Odd Dive behaviour on Rotary Tube

Problems and questions
1

I’ve been using Sheetcam for what seems like forever.
We now have a rotary profile cutter that we cut HSS square tube on.

I’m having an odd issue with how the code handles the corners.

Odd dive
Odd dive1131×526 97.8 KB

As you can see in the red circled areas, the Z axis dives to cut height while still in the corner.

I’ve found the section of code where this occures:

N57610 X1.7346 Y0.9005 Z4.1234 A-121.6125
N57710 X1.7325 Y0.9515 Z4.1109 A-120.8386 ← I’m still at 120 deg
N57810 Y2.6985 Z3.0900 F9963.9528 A-90.0000 ← This is the issue , it dives from 4.11" to 3.09" while rotating 30 Deg
N57910 Y-2.7250 F59.7073
N58010 Y-2.5141 Z3.3072 F14961.4939 A-85.2744
N58110 Y-2.2817 Z3.4894 F13782.6132 A-80.9223

I can supply the full code if needed. I have 2 of these events on this one part.

Here is the other:
N64510 X5.5389 Y0.5987 Z4.1827 A53.8552
N64610 X5.5433 Y0.8282 Z4.1399 F13234.6662 A57.2957
N64710 Y2.7119 Z3.0900 F10877.1486 A90.0000
N64810 Y-2.7377 F59.717
N64910 Y-2.5021 Z3.3174 F36934.0906 A94.9575
N65010 Y-2.4047 Z3.3966 F13782.6257 A96.8073

This one starts to dive Z with over 32 degs of rotation left.

Here are how the other corners look like, these are fine.

59610 X1.8709 Y2.2741 Z3.4949 A-9.2152
N59710 X1.8847 Y2.5402 Z3.2849 A-4.2205
N59810 X1.8986 Y2.7466 Z3.0900 F9830.8464 A0.0000
N59910 X5.4005 Y-2.7453 F59.8311
N60010 X5.4035 Y-2.5001 Z3.3190 F53516.1595 A4.9947
N60110 X5.4065 Y-2.4432 Z3.3658 F11634.3889 A6.0805

The Z dive to cut height only has to move 4 degs on this one.

I’m stumped as to why it’s doing this. I tried changing the starts to make the rotation clockwise , but i’m still getting the issue.

2

What post processor are you using?

3

I’m using both my own post and your PlasmaRotary Mach3 post.

I’m putting together a package with all my testing to send to you, if that’s alright.

The basic issue is always the same in both posts, when it wraps around one corner, only one corner now, it jumps the A rotation by a large angle. For example, when wrapping around a corner, the segments all span between 3 and 5 degrees, but when we see the error , there is a large jump in degrees, one of the patterns we have been seeing with our cut pattern is either + 43 degree or -43 degree jumps at the corner.

Ex:
N21010 X2.2904 Y-0.1397 Z2.8234 A-48.1818
N21110 X2.2891 Y0.0646 Z2.8265 A-43.5291
N21210 X2.2456 Y1.7734 Z2.0900 F14597.3481 A0.0000

Also to note: The simulation in sheetcam is always fine, but the error is visible in the generated Gcode and in the toolpath generated by mach3

Thanks.

4

Sure, send your test files to me. Sim is basically just a custom post processor so it’s odd that the issue doesn’t show up in simulation.

5

I always found it off that the Sim was showing very different behaviour than the actual code. Good to know.

Where would you like me to send them?
I can send you the Gcode that we produced, my post and the DXF file.

6

Sorry for the delay in replying. Could you email them to me: les@sheetcam.com

7

Well, that took some finding. I will try to upload a fix in the next few days.

8

Hey Les,
Any update on the fix?

9

There was a new update to the development version out a couple of hours ago. That should solve the problem.

10

I will give it a go!

12

I ran 7.1.35 to get a baseline, it generated 509 lines of code and obviously had the dive issue.

I updated to 7.1.36, ran the exact same file and start locations. It generated 2877 lines of code, I see absolutely no degree jumps . Looks good!

