G41/G42 discussion.

[cncdrive]

Terry, the problem is not tight corners.

[ger21]

Perhaps there are better algorithms out there?
https://www.sciencedirect.com/science/a ... 851730009X

[cncdrive]

Well, they write "Since the algorithm uses the exact part geometry", which suggests that it requires the part geometry to be known and so it is like a ray tracking algorithm.
G41/42 does not know the part geometry in advance is the problem... the paper description talks about using the algorithm in CAM softwares which is fine, there are much more advanced algorithms ofcourse when the algorithm knows the whole path and when there is enough time available etc. And finally this type of algorithms requires a closed path to be offset or if the path is not offset then the algorithm can only walk around the opened path offsetting it on both sides (around). If you check the figures on the linked site you can see approximately how it works, it works the way like a ray tracking algorithm and not like a G41/42. Again, there is a huge difference in what datas are available and what are the requirements for the algorithm to work and what are the results.
I know what I'm talking about, because you know we also made a basic CAM module for the UCCNC which uses ray tracking algorithm, but that could not work as a G41/42, because as said the requirements for an algoritm like that is different than what is available for a task of G41/42 and also the result is different, because it can only fully offset a path around and not running along the path with an offset.
I know it is not easy to see and understand the differences without actually implementing both these algorithms or at least studying these deep enough and it is also hard for me to properly describe these things in English, but beleive me that I know what I'm talking about, because I already implemented both algorithm types.

[cncdrive]

It will be hard without having the code and seeing the problem.

[Battwell]

the sample i showed above is the same profile i cut using the biesse. originally all my coding was parametric so even angles were specified as trigonometry. so the control didnt know what it was going to cut until it got to that line and calculated the trig etc. (point to point controller- no cv on it)
but it did offset that part properly (part was a lot more complex - just a snippet that fails was shown)
so - basically- the italians had worked out how to do it in the early 90s.
if i broke say a 12mm cutter i could just select another- eg 14mm from my tool table and re run the program fine.
referring to my sample pic above:
when it came to the lower segment (where you see it tapering in as a gouge- the original control would follow the line then create a smooth radius out to avoid the inside corner.( smaller segment than tool radius) rounding the next corner (creating the 90deg sharp corner) and carrying on the straight section as normal
this makes me think they must have been using the circumference as the reference - not the centre point- if that makes sence

[ger21]

I know what I'm talking about,

I know that you know much more about this than I do, and more than I want to know.

But Battwell and I have used machines that were capable of doing this, so we know that it's possible.

[ger21]

Balazs here is teh sample gcode. This problem has been a BAIN to DIY cnc controllers sense teh beginning. ALL of them have teh same problem. They do not compIN correctly and it leaves a small uncut portion of teh contour. NOW it does compOUT correctly.

I would argue that it's just coded wrong, and it is comping in correctly.
There are rules you need to follow when using G41/G42, and if you don't follow them, it won't work.

[ger21]

I disagree. I'd never comp in at 90° to the part. But my industry is very different from yours. If I was going to come in at 90°, it use an arc to enter the part tangentially.

but you could also assume that your gouging example was coded wrong as well.

No, it's different. You're asking me to change the shape of my part. I'm asking you to change the leadin.
Comp should be able to follow the part, without gouging, regardless of the part shape or geometry. If it can't reach certain places, that's fine.

[ger21]

Changing tool diameter is not an option, as it triples run time.

[Battwell]

my biesse control allowed an arc lead in as gerry showed- common for inside profile-
imagine an inside rectangle- your not going to start in a corner- always along an edge.
in woodwork - we use quite large diameter tools for profiling- mainly 12mm to 20mm diameter (1/2" to 3/4"+ sort of sizes) as we have to cut fast- and you cant cut say 50mm oak with a 1mm tool diameter!
so its quite easy to get a step in the profile which is less than tool radius.