![]() |
#1
|
|||
|
|||
Pulling whatever is left of hair and no solution
Hello fellow MechMaters,
Started cutting some circles, and noticed they are not the size they are suppose to be. Circle is real circle, not ellipse, but it is a bit short in diameter. Tried to re-calibrate at the length of 80" and got same bad results. For cutting I used 1/2" bit, just to be sure, there is no bending (conventional cut). Then, today I tried to do more investigation. Tighten all nuts and bolts on the Y axis, did some more cutting and got some strange results. For a purpose of measuring, I cut two lines with known distance between them. I changed the distance, and then measured what have I got. If the distance between the lines is 1.5" error is 1mm or 0.04". With 3" distance, I've got error of 0.8mm or 0.031". With distance of 5.5" error is 0.3mm or 0.012". With distance of 11.5" between lines error is 0.1mm or 0.004". With distance of 37.5" error is almost zero. So, conclusion is as distance increases, error decreases. I would understand that it is other way around, but not this way. Any suggestion where to look, what to do? Why is it happening? Thanks Milosh |
#2
|
|||
|
|||
How about acceleration ( or deceleration )? Is the error always negative, or positive?
|
#3
|
|||
|
|||
![]() Quote:
Error is always negative. In previous post, I should write that those test lines are about 3" long and they are parallel to each other. When they are cut, I just measure the distance between them. And the distance is getting bigger error is getting smaller. It has to be something related to cutting. I used V bit and engraved 3" circle and square. As much as I can measure, they are exact. Have no idea what may be the problem. Thanks |
#4
|
|||
|
|||
I don't have any great quick answers, so I'm going to start throwing ideas in case one of them triggers thoughts:
Are the parallel lines cut in the same direction, (left to right, left to right), or in opposite directions? ( left to right, right to left)? If you switch from conventional to climb does the error toggle from negative to positive? If you chuck up a piece of metal rod, are you able to wiggle the end of it 0.5mm or so in the direction of your error? What level of microstepping are you using with your drives? |
#5
|
|||
|
|||
Quote:
Tried before to cut circle upcut, and got positive error. Hmm. Wiggling with a rod. This is tricky one. I have 203V drivers and they loos their lock after 0.7 sec. In a state like that, (not locked) I am able to move a bit of gantry or Y car. What is moving is a timing belt but not the motor. One side is getting tense while other is getting loos a bit. Question is how tight timing belt should be? Maybe it is stretched a bit. I can try to tight it a bit and check it out again. Maybe, this is a problem. ![]() Microsteps: I didn't calculate steps or changed anything. I just used the Mach calculation and entered it into the motor settings. Thanks |
#6
|
|||
|
|||
Use a couple of clamps to immobilize the gantry and car. We're looking for slop in the bearings, motor mount, z axis, belts, etc. If those two can't move, and you get a wiggle, you'll have a place to start hunting.
If the parallel lines are cut in opposite directions, it will tend to maximize any offset due to rotational forces. If you go from conventional to climb and the offset goes the other way, or if you cut the same direction and the offset goes away, it tends to confirm that the tool is wiggly enough to offset by 0.5mm. Then the question is why, which gets to that list above. Timing belt backlash is certainly a possibility, depending on type of belt. The fact that your circle size changed when you went from conventional to climb tends to confirm that something is allowing some slop, since you aren't removing enough material to really justify tool deflection. Any of our experts have any ideas here? |
#7
|
|||
|
|||
It is strange that clamps didn't come to my mind, to use them to fix it. Tomorrow will try that as well as to tighten timing belt. Have to check what type of timing belt I ordered.
You know, what is strange is that, if there is a slope in any part, why it is not the same but it differ with distance? This is what I can not understand. Number of steps? Simply, don't know. Tomorrow is a new day. Will try to do something new. Good night and thanks for ideas. |
#8
|
|||
|
|||
Digger,
I have seen similar "oblong" results on a few of my machines over the years and below is the list of the things I found to caused this. - VEE bearing holder on the z-slide caused the tool to move under load in one direction - VEE bearing on y car or gantry slightly out of adjustment which pushed the car "up out of plane" under load - Check the circle on different parts of the table.....I had a shim under the rail off once and caused the machine to have a "bump on the road" just enough to cause the circle to be out of whack on only one area of the table - hold down springs *TOO TIGHT* and bind in one direction and caused motor to slightly stall in one direction under load |
#9
|
|||
|
|||
To me this looks like rack non linearity.
In any case the rack pitch is used to calculate the steps. Doing calibration introduces parts of rack which may have error and may add up or deduct from total step count needed for that 80" move. Also what is used to measure 80" ... tape measures also have error non linear error... etc |
#10
|
|||
|
|||
Was busy lately and out of garage but I was able to do something. Tighten all timing belts and make more lines.
![]() Also I tried to measure the flex of gantry and a car. With the spindle collet almost touching the table, dial indicator about a few inches above it and applying force of 14kg I have a flex of 0.5mm. (I know that a unit for force is not kg but have nothing else to measure how much I pull). Is it too much of a flex? It looks like it is. Have to figure it out where it is coming from. BTW gantry and car were clamped. Hope, over the weekend will try to see is there any disbalance in collet or spindle as well. Keep you posted guys. |
#11
|
|||
|
|||
Is that 0.5mm flex the same in all directions, or does it have a definite orientation relative to the Z-slide? If no orientation, I'd be looking hard at the spindle. If oriented, look at the z-slide.
|
#12
|
|||
|
|||
So, here is update.
As mentioned before, first I tighten the timing belts. Think, it changed something. Then, when I cut those lines I realized that their width is not 1/2" but a bit more. I can not recall that I ever measured runout of spindle. Got nice dial indicator and measured that runout is about 2 thou. Got a piece of grounded rod, and measured. Here is funny part. I've got about 15 thou runout. at 2" from spindle. This is too much. Then I tried to figure out what is a problem. Is it a collet or rod? Played a bit with it, and couldn't conclude anything. Then, I decided to change collet nut. Noticed that collet is sitting differently and the runout is about 7 thou. Put back original nut, but this time more carefully, and measured it again. About the same runout. Then, I cut same file, same lines. Width is about 0.2mm wider as the distance between the lines. So, my conclusion is that collet didn't seat properly on the nut and it induced that bit runout. Spindle is 3hp air cooled. China made, as well as collets and nut. Will get some better collets and nut, and will try what will be the runout with them. Beside that, have to see what are the options regarding the runout of the spindle taper. Hope that this story will help somebody who is trying to figure out what is wrong. Another place to look into. Guys thank you very much for all your comments and sugestions. As always, MM community is great one. Milosh |
![]() |
|
|
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
a solution for gaining 10 cm more space on the z axis | equilibrarte | 10. Base Table | 9 | Wed 17 December 2014 00:07 |
Compression spiral LEFT hand bits | danilom | Cutting various materials - bit selection, feeds, speeds | 12 | Sun 25 September 2011 01:10 |
Can the X,Y zero be moved from right to left front? | Hans | 10. Base Table | 9 | Wed 08 June 2011 22:34 |
Keyboard/pendant solution | Art | Human Interface Devices (HID's) | 7 | Fri 19 June 2009 06:41 |
Bob CampBell's control box components and full box solution | Gerald_D | 70. Control Systems | 7 | Sat 18 August 2007 18:38 |