GearHeads Corner

Hi Guys:

  New version has a reattempt at a fix for the geaartrain calculator to give more results.
It also has another version of Auggie ( 1.0002)  that has more crash sources repaired.

Thx
Art

Thanks for the geartrain enhancement, it works great!

Mark

Mark:

Thx for the confirmation. :)

Art

windows 10 does not like geatotic had to put a exception in, once that's done it's fine.

That's the we're all gonna die of virii if you install this abomination from an untrusted source popup, it's normal unless you go through the reportedly expensive (buying a key iirc) M$ tax.  Since we all trust Art and shake our heads at the warning, and can't wait for the next major wonderfulness, it's safely ignored.

Once past the red screen of impending doom it installs and starts fine on my laptop, W10 64b insider.  Didn't try anything, got roped into doing some 3D printing, which involved rebuilding and reflashing the printer's firmware, ran out of time for poking around.  Darn.  At least managed to lay out a driver interface board yesterday, gave me a chance to try Fritzing, which was actually kinda fun.

Is there anywhere I can go to learn more about Tempest, now I've read through the mach3 forum discussion from 2009?

Kirk

DanL:

AVG Antivirus doesn't like it either.  Like you, I had to allow exceptions with Win 10.  My computer didn't explode and didn't get sick so I guess this Gearotic software is OK.  :D

for me AVG had no problem, just win10. Art will not put too many virus on our computer's just one to make it explode if we piss him off. ;D

there is nothing worse than a virus that will blow a computer up had that one, lucky for me it was winter so was able to pull the plug before I had magic smoke.

Kirk:

>>Is there anywhere I can go to learn more about Tempest,

  No. Tempest was my second attempt at higher order motion planning after I retired. I never finished it but released it
as a tech demo in a special version of Mach3 for heavy machines. It was completed only for running, but it couldnt pause or stop,
it had to run the entire file. It was finished only last month with the addition of the continue processor. It was used by several
people over the years since it was written but only by those that could accept its limitations.  I figured it was a no brainer to use it
as I had probably spent a year writing it..at the time EMC was planning on a quintic planner ( or so went the rumor) so I
investigated and wrote Tempest in case I would one day need it. Most I think couldnt see a difference from Mach3, but those that could
saw a big difference. I had many requests for its completion.

  As its late, and Im tired of coding tonight, , Ill make a few notes on it for interests and archival sake..

Tempest is 9 axis, we're using 4. The 9 axis are divided into 3 sets of 3, x,y,z, - a,b,c and u,v,w.

There are 3 multiaxis modes..

1) Orientation mode.
  The speed planning is done for x,y,z with A as an orientation axis, so the planning is just x,y,z , and then A is planned at the same feedrate. If A takes longer than x,y,z, then x,y,z are stretched to fit A's timing, otherwise A's tiem is stretched to fit x,y,z.

2) Feedrate on circumferance mode

    The linear distance of the rotary axis is computed using the Z axis as the distance from rotary zero, and the speed is then planned
as multidimensional distance on the set feedrate for all axis

3) Feedrate on Set Radius Mode

  as above but the Z axis is ignored and the radius is set in the runtime config bar.

  When tempest plans moves, it replans all moves except the one in motion when any move is added to it, this is necessary as each
change in speed is controlled by jerk, so max speeds can only be realized by re-planning back and forth in a wave motion across
the queued up moves to reach maximum interwaypoint velocities. Due to using sinusoidal accelerations there is no easy way to
predetermine motion times so when a bottleneck is seen, its this planning process where it will be found. Each motion is a combination of up to 7 quintic coefficients depending on the complexity of the speed profile. Tempest's motion should compare well to any system
of motion on a modern machine, I would think. No-one sane would go higher than 6th order smoothing.. well, maybe some
electron microscope slide system or something... :) 

  When lines are added to lines, they may be joined by a 6th order polynomial, or a hermite curve, this is locked to polynomial presently,
and any lne/arc, arc/line or arc/arc blending is done only by 3D hermite curves. The planner plans in 1ms trajectories using a
taylor series expansion of the quintics precalculated on each of the optimisations done during the addition of
motion commands. So planning can be a bottleneck, but actual running of the path is quite quick.

  I advise not using too high a look ahead as a system may get bogged down in optimization stages of the planner.

    All that having been said, tempest is young in terms of user numbers, so Ill watch it with interest to see how well it works
across systems.. Id hate to have to troubleshoot it, Im not sure I fully understand how it works anymore...like Mach3's
printer port driver, I look at the code and wonder what I was thinking...

Art




 

[quote author=ArtF link=topic=1432.msg10050#msg10050 date=1451441031I look at the code and wonder what I was thinking...
[/quote]
Back when I really was a software developer (in the days of tractor feed paper and when talking about 026s didn't get blank stares) I'd have that problem the next day.  Sometimes the same afternoon...

Now I have to go learn about quintic functions :)

Thanks for the write up, Art!

Got a first pass on the driver board, ordered up a few more parts (decided to use connectors rather than cut and solder wires from the ribbon cable), postponed the date with the magic smoke.  Still have the kids around slowing down any actual Auggying, although they did coerce me into reflashing the 3D printer so the low friction trapezoidal screw mod would work.

Kirk