My CNC Router Adventures

[Lane]

Not sure what tools Alibre has so I can not do a proper comparison.

I do know Rhino has some mesh tools and more are being added in the next release. I know you can import .obj files into Rhino and IIRC convert them into solids then modify with regular tools. IIRC they also have a "point cloud" to mesh conversion.

Parametric modeling does have some advantages. Your drawing does not have to be perfect first time. Instead you do a rough sketch then adjust parameters. One line may be horizontal, another perpendicular to that, a third line is at 23 deg to the second and has a line at 45 deg to it, a series of holes is xx distance from the edge and centered. You can change a parameter and the drawing is adjusted instead of manually redrawing.

[NoseGoblin]

[mangozac]

Thanks for the "review" NoseGoblin. For the time being I don't intend to be doing any proper 3D, so I'll just keep an eye on things and maybe give the respective packages a test drive at some point.

For now I'm just excited to be cutting precision 2D forms such as this:

It's a small PCB for work onto which that green connector directly fits. I did a test cut in MDF yesterday (during which I discovered that I had calculated the steps per revolution incorrectly) and this morning decided to bite the bullet and cut into some FR4 copper clad board. The 2mm diameter carbide cutters sent to me by Bar worked fantastically for this. I used an engraving cutter that came with the machine for making the copper isolation cuts, but I need to get some with a sharper point for doing real (finer) PCB work.

The board was drawn in Altium designer, printed to PDF and imported into Inkscape where it was cleaned up and an offset path generated as the toolpath (a 1mm offset corresponding to the 1mm radius of the cutter). From there I converted the SVG to G-Code using PyCAM.

This process is about to become WAY simpler as I have been speaking to Andrej and he is adding board outline milling to his Gerber import process. He had a beta version of USB CNC with this addition available less than 12 hours after I suggested it to him!

It turned out very nicely - now I have to make 50 of them. I'm tossing up whether to just duplicate 50 times and generate the G-code from that, or learn how to do subroutines and have the G-Code generate an array of offset duplicates.

[blind pig]

I *really* need to get that book out that I ordered on CNC Routing and learn what the hell you are all talking about!

On a much better note, my 40W laser cutter/engraver arrived today (somehow slipping past Customs and $700 worth of Import Tax and Duty) so I'm very happy about that!

it is a large box that needs two to lift the thing and that is where it'll have to stay at the moment, as I'm running a Memorial Warmachine Tournament for my partner tomorrow and then have to make the workbench/desk (I threw out the idea of using a pre made cubboard) on Sunday, after a sleep in and a clean out of my garage.

I have bought all the demineralised water for the cooling system and am going to use an old bar fridge as a cooling system for the water. I think I'm gonna need it come summer here in the sub-tropics. I also got the ducting and fan from a hydroponic place, so I don't choke to death cutting plastic and MDF. An electrician is coming next week (a mate from my games club) to add the 15Amp power for it all.

[Lane]

[Kroko]

I would make loop with O code

and then use G92 offset with parameters within loop

[mangozac]

[Bar]

Great work. Loving that PCB!!!! And i am very glad those cutters are working out for you.

[blind pig]

I'll post pics of my work area and the laser in coming week or so.

I didn't get much (or any) work today after a massive day yesterday and a huge sleep-in today.

[blind pig]

Any update on the CNC router work Zac?

[edot] for drunk typing

[mangozac]

[Lane]

You should be able to check for backlash without instruments.

Mount a piece of scrap wood or plastic on the table.
Use an endmill to make a spot near one corner. Record this position as X1.
Move in X to the adjacent corner and make a spot. You now have the endpoints of a line. Record as X2.

Offset in Y slightly greater than the tool diameter and make a third spot. This will give you a measurable reference point.

program a loop to make +/- Y moves and single direction X moves from X2 to X1. This will be air cuts, not material removal.
ie.
Y10
Y-10
X-.5
make a spot at same Y position at X1 as you did at X2.
measure distance between your reference spots and the endpoint spots for the Y pattern.
That will give you a measure of how bad backlash is.

Offset in Y again and make a spot at X1, move back to X2 and make a spot.
Repeat your pattern making a light to medium depth cut at slow speed.

Repeat above with same depth but higher speed.

Repeat as above but heavy cut low speed and again with heavy cut high speed.

These cuts will give an indication of any lost steps and what conditions they occur at.
Remember this is an open loop system, the controller assumes that the tool moves when told to.

You should also do a series of circular test cuts using air cuts and various depth/ speed combinations. If your acceleration is not balanced between axis the circle will distort.

[mangozac]

Thanks Lane. I was aware of that kind of technique for measuring backlash, but I figure it's easier to just set up the dial gauge

Hopefully my friend should drop it off today, in which case I'll do some measurements late this afternoon.

[mangozac]

Well I did acquire the dial gauge, but haven't had a chance to use it yet! Maybe tomorrow after work...

In other news, I was today playing with the evaluation version of Rhino. I don't know if I've mentioned it elsewhere but at work I'm currently going through the process of getting a part for a new product injection moulded. So I figured this would be a good test for trying out the CAD package. I can't post a pic of the part, but it's effectively a replacement rectangular lid for a metal enclosure that features three bulbous lenses on it.

Rhino I have a very neutral feeling about. It seems extremely stable and certainly easy enough to use. Yet I still didn't quite feel like it was the optimal tool for the job. I had issues doing boolean operations on a few parts (with Rhino simply stating that the operation had failed), which left me making some sloppy short cuts. Also, the trial version of Rhino has copy/paste disabled, which is very, very frustrating. If they're going to limit the trial to 25 saves then why cripple it quite badly by taking away copy/paste?

Right now I'm downloading a trial for Alibre as I'd really like to try it out (the personal edition). It certainly sounds like it would suit me perfectly - in particular the parametric design features. I figure the NURBS and really complex surface stuff provided by Rhino is a bit over the top for applications I'll be using it for: at work things remain pretty geometric and for hobby use I don't plan to use it for any exotic shapes....

So watch this space - I will report back on Alibre!

[mangozac]

OK after another morning of 3D CAD testing I think we have a clear winner! While I'm sure Alibre would be great for mechanical engineering projects with moving parts and high tolerances, the parametric aspect of it just isn't intuitive enough for designing models. I felt that the learning process was taking twice as long for Alibre as it did for Rhino, and the processes involved just don't feel natural.

So I will email Rhino about getting a student copy sorted so I can get some beloved copy/paste action going

Oh and I did some more reading up about boolean operations, and now that I understand it some more I'm sure I'll be able to use it better and without getting failed operations.

Finally, I did a little bit of backlash measurement this afternoon. I wrote 2 G Code files, one for testing the x and y axis separately. I simply had it move forward 10mm then back 10mm, ten times in a row. I set up the dial gauge so it was zeroed at about half its travel and then noted the difference after the program was run. So far I have only tested the X and Y axes, but neither showed any measurable amount of backlash. I might repeat the experiment and get it to do like 50 movements tomorrow...