I believe these sets had only 6 passes to cut through 1/4". The ones on the right were 8 passes. I was being stupidly conservative with the Z lead in. If it was losing steps it would look like this and got a lot worse very quickly. I can hear when it misses a step because it sounds like a crashing hard drive. It just clicks clicks clicks then continues on.
So each pass was incorrect in exactly the same way? That does sound like backlash, or possibly the torque of the tool against the piece flexing the gantry in a way which differs depending on whether the cutter is cutting 'up' or 'down' -- have you measured the backlash? Incidentally, here's a great resource for anyone who does CNC: http://lcamtuf.coredump.cx/gcnc/
Yep, each one was off exactly the same way as the previous pass. I do not have a dial indicator to adjust for backlash, but from the previous runs, the accuracy was acceptable/close enough to be negligable. I am suspecting the controller board over running the axis. I was advised on CNCZone and the ShapeOko forums to turn the amps down to 75% to cut down on the line noise.
Well, resetting the board to 75% proved no difference. I was smart this time and used 3/16th foam board to test instead of the birch ply When I was jogging the Y axis I noticed some buzzing from the belt and when I went to investigate, the tensioner had pulled away about 1/16th of an inch (I could see in the saw dust where it was clear behind the tensioner bracket) I retensioned the belt and REALLY cranked down on the lock screw. It has been running on the foam board for an hour now and every side is exactly as it should be, and they are exactly as the plans had laid out. I still have SO much to learn about these CNC routers. I also fixed my feed rate problem. The CAM software I am using sets a default FR of 4 IPM, so no matter what I have it set to in LinuxCNC, the GCode is telling the controller that it will not travel at more than 4 IPM. I currently have it at 12 IPM over ride to 14.5 and she is singing along happy. ----- Back to cutting wood at 14.5 IPM and the bit isnt terribly happy (very buzzy which means its vibrating) but I dont think that is a speed issue. Ive got it spinning at 30k and if I slow the travel all the way down it is still buzzy. ON WARDS!!!
Great news -- there is a lot to learn. When I got my CNC router I got a wonderful result the first time. Then I made a configuration change, and my cuts became awful. After trying all sorts of adjustments I realized that because of the change I'd made my spindle was now running backwards -- hard to spot at 18,000RPM, but not what end mills are expecting!
I spoke too soon. As I was typing that out last night, I guess I forgot to double check the profile advance Z-Link and the machine decided to cut an 8" long, 1/4" deep path through the piece I had just made to start on the next one, which ruined the stock for that one as it now had a through cut, through it. I can probably repair the board and still use it with a little epoxy and saw dust to make a filler and binder. The good news is, the machine is able to cut through in a single pass, the bad news is it broke the bit when it tried to jog back home after I released the Estop and powered it back on.
Well, now you know how deep to you can cut I suspect the buzzing is the Dremel bearings. I have an older model Dremel, but my bearings didn't last long at 30,000 RPM. As the bearings wear, the run out on the bit will get more and eventually you'll start breaking bits. I like to make at least two passes, even if I can cut the material in one pass. I'll leave at least 0.05" for the final pass. This helps in two ways - first the load is lighter on the hold downs, so you are less likely to move the material, and second, the material left at the bottom holds the part securely as it goes around. On the last pass, the part will move a little when it gets cut out, but you will only have a small tab at the bottom rather than a tab that extends the entire width of the part. Awesome setup, good luck cutting the rest of the parts.
The epoxy repair worked! BUT it is my controller board causing the real issues. At anything over 25% amperage it just randomly drops the X or Y axis. I am tempted to either fix it (cheap, but will require buying a lot of small parts) or replace it with one of higher quality (more expensive, but more reliable)
Its based on the TB6560. Youve probably seen them on ebay. The cheap, chinese blue 3/4/5 axis boards.
That's frustrating, one of the reasons that after numerous issues and smoked boards that I now have separate driver boards for each axis. A couple of things to check: Make sure that the power supply isn't dropping out and make sure there is plenty of active cooling (fans). One other thing to check is the voltage rating of the capacitors. I was looking at a RAMPS 1.4 driver board this week that is rated for 35 volts, but I can see in the picture that they only used 16 volt capacitors so 35 volts obviously will not work.
That is my next step if the fix doesnt work. ODDLY enough I have the buffer chip already, so I just need a few resistors and a capacitor. Trying to find a controller with individual axis that will work with 4wire steppers. The only ones I have found are 5/6/8 wire and expressly say "NOT FOR 4 WIRE"
My G540 has been flawless so far, running at 48V and 3A -- of course that is more expensive. (I might have to drop the current -- my stepper motors get almost too hot to touch on long jobs). I use a 7.3A power supply for 3 axes. (this one http://www.homanndesigns.com/store/index.php?main_page=product_info&cPath=26&products_id=82)
I second the 540's.. I've built xylotex systems and ones with G540's, and the improved ability to deal with the midband resonance issue is enough to make the 540's worth it, even if they are overkill on the power side of things.
Small update. I took the pieces that were properly cut out and compared them to the plans. Everything is off by 1" or so. Something is wrong. So i made a 1"x1" calibration box in CAD, send the GCode to the router and ended up with a .95"x.90" box. That means that over a 9" rib span Im losing almost an inch of length, and 3/4" of height. I recalibrated it last night, and now I only lose 1/8" on a 12" travel. Still not good, but a lot better then losing an inch plus on the Y axis! I also added an auto-home to my controller so the results are a lot more accurate.
When you say you are loosing 1/8" is that repeatable or is it lost steps.... E.G. if you cut a square 100x over the same spot does its position or is the size just off... if size, you are good, you just are off in your steps/inch setting.. if it drifts, then you are missing steps.
I set the max (limit) travel to 12" on the Y axis, called it home (.01"), set a 12" ruler at the edge of the bed (0,0) and told it to find the LIMIT switch. When CAM read 12" traveled the bit was pointed right at 11 7/8" But, now that I think about it, that would almost make sense as the button full push is 1/8" and it stops as soon as it TOUCHES the button at HOME and LIMIT (very sensitive pushbuttons I guess). It is 100% repeated, and as I was not cutting anything, using the conservative 25% amp rating gives no controller-induced missed steps. I did the test several times, tweeking the lead screw ratio to try and get it to read the proper gantry travel. Im not aiming for .0001" accuracy out of a desktop CNC router, but when the CAM says "12 inches Y Axis" it would be nice for it to have actually traveled 12". The really odd thing is, both the X and Y axis have the same travel and values, yet there is no inaccuracy with the X axis. I have 12" marked out on the bed and the bit will go from .01" inside of one line to .01" inside the next (where it hits the limit) and the CAM reads 11.98". I will have to play with it some more tonight.
Ok, so, 11.75" travel is 11.76" actual. I will take .01" accuracy. It DOES help to use the correct tooth count on the pully. 18 teeth is not quite the same as 20. ONWARDS TO CUTTING PARTS!
