View Full Version : Accuracy Problem

15-06-2013, 12:39 PM
I bought a 3020 mill primarily for making PCBs and small panels and I think I might have an accuracy problem creeping in somewhere in my operation or setup. Here is my routine:

1. Export tracks HPGL milling file from CAD program.
2. Export outline HPGL milling file from CAD program.
3. Export Excellon drilling file from CAD program.
4. Import tracks milling file to SheetCAM, set outside contour routing operation and save job then save TAP file from post processor.
5. Import outline milling file to SheetCAM, set outside contour routing operation and save job then save TAP file from post processor.
6. Import drilling file to SheetCAM, save job then save TAP file from post processor.
7. Load tracks milling TAP file into Mach3 and do.
8. Load drilling TAP file into Mach3 and do.
9. Load outline milling TAP file into Mach3 and do.


Is this something that a Smoothstepper would sort? Or some other problem?

15-06-2013, 02:12 PM
Not sure what's going on but the drilled holes all look offset to the right!
Maybe a zero positional error when you loaded the drill!
Just a thought!


15-06-2013, 02:47 PM
Try exporting all as one file out of Cad then make your individual tool-paths in sheet cam move tap files to Mach3 keeps everything in same 0 position.

Errors can creep in if you split up the cad file.


15-06-2013, 04:44 PM
I've done two boards, first had holes offset mainly too low, this board I added 0.25mm offset but it also had a right offset it seems too.

When changing tools or loading TAP files, the same instance of Mach3 is running so should still be in the same place?

15-06-2013, 04:58 PM
Why use an offset? if you create your tracks in cad with the given clearance between and place holes where needed then export or import into sheetcam, select all holes make toolpath, select all tracks make toolpath, export individual taps files into mach job done they all reside on the same Zero centre.


15-06-2013, 05:08 PM
Just done a google on Excellon Ref: Excellon format - Wikipedia, the free encyclopedia (http://en.wikipedia.org/wiki/Excellon_format)

(Furthermore, the implementation of Excellon output in CAD is often poor, resulting in problems such as poor registration between the drill and the image.)

Could be your problem.


15-06-2013, 06:10 PM
Thanks for that. The ref point seems to be the same across all 3 files. I didn't know that about Excellon files not being that accurate, and with the drills moving seemingly randomly it could be the answer.

I'm not sure if I can export a Gerber file showing drill data though. It's only a basic CAD package.

15-06-2013, 06:23 PM
OK, after a quick check my software will export Gerber RS-274X which can includes hole data if I tick the box on exporting. Now my problem is how to I convert this into a drill file for my mill?

16-06-2013, 11:41 AM
Thinking about this overnight, I have made similar boards previously using the exact same method on two different friends mills, and those did not have the problem. Could it be an issue with my drive controller?

16-06-2013, 12:54 PM
Could it be an issue with my drive controller?

Chances are it won't be a drive problem. Looking at the holes offsets and the fact they offset consistently it's More likely to be either a lose coupler slipping or you have acceleration set to high in Mach causing it to jump the odd step on direction changes.

Post the tap files and will check for you. Will combine into one G-code file and then you'll see on the screen if holes are offset or whether the machines causing the problem.

17-06-2013, 06:33 PM
Many thanks Jazz. Here are the 2 TAP files.

tracks (http://www.readresearch.co.uk/personal/mycncuk/amp_tracks.tap)
drills (http://www.readresearch.co.uk/personal/mycncuk/amp_drills.tap)

And here are my settings.


Note that after some advice on here I changed Step Pulse to 5us.

The settings above are somewhat different to what my machne setup guide suggested, which was 400 steps, 2000 velocity, 200 accel, 10 step and 3 dir. The machine did not move to the correct scale with these settings.

17-06-2013, 11:55 PM
Ok well you don't look over tuned. 1800mm/min velocity isn't overly fast but it could be at the upper limit of tight or worn machine, esp these low power machines with crappy electrics. For a test you could drop velocity to 1000mm/min.
Personally don't think this is the problem has the code use's very few G0 rapid moves, plus most are Z axis moves and the G1 commanded moves are all low feed moves.

The G-code is fine and everything lines up. I've Attached combined files. 9079

First for simplicity's sake I would check the mechanics of the machine.
Check for loose couplers or signs of slippage.
Check ballscrews for end float and ballnuts for backlash.
Check all axis for binding screws or linear bearings things like this can easily rob steps from a low power machine.

One problem I've encountered with these machines is very poor quality wires and connectors, esp the stepper wires. Infact the whole electrics are shit.!!
Go over the wires and connections with fine tooth comb. To be honest For the price of few meters of shielded cable it's worth replacing the stepper cables anyway. Electrical noise could be affecting the motors and shielded cable grounded at the controller end only will help reduce this.

If the electrics check out ok and there's no binding etc in the mechanics then can't suggest anything else other than trying on a different PC in case you have parallel port issue.?

After this if your happy with the machine and it's mechanics(so long has it use's ballscrews not lead screw) then I'd throw away the crappy all-in-one drive boards and put some decent motors with 50v drives on it along with 45V PSU.
This will sort any electrics problems and make the machine fly.

Good luck.!

09-09-2013, 11:24 AM
This turned out to be connected to the y axis stalling issue. The machine was loosing steps because the clock speed was set too too high.