Re: [Emc-users] correct x error from bed wear

[Chris Albertson]

This has already been done on Merlin.  Merlin is the software that
interpret g-code for most 3D printers.

The way it works is there is a distance sensor on the print head and it
probes the entire bed in a grid pattern.  It ALWAYS finds that the bed is
not level and not flat.   There are several kinds of probes sorted.   One
is a simple microswitch.

Then after probing the bed and seeing that  it is not flat or square it
computers a transformation matrix and then later all X,Y,Z values in the
g-code are transformed.  In the end you have a part that is perpendicular
to the bed but maybe not vertical.Depending on where on the bed your
part is it might be leaning in a different direction.

It works well within reason.  Actual errors need to be only a few
thousands.  In my case the bad is slightly bowel shape with the =center
being about 0.2mm deeper than the corners.  This level of error is easily
corrected.

On Sat, Sep 8, 2018 at 12:23 PM Gene Heskett  wrote:

> Greetings all;
>
> I am about to give up on this bedwear comp project.
>
> How I determine the amount of correction needed has been tried by
> watching the dial as Z is moved, taking notes as to which way the center
> of the wobble (the spindle is running about 15 rpms) moves, and putting
> that DRO's RAD in the hal files lincurve "setp" list.  Makes it worse,
> change sign of lincurve y-val, still worse.  Seems like the correction
> is being multiplied by 3 or more.
>
> I have run it to a lincurve X-val-nn point, and using the jog dial,
> centered the dials wobble on zero, then put the obtained rad into a
> y-val-nn, again making it worse with either sign.
>
> So how do you folks derive the correction needed?
>
> I'm assuming the offset itself is in radius, not diameter. In which case
> the needed radius correction max's at about 2.5 thou. And that the sign
> is the "tricky" part.
>
> Thanks everybody.
>
> --
> 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 
>
>
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>


-- 

Chris Albertson
Redondo Beach, California

___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Gene Heskett]

On Saturday 08 September 2018 18:30:42 Chris Albertson wrote:

> This has already been done on Merlin.  Merlin is the software that
> interpret g-code for most 3D printers.
>
I see, but that also is dealing with what effectively is a plain one 
dimension rendered in 3d.

A lathe is a slightly different critter, and the major src of the error 
is the tilting of the carriage in and out due to different amounts of 
wear front and rear, so the correction consists of moving a dial along 
the centerline height, and measuring this in and out wobble the tilting 
creates and applying an in or out correction of 3 thou maximum, so the 
cutting tool effectively follows a straight line.

Due to the wear createing a height of tool motion, its acknowledged but 
unless working on a eighth inch or smaller part, this height change is 
surely under 0.001" in effect on the part unless its super teeny.
 
> The way it works is there is a distance sensor on the print head and
> it probes the entire bed in a grid pattern.  It ALWAYS finds that the
> bed is not level and not flat.   There are several kinds of probes
> sorted.   One is a simple microswitch.
>
> Then after probing the bed and seeing that  it is not flat or square
> it computers a transformation matrix and then later all X,Y,Z values
> in the g-code are transformed.  In the end you have a part that is
> perpendicular to the bed but maybe not vertical.Depending on where
> on the bed your part is it might be leaning in a different direction.
>
Again, an effect that on a lathe, that is quite miniscule, and with only 
stepper motors to microstep control, plus the lash in the ball nuts and 
thrust bearings, is quite low. So low that its a non-problem if not 
working on the hubble mirror.

> It works well within reason.  Actual errors need to be only a few
> thousands.  In my case the bad is slightly bowel shape with the
> =center being about 0.2mm deeper than the corners.  This level of
> error is easily corrected.

I assume you meant "bowl" shaped. :)
 
Do we have a receiver/logger module for the digital output of a scale I 
intended to put on the tailstock barrel, one of Shar's lower cost models 
with a remote display. This, fed to some math to detect the high and low 
spots and how high or low which would allow the error to be logged. I 
could mount it in a tool holder and put a teeny ball bearing to ride the 
calibration rod, and rubber band its slider toward the rod.  Its a 
thought anyway. I'd have to obtain the data cable, or just log it from 
its own display every 1/4" if no recording module is available.

At least I would have a good starting point, which may in fact be "good 
enough for the girls I go with".

