Backlash or CV mode too loose?

[Jay C]

Have a look at a microphoto of a test circuit. The lines are 16mils and the isolation set to 20mils for the think lines. It looks like backlash to me, but then again, it could be the CV Dist tolerance. What do you think?

Copper is 3oz, I used a 45-degree v-bit at 14ipm and 20K RPM.

http://picasaweb.google.com/jay.cout...76763024161666

[harryn]

Hi Jay,

Just guessing here, but it doesnt seem quite like normal backlash to me. For some reason, I reminds me of what happens when something has come loose near the head.

[Drools]

I wish I could provide some feedback but I'm no where near experienced enough to comment. However watching threads like this is where us new guys can get some troubleshooting experience.
Your milling looks somewhat close to the picture but not like it is supposed to. What changed in your setup from your last posting of the sensor?

[Jay C]

A lot has changed actually.

• I upgraded to a P4 1.6GHZ laptop from a P3-700 laptop.
• Running the kernel at 60KHz (after running the drivertest).
• New 370Oz-in stepper to replace the X and Y 270 oz-in stepper. Had to wire them in series versus using parallel due to the current limits of the Geckos (3.5A).
• Running Mach3 instead of Quantum and thus CV mode with 200 lines look ahead.
  [Using gOpt optimize instead of Opti, but the Gcode looks fine (see attached).
• I slows the feed to 14ipm from 20ipm
• 3oz versus 1/2oz
• 45 degree bit versus 60-degrees

More testing scheduled for today

[Drools]

wow not like a single change you can point your finger at...
If you like I can run the brd on my mill and compare. We used to have a similar setup.

[Jay C]

Brett, after running some new tests, I think the issue is the Z axis. It's "floppy" and I think te soft copper is just pushing to one side and not cutting. I'll email you the file ... do you have some 3oz copper?

[Drools]

I have some 2oz scraps, does it have to be 3oz?

[Drools]

I searched the Mach3 forum and they are not at all keen on using laptops for Mach3. I guess it is hit and miss on getting one that works as expected.

[Jay C]

Well, being that condition is repeatable, I feel confident that is not the issue. I changed over tot he Wolfgang spindle. I also feel confident that it's a Z-issue exacerbated by the thicker copper. There is movement around the Z due to some slop in the X carriage. It magnifies to visible displacement at the tip. Slowing way down helps against this. I have some more testing to see how slow.

It's like your car tires catching a rut and slipping into it after a big snow ... it's the path of least resistance.

Jay

[Drools]

I did not realize just how much you are talking about. My original Z axis was based upon brass bushings/rods salvaged from an old dot matrix printer. They were very heavy duty but I could not get them to the point where there was no movement when I grabbed the end of the Wolfgang spindle. I switched to THK linear rails and the problem was solved. I have heard users talking about the backlash setting in M3 could that help if set correctly. I guess the real fix would be to remove the slop on all of the axis. Sorry Jay I'm just not much help on this one.

[Jay C]

Update:

I have slowed my machine down to 8ipm for the results you see below. The biggest issue I see is the intersecting 16mil line is missing. The coil and the diagonal zigzag look great. I also measure the traces with my calipers and they were .0165".

I have ordered some Precise Bits 60 degree cutters to compare.

Brent, no worries. The slope is negligible for larger items. Plowing through 3oz of PCB is showing a limitation of the bronze bushings in this very extreme case.

Jay

[Drools]

Here is what I ended up with.
Copper: 2oz.
Default Isolation: .001
Max Isolation: .03

Tool Size: .005
Step: .0045

Z down: -0.0045, -0.005 and -0.006

Speed 30ipm

The bit used is this one: EM2E8-0625-60V, Tip size: .005
I have attached pictures of the bit tip as posted in a previous thread.

The one that looks nice and clean was done at Z-down -0.0045 but 2 passes, then it was -0.005 and -0.006 one pass each. I'm going to try the same test with a simple 45degree bit from Curt.