Kb58's granite CNC router build

[kb58]

The meeting with the servo manufacturer went great. The Support Engineer remoted into my laptop and adjusted various parameters, while I jogged the axis back and forth. What was cool is that the controller software includes a virtual oscilloscope, so he had that running and watched the servos response to step inputs. Once one servo was calibrated, its settings were copied into its twin at the other end of the gantry.

With that complete, now starts the long process of calibrating axis movements and testing alignment. The display on the left is the dedicated router controller, and my laptop on the right is for viewing YT videos on setting these things up. A big part of the unknown is simply knowing what all the controller buttons do. I have the manual but at 472 pages, it'll take a while to get through.

One issue I've already run into is a consequence of using the "paired drive" option on the controller. That is, feeding the same command to two servos. This is needed because the left and right servos work as one to move the gantry fore and aft. Here's the catch: what if when you command a Y movement of, oh, 10 cm, and one side moves 99.8 mm, and the other moves 100.3 mm, both due to component tolerances? This means that a rectangular part you're machining will be narrower on one end than the other. The controller has adjustments to scale commanded-vs-actual motion exactly, but when paired, it still sends the same command to both - they can't be separated. All I can do is measure both motions and set a correction that minimizes the overall error.

 

[kb58]

Asked for and received helpful advice on the Centroid CNC forum, and the happy face/pumpkin successfully completed. Of course I'm not showing the pile of failed attempts.

This was just for fun. Still ahead is the ugly business of ensuring final alignment before the router is used for real. Thinking about how to ensure that the gantry and side rails are square to each other seems like such a simple thing but isn't. What I really need is a huge precision square, and while they certainly exist (in granite!), I can't justify the expense, size, and future need. I could draw a huge rectangle and measure it diagonally, but that's not very precise.

"Typing out loud", I'm thinking a more practical, yet still accurate way is to get a large and known-accurate framing square, placing one edge along one linear rail and measuring off that to the other. Steel is nice because a magnetic base will stick to it, though I could just stick the magnetic base to the linear rail instead. The X and Y rails are offset vertically about 120 mm, making measurements less straightforward, but should still be doable. Large carpenter squares are fairly inexpensive, but they're a bit of a gamble for real accuracy. I have one and while it's great for wood projects, both outside edges curve inward slightly, with the high point being roughly mid-span. I'll check out some other brands and maybe they'll be better. With the router built on a surface plate, it's easy to check for straightness along at least one edge!

Yet another way would be to suck it up, wire the spindle, get some cheap cutters and sacrificial material, cutting large rectangles and measuring the results. I'm not sure this is much better, as it still means using a steel scale and eyeballing it. It makes me wonder how Piotr made his machine so precise - on the order of microns. I do recall him using large granite reference triangles though... baller.

 

[kb58]

Bought a brand new framing square, one made by a different manufacturer. If I choose to see the humor in this, it's that it has the exact same problem. Curious, it was measured on the surface plate. Turns out that the edges on the inside of the square are very straight, but the outside edges, not so much. Pressing down on one end, a 0.017" feeler gauge fits under it, suggesting that the high point - about 9" from the inside corner - is 0.0085" higher than the ends. Again, it matters not when building a house, but once again, me being a selective cheapskate means burning the gas to go back and get a better one - if that exists.

Edit: Well that kicked my butt, using cheap tools doesn't work... imagine that. Even their "best" framing square simply isn't appropriate for this, something I should have expected. I guess I can measure it by doing, cutting large pieces of something cheap and measuring it for squareness. Hmm, I can cut something then measure it with the mill's DRO, that would work up to its full travel of around 560 mm. Question becomes: what cheap material to use for test cuts? Cheapest would probably be from home supply centers, something like processed wood maybe, like Melamine? Plywood is out due to spring-back and no doubt a lumpy cut edge. I suppose I could cut some thinner aluminum.

