Fairly experienced user having issues with a self-built Sphinx 1050 machine. I have successfully calibrated and trammed the machine to what I consider satisfactory (I have a background in precision metal machining, so I'm likely on the picky side). I then tested the accuracy by cutting a 100mm square, and achieved results of 100.05 along the X axis and 100.07 along the Y axis, which is accurate enough for the work I plan on doing. Then I start cutting my first project, which is hexagonal (6 side) shaped. The hexagon is aligned such that the points are on X- and X+, with flats on Y+ and Y-. This means that the "north" and "south" segments of the hex are cut only by travel of the machine in the X axis, while the other 4 segments are a combination of X and Y travel. When the project is complete, I use calipers to measure the "flat to flat" dimension in all three directions. The Y- to Y+ distance is about .75mm smaller than the two "diagonal" measurements, with those two dimensions within .05mm of each other. Compared to the drawing dimensions, the diagonal dimensions are about .5mm under-sized, and the Y to Y dimension is about .25mm over-sized. As a reference, the hexagon is about 100mm from point to point, and roughly 88mm from flat to flat. I'm using Carbide Create for the drawing, then OpenBuilds software for both G code generation and also program execution/control (BlackBox 32). Can anyone help me with what the problem might be, please?
How did you calibrate your steps/mm? You need to do it over the longest distance you can accurately measure. You also need to take any backlash out of the measurements. Calibrating your cnc Alex.
Calibrated per those very instructions. I used 250mm in Y and 600 in X (longest measuring tool that would lay in the bed). No backlash in the measurements.
Quote: Compared to the drawing dimensions, the diagonal dimensions are about .5mm under-sized, and the Y to Y dimension is about .25mm over-sized. If that's the case, I can't get my head around how... {quote} Y- to Y+ distance is about .75mm smaller than the two "diagonal" measurements. Can you measure the 3 "point to point" dimensions as well?
Here is a drawing of how I have the hex oriented...X+ is to the right of the drawing and Y+ is to the top. There are the three dimensions for flat to flat and three for point to point. One thing that I find odd is that the flat to flat have more range (.045") than the point to point measurements (.021"). Given that the point to point measurement is geometrically 1.1547 times the flat to flat, I would have expected the results to be worse by about 15%. Also (potentially) of note. This hex has another hex pocket inset 3mm inside of it. The inset dimension is pretty accurately 3mm on all 6 sides and, as a result, all flat to flat and point to point dimensions more or less mirror the pattern of the outer profile. Lastly, the hexagon, as drawn, is 4" from point to point and 3.464" from flat to flat.
Check the grub screws in the flexible couplings are nice and firm - especially the Y axes. When a machine is under load, there is the possibility of some slip in those couplings which won't necessarily show up during a calibration test. After that, try cutting a circle.
What am I looking for when cutting a circle? I can tell you that I noticed that cutting a circle at 300mm/minute or higher, the circle came out oval shaped. Also, what are you referring to as grub screws? Obviously the flexible couplings are flexible, but how much play should they have?
Grub screws = the small screws fixing the coupler to the motor shaft (make sure it is gripping the flat section of the shaft) and the leadscrew - the latter is the most likely to slip. 300 mm/min is not fast. It does sound as though something is slipping. When you noticed the oval "circles" was the longer dimension exactly aligned with the X or Y axis (more likely a calibration issue) or at a slight angle (backlash caused by something slipping - motor is turning but gantry is not moving - "catches up" when there is a change in direction). Alex.
Yes, they are definitely easier to tighten on the flat of the motor shaft than on the lead screw itself. I'll check that after work. I agree that 300mm/min isn't fast at all. However, slowing down to 100mm/min seems to remove most of the oblong. The project I'm trying to cut has a total of 6 circular pockets that are 5/16" diameter and 3/32" deep. The pockets would come out oblong at an angle as opposed to in 1 axis, and also weren't consistent as to where they "bulge"...each looks slightly different. Based on what you're saying, I would assume that's a sign of slippage. If the slippage isn't in the flexible connectors, what is the next place you'd suggest looking?