Thanks Chris.
>
> On Sat, Sep 8, 2018 at 12:23 PM Gene Heskett  
wrote:
> > Greetings all;
> >
> > I am about to give up on this bedwear comp project.
> >
> > How I determine the amount of correction needed has been tried by
> > watching the dial as Z is moved, taking notes as to which way the
> > center of the wobble (the spindle is running about 15 rpms) moves,
> > and putting that DRO's RAD in the hal files lincurve "setp" list. 
> > Makes it worse, change sign of lincurve y-val, still worse.  Seems
> > like the correction is being multiplied by 3 or more.
> >
> > I have run it to a lincurve X-val-nn point, and using the jog dial,
> > centered the dials wobble on zero, then put the obtained rad into a
> > y-val-nn, again making it worse with either sign.
> >
> > So how do you folks derive the correction needed?
> >
> > I'm assuming the offset itself is in radius, not diameter. In which
> > case the needed radius correction max's at about 2.5 thou. And that
> > the sign is the "tricky" part.
> >
> > Thanks everybody.
> >
> > --
> > 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 
> >
> >
> > ___
> > Emc-users mailing list
> > Emc-users@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/emc-users



-- 
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 


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[jeremy youngs]

Gene , can you mount a probe on the carriage and probe a lathe alignment
bar ( or drill rod )?

___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Phillip Carter]

On 9/9/18 5:18 am, Gene Heskett wrote:

Greetings all;

I am about to give up on this bedwear comp project.

How I determine the amount of correction needed has been tried by
watching the dial as Z is moved, taking notes as to which way the center
of the wobble (the spindle is running about 15 rpms) moves, and putting
that DRO's RAD in the hal files lincurve "setp" list.  Makes it worse,
change sign of lincurve y-val, still worse.  Seems like the correction
is being multiplied by 3 or more.

I have run it to a lincurve X-val-nn point, and using the jog dial,
centered the dials wobble on zero, then put the obtained rad into a
y-val-nn, again making it worse with either sign.

So how do you folks derive the correction needed?

I'm assuming the offset itself is in radius, not diameter. In which case
the needed radius correction max's at about 2.5 thou. And that the sign
is the "tricky" part.

Thanks everybody.

This lathe bedwear may be a good candidate for the 
dgarr/external_offsets branch.


Cheers, Phill



___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Gene Heskett]

On Saturday 08 September 2018 22:03:21 jeremy youngs wrote:

> Gene , can you mount a probe on the carriage and probe a lathe
> alignment bar ( or drill rod )?

I don't have a probe per sei on it. What I'm doing is sticking a mag base 
to the top of the toolpost holder base that has replaced the mangled 
compound, with atm a dial indicator riding a 21" bar of .500 diameter A2 
rod. Theres a thou or so wobble right at the er40 adapter, and the 
center dimple on the other end is off center about 2 thou riding a live 
center in the tailstock barrel. The A2 is slightly bent so I get about a 
10 thou wobble in the middle of it, but with the spindle turning, the 
true center is easily seen. I have used the lathes x motor to try and 
straighten it better, but by the time I  get enough bow in it to make a 
diff, the rod shortens and pops the end off the live center.

I would do it according to demagnetizeing rules by pushing it far enough 
to bent it back, with the spindle turning, and once I have evidence that 
I am actually bending it, the let it run while I slowly back off the 
push, which should make it run pretty true. But w/o a socket instead of 
a center in the tailstock, I can't get up enough pressure to actually 
bend it straight. A half inch of A2 20+ inches long is still pretty 
strong stuff.

I should buy another 36" piece of 3/4" A2 which would obviously be 
stiffer. But I wouldn't wager it would be straighter. So I'd still have 
to deal with the wobble. And that much A2 isn't cheap.

Setting up a contact probe would trigger on the peaks of the wobble, and 
that measurement would be meaningless w/o knowing the p-p wobble at each 
probed point. I can get that accurate with a dial. I think...

>From what I have observed so far, I think the huge majority of what I 
need to correct is in the first 8" from the spindle, the rest looks like 
a fairly straight line, IF the tailstock is on center. The major error 
once the sled is pulled to center, is vertical. 50 thou off at least.

Thanks Jeremy.
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users



-- 
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 


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Gene Heskett]

On Saturday 08 September 2018 22:34:06 Phillip Carter wrote:

> On 9/9/18 5:18 am, Gene Heskett wrote:
> > Greetings all;
> >
> > I am about to give up on this bedwear comp project.
> >
> > How I determine the amount of correction needed has been tried by
> > watching the dial as Z is moved, taking notes as to which way the
> > center of the wobble (the spindle is running about 15 rpms) moves,
> > and putting that DRO's RAD in the hal files lincurve "setp" list. 
> > Makes it worse, change sign of lincurve y-val, still worse.  Seems
> > like the correction is being multiplied by 3 or more.
> >
> > I have run it to a lincurve X-val-nn point, and using the jog dial,
> > centered the dials wobble on zero, then put the obtained rad into a
> > y-val-nn, again making it worse with either sign.
> >
> > So how do you folks derive the correction needed?
> >
> > I'm assuming the offset itself is in radius, not diameter. In which
> > case the needed radius correction max's at about 2.5 thou. And that
> > the sign is the "tricky" part.
> >
> > Thanks everybody.
>
> This lathe bedwear may be a good candidate for the
> dgarr/external_offsets branch.
>
Humm, can't say I've ever heard of that.  URL?