Before going there, I have to address flying-chip concerns. I have a solution for chips wanting to fly out the back of the router bed but up front it's not so easy. Shifting mindset to "If I was a flying chip, where might I land", shows many annoying places, including on the tops of the X linear rails. Worse, because these are roller-type rails (not ball bearing), there are V-grooves along the sides, forming wonderful catch-troughs for flying chips. See pic below. There are also a lot of fun places on the Z assembly. Some of this can be addressed by simple shields or bellows-type covers. More food for thought.

Oh, and I was asked what I was going to do for a hold down system... sigh.

 

[kb58]

Connected the controller to the VFD and verified that the spindle could be turned on and off, and turned the right way. Speed setting though threw me a curve ball. In short, EMI (electromagnetic interference) generated by the VFD was getting into the control cable and feeding back into itself. As a result, a speed would be commanded, the VFD would start to speed up, but then the speed would start varying all over the place. Grounding the shield in the control cable fixed it, but I think the VFD also needs a filter, perhaps one on the input side and output side.

 

[kb58]

The first real job for the router may be a bridge too far: drilling anchor holes for many 9.5-mm threaded inserts in the granite bed. The potential problem is tool speed, recommended as 1,250 rpm. Routers work with small cutters spinning very fast, so spindles aren't designed for low speed and torque. I've sent the question to the spindle manufacturer but am not very hopeful. I really don't look forward to endless hours of drilling by hand, but there's no alternative if the router can't do it. Even if it can, there are two rows and columns of anchors outside the accessible workspace, so those have to be done by hand regardless.

I'm not even sure a CNC router is a good tool to drill these holes. It seems like feed speed would be an issue, as going too fast means that the bit overheats, or worse, the bit isn't keeping up and the Z axis just keeps pushing harder and harder until something gives. In addition, figuring out the feed rate for a new diamond core drill is going to produce some number, but with each hole drilled, it wears, so the feed speed needs to be reduced by "some" amount. Thinking it over, it's starting to sound like a bad idea, but we'll see what the spindle manufacturer says.

An ordinary hand drill could be swapped in for this, but they're such an odd shape, the adapter could take a long time to fabricate, and this still doesn't get around the feed rate issue or the rows that have to be done manually anyway. I think I just have to start ("how do you eat an elephant), drilling as many as I feel like, then sitting it aside for the next session. The good thing is that they don't have to be super accurate, only holding down a spoil board, probably a sheet of MDF.

 

[kb58]

I'm going to try having the router drill one hole in the granite, making adjustments and seeing how it does. If it completes without ruining the drill bit, then more holes will be tried - CNC machines are great at doing the same thing over and over. Worst case, the bit is destroyed, and I have to drill them by hand, but I"m already facing that if I don't try.

 

[kb58]

With great trepidation, "conversational G-code" was created to drill one hole, and the process stepped through slowly. It's very intimidating watching the spindle moving fast straight down toward the granite, worrying that it'll keep going due to some programming goof and at best, destroying the drill bit.

The awkward thing about drilling granite with a CNC machine is the material's hardness. No matter how slowly it's fed, each time the Z axis advances into the stone, the diamond drill can't instantly drill into it like a normal drill in steel or aluminum would. With each advance, the Z assembly and gantry deform slightly, until the weight drives the drill bit in to relieve the pressure. That was the primary reason that the feed rate was set to a very slow 1.2 mm / minute. Could it have gone faster? Probably, but I didn't want the risk.

This was something of a worst-case first customer for the spindle, and it was a little disappointing to feel the spindle vibrating during the drilling. I don't know if it's a bent drill bit, or flexing in the entire assembly. For the next hole, the drill bit will be recessed much further into the collet, as there was way too much of it hanging out and could only be contributing to the vibration.

So all that worry aside, it actually worked - slow, but it worked. A few settings in the drill sequence will be adjusted to save some time, then I'll try setting it up to do an entire column - carefully.

 

[kb58]

Regarding drill quality, small diamond core drills are pretty bad. Put it in the lathe chuck and tapped it a few times with a rubber hammer to "bring it around."