Motor couplers is probably the commonest cause, but there are many other things it could be. Only someone standing next to the machine will be able to spot them as the amount of movement is usually small. With the steppers powered on, starting at the bit, wiggle things with reasonable force looking for anything moving that shouldn't. Alex.
I checked the couplings and other movable parts and made sure everything was tight, and didn't seem to have any sloppiness. I ran the item again and got the same results. Picture of project attached (dice box). I then measured the 7 hex shaped pockets and they display the same measurement error as the overall hex cutout. The measurement is smallest from top flat to bottom, while the two diagonal measurements are larger and roughly equal to each other. These results don't indicate slop or slippage to me, they indicate everything is doing what it's told. So, drawing error or G code error (neither very likely) or calibration error. Here's what I'm thinking. I used a cheap ruler to calibrate, 250mm across Y and 600 across X. Let's say the ruler is off by 1% equally along the length, it would be more out the further along you go. So using it to calibrate would mean both axes are wrong, but that X is "more wrong" than Y. This would result in two conditions, both of which I'm seeing. First is that neither axis would be accurate to what is programmed. Second is that neither axis would match the other, therefore resulting in distortion. I'm going to grab a 500mm Starrett rule from work to calibrate with. I'll either find out I had an error or that I'm right on calibration, but I'm sure the Starrett can be trusted. Does the consistency of my error give you any other ideas besides what I mention here?
Ok, several things here. I brought home a 600mm Starrett rule and it matched up well with what I had already calibrated. Went back through Y and X calibration steps again and settings were good. Checked flexible couplings and they were tight to both the flat of the stepper motors and the lead screws. Next checked for motion and play wherever I could find it. The left-hand Y axis lead screw could be wiggled horizontally, so I re-tightened the collars against the spacers and bearings and removed that play. Continuing to check all axes and places for potential motion led me to the long X axis rail. The Z axis assembly can be wiggled side to side on the rail, along the X axis, fairly easily by hand. I checked for longitudinal wiggle in the lead screw similar to what I found in the Y axis, and there was none. I feel that this means there is wiggle in the connection between the X axis lead screw and the anti-backlash nut. How do I go about adjusting that? While I was at it, I can also grab the spindle and it has a slight bit of play as well. Essentially it feels as if it pivots a bit rotationally around the X axis C beam, if that makes sense. My guess is that can be taken care of by adjusting the eccentric spacers and tightening the rollers. When it comes to adjusting rollers, I know that the eccentric spacers get turned with a wrench in one direction or another to loosen or tighten the connections. One question I have is how do you tell when you have the right tension on the wheels? Second, what do you do if you feel you've tightened the eccentric spacer as much as you can and there still seems to be some wiggle in them? What else could lead to the play along the X axis, and what could cause the Z axis assembly to have some play back and forth in the Y-/Y+ direction? Anything else I'm missing that needs checked for tightness, play, perpendicularity, or parallelism?
First check that bolts holding nutblock to carriage is properly tight, if left loose it can slide there a little. While in there you can feel for movement of nutblock to leadscrew too and if needed...
