4 thou error on a small part

Hello,
I have a Precision Matheus 728, with Linuxcnc 2.9.2, Mesa cards, and gmoccapy GUI

My milling vise came with a slot key (0.5000"), that does not fit my milling machine (12mm slot).I designed a new key on FreeCAD 1.0. Generated G-code. Used it without issues.

But the new key did not fit the milling machine slot.

Using a micrometer, the base of the key measures exactly 0.5000"
But the slot side measures 0.476x (It should be 0.4724).
I am using a 3/8" end mill (carbide, new)
Sotck is 1018 steel, Spindle 1500, Horz feed = 0.5 ipm

I don't understand why the first operation produces the expected result, but the second is 4 thou thicker than it should.
Does anyone can give me an insight on what could be wrong ?

Details:
- The 3/8 endmill is loaded in the beginning of the program.
- The first operations makes a full profile around the stock and produces a dead on 0.5000 depth. (Clockwise, 3 thou depth of cut on last pass)
- All relevant operations are executed using the same tool, same speed and feeds. There are a few M codes between operations though.
- I milled a second part with exact same results (one dimension dead on, the other 4 thou off)
- For the second part I also used a "profile" operation (also Clockwise, 10 thou depth of cut on last pass). The first was done wirh an "adaptive" operation
- The attached G-code is the one I used in the second part ("profile" operation)
- As a sanity check, I run the program again cutting "air". Annotaded the DRO values for Y at the relevant moments, and everything seems correct.
DRO Y abs:
5.4375
4.5625 -> less tool -> 0.5000
and
4.5901 -> less tool -> 0.4724

Is it possible the 10 thou depth of cut on last operation (versus 3 on the first operatiion) could cause a deflection of 4 thou in the column ?
I can only measure a few tenths using my hands
Thanks in advance for any suggestions.
Edmar
 

(for context for anybody not reading the gcode: he's doing a profile to 0.500 wide all around the part, then machining to 12mm by taking another 2 cuts at a higher Z level- one roughing, one finishing, taking all of the 0.7mm from one side of the part)

First thing that strikes me is 0.5 IPM feed seems really low at 1500 RPM. If you're using a 4 flute cutter, that's 0.00008 IPT (not a typo- under a tenth chip load). Possibly nothing to do with anything, but I'd bump up the feed quite a bit. I often finish at 10x that chipload (albeit in aluminum, but that's not going to change the scallop height), roughing obviously at more. Seems like you're going to burn up endmills like that. Check out a speeds/feeds calculator (GWizard, FSWizard, HSMAdvisor, etc.) if you haven't.

Does the DRO agree when you're cutting metal, not just air?

Do the DRO scales feed back into LinuxCNC?

What size stock are you using? I guess the 1018 is cold-rolled? 1018 CRS is notorious AFAIK for having internal stresses that cause the part to "banana", especially if you remove more material from one side than the other. I don't know how you're measuring, but if you're putting caliper jaws across the entire length of the 12mm dimension, that could be throwing it off.

Do you get closer to your desired result if you re-cut the same part (eg, run the program twice without changing the stock/setup)? Or does it cut a little more off the 12mm dimension? You could alternately add a "spring pass" to your gcode- just repeat the 12mm finishing cut.


The gcode looks OK to me.

You can try cutting in a different/easier material (eg, plastic or aluminum, adjusting the speeds/feeds accordingly) to see if the dimensions come out the same- that would presumably help narrow down deflection vs some other problem.


Finally, note that especially steel-cutting end mills are commonly up to 0.002" under the nominal diameter (eg, www.destinytool.com/raptor-dvh). May or may not have anything to do with anything, given your 0.500" dimension is right... but it could be an undersized end mill + backlash cancelling each other out. When you cut on one side for the 12mm dimension, you wouldn't have that effect.