> Cheers, Phill

Thanks Phill.>
>
>
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users



-- 
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 


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Phillip Carter]

On 9/9/18 1:14 pm, Gene Heskett wrote:

On Saturday 08 September 2018 22:34:06 Phillip Carter wrote:


On 9/9/18 5:18 am, Gene Heskett wrote:

Greetings all;

I am about to give up on this bedwear comp project.

How I determine the amount of correction needed has been tried by
watching the dial as Z is moved, taking notes as to which way the
center of the wobble (the spindle is running about 15 rpms) moves,
and putting that DRO's RAD in the hal files lincurve "setp" list.
Makes it worse, change sign of lincurve y-val, still worse.  Seems
like the correction is being multiplied by 3 or more.

I have run it to a lincurve X-val-nn point, and using the jog dial,
centered the dials wobble on zero, then put the obtained rad into a
y-val-nn, again making it worse with either sign.

So how do you folks derive the correction needed?

I'm assuming the offset itself is in radius, not diameter. In which
case the needed radius correction max's at about 2.5 thou. And that
the sign is the "tricky" part.

Thanks everybody.

This lathe bedwear may be a good candidate for the
dgarr/external_offsets branch.


Humm, can't say I've ever heard of that.  URL?


https://github.com/LinuxCNC/linuxcnc/tree/dgarr/external_offsets

Cheers, Phill

Thanks Phill.>


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users







___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Gene Heskett]

On Sunday 09 September 2018 00:17:32 Phillip Carter wrote:

> On 9/9/18 1:14 pm, Gene Heskett wrote:
> > On Saturday 08 September 2018 22:34:06 Phillip Carter wrote:
> >> On 9/9/18 5:18 am, Gene Heskett wrote:
> >>> Greetings all;
> >>>
> >>> I am about to give up on this bedwear comp project.
> >>>
> >>> How I determine the amount of correction needed has been tried by
> >>> watching the dial as Z is moved, taking notes as to which way the
> >>> center of the wobble (the spindle is running about 15 rpms) moves,
> >>> and putting that DRO's RAD in the hal files lincurve "setp" list.
> >>> Makes it worse, change sign of lincurve y-val, still worse.  Seems
> >>> like the correction is being multiplied by 3 or more.
> >>>
> >>> I have run it to a lincurve X-val-nn point, and using the jog
> >>> dial, centered the dials wobble on zero, then put the obtained rad
> >>> into a y-val-nn, again making it worse with either sign.
> >>>
> >>> So how do you folks derive the correction needed?
> >>>
> >>> I'm assuming the offset itself is in radius, not diameter. In
> >>> which case the needed radius correction max's at about 2.5 thou.
> >>> And that the sign is the "tricky" part.
> >>>
> >>> Thanks everybody.
> >>
> >> This lathe bedwear may be a good candidate for the
> >> dgarr/external_offsets branch.
> >
> > Humm, can't say I've ever heard of that.  URL?
>
> https://github.com/LinuxCNC/linuxcnc/tree/dgarr/external_offsets

Thanks Phill.

Quick inspection, looks like Dewey is 86 commits behind even an uptodate 
2.7.17, and I'm already playing canary in the coal mine now, running 
Master on all 4 machines all the time. Its the least I can offer this 
project, hoping to catch typu and thinko bugs to keep them out of 
2.7.18. And due to a lack of time since I'm doing it all here at the 
coyote.den, including digging up and fixing a broken water line this 
past week, likely not doing a very good job of testing. I still have 
aches and pains from that. Part of the reason I'm awake in the not so 
wee hours.

