[mrscheider]

All,
I'm seeing something new in my Taig CNC mill and want to verify the required actions. Using Mach3 and a 1/4 inch flat end mill, I find the edge of my stock and set the Y axis to -0.125. I jog over to the positive side of Y and do the same on the X axis. Then if I command the Y to move to -0.125, it's off slightly. Machs DRO reads -0.125 but after finding the edge again, the Y axis now reads -0.1261. After reseting I noticed the following:
If I move Y from -0.125 to -0.5 and use the command to move back to -0.125 it looks accurate, if I move Y from -0.125 to 0.5 and again move to Y-0.125, it's off by 0.0011. Furthermore, if I start Y at -0.125, move to 0.5, then move to -0.5 and return -0.125 in a continuous motion, it's accurate.
So basically once I move the Y table towards me, it doesn't move away from me for the first 0.0011 that the motor turns. This is not related to something I changed recently in the software or to the machine. Last week all three axis were accurate, last night, I bagan seeing the error in my finished parts and measured tonight to confirm. X and Z appear accurate. What adjustment would be nessesary to resolve this?
Thanks so much for any sugestions.

Mike

[TOC Admin]

that is a backlash problem......though .0011 is not real bad but you can probably adjust at least half of it out of it...

[escott76]

All,
Using Mach3 and a 1/4 inch flat end mill, I find the edge of my stock and...

How sure are you that the end mill is .2500" ? And how are you using the tool to find an edge? I'm not saying you don't have any backlash in the system, but you have to account for all variables.
.0011" isn't all that bad in a lead screw machine anyway. You may be able to get it lower, but be aware that the tighter you set the lash, the faster things will wear, and you'll be right back to where you started in no time.

[mrscheider]

Hello, thank you for the replies. I tried my best to work in the variables because of the inaccuracy being so little. I measured the tool with a dial indicator trying to find the widest part of the flutes and that read 0.2497
When finding the edge, I used a backlight until the edge of the flute blocked the shadow. Now, I know thats not the most accurate, but regardless of how close to the edge I do get, I should still be able to return to that same spot.
And while 0.0011 is a very small inaccuracy, it shows up quite noticably on the part because I'm coming down Z 50%, flipping the stock and coming down the other 50%. Yes, this method in itself introduces more variables but keep in mind, It's been reliable this long using the same method, G-code and material. This a completely new issue. To give you an idea of the amount of error, it's about the thickness of an exacto knife blade. Very small indeed.. but if the surface is expected to be one smooth plane, that little change in thickness is very noticable and I have to admit.. a little annoying.
The problem is then compounded by having to sand off that error which changes the thickness from one part to the adjacent part mounted next to it. Again, all expected to be the same level plane.
So should I try to work out this error or is it too small already to improve without hastening wear? A good point which I didnt consider, btw.. so thank you.

Regards,
Mike

[escott76]

Hello, thank you for the replies. I tried my best to work in the variables because of the inaccuracy being so little. I measured the tool with a dial indicator trying to find the widest part of the flutes and that read 0.2497
When finding the edge, I used a backlight until the edge of the flute blocked the shadow. Now, I know thats not the most accurate, but regardless of how close to the edge I do get, I should still be able to return to that same spot.
And while 0.0011 is a very small inaccuracy, it shows up quite noticably on the part because I'm coming down Z 50%, flipping the stock and coming down the other 50%. Yes, this method in itself introduces more variables but keep in mind, It's been reliable this long using the same method, G-code and material. This a completely new issue. To give you an idea of the amount of error, it's about the thickness of an exacto knife blade. Very small indeed.. but if the surface is expected to be one smooth plane, that little change in thickness is very noticable and I have to admit.. a little annoying.
The problem is then compounded by having to sand off that error which changes the thickness from one part to the adjacent part mounted next to it. Again, all expected to be the same level plane.
So should I try to work out this error or is it too small already to improve without hastening wear? A good point which I didnt consider, btw.. so thank you.

