The only file I have run so far that has caused accelleration overages 
is tort.ngc.  (this is a lot of arc-arc arc-line)

sam
On 11/22/2013 7:48 AM, sam sokolik wrote:
> Ok - initial testing.
>
> I ran spiral, splash, and 3d_chips.  (the slow feeds that I was seeing
> on 3d_chips seems to be gone)
>
> No constraint violations yet so far in all of them (I thought I was
> seeing violations in the splash screen before because of arc-arc but
> maybe I was smoking or it was actually something else.....))
>
> So - awesome!!  nice work!
>
> For a comparison
> -3dchips with the original blending takes 4min 12 seconds - new takes
> 2min 11 seconds.
> -Full spiral takes 1 min 15 sec  - new takes 23seconds.
>
> So - wow.
>
> Bad news .  single block still isn't right - but it is better.  when you
> pause the program - then click - run next line - it just runs till the
> end of the program.
>
> It looks like you have just optimized for 100% feed rate and don't allow
> anything higher.   That seems like a good solution until the bugs are
> worked out.  The programs so far seem to end normally with FO set to
> something like 200 also.
>
> Questions.
> -It looks like arc-arc you are falling back to parabolic blends.  (I ran
> arcspiral.ngc  and it runs exactly the same speed as the old tp) - is
> this something you hope to also tackle? (iirc it is on your list)
>
> -I don't have a way to test it right now - but I wonder how output
> control works now ie
> http://www.linuxcnc.org/docs/devel/html/gcode/m-code.html#sec:M62-M65
> and
> http://www.linuxcnc.org/docs/devel/html/gcode/m-code.html#sec:M67-Analog-Output
>
> Again - nice work!
> sam
>
>
>
>
> On 11/22/2013 1:21 AM, Robert Ellenberg wrote:
>> That's unfortunately an issue with parabolic blends, so my fixes wouldn't
>> help there. I wouldn't hold back though, anything you can do to break it is
>> a good thing!
>>
>>
>> On Thu, Nov 21, 2013 at 10:00 PM, sam sokolik <[email protected]>wrote:
>>
>>> does this work arc-arc also - or should I stay away from those
>>> programs?  (or did that get fixed with teh split-cycle method?)
>>>
>>> can't wait to bang on it
>>>
>>> thanks
>>> sam
>>> On 11/21/2013 04:19 PM, Robert Ellenberg wrote:
>>>> On Nov 19, 2013 12:18 PM, "sam sokolik" <[email protected]> wrote:
>>>>> Heh - Ok.  I have been playing with it and the motion is a lot
>>>>> smoother.  The velocity on the spiral is a strait line (used to be
>>>>> scaloped..)
>>>>>
>>>>> Couple more little issues..  :)
>>>>> On the spiral - if you pause it - then single block it.  (try it - I
>>>>> dare you) ;) - it will take off in some random direction - or do a
>>>>> circle over and over.  (kinda looks like the maybe the last blend move
>>>>> gets extended...  )  Then when you take it out of single block - you get
>>>>> a large step change in acelleration/velocity as it goes to where it
>>>>> supposed to be instantly and continues on.
>>>> This should be fixed in the latest push.
>>>>
>>>>> I don't know how the feedrate override should act.  I had mentioned that
>>>>> it didn't violate the constraints when I want to 120% - (its speed went
>>>>> up also 290ipm vs 350ipm.  But then I thought - ok - so then I should be
>>>>> able to set the FO max to 100% in the ini and get 350ipm...  but It
>>>>> doesn't - I still get 290ipm.  (I was thinking it calculated worse case
>>>>> for the FO)  So there must be some headroom in there.
>>>>>
>>>>> Ok - thinking out loud...  I just set the FO in the ini to 200%.  I was
>>>>> thinking I would break it being able to go over - but you have some sort
>>>>> of checking in there (more like the original planner) where it seems to
>>>>> stop at 350ipm and doesn't go over acceleration constraints.  That is
>>>>> cool.  Maybe you have 120% hard coded?  huh? Do you?  ;)
>>>>>
>>>>> Well - one more thing - setting the feed rate override to 200% cause the
>>>>> program not to end like before.  (the motion ends but axis still thinks
>>>>> it is running the program - the play button is pushed.) ie..
>>>>> http://imagebin.org/277887  Notice though - the constraints are
>>>>> obeyed...  Yay