-- 
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 


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Chris Albertson]

No,  the Software in Merlin is probing the bed using a grid pattern and
modeling the bed's actual shape.. Iet to choose how dense is the sampling
grid.   I have a capacitive sensor but I'm going to hav e to change to a
contact switch because I'm also changing the bad from aluminum to
borosilicate glass

On a lathe I think the hard part is finding a straight reference for X and
Y.  A laser beam would be a VERY straight line

What if there was a 1mm diameter laser beam that shoots through the hole in
the spindle?   Then you mount a camera on the tool post and move the tool
post.  If the lathe is not 100% perfect the laser spot will move relative
to the camera sensor's CCD image sensor. The camera needs to be more
like a microscope then wide angle webcam.   But you can buy a USB interface
microscope now for under $20.  Mount one of those to the tool post.

As for the laser, I actually own one that would work and I've actualy used
it on my lathe.  It has designed for gun owners.  It is the same size as a
.38 round and the idea is you can chamber the laser in the gun and then a
green laser beam shines down the barrel and you can use it to align a
gunsight.   It is powered by a tiny battery. You can make an adaptor to
allow you to "chamber" the round in the spindle hole.

If you mount some ultra-fine graph paper on the tool post and look really
carefully in theory you can see how the bed moves in X and Y vs. in Z.
But there are to practical issues
1) It is really hard to see such small movements.   I literally need a
microscope to see.   I do have a micrometer grid target.  Youcam buy them
in eBay for $13 and yes you need a good microscope to see uM size grid
lines.  The lines are laser etched on glass.   $13 is not muchIt
intended use is for measuring "stuff" you see in a biological microscope.
 "How long is that bacterial?"

2) the cheap laser-bullet thing is not well made.  It is good enough fo
make a gun owner happy but not a machinist who cares about "thousandths" in
other words the runout is horrible.  But you work around this  DOn't bother
to align it. The laser will tract a donut shape on a fixed target.  and
this is good enough because yo know the exact centerline is in the center
of the donut hole.And actually you WANT a donut and not a dot because
with more pixels you get less noise.   The "centroid" is what you are
looking to find.   You can find the center to about 1/10th of a grid line.

No not bother to make the mechanics perfect.  It is easier to just get
"close" and then measure the imperfection.  expect runout and that the
spinning laser will trace out a donut.

Of course you are not actual looking at a laser dot in practice you are
photographing the dot and using software to find the actual centerline.
You use the laser etched grprid only once to calibrate the camera/optics.

Lasers are like cheating when it comes to measurement and alignments.   I
used on two years ago or the first time to make a block wall.   Never again
will I use a mason's line.  lasers don't sag or blow in the wind and even a
rank amateur like be can set blocks to near perfect plumber and level




On Sat, Sep 8, 2018 at 4:18 PM Gene Heskett  wrote:

> On Saturday 08 September 2018 18:30:42 Chris Albertson wrote:
>
> > This has already been done on Merlin.  Merlin is the software that
> > interpret g-code for most 3D printers.
> >
> I see, but that also is dealing with what effectively is a plain one
> dimension rendered in 3d.
>
> A lathe is a slightly different critter, and the major src of the error
> is the tilting of the carriage in and out due to different amounts of
> wear front and rear, so the correction consists of moving a dial along
> the centerline height, and measuring this in and out wobble the tilting
> creates and applying an in or out correction of 3 thou maximum, so the
> cutting tool effectively follows a straight line.
>
> Due to the wear createing a height of tool motion, its acknowledged but
> unless working on a eighth inch or smaller part, this height change is
> surely under 0.001" in effect on the part unless its super teeny.
>
> > The way it works is there is a distance sensor on the print head and
> > it probes the entire bed in a grid pattern.  It ALWAYS finds that the
> > bed is not level and not flat.   There are several kinds of probes
> > sorted.   One is a simple microswitch.
> >
> > Then after probing the bed and seeing that  it is not flat or square
> > it computers a transformation matrix and then later all X,Y,Z values
> > in the g-code are transformed.  In the end you have a part that is
> > perpendicular to the bed but maybe not vertical.Depending on where
> > on the bed your part is it might be leaning in a different direction.
> >
> Again, an effect that on a lathe, that is quite miniscule, and with only
> stepper motors to microstep control, plus the lash in the ball nuts and
> thrust bearings, is quite low. So low that its a non-problem if not
> working on t

Re: [Emc-users] correct x error from bed wear

[Gene Heskett]

On Sunday 09 September 2018 18:14:52 Chris Albertson wrote:

> No,  the Software in Merlin is probing the bed using a grid pattern
> and modeling the bed's actual shape.. Iet to choose how dense is the
> sampling grid.   I have a capacitive sensor but I'm going to hav e to
> change to a contact switch because I'm also changing the bad from
> aluminum to borosilicate glass
>
> On a lathe I think the hard part is finding a straight reference for X
> and Y.  A laser beam would be a VERY straight line
>
Agreed there!