Tonight I took the X axis C beam off of the Y axis mounts, convinced that I needed to tighten the X axis nut. Upon removal of the beam and pulling the Z axis assembly from the beam, I realized the X axis nut mounting screws were loose. (They were not loose upon installation). I tightened them down, then also snugged the grub screw on the nut for good measure. I then set about setting proper tension of the Z axis assembly wheels that ride along the X axis, as I felt the right-hand side (X+) turned too easily (i.e. I could turn it by hand but it would not move the assembly down the beam). With the C beam still disassembled, I felt that I had the wheels at the right tension. Now, with the ballnut mounted tight, tensioned to the lead screw well, and the wheels having what feels to be the proper tension, I reassembled everything. Now the X axis will not turn! I set the machine control to .01" and the stepper motor attempts to move, but does successfully move (rotates slightly forward and rolls back) and the X axis lead screw does NOT rotate. I even powered the machine down so as to unlock the steppers, then attempted to turn the lead screw by hand and it will not turn. Thinking perhaps I had overdone it on tightening the grub screw on the locknut, I loosen it bit by bit but still get no motion. Then I turn my attention back to the wheels, wondering if they're too tight. As I had noted before, the Z axis assembly seemed to have some flex around the center-line of the X axis C beam, which I took as being loose wheels primarily. At this point, the front (Y-) wheels all turn fairly easily by hand, but not to the point of gripping the C beam and rolling. The back wheels (Y+) all seem very snug and don't turn by hand, though that could be related to the overall Z assembly just not moving period. At this point, I'm at a loss as to where I've gone wrong. What do I have too loose/too tight/out of alignment? I feel like I followed the assembly video quite well. However, the videos were quite descriptive about what piece goes where and what it attaches to, but VERY lacking in terms of the intricacies for accuracy. By that I mean that the videos did not suggest use of a square to check squareness, ruler to check parallelism, checking for slop along the way, how tight/loose various components (nut blocks, retainer rings, wheelse) should be, etc. Your group might think some of these things are straight forward, but my thoughts are around "rigidity", therefore my mindset is tightening things very tightly. Additionally, things that were tight at assembly have since been found to be loose. I can program CNC lathes long-hand (meaning from a paper drawing with no G code generator) and build hydraulic presses for a living, so I'm not exactly a dummy along these lines. Am I this inept?
I backed it out slowly and tested it as I went and the motor still wouldn't turn it. Backed it out until it almost fell out and was not in contact. Then got to thinking that maybe that was just as bad and that it needed SOME tension, so I slowly worked it back in and it wouldn't turn at all still. So do I maybe need to back it all the way off, take the lead screw back out, loosen the nut, insert the lead screw, then work up to proper tension on the nut? Apologies, I'm BY FAR a more visual person, so I need very specific written instructions.
It sounds to me like you are doing it right. Did you try hand turning it both directions or just one? On time on my lathe, I dropped a screw into the channel the wheel runs on and I did not notice. When I went to jog the machine it made a horrible noise and would not move. It took me awhile to realize it was a physical obstruction causing the binding. It is probably not your issue, but it may be worth a quick look.
Remove the X stepper and the flexible coupling. Remove the bearing{s} from both ends of the leadscrew. Loosen the lock collars from both ends of the leadscrew and ensure they are free to slide along the leadscrew. Now grab the Z assembly and push it back and for along C beam - it should move with some freedom. If it does not, you have a problem there to investigate. If it's ok, then try turning the leadscrew by hand. If you can't , then you have a nut block problem. Check bearings run freely while they are out.
