[lcvette]

Can you do a quick video of you setting your axis steps per method? I do mine by just jogging the indicator to the spindle until it barely touches, zero the indicator face, zero the dro in mach, go to steps per setup for that axis and set the distance to travel for nearest whole number just under the max travel of your indicator. For example my indicator plunger safely goes 1" so I use a 1.000" move to set steps per.

Be sure backlash comp is off and only measure in the same direction you last jogged to eliminate any backlash mucking up the readings. Also dinnertime setting your speeds slower eliminates overtravel of the axis into its backlash area past where the motor stops turning. I usually tune using 15ipm to ensure this.

Let me know how it goes!

[onocyclone]

What you said above Is exactly what I do. I've got that down to a tee.

I have no idea. When I code with G91 it is exactly the same results.

My only suspicion is the spindle or the spindle adaptor. The tool is capable of plunging down and allowing a ground dowel to be pressed into it. I will try again soon with a different end mill.


Also in the tool table I may have put the diameter of the tool in the wrong box. Are there any offsets in that tool table ?

[lcvette]

Not unless you are using the tool wear settings. But your off a pretty good bit. if your all trammed square to the table and the column is perpendicular and not angled, then it's either in your steps per, backlash comp numbers or tool diameter error (but this would show equal error on both sides). So..... Not sure what else to tell ya except try rechecking your backlash figures and verify your going about it all the right way. The fact that your of different amounts on both x and y tells me you are getting incorrect figures during your setup. Bad numbers entered in will yield bad numbers coming out in the work.

You may also try checking the sherline half pulse mode, I ran that checked when still on a parallel port.

[onocyclone]

The machine is trammed very well. So much so it took a long time to square the head up. Within 0.01mm I would say. And the squareness of the table is bang on. I've machined a square block and without Moving it I ran the DTI along it and it was perfect. No issues what so ever.

[lcvette]

The block dimensions were perfect? Than I'm a little lost, what's the issue? I thought it was your parts were out of spec in x and y axis?

[onocyclone]

Parallel accuracy not dimensionally accurate.

[lcvette]

Well, as mentioned above, maybe loosen your Gibbs a bit to make sure it isn't Stiction, and reset your steps per as described above. Then test another part.

[louieatienza]

Parallel accuracy not dimensionally accurate.

Haven't tried conventional cutting for finish pass yet to see if that helps?

[onocyclone]

Hi guys,

I've got something interesting to report,

I machined some small parts with a 2mm end mill and these are the results:

Attachment 273460Attachment 273462Attachment 273464

It's quite interesting how internal features are more accurate but all of it is programmed to climb mill !!

Any ideas guys ?

I'm going to investigate the 12mm tool I was using before, see if it is in the collet correctly.

Cheers,

[LeeWay]

Cheap calipers will not be very accurate repeatably. I would use a micrometer, height gauge or something with a little better resolution. Even good quality drill bits can be used sometimes.
Also check that the end mills are indeed what they are spec'ed. I often find that much discrepancy between nominal and actual size.

[onocyclone]

I may have found the issue guys !

Some bad damage to the tool !

[louieatienza]

Hi guys,

I've got something interesting to report,

I machined some small parts with a 2mm end mill and these are the results:

Attachment 273460Attachment 273462Attachment 273464

It's quite interesting how internal features are more accurate but all of it is programmed to climb mill !!

Any ideas guys ?

I'm going to investigate the 12mm tool I was using before, see if it is in the collet correctly.

Cheers,

This case is a bit inconclusive since you're technically slotting.... so the tool is "supported" all the way around. You sure the internals are climb cut? That would mean counterclockwise for internal pocket.

[lcvette]

What were your backlash figures again? It very well may be that unless you can get your backlash down some more that you'll have to contend with the error. While climb cutting, the tool can pull the work by a portion or whole amount of the backlash. And when you add that to your backlash comp, sometimes it can overshoot or undershoot creating inconsistency. Same reason why backlash comp is not a good idea on the z axis (weight of the head takes up the backlash).

I have issues too, but generally when starting a cut it will pull the tool into the part at the lead in leaving a little cut mark. I just ordered the double ballnuts to try and eliminate the issue. At some point you will hit the wall where you need to add higher quality parts to get better results. If the error is repeatable, you could try adjusting the backlash comp to get it as close as possible to split the difference of backlash slack take up from tool pull way slip and controlled backlash comp.

That off course will create other issues, but you'll have to play with it to get it narrowed down as close as possible.

Ideally, the less backlash you start with the better off you are!

[louieatienza]

What were your backlash figures again? It very well may be that unless you can get your backlash down some more that you'll have to contend with the error. While climb cutting, the tool can pull the work by a portion or whole amount of the backlash. And when you add that to your backlash comp, sometimes it can overshoot or undershoot creating inconsistency. Same reason why backlash comp is not a good idea on the z axis (weight of the head takes up the backlash).

I have issues too, but generally when starting a cut it will pull the tool into the part at the lead in leaving a little cut mark. I just ordered the double ballnuts to try and eliminate the issue. At some point you will hit the wall where you need to add higher quality parts to get better results. If the error is repeatable, you could try adjusting the backlash comp to get it as close as possible to split the difference of backlash slack take up from tool pull way slip and controlled backlash comp.