I will be back in my shop tomorrow, I will run the generated Gcode on my new rotary and get back to you with results! To me this looks like it will produce a smoother cut file.

Thanks for working on this!

13

Was that 509 lines correct? The new version should produce fewer lines, not more.

14

Well then…

The code generated on 7.1.36 is 2877 lines, It is correct.
The same file processed on 7.1.35 is 509 lines and is incorrect.

I can send you the gcode in an email if you would like. I cannot send the job file from 7.1.35 because it wasn’t saved and i’m updated.

This is a snip of code showing the movement from 80 Deg to 73 deg.

Code from 7.1.36
N2110 Y1.5571 Z2.2637 F13763.8093 A83.9871
N2210 Y1.5480 Z2.2710 A83.7187
N2310 Y1.5390 Z2.2782 A83.4503
N2410 Y1.5299 Z2.2854 A83.1819
N2510 Y1.5207 Z2.2925 A82.9134
N2610 Y1.5116 Z2.2996 A82.6450
N2710 Y1.5024 Z2.3067 A82.3766
N2810 Y1.4931 Z2.3137 A82.1082
N2910 Y1.4839 Z2.3207 A81.8398
N3010 Y1.4746 Z2.3276 A81.5713
N3110 Y1.4653 Z2.3345 A81.3029
N3210 Y1.4559 Z2.3413 A81.0345
N3310 Y1.4465 Z2.3481 A80.7661
N3410 Y1.4371 Z2.3549 A80.4976
N3510 Y1.4276 Z2.3616 A80.2292
N3610 Y1.4181 Z2.3683 A79.9608
N3710 Y1.4086 Z2.3749 A79.6924
N3810 Y1.3991 Z2.3815 A79.4240
N3910 Y1.3895 Z2.3880 A79.1555
N4010 Y1.3799 Z2.3945 A78.8871
N4110 Y1.3702 Z2.4009 A78.6187
N4210 Y1.3606 Z2.4073 A78.3503
N4310 Y1.3509 Z2.4137 A78.0819
N4410 Y1.3411 Z2.4200 A77.8134
N4510 Y1.3314 Z2.4262 A77.5450
N4610 Y1.3216 Z2.4325 A77.2766
N4710 Y1.3118 Z2.4386 A77.0082
N4810 Y1.3019 Z2.4447 A76.7398
N4910 Y1.2921 Z2.4508 A76.4713
N5010 Y1.2822 Z2.4569 A76.2029
N5110 Y1.2722 Z2.4628 A75.9345
N5210 Y1.2623 Z2.4688 A75.6661
N5310 Y1.2523 Z2.4747 A75.3976
N5410 Y1.2423 Z2.4805 A75.1292
N5510 Y1.2322 Z2.4863 A74.8608
N5610 Y1.2222 Z2.4921 A74.5924
N5710 Y1.2121 Z2.4978 A74.3240
N5810 Y1.2019 Z2.5034 A74.0555
N5910 Y1.1918 Z2.5090 A73.7871

Code from 7.1.35
N1910 Y1.4546 Z2.3423 F13763.804 A80.9963
N2010 Y1.3110 Z2.4391 A76.9879
N2110 Y1.1611 Z2.5256 A72.9794

7.1.35 was moving about 4 degrees per line.
7.1.36 is moving about 0.4 degrees per line.

15

Could you send me a copy of your post processor. In particular I’m interested if you set the maxRotation value and if so, what value you use.

16

I do not have a maxRotaion valuer in it.

I emailed my post to you now.

17

Before you go too far into anything , i’m running some tests with a modified dxf .

I’m about to run it on my machine to see how it looks and performs. There might not be a problem here.

18

I’ll wait for your results. As I mentioned before I would have expected fewer lines of code with more evely spaced angles.

19

I guess the one file was an odd fluke at 2000 lines of code.

I made a new file, ran it with 7.1.36 and it cut very smoothly and is only 452 lines.

I think this is a win!