Between yesterday and today I got everything back in order. It looks like I had tightened the anti-backlash nut too much trying to ensure there was no play, which pinched the lead screw. Everything is tight with no wiggle. Re-calibrated the X and Y axes according to the methods mentioned above. I then generated G code for a 2" square and cut it 7 times using the same exact material, tool, and cutting parameters. Below is a chart of the measurement results. The measurements from X- to X+ ranged a total of .004" and averaged 2" on the dot, which I find acceptable. The measurements from Y- to Y+ ranged a total of .009" and averaged 1.9937", neither of which I'm pleased with. I then re-ran the hex-shaped project referenced above. The measurable results were a bit better than what I've been getting, but still very much in the unaccebtable range to me. The measurable dimensions from flat to flat of the hex are as follows. Top to bottom (Y+ to Y-): 3.471", top left to bottom right is 3.485", and top right to bottom left is 3.465". At this point it seems I have a Y axis problem more than an X axis problem. My two questions at this point are: 1. Would you agree that my X axis numbers above (.004" range, 2" average) are acceptable for this machine? 2. What am I missing on the Y axis? I feel I've checked every moving part possible and find no signs of wiggle, certainly nothing to indicate it being different than the X axis. I'm not sure I'll get any shop time in tomorrow, but may do some experimenting in Autocad. At this point. the .020" total variation I'm still getting on the hex cut seems MASSIVE. But that hex shape is twice the size of the squares I was cutting today. I may toy around in Autocad extrapolating the ranges/errors I was seeing in the square cuts across the hex profile and see if it draws out with the same variation/range I'm seeing in the hex cut. Square X- to X+ Y- to Y+ #1 1.998 1.994 #2 2.000 1.999 #3 2.002 1.991 #4 2.000 1.995 #5 2.000 1.995 #6 2.000 1.992 #7 2.000 1.990 Avg. 2.000 1.9937 Std. Dev. 0.0012 0.0030
I also cut a circle yesterday while toying around. It was 50mm diameter and I measured it with a pair of micrometers at various quadrants. There was variation all the way around with a noticeable high spot and low spot. Obviously that tells me that something is out of whack, but how does this lead me to knowing where to look?
Is the high spot noticeable in a picture if you were to provide it and label the axis directions? Also, are you doing both a roughing and finishing pass? If you do a roughing and leave about 0.3mm to cut with a full depth finishing pass, you will probably have more accurate results. Sometimes things like knots and other changes in density can "grab" the endmill and cause it to deviate slightly. Especially if using smaller diameter endmills.
It is not visibly noticeable, really. Just a single profile pass with 1/4" end mill, 3mm depth per pass.
That picture @Giarc suggested, with the axes and the high/low spots labelled will help us tell you where to look. Alex.
As @Giarc says, you need to account for tool deflection to achieve the levels of accuracy you are seeking. This means addressing the machining strategy in CAM - it is standard practice for accurate work to use a roughing cut to remove the bulk of the material followed by one or more finishing passes. Cutting a slot around a profile is a roughing operation - it engages 100% of the tool's diameter and generates high cutting forces at right angles to the travel. Materials like plywood will cut slightly differently depending on the grain direction and natural varation in wood density, so the effect will vary along the cut. I would humbly suggest trying a two or three step machining strategy: Remove the bulk of the material to your finished depth (using how ever many passes it takes), but leaving (say) 0.25mm / 0.010" of stock material (roughing pass). Then make a contour cut, at the full depth, to the finished size (finishing pass). Repeat the finishing pass with no additional stock removal (spring pass). This should substantially remove tool deflection from the equation. That is typical of tool deflection - slower speeds generate less cutting force.
I will try the roughing vs. finishing/spring passes. That is not a new concept to me from my days of programming/setting up/operating CNC lathes in metalworking (hogging away material with one tool and then using a smaller tool for finish pass). All of the trial cuts I've made have been with 1/4" end mills, three styles (up cut, down cut, compression), of what looks to me to be a pretty good tool (Spe-tool or something like that). My first CNC router was a cheap Genmitsu 3018 that I had to assemble, and it went right together and did not have this sort of variation; it was a MUCH less rigid machine, as well. So, next chance I get in the shop I will try the rough/finish method. Although I still maintain that my square cut test showing less variation in X than in Y indicates that I still have a Y axis issue. I also watched a video this weekend about using the 3/4/5 method to check squareness of X axis to Y axis. I will do that when I get a chance as well. Would being out of square potentially lead to the measurable variation I'm seeing? In spite of all, I'm running pretty light parameters overall (1/4" end mill, 3mm depth of cut, ranges from 150-300mm/min feed, 50-100mm/min plunge, 12,000 RPM) in a variety of wood (cherry, walnut, maple, cheap plywood). I still maintain that the machine is not performing to what I would expect it to be from a repeatability and accuracy standpoint. For now, here are the dimensions requested for the circle test, showing the actual measurements across the Y axis, X axis, and the low and high spots (measured with 1-2" micrometers).
