Hello - I am getting some odd results when tramming. I am using a surface plate that I level first level to the machine axes with a dial indicator that I magnet onto the spindle. It is within .001 between the 4 corners of the plate.
Here is a photo of the tramming tool from the right (within .002):
I feel like these should be giving the same reading, I am not sure what I am missing here. My first thought was that it was a bad collet, however I checked the runout on the spindle and on a blank bit with a dial indicator and it is within a thou on both.
Anyone have any ideas about what I am doing wrong/what could be happening?
That means your Z needs to be shimmed in the Y direction; you have “nod”.
There are two things to check:
Put a test indicator in the collet and reference a precision square as you move Z up and down. Any deviation is fixed by shimming between the carriage and the Z lift (the red part)
Put a precision rod in the collet and mount the test indicator on a fixed point. Any deviation here is fixed by shimming between the Z lift and the spindle mount.
Yes I did. 180 degree spindle rotations should be identical if the dual dial indicator rig is working properly unless I am missing something. I thought it was working properly, and as you can see in my original photo it is close to perfect tram one direction, but after you rotate 180 its off. I didn’t check the full revolution the first time I trammed with this dial indicator holder, only x and y.
I just took another pass at it, my first solution was to just use only one side of the tramming dial indicator holder. I think I got some better results, however the x-direction cut results are pretty confusing. Appears to be out of tram in one direction at the bottom, and out of tram in the other direction at the top, fading in between the two.
X-axis test cuts on the left, y-axis on the right. 2" surfacing bit, 1.9" stepover to maximize visibility of the tram. 80ipm .01 doc if it matters, running slow to minimize machine deflection.
Thanks for the tips @djdelorie, I will try those methods as well.
Anyway, this is a solvable problem. Just have to find the culprit. On the forum so far, I’ve seen loose bolts, needing a shim, measuring with a square that wasn’t actually square, loose/dirty collet, badly machined collet, bent tool and probably more I can’t recall.
Thanks David, not sure exactly but probably a couple thousand hours.
I am taking apart the x axis and z axis right now to see if I can find anything loose. Nothing yet. Spindle and collet runout appear to be within a thou.
There’s nothing wrong with the gauge that’s how it’s supposed to work if you had .000 in both directions no shims would be needed and move on to the next step
I disassembled and reassembled the Z axis down to the screws that attach it to the x-axis bearings. No loose screws. Trammed to the point that the indicator read within a thou for a whole revolution.
Did my tram test file again, pretty much the exact same results. Going to see what avid recommends on Monday, thanks for the help guys.
The tramming process is what makes them identical. First change one dial’s setting to read zero, then rotate the tool 180 deg and change the other dial’s setting to read zero at the same spot. Then adjust the machine so that the two dials match.
I’m no expert, but it looks like the spindle is slightly “nodding” with Y movements. I suppose it’s possible that the linear rails could deflect. If that’s the case, tramming at one Y position and checking it at other positions would yield different results, something you can test.
As a sanity check, it makes sense that the other direction wouldn’t show this, as the moving mass is lower in that direction, the support beam is bigger, and there are two rails. You still might see the effect if you trammed near min X and checked it near max X.
You might just be pushing the limits of the machine’s accuracy.
@JeffsDesign I think I was misinformed about the proper way to use that tool, sorry for the confusion. You are correct, adding more shims to the bottom got me closer
@Stephen That method would work better. I think that my single dial indicator method trammed it very close to flat at this point. Your method would account for the error I was seeing between dial indicators as they would both be measuring off a common point on the surface plate instead of based on their unengaged zero value, makes sense.
I agree, based on that surfacing result there is definitely some uneven nodding. Fully disassembled the Z axis down to the points where they attach to the X-axis plate, no loose screws anywhere. Going to run some more tests to see how it performs in different areas of the bed as you suggest.
I hope I am not hitting the limits of the machines accuracy. I can barely grab the tramming error with my fingernail, so not the end of the world if that is the case. If it is flexing, it seems like a possible solution could be adding mass on the opposite side of the Z-axis to counteract the twisting action on the gantry.
@djdelorie do you see how the tram is uneven throughout that photo? Did you have similar results that you managed to fix? It seems to go from bad tram in one direction to bad tram in another direction throughout the cut.
If your “nod” is different depending on which way you’re moving, something is loose. Think of which way the bit will be pushing the spindle around when it cuts, and see if it corresponds to the errors you see.
Also, check the nod at both ends of the gantry, in case one end is twisted. I doubt this is the case, but it’s easy to check.
Also, while the system is on but the spindle is off (i.e. the steppers are “holding”), grab the spindle and try to move/tilt/nod it. That might expose some looseness in the system.
This has been mostly solved pending further testing after discussing with Avid support. It looks like I had a 1/8" valley in the machine frame, leading to uneven tram. I have gotten much more consistent results after re-leveling and reinforcing the machine.
I added extra supports between adjacent legs, and added diagonal members to every leg/bed connection. The amount of machine vibration I feel when the cnc makes turns is significantly lower.
Using the surface plate on the spoil board maybe be giving you false readings. There isn’t any guarantee that both sides of the plate are parallel to each other. This can be tested by rotating the plate around 90 degrees and then 180. Surface plates give you a single flat plain and are expected to be leveled.