That off course will create other issues, but you'll have to play with it to get it narrowed down as close as possible.

Ideally, the less backlash you start with the better off you are!

Climb cutting PUSHES the tool AWAY from the cut line. Conventional cutting PULLS the tool INTO the cut. Which is why it's called conventional cutting; when you work manually you want the tool to "pull in" to take up the lash. On the CNC it's the opposite; the tool will push away from the line to take up any slop, this leaves a slightly PLUS tolerance part. The finish pass takes off very little and will not pull the tool in significantly. When I rough, I always use climb cutting, and rely on the "spring" of my tool and machine so any error is on the plus side. I'll then conventional cut to tolerance if necessary, but since I mostly machine aluminum, I can usually climb cut the finish pass and it leaves a very smooth surface.

Probably the only condition where a climb cut will pull is an inside corner, since the chipload literally multiplies. But if you have CAM with trochoidal toolpaths it will "peck" into corners helping eliminate that chipload "overload."

[lcvette]

I respectfully disagree, climb cutting pulls the tool forward, that's why you should manually conventional cut to keep the tool from running/climbing. On a manual machine you can feel the tool pulling on the leadscrew. Conventional cutting keeps the tools tendency to run in the opposite direction fighting the driving force of the lead screw. However it's not the best way to cut with a cnc machine, climb is. The finish pass can be done as climb to help but it is still a stop gap measure. In the end the best solution is mechanically eliminating backlash.

[louieatienza]

I respectfully disagree, climb cutting pulls the tool forward, that's why you should manually conventional cut to keep the tool from running/climbing. On a manual machine you can feel the tool pulling on the leadscrew. Conventional cutting keeps the tools tendency to run in the opposite direction fighting the driving force of the lead screw. However it's not the best way to cut with a cnc machine, climb is. The finish pass can be done as climb to help but it is still a stop gap measure. In the end the best solution is mechanically eliminating backlash.

Yes it pulls the tool forward, but AWAY from the cut line. If there's any flex in the system, including the tool itself, you'll ALWAYS have a plus tolerance OD when you climb cut.

[lcvette]

I think you're discussing a different issue, there isn't that much flex in these mills as his posted figures. These are backlash related issues. If his Gibbs are snugged properly, tool deflection shouldn't be more than a couple tenths. And if he is running the machine within its capacity, I don't think he'd see that even with a small endmill. The bigger issue is the slop in the screws. And like I said, conventional milling will keep the driving pressure against the screw better than climb milling. However it's a less efficient way to cut using CNC so it may be beneficial to do your heavy material removal as climb and finish with a conventional cut for higher dimensional accuracy.

[109jb]

You keep posting pictures where you take measurements with a digital caliper, but never say in the post what the dimension is supposed to be.

If inside dimensions are coming out too big and outside are too small then the tool is cutting oversize and you program to compensate.

Likewise if inside dimensions are small and outside are too big the tool is cutting undersize and you program to compensate

Both of the above scenarios are where using tool diameter compensation would help.

Also. You steps per of 321.2691707 doesn't match the theoretical value of 320. Lets just speculate that your screws are perfect. If you applied this wrong then a 50mm desired outside dimension would be cut a 50.20 simply because of the steps per setting being wrong. That would only apply to that axis.

I personally don't think that entering anything but the theroretical value does any good unless you are going to map the entire screw. The reason I say this is because the screw accuracy values stated say something like no more than 0.006" per foot of screw length. However, all of that 0.006 can be in a very short distance. Or, it can be off +0.006 in the fist 6 inches and off -.006 in the last 6 inches and it still meets the spec. Now if you are going to map the entire screw and have a controller that can use the screw mapping then that would be different. Screw mapping is where you measure the actual distance moved vs the programmed distance at many points over the entire travel of an axis. This then gets entered into a table that the CNC software uses to compensate for the measured error. LinuxCNC can do this but I don't believe Mach3 can. The difference here is you are not taking one measurement and applying it to the entire travel of the axis.

Also, I commented earlier that your video of how you checked backlash does not actually check backlash but just repeatability. Backlash can only be checked by first moving one direction and then reversing. Your video only showed moves in one direction to get the indicator reading you showed.

There are simply way too many unknowns in your setup. You will have to do much more explaining of EXACTLY how you determined things like your steps per and your backlash before anyone can give you any meaningful advice.

[onocyclone]

The dimensions of the latest part should have been :

18.00 and 6.00mm

I got 18.08 and 6.02mm

In regards to the steps per, I initially started with 320 but the mach 3 axis calibration setting changed it to that 321...... Etc

I am extremely surprised that the backlash comp is still so high even with Oldham couplers.

[lcvette]

How many times did you run the axis calibration? I had to run mine quite a few times before it was repeatable, perhaps you are done calibrating yet?

And I agree, we are posting quite shoot of helpful info but it seems you aren't reciprocating very much detailed feedback, that's the only way we'll be helpful to you. Perhaps you could take a few moments?