Levelled, but still air cuts

[vadim]

Hi guys, I have a strange problem...
I've had one tremendously successful result first time I made a PCB with autoleveller. Used 0.1mm bit etching to 0.1 depth.

However, for some reason, I can not repeat that result anymore. Now whenever I try, i still get either a proper aircut somewhere, or the depth just is not enough. Most of the board seems ok, it's just there're some localized place(s) where depth of cutting just isn't sufficient.
I'm using a double sided tape to fix the board. Before I'd make sure that it's secured very well, now, however I grew careless and assumed that autoleveller will take care of it. So if there's a small kink in the tape - I don't bother with it.. Could that be a problem? I reasoned that at 10mm spacing and 0.1mm depth, spacing is 100 times the depth, so a lot of "noise" can happen between probe points, particularly with kinks in the tape?

What concerns me though is that I tried to level out a piece of plastic using the "tin foil" method, however, after removing the foil - bit was air cutting all the way. E.g. didn't touch the surface at all. I got around it by jogging down to zero, and the slightly below it and reset DRO there. Kind of offset the zero. Then it was all well and levelled.


So my question is - are my problems most likely related to errors in the setup. Thickness of foil (And yes, it was smooth & flush with surface - used water under it), or kinks in the tape and so on?
Or is it possible my Mach3 setup is somehow stuffed somewhere? E.g. Is there an offset setting for Z or something like that?
Only other alternative to the above 2 would be a CNC controller problem, whereby it's missing steps in Z axis??? I doubt it's that though...

[vadim]

After browsing through this forum, another possible answer could be related to arcs vs straight lines?
In both examples above(plastic & PCB) I have a lot of curved and round shapes, but I'm not sure atm if they're broken down into lines or not.

Update: I had a look at the PCB G-Code, there're no arcs G2 or G3...

[daedelus]

Hi Vadim,

Only 0.8.3 supports arcs properly, i.e. breaks down arcs into smaller segments. But as you said, you dont have arcs in your gcode so thats not the problem.

I used to find with DS tape, the kinks did cause problems as the probe only lightly touched the surface, whereas the tool exerted a lot more pressure when it performs the etch. If you have kinks in the tape, this creates a see-sawing effect during etching. So yes, it is important to tightly secure your piece to the table before you start. Try and get rid of those kinks, or use some other method of securing the piece such as a vacuum table or clamps in all 4 corners: http://www.autoleveller.co.uk/table-setup/. The point is, you need to make sure nothing moves even slightly between probing and etching.

With regards to the tin-foil, I would try to compensate for this in your CAM program, i.e. add an extra 0.2mm to the depth (or however thick the tin foil is).

These are my suggestions anyway.

best of luck,

[vadim]

Thanks for your answers, James. It makes sense (about kinks) now and also explains why first time I had a perfect result - I was still carefull from the times before autoleveller and stuck everything quite well
I'll be giving it another go this weekend. Will post if I find something else I can't quite explain and solve myself

Let me also take this opportunity to thank you for this great software which is available free of charge and also for your support on this forum

[vadim]

I'll have to resurrect this old topic unfortunately as I've just recently been playng around with making PCBs and there're definite air cuts. This time I made sure there're no kinks or anything like that. Here's the whole story.

A PCB size approx 100mm by 100mm. There're 100 something probe points. Probing goes smoothly, but when the cut begins it doesn't even touch the PCB.. I manually fiddle with the Z by moving it down by about 0.2mm and then all seems to go fine. However as the cut progresses I clearly see that it goes deeper. Long story short by about "half the board done" Z is back by +0.2mm (e.g. original position). Now I would have though that Z axis might be missing steps or something, but then the problem would be progressively worse (statistically, diviate more and more) along the probe lines which go zig-zag across the board. However, what I see is that the cut is deep(er) even on the first probe line.

Again, long story short, I don't think Z axis steps/precision is a problem here.
From what I can see, it appears that levelled GCode starts off too high but then eventually goes down to correct hight by about "half the board done" metric.
I know for sure that I had this working perfectly at some stage, but I don't remember which version of leveller I used. All current hurdles are with the latest free version.
James, any chance what I described makes sense in terms of bug fixes you might have done in the few last releases of the free versions?

[daedelus]

Hi Vadim,

Whilst I cannot say for certain what caused the issue, there does seem to be some difference between the free 0.7.7 version and member versions

Back in October I think, I ran some experiments to check for differences between free and member versions, and posted the results here:

viewtopic.php?p=402&sid=c7feb67ec460543 ... b322e#p402

Unfortunately I dont know which or if an earlier free version is better (and am not convinced there is much of a problem with the free version) but the member version gives the best results according to this.

In light of my experiments I do plan to upgrade the algorithm for both the member version and free version so that both versions are consistent and give a good result, and this is on top of my list of changes.

[vadim]

Hi James,
Thanks for sharing your thoughts. I'm going to try a current version next time (might be some time though) and will post the results here.

[booski]

Just out of curiosity, what machine are you using? Doesn't necessarily sound like a code problem but perhaps an issue with backlash in the z axis

[vadim]

A cheapy "3020"(Whatever the brand on ebay.. )
If by backlash you mean the "slop", I doubt it, it's too big and can be reliably eliminated. Although I wouldn't dismiss it altogether. However, I didn't have that problem before. And in fact consistent depth increase I noted before going left to right(after I corrected for the gap) does actually point towards either a code error (precision?) or levelling issue on my side...

Once I test with the current version I'll post my results. If it's still the same, that very likely points to some problems on my end

[Mr Mayhem]

I just tried using Autolever version 0.8.6 on a 12 inch x 12 inch PCB using the v-tool as a probe, default settings except for grid area and offset, in the probe-and-cut in one step mode (original way of using Autoleveler before persistent probe files were added). I am using mm in both Autoleveler and Linuxcnc. I got air cuts from the very start of cutting, where it seemed to be too high very close to 2mm. I determined this by stopping the program within 10 seconds or so from this beginning of cutting stage, and jogging down to touch the board. It seems very close to 2mm. I will do more examination if I run into the issue again using the saved probe file mode. I noticed 2mm is one of the default settings (for probing retract height), so maybe this is related to the bug? I also want to know when running a gcode program that was leveled using the saved probe file, how do I touch off correctly at the start of the leveled gcode file to insure accuracy, esp if I probed using a probe instead of the tool tip? I suppose if the log file is using relative numbers above and below zero, zero being equal to the first probe touch, I can swap in a tool in place of the probe, and zero it on the same place manually, and everything will jive? I would be properly zeroed in z for the leveled code to run?