>>>> Yeah, feed override needs more work. The current version should at a
>>>> minimum not fail under feed override, but it doesn't give optimal
>>> behavior
>>>> either. If we can get to the point of there being no acceleration
>>>> violations under any circumstances, the next step is to improve the
>>>> override. I just don't want to build additional stuff on top of a buggy
>>>> base, then have to debug both at the same time.
>>>>
>>>> The good news is, I found a way to eliminate overshoot of trajectory
>>>> segments. Previous versions had a bunch of band-aid fixes to prevent
>>>> overshoot from causing too much damage, but made if very difficult to
>>>> control acceleration during tangent "joints". The new "split-cycle"
>>> method
>>>> uses the exact time that a segment ends, and updates the next one with
>>> the
>>>> difference.
>>>>
>>>> If you have some time, could you give it a quick sanity check? I don't
>>> need
>>>> exhaustive testing for another few days, but a quick check to see if you
>>>> can break it would be very helpful. Stepping, crazy feed overrides,
>>> spindle
>>>> start/stop, coolant, whatever you can think of to trip it up is fair
>>> game.
>>>>     My goal is to get this on hardware ASAP.
>>>>
>>>> Thanks!
>>>>
>>>> Rob
>>>>
>>> ------------------------------------------------------------------------------
>>>> Shape the Mobile Experience: Free Subscription
>>>> Software experts and developers: Be at the forefront of tech innovation.
>>>> Intel(R) Software Adrenaline delivers strategic insight and game-changing
>>>> conversations that shape the rapidly evolving mobile landscape. Sign up
>>> now.
>>> http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/ostg.clktrk
>>>> _______________________________________________
>>>> Emc-developers mailing list
>>>> [email protected]
>>>> https://lists.sourceforge.net/lists/listinfo/emc-developers
>>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Shape the Mobile Experience: Free Subscription
>>> Software experts and developers: Be at the forefront of tech innovation.
>>> Intel(R) Software Adrenaline delivers strategic insight and game-changing
>>> conversations that shape the rapidly evolving mobile landscape. Sign up
>>> now.
>>> http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/ostg.clktrk
>>> _______________________________________________
>>> Emc-developers mailing list
>>> [email protected]
>>> https://lists.sourceforge.net/lists/listinfo/emc-developers
>>>
>> ------------------------------------------------------------------------------
>> Shape the Mobile Experience: Free Subscription
>> Software experts and developers: Be at the forefront of tech innovation.
>> Intel(R) Software Adrenaline delivers strategic insight and game-changing
>> conversations that shape the rapidly evolving mobile landscape. Sign up now.
>> http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/ostg.clktrk
>> _______________________________________________
>> Emc-developers mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/emc-developers
>>
>>
>
> ------------------------------------------------------------------------------
> Shape the Mobile Experience: Free Subscription
> Software experts and developers: Be at the forefront of tech innovation.
> Intel(R) Software Adrenaline delivers strategic insight and game-changing
> conversations that shape the rapidly evolving mobile landscape. Sign up now.
> http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/ostg.clktrk
> _______________________________________________
> Emc-developers mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/emc-developers
>
>


------------------------------------------------------------------------------
Shape the Mobile Experience: Free Subscription
Software experts and developers: Be at the forefront of tech innovation.
Intel(R) Software Adrenaline delivers strategic insight and game-changing 
conversations that shape the rapidly evolving mobile landscape. Sign up now. 
http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/ostg.clktrk
_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers

Reply via email to