[Emc-developers] Got a bug.

[Gene Heskett]

Greetings everybody;

Running LINUXCNC - 2.8.0-pre1-3215-gf53ece1, on the pi-3b.

Tonight I started to rewrite this square thread program, and thought I'd
be cute and specify the pre-turn to size before actually cutting the
thread, using a z feed per revolution, aka G33. Doesn't work, sits at
the g33 line forever w/o moving.

Another routine that uses G76 to carve the thread, works as expected. I
think that validates my hal connections to the encoders index, but its
running in quadrature mode, not single channel counter as it says in the
docs currently shipping. So are the other 2 machines that have encoders.

2 other machines have run the G33.1 for rigid tapping, but I was hurting,
both feet and back by then, so didn't have the sense to check them to
see if it still works on them.

The armhf version has a few other gotchas. There are pins that a
halmeter or halconfig cannot see, but the halscope can, and vice versa.
No clue how hard it might be to smack a bug like that though.

Is there something in the hal file dept that G33 needs but G76 doesn't?

Thanks guys.

Cheers, Gene Heskett

[andy pugh]

Post by Gene Heskett
thought I'd
be cute and specify the pre-turn to size before actually cutting the
thread, using a z feed per revolution, aka G33. Doesn't work, sits at
the g33 line forever w/o moving.

For turning work it would be more usual to use the G95 feed mode for
feed-per-rev.

Possibly the spindle speed was too high for the acceleration limits?
(G95 does not trouble itself with accurate position synch)

[Gene Heskett]

Post by andy pugh

Post by Gene Heskett
thought I'd
be cute and specify the pre-turn to size before actually cutting the
thread, using a z feed per revolution, aka G33. Doesn't work, sits
at the g33 line forever w/o moving.

For turning work it would be more usual to use the G95 feed mode for
feed-per-rev.
Possibly the spindle speed was too high for the acceleration limits?
(G95 does not trouble itself with accurate position synch)

300 rpms? This is plastic water pipe, don't want it to get hot. But
I'll check G95 out as this isn't thread cutting.

The docs say motion.spindle-speed-in must be connected. I thought it was
but at 50 revs, a halmeter says 0.0000. What does it expect to be
connected to? I've a filtered feed to the pyvcp tach, in rpms? Or does
it need rps?, I have both available.

Cheers, Gene Heskett

[andy pugh]

Post by Gene Heskett
The docs say motion.spindle-speed-in must be connected. I thought it was
but at 50 revs, a halmeter says 0.0000. What does it expect to be
connected to? I've a filtered feed to the pyvcp tach, in rpms? Or does
it need rps?

RPS

[Gene Heskett]

Post by Gene Heskett
The docs say motion.spindle-speed-in must be connected. I thought it
was but at 50 revs, a halmeter says 0.0000. What does it expect to
be connected to? I've a filtered feed to the pyvcp tach, in rpms?
Or does it need rps?

RPS

Which would be hm2_[HOSTMOT2](BOARD).0.encoder.00.velocity,
which is noisy, 40% peak to peak running at about 25 rpms. That will make
Z sound very rough, but if it can keep up, I should get a better thread.

I've an idea how to not average, but to adjust the gain of a sum2 with a
4 cycle counter initially preset by the index pulse so the counter is in
synch with the 4 step noise.

I do have an averaging lashup that effectively divides this noise by 4,
in place in the TLM and G0704. But the lag that creates is an offset in
the taps depth that makes the threads a looser fit. An additional thou
in the backlash might help that, but that has not been tried.

If the addf's are in the right order, the lag of this idea, if I can make
hal code out of it, would not be a factor in a G33.1 turnaround. I would
need to make the 4 step counter run backwards then the spindle's
absolute motion was backwards.

Anyway, go see if the g33 now works, but first, make a new tool or
regrind this one for a longer left edge as it needs to leave a clean
shoulder at the left end of the first cut. The right edge only needs
the depth of the thread, and thats only .075". Barrel and action
wrenches s/b here tommorrow so I'll be able to test fit this pvc pipe.
If it fits, I do it to the end of this SS barrel. This being the 4th,
its a national holiday on this piece of dirt. I already put out the
flag. :)

Is anyone selling steady rest sliders with copper or perhaps oilite
contact surfaces on the tips that could be used in place of the steel
ones? All I can find today are rolling bearings or brass and the brass
ones are a bit small, 55mm long for a smaller myford. Rolling bearings
I have used but swarf can knock then out of kilter.

Thank you Andy.

Cheers, Gene Heskett

[andy pugh]

Post by Gene Heskett
Which would be hm2_[HOSTMOT2](BOARD).0.encoder.00.velocity,
which is noisy, 40% peak to peak running at about 25 rpms. That will make
Z sound very rough, but if it can keep up, I should get a better thread.

G33, G33.1 and G76 use motion.spindle-revs, and slave the Z to the
spindle position.

G95 feed mode uses motion.spindle-velocity-in.

Two different things. I think we are talking at crossed-purposes.

[Gene Heskett]

Post by andy pugh

Post by Gene Heskett
Which would be hm2_[HOSTMOT2](BOARD).0.encoder.00.velocity,
which is noisy, 40% peak to peak running at about 25 rpms. That will
make Z sound very rough, but if it can keep up, I should get a
better thread.

G33, G33.1 and G76 use motion.spindle-revs, and slave the Z to the
spindle position.
G95 feed mode uses motion.spindle-velocity-in.
Two different things. I think we are talking at crossed-purposes.

Entirely possible, Andy. And I still haven't checked, got side tracked
trying to grind a tool, and I think I've wasted my time, burning up a
5/16" piece of tool steel because the alox wheel got dull and heated it
up to bright red, several times. I cleaned a dressed the wheel several
times, and it wound up exploding while just spinning free. And I am not
at all sure a .0485" wide tool with a .200 stickout, will survive
makeing threads .075" deep in SS. I don't figure on any more than 5 thou
per pass. And a tool I could shave with. I need a tool grinder but I
can't see the $750-$850 for a well equipt one with a diamond cup wheel.

Dremel makes a nice disk, but the %$#@^@& mandrel and disk are special,
and flex like cooked spaghetti. I've got 2 of them stacked for improved
rigidity but the dremels lowest speed still ablates the diamond in just
a few minutes run time. And they are $20 a disk! Sigh...

Cheers, Gene Heskett