requires the Ultimate license which I think starts at $2,000. I only have the Router Pro license.

[tblough]

IIRC, the upgrade from Pro to Ultimate is only $299.

[cncsnw]

I do not have time to review the volumetric compensation manual for you in detail right now. With luck, cnckeith or someone else who has worked with it can comment. But...

I do not think that volumetric compensation, as currently proposed, will do what you want. If I understand it correctly, it would add one Z correction amount for every X interval (in cnc-x.tab), and one Z correction amount for every Y interval (in cnc-y.tab).

If you want to map out a randomly undulating table surface, you don't just need X + Y compensation amounts; you need X * Y compensation amounts.

In other words, considering a 10" x 10" square of table surface, sampled at 1/2" intervals, the current implementation would give you 21 corrections based on X position, and 21 corrections based on Y position. You want 441 correction amounts based on an XY grid. The current implementation would work for a table surface that was perfectly flat, but not level in either direction; but it would not work for a table surface that is compound curved.

[cnckeith]

Ultimate License for Acorn is $399
https://shopcentroidcnc.com/acorn-cnc-controller/