On Friday 02 October 2015 07:40:36 EBo wrote:

> On Oct 2 2015 2:54 AM, Gene Heskett wrote:
> > On Thursday 01 October 2015 19:56:42 Gene Heskett wrote:
> >> > On 10/1/2015 5:34 PM, Gene Heskett wrote:
> >> > > Greetings all;
> >> > >
> >> > > This g-code to carve the joints in a blanket chest, got
> >> > > somewhat refactored over the last couple of months as I tried
> >> > > to move the math heavy stuff out of the subroutines in favor of
> >> > > doing  some
> >>
> >> of
> >>
> >> > > it in the initilazation phase, setting vars that the
> >> > > subroutines can use. And I am not done just yet, there are
> >> > > still some math
> >>
> >> ops
> >>
> >> > > But in the process, the backplot has gotten a little funkity,
> >> > > drawing its red lines during the program's execution at about
> >>
> >> the
> >>
> >> > > equ of the tool #1 radius, (0.125") down and left of the
> >>
> >> original
> >>
> >> > > wireframe drawing.
> >
> > Actually, I did some rudimentary measuring using the onscreeen dro.
> > the x
> > offset between backplot and actual red execution trace is -0.0625".
> >
> > The y offset is 2x that, -0.125
> >
> > I took all the G9x.x out of the preamble in my code, and restarted
> > the
> > sim version just now to reset it to a known state, reloaded the code
> > and
> > ran it again, with exactly the same offset results.
> >
> > At least its consistent. :(
> >
> > Thanks John.  Any more ideas?
>
> consistency is good.  It will allow us to actually debug the thing.
>
> One idea is to remove as much as possible from the code to get a
> minimal replication set.  The smaller the code which replicates the
> behaviour, the easier it will be for someone to trace through it.  Can
> you get it down to a 5 line program?  10 lines? 100 lines?...
>
>    Just a thought.
>
>    EBo --

Currently 341 LOC if the comment lines are included.  So not very easily.

As I was waking up, I had a thought.  This code uses 2 different tools, 
and I am g38.2 measuring the side of the tool in both X & Y, and thats 
figured into the seed_value that establishes the upper left corner of 
the pattern carved.  I should offset the tool_radius into the values I 
set into G55 with an L1- L2 P2 setting, which should then center the 
tool trace on the backplot as I am NOT applying the tool table offset 
anyplace. I haven't had any great amount of luck trying to use a too, 
compensation move as even if it has inches to do the initial move, the 
error message is always that there is insufficient room to do the move.  
So I do climb cuts only in most code I write, and do my own tool path 
offsets in  what gets sent to the planner.  And unlike tool comp, it 
JUst Works(TM) if I don't muck it up in my math.

Its the second, roundover bit that will turn into a problem child with 
that approach but thats not something I cannot solve because there is a 
different code path for it in the subroutine o<find_jig>.  Its effective 
tool_radius at the tip is maybe 50 thou, and hard to measure. And 
whereas the tool_radius is set at about 0.122" for the 1/4" tool, which 
adjusts the size of this in a manner that leaves a glue joint fit when 
the code is inverted to carve the ends of the chest with mating Green & 
Green style fingers, effectively huge box joints.  The pine test wood 
seems to have a rebound that tightens the joint fit, whereas the real 
mahogany gets about the same fit if the tool_radius is set about a thou 
closer, but again, these are details to be set as vars in the preamble 
and something I've not yet made automatic but probably will before I 
actually start carving mahogany.  They will also change some with the 
sharpness of the tool as a dull tool needs to be pushed harder to get 
the same result.  An amplified version of what happens in much more 
rigid metal I think.  The 1/4" carving is done in 3 steps, but I figure 
the tool is getting dull when I can see the step in depth in the 
finished cut.  The roundover bit pass should only need one pass as is 
now programmed.

There's more to consider of course, like its path going around a corner 
the 1/4" tool carved with a 1/8" radius where in inside corner is 
straight 90 degree path, but the tool carves a tool_diameter radius arc 
for the ends of the fingers, giving them a nice roundover that is the 
same outside radius as that of the tools minimum arc as it makes a sharp 
90 at the inside end  of the of the pocket. So the radius of that arc 
the roundover bit follows must be expanded by its tip radius so it 
effectively follows the the radius left by the 1/4" tool.

This code is complex enough that when I had it doing the math in the 
carving motions subroutines, pauses in motion could be seem because it 
was too busy doing the math.  So I am gradually moving to a model where 
all the math for these offsets is being done in the inital pass through 
o<find_jig> and all the actual machine moving code has to do is lookup 
the #<_vars> holding what its supposed to be doing.

So that is what I'll be working on today, in addition to seeing what I 
can get done with flipping the jigs gauging bar with its G38.2 contacts, 
in and out of position with an m7-8-9 command eventually.  This doesn't 
currently have coolant or mist, so the mist and flood buttons can be 
re-assigned in hardware to both flip the jig as needed, and start/stop a 
vacuum to help keep me from drowning in sawdust.  The jig must be clean 
when the gauging bar comes down else it sitting on just one bit of wood 
swarf and the precision goes out the window.  I need a smaller version 
of Oneida's Dust Deputy to interpose in the vacuum hose, one sized to 
fit the suck of a Bucket Max vacuum head.  Those things are amazingly 
efficient but huge. I've had one hanging on the side of a 12 gallon 
shop-vac, and have dumped its 5 gallon catch bucket several times while 
hooked up to the table saw's rear port, and the bottom of the 12 gallons 
big bucket has remained clean enough to see most of the bottom.  The 
Dust Deputy caught 99.95% of the stuff from the saw.

But I've not yet carved up a vacuum nozzle to fit the mill either.

I also found some more cameras that I might be able to use with camview. 
Cheaper than the so called colonoscopy things they sold for $22 a couple 
years ago on fleabay.  Based on history, I bought 4 of them so's I'd 
have a spare  These are small enough I can hide it inside the head 
casting where it should be protected from errant hold-down bolts, unlike 
the teeny one I had on the toy mill till this jigs right hand end board 
clamp bolt knocked it off.

Some days, its just not worth it to chew thru the straps and get up to 
pee & make coffee...  Buying a new mill is like getting a pet, they need 
care, feeding and petting on a constant basis it seems.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>

------------------------------------------------------------------------------
_______________________________________________
Emc-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-developers

Reply via email to