"rapid retract" - accel/decel interpolation issue?

  • Der3er
  • Der3er's Avatar Topic Author
  • Visitor
  • Visitor
30 Mar 2010 10:55 - 30 Mar 2010 11:08 #2476 by Der3er
hello!

this may seem simple to advanced users, but... i just can't seem to figure it out myself

i run code usually at approx. 80 ipm, using emc2 for bot control. at first glance, everything seems to work quite ok. BUT: depending on safe z height (or so it seems), my machine seems to mess up every time it reaches the END of any given toolpath. emc2 shows both g-code and the actual -as is- toolpath, representing actual machine travel. now these two differ!

while ramping in works fine, interpolation seems to mess up the end of my toolpaths. where a straight move to the end of the toolpath should have been executed, in fact some sort of interpolation between last target xyz on the toolpath and the next target that includes rapids seems to occur. what this does is: start of the toolpath is precise and fine, end of the toolpath is not. z-moves at the end of the toolpath are visible where there should be none. it seems to relate to the settings for safe z. i did try a few different settings, but have not been able the elimate the actual cause for this. i can get it down to "quite unnoticeable" but now that i'm aware of it i found traces of this phenomenon in all my workpieces.

does this sound familiar to anyone? am i missing sth.? my guess would be that this behaviour must be related to acceleration and deceleration and that there must be a setting that controls this? any help will be appreciated.
Attachments:
Last edit: 30 Mar 2010 11:08 by Der3er.

Please Log in or Create an account to join the conversation.

  • Der3er
  • Der3er's Avatar Topic Author
  • Visitor
  • Visitor
30 Mar 2010 11:11 #2479 by Der3er
...and this is how it looks like
Attachments:

Please Log in or Create an account to join the conversation.

More
30 Mar 2010 11:18 #2480 by BigJohnT
If you saying it is cutting corners then that depends on your settings:
www.linuxcnc.org/docview/html//common_User_Concepts.html

I can't tell from your backplot what your describing.

John

Please Log in or Create an account to join the conversation.

  • Der3er
  • Der3er's Avatar Topic Author
  • Visitor
  • Visitor
30 Mar 2010 11:39 #2481 by Der3er
Thanks! Then it seems that preventing this behaviour would need a G64 P0.10 to ensure a path accuracy that is always at least 1/10 of a millimeter?

Please Log in or Create an account to join the conversation.

More
30 Mar 2010 11:42 #2482 by BigJohnT
Yes, using G64 Pn will give you best possible speed within Pn tolerance. I have to use it on my plasma cutter for the naive cam detctor as plasma files are usually a bunch of little lines.

John

Please Log in or Create an account to join the conversation.

  • Der3er
  • Der3er's Avatar Topic Author
  • Visitor
  • Visitor
30 Mar 2010 11:50 #2483 by Der3er
Great. Thanks a lot for your help on this one.

Please Log in or Create an account to join the conversation.

  • Der3er
  • Der3er's Avatar Topic Author
  • Visitor
  • Visitor
02 Nov 2010 10:58 #5016 by Der3er
I might add: a G64 P0.01 works well most of the time. Although emc 2.3.0 will ignore this at times, stating in MDI that G64 is active. But instead, it seems like it is working in -unprecise- default mode. The toolpaths differ significantly from original design e.g. holes will not be round and the likes. Rather annoying. When this happens you just need to trash your workpiece and enter a G61. This seems fully functional all of the time. But is really slow to work with. Maybe it is time to install 2.4.4?

Please Log in or Create an account to join the conversation.

More
02 Nov 2010 11:52 #5019 by BigJohnT
EMC does not ignore settings in the program. Enter G61 in the MDI box and you will see the Active G-Codes change to G61. Run a program with G64 and you will see the Active G-Codes change to G64. At least in 2.4.5 I know it does but I don't have an old version like 2.3.0 to test it on... I seem to recall that G64 was in the Active G-Codes no matter what was active in some old version of EMC... but that is from memory so it may not be correct.

Holes not round can be from several sources including mechanical. You should have your trajectory set up in the preamble of each g code file. If your in G64 mode without the P(best possible speed) and your acceleration is slow then your actual path can be quite different at each direction change than programmed on machines with slow acceleration.

There has been a lot of improvement in EMC since 2.3.0 at the least I would update to the newest 2.3 version at the least as there is usually some bug fixes right after a new release. 2.4.5 is quite a bit improved over 2.3.x...

G64 (best possible speed) is the default on start up of EMC but it does not mean lack of precision.

G64 without P means to keep the best speed possible, no matter how far away from the programmed point you end up.


Yea, G61 is slow as it comes to a complete stop at the end of each move.

Keep in mind that the backplot is not as precise as the axis movement as it is not updated in real time so the path shown around corners may not be the actual path that the axis takes.

John

Please Log in or Create an account to join the conversation.

  • Der3er
  • Der3er's Avatar Topic Author
  • Visitor
  • Visitor
04 Nov 2010 17:51 #5081 by Der3er
Thanks a lot John. Fully understood. What I was referring to, though, was a specific behaviour of the g64 command: Surely a change from g61 to g64 is reflected in the display of active g-code. What is not displayed is the precision level of the g64 px.xx. So while I enter a g64 p0.02 I would expect that precison. The issue is: There is no way of knowing what precison level is active (until you see the first erratic move, that is). Now sometimes a p0.02 works (workpieces fitting together nicely) and sometimes it don't. Same g-code, same machine. Usually, running g-code at a precison of 0.02 works fine for my workpieces. What I was trying to point out was that while a g61 and a g64 p00.2 should produce almost the same result, in reality the g64 p0.02 does not work every time. sometimes it does, sometimes it don't. Hope this issue is resolved in 2.4x - I'll be updating soon...

Regards

Please Log in or Create an account to join the conversation.

More
04 Nov 2010 19:34 - 04 Nov 2010 19:35 #5083 by BigJohnT
I see this in the release notes for 2.3.0 "fix problem with full circles with G64P- " and that is the only entry for G64 that I can find.

I've looked before at the Active G Codes window thinking I might find the G64 Pn.n but only found the G64. Maybe someone would add it to the DRO tab in 2.5 as there is a lot of info on that tab including offsets that is very handy. You should still put G64Pn.n in the preamble of each file.

I don't recall any path issues with G64 Pn.n so after you upgrade to 2.4.5 and you still have part problems I would investigate the mechanics of your machine.

Don't miss the upgrading instructions from 2.3 to 2.4

wiki.linuxcnc.org/cgi-bin/emcinfo.pl?UpdatingTo2.4

John
Last edit: 04 Nov 2010 19:35 by BigJohnT.

Please Log in or Create an account to join the conversation.

Time to create page: 0.141 seconds
Powered by Kunena Forum