Good progress was made yesterday, drilling 16 holes with the router and one by hand. Part way through drilling I thought, "this thing is pretty awesome and there will be many uses for it. Even better, while it's doing its thing, I can be doing something else." It's only drilling one hole at a time because since the drill bit is hollow, once it gets to a certain depth, it's best to stop it and snap off the column of stone in the center. If it's not done, it eventually breaks off on its own, tips sideways, and the next time the drill bit backs out to clear itself and heads back in, it can hit the broken off piece.

I fully expected to get the rest done today, but shot myself in the foot, committing a software sin that I wouldn't have done had I not been overly excited. Last night I upgraded from the free version of the control software to enable more features, plus it's more oriented toward controlling a router instead of being generic (lathe, mill, router, and plasma cutter). In addition to the upgrade, they noted that I was running the previous version of released software, so that was updated at the same time. The mistake was installing the upgrade before finishing all the holes, and between the two changes, it erased the hardware configuration files. Yes, that's on me to have saved that information off, though ironically, it wouldn't have mattered if I'd done it the "right way", generating a report file that contains the system settings. The new version is incompatible with the previous one, so unless the numbers were written down or pics taken, they'd be lost regardless.

Entering the system values again took much of the morning, then the same drill program used yesterday was reloaded, figuring it would be fine... nope. Here's the really strange part. Running the same program moved the drill bit to a slightly different position! It was close, but for a CNC machine, it was miles off, about 0.5 mm. It was as if both the X and Y home switches had shifted, but they are all tight. I still don't get it. Anyway, reality was dealt with by adjusting the hole coordinates to match an existing hole. Come to think of it, I haven't yet double checked that offsets to the adjacent holes are correct. Thankfully, this hole pattern doesn't need to be super accurate - it's just for mounting either a spoil board or fixture plate. The fixture plate on the other hand, needs to be precise, so there'll be no messing about with sw in the middle of that job!

 

[kb58]

About half done drilling the holes. The router is doing great; the slow progress is due to manually drill the holes around the perimeter where the router can't reach. (I picked up that tip from a YT video, that having that area available for work hold-down is very handy.)

Shown is one of the perimeter holes being drilled, after locating it with a jig placed on adjacent holes. Toward the end of the day, the manual drilling seemed to be taking longer and longer, probably due to being tired of doing it - and the drill bit getting dull.

 

[kb58]

Well that's done, and something I don't want to do again! The router was able to drill the holes that it could reach, so all the ones around the edge had to be done manually - that took forever. That pile of granite "droppings", and an even bigger pile of rock dust, represent having drilled through 1,512 mm of granite! The three diamond core drills gave their lives to accomplish this - totally worth it.

Turns out that these drill bits are notorious for drifting. The problem is the design: the abrasive grit is in a big lump around the bottom end of the drill. As a result, as it drills deeper, the drill bit above the actual cutting surface is not touching the walls of the hole being drilled, so nothing is guiding it; nothing prevents the bit from shifting laterally. Sure enough, many of the manually-drilled holes are angled, even though the drill remained square to the work. The close up pics show how different a hole drilled by the router is to one drilled by hand; you can see the lateral drifting of the hand-drilled one. Thankfully it doesn't matter much - here - because the clearance holes in the plate above cover for any alignment errors.

After going back and forth a few times, an aluminum tool plate will be bolted to the granite, with many threaded holes for whatever fixture goes on top of that. I'm not yet sure how the holes in the tool plate will be drilled. While the obvious answer is "the router", I'll have to try it. My concern is it throwing chips everywhere before I'm ready to deal with that. That said, I don't think it should be an issue for simple drilling. Milling around the perimeter of a part with a side cutter will throw stuff everywhere, but that's not being done here. If it doesn't work out, the router can at least mark the holes with a center drill, then drill them manually.

There are some really low profile vice-like... things that attach directly to fixture plates and seem really useful. There may also be provisions for more traditional vices, but the point is that it's very versatile. For other stuff, an MDF sacrificial sheet will sit on top.

I really appreciate the YT Intro to CNC milling videos produced by JBWorx: https://www.youtube.com/@JBWorx/videos