Regards,
Mike

I just mic'd a scalpel blade (.016") and a single edge razor blade (.009") which are both thinner than an X-acto blade. 0.016" is WAY more than you are talking about, and no offense, but it's gonna be tricky to feel .0011 much less have it be the thickness of a blade.
Backlash is an undeniable fact of life for machines with lead screws like this. Without pre-loaded nuts, you are going to have a really hard time getting rid of it. It's also not the only factor in creeping errors like this. You don't mention work holding, and how square your method of work holding is to the rest of the machine. It all stacks up.
If it makes that much of a difference find a better method of cutting the part in one shot, otherwise make plans to get rid of the extra stock in deburring/sanding/post finish.
Chasing .0011 of backlash on a Taig is going to be a game of diminishing returns, and if your reports of offset are in any way accurate you have bigger problems somewhere else.

[TOC Admin]

Just for reference ...Check that your Stepper Motor Coupler is tight...could be slipping on the shaft a bit

[mrscheider]

escott76,
You are absolutely correct! My x-acto blade measures 0.015" and it's ever so slightly larger than the amount of error the machine is cutting. I was obviously calculating this wrong.
In an effort to figure out where I was miscalculating, I realized that I went to the negative side of the edge and returned from the positive side which took up the backlash I was supposed to be counting. So what I did this time was set the tool to the edge. To make it simpler (for me) I called this position 0.
Now, I jogged to 0.5 manually, then told Mach3 to move back to 0. Again it's off the described amount.
THIS TIME, I manually jogged back to where the edge should be, being careful not to pass that beginning point (passing and returning to this point originally caused me to mis-calculate the distance) and now I'm seeing 0.0135. That is the correct amount of inaccuracy Im seeing and I apologize for not having it together before posting.
So what do you guys think? What should I try to resolve?
Thanks again for the help getting me straight!
Mike

[Hirudin]

I'm sorry, maybe I'm not reading this right, but first you said the machine was off 0.0011 ("eleven ten-thousandths") now you're saying it's off 0.0135 ("one hundred thirty five ten-thousandths").
0.0011 is pretty small - roughly the thickness of... I couldn't find anything. Thinner than run of the mill printer paper.
0.0135 is rather large

As for getting rid of the backlash... If it's only 0.0011 you might be able to adjust it away, but it might not really be worth while (because of the additional wear). I'll let the experts chime in on this.

If it's 0.0135 check out the instructions for setting up a Taig on cartertools.com ...
http://www.cartertools.com/millset.html
http://www.cartertools.com/millset2.html (<- look on the bottom of this page for a link to another page)

[mrscheider]

Hello Hirudin,
No, you're reading it correctly. LIke I said, I mis-calculated the first time when measuring. So forget the original number.
Lets put this whole thing another way, shall we?
If I set my mill to jog at 1 ipm and jog the Y axis, the DRO's show a change of 0.0135" BEFORE the table actually moves at all. Last night I experimented and it's basically anytime I move Y in an opposite direction.

My only question is what is causing this? Is this backlash? If so I can work with the backlash adjustment screws to try to resolve. But can backlash come out of the blue like this? One night its fine, the next night its not?

Thanks,
Mike

[escott76]

Hello Hirudin,
No, you're reading it correctly. LIke I said, I mis-calculated the first time when measuring. So forget the original number.
Lets put this whole thing another way, shall we?
If I set my mill to jog at 1 ipm and jog the Y axis, the DRO's show a change of 0.0135" BEFORE the table actually moves at all. Last night I experimented and it's basically anytime I move Y in an opposite direction.

My only question is what is causing this? Is this backlash? If so I can work with the backlash adjustment screws to try to resolve. But can backlash come out of the blue like this? One night its fine, the next night its not?

Thanks,
Mike

It sounds like backlash, but it's hard to determine the cause of the error not having things in front of you to test. To tell where things are coming from you need to start using your DTI and measuring what is moving. Check for play in the saddle. Set it up so you can measure the deflection of the saddle, and try to twist it from above (as if you were twisting things about the Z axis. See if you've got movement there as that will be the first thing to correct. Then start looking into lash in the screw/nut itself.
As to weather or not this can happen in one night sure. Things come loose all the time. Vibration, crashes, all kinds of things can cause something that was once tight to now suddenly let go.
Measure carefully to find where the source of the problem is and then eliminate it. Others have posted guides as to how to adjust and set up the Taig.

[mrscheider]

Sounds good.
Excellent advise. Thank you Escott76. I will search for a few threads of similar issues and see if I can get it worked out.
Really, thanks a million everyone!
Kind regards,

Mike