Page 2 of 2

Re: Trace width inconsistencies

Posted: Thu Apr 02, 2015 8:15 pm
by aiden1015
I think it definitely takes a lot of tuning to get working properly. My main issue with the 3040's is that in order to get those working, it seems like you have to throw out everything except the frame. The parallel port is too outdated for the controller, I've heard nothing but bad things about the stepper drivers, and the motors seem to be hit or miss on quality.

I did discover something interesting today though. I measured the actual isolation gap of my vertical trace and measured it against my horizontal trace. I forget the numbers, but the isolation gap on the horizontal traces are wider than that of the vertical.

I'm still confused as to why this is happening though, clearly my v-bit is going deeper in the horizontal than the virtical...but I have no idea why. I'm going to look at the g-code and see if there's anything funny going on there. It seems like it must be in the g-code and not the auto-leveling because no matter where I on the copper clad, I get the same result. I might need a sanity check on my logic here...


It looks like I was just chasing a red herring, this last test shows the isolation widths are the same.
IMG_1138.JPG (121.77 KiB) Viewed 1790 times

Re: Trace width inconsistencies

Posted: Thu Apr 02, 2015 11:39 pm
by jlauer
My conclusion on some of the stuff you're pointing out is due to the rigidity of the spindle on Shapeoko--not backlash. The spindle/Z-axis is rigid on the Shapeoko going left to right (X dir), but has a lot of flex moving forward and backward (Y dir). So, depending on how you enter a cut, you get a drag effect that either raises the endmill up out of the material or digs it deeper into the material until the flex is all used up. So, it's not timing belts, it's just basic rigidity. For example, on my Shapeoko, if you are towards the front of the machine and cut moving towards the back (moving in the positive on Y) you get the endmill raising up out of the material and you get a shallower depth cut. If I'm on the right side of the machine moving left (negative into the X direction) I get a nicer rigid cut.

BTW, you could add an algorithm into ChiliPeppr that reworks your Gcode moves to always enter cuts from a certain direction to minimize backlash and rigidity. If you pulled this off I think the community would throw you a party, but from what I've read it is possible to do through math and rejiggering of the tool paths.

On the 3040, the quality does depend who you buy from. Given that it's generic you get higher-end providers and lower-end providers. Check out my video on my ChiliPeppr-izing of a 3040. I've been super happy with it but I got it from a higher-end shop. I'm finding the steppers to be quite nice. Of course I tossed the parallel port guts though.

Re: Trace width inconsistencies

Posted: Fri Apr 03, 2015 1:18 pm
by aiden1015
That sounds plausible. But I would think slowing my feed rate down would prevent rigidity issues, since the end mill is able to keep up and get material out of the way fast enough that I wouldn't notice.

Maybe the best solution here is to accept that 32-TQFM (SMD Arduino IC package) can be done for the most part and not to go any smaller than 0.5mm traces and accept the my axis limitations. Anything finer, and I'm better off dealing with the 10 day turn around times of Oshpark for my PCBs.

Re: Trace width inconsistencies

Posted: Wed Apr 15, 2015 1:25 am
by aiden1015
In case anyone else runs into a similar problem, I just wanted to update and show that I fixed it. The problem was backlash from the pulleys as cvoinescu pointed out. For me tightening wasn't enough though. What I did was to physically remove the Y-Axis motors and remove the pulleys. I took my Dremel and used the grinder to flatten one side so that one of the set screws had something to chew on (trying to create this: ... t_flat.jpg) this prevents the pulley from rotating very much and gave me a nice and even trace widths in both the x and y axis.

Again here's the before of my test which resulted in a 0.5mm and 0.34mm trace
FullSizeRender (1).jpg
FullSizeRender (1).jpg (396.84 KiB) Viewed 1697 times
And here's after. Both are nearly exactly 0.5mm (I purposefully didn't go through the copper all the way)
FullSizeRender.jpg (233.06 KiB) Viewed 1697 times