> What if there was a 1mm diameter laser beam that shoots through the
> hole in the spindle?   Then you mount a camera on the tool post and
> move the tool post.  If the lathe is not 100% perfect the laser spot
> will move relative to the camera sensor's CCD image sensor. The
> camera needs to be more like a microscope then wide angle webcam.  
> But you can buy a USB interface microscope now for under $20.  Mount
> one of those to the tool post.
>
> As for the laser, I actually own one that would work and I've actualy
> used it on my lathe.  It has designed for gun owners.  It is the same
> size as a .38 round and the idea is you can chamber the laser in the
> gun and then a green laser beam shines down the barrel and you can use
> it to align a gunsight.   It is powered by a tiny battery. You can
> make an adaptor to allow you to "chamber" the round in the spindle
> hole.

Or it could be chucked in an er40 collet, with the "cartridge" rim pushed 
against the rear of the collet. The problem there is angular since an 
er40 collet has proved to me that it can and will, tilt the equ of 5 
MOA. So the spindle has to be rotating so the beam draws a circle. I 
have a pointer, but its nowhere near straight like the gunsite 
boresighter is expected to be.  As for the camera, I have several that 
might serve if the lens was removed, exposing the ccd to the beam 
directly. Older cameras would be better because the ccd's have shrunk to 
where the beam may even overflow the edges unless it can be tightened up 
beam size wise from a whopping big millimeter.

Its a thought I've explored while drifting off to sleep nights, several 
times. But I've always had a bottleneck shaped "cartridge" in mind, but 
now I'll look for "38" type since they are obviously made.

But right now I just got an order from the missus for a coke float. She's 
dying of COPD, so I always see that famous 2 rule sign about the 
boss. :)

Now, I have just come in from trying to map it again, but I am seeing a 
problem in that the tool location may not equal the centerline of the 
crossfeed, and that can probably effect the correction needed. Humm, but 
thats an absolute carriage position that only changes if I diddle the 
homing switches location. Not terribly likely in the remainder of my 
ride here since I'll be 84 in 3 weeks.

I started out about 1/8" from the er40 collet turning the rod, which is 
about -5.8" from the home switch, touched x off to 0. and logged the 
RAD reading every .2" to about 7" away, at which point I switched to 1" 
steps as I was off the higher wear by then. Finding the tailstock 
off -6.5 thou, which I tried to correct, and found that a cast iron 
bitch to get under a thou, so I need to assume the 6.5 thou error there 
that the tailstock, and scale it down to zero by the time I get back to 
the spindle, then put all those figures into the lincurve, which if I do 
it right, should get me within a thou, at least for the 20" length of 
that A2. And that WILL get it good enough for the girls I go with.

> If you mount some ultra-fine graph paper on the tool post and look
> really carefully in theory you can see how the bed moves in X and Y
> vs. in Z. But there are to practical issues
> 1) It is really hard to see such small movements.   I literally need a
> microscope to see.   I do have a micrometer grid target.  Youcam buy
> them in eBay for $13 and yes you need a good microscope to see uM size
> grid lines.  The lines are laser etched on glass.   $13 is not much   
> It intended use is for measuring "stuff" you see in a biological
> microscope. "How long is that bacterial?"
>
> 2) the cheap laser-bullet thing is not well made.  It is good enough
> fo make a gun owner happy but not a machinist who cares about
> "thousandths" in other words the runout is horrible.  But you work
> around this  DOn't bother to align it. The laser will tract a donut
> shape on a fixed target.  and this is good enough because yo know the
> exact centerline is in the center of the donut hole.And actually
> you WANT a donut and not a dot because with more pixels you get less
> noise.   The "centroid" is what you are looking to find.   You can
> find the center to about 1/10th of a grid line.

That pretty well agrees with my thinking on this, but if a mark for 
angular reference is scratched into the front of the cartridge a sliver 
of reynolds wrap superglued to the "low" side ought to reduce the size 
of the donut if i

Re: [Emc-users] correct x error from bed wear

[Chris Albertson]

The lasers are easy to find on Amazon but be warned the laser is not
perfectly aligned with the housing.But as I said, this does not matter
if you spin the spindle.

I think a laser is the only option.   If you are measuring the bed your
reference needs to be at least 10X straighter then the bed.  Precsion rods
are not good enough but a $20 laser is literally perfect (as long as you
rotate the spindle.)

With effort you can use the same laser to measure the error with the lead
screw pitch. Use it as a laser range finder or laser inferometer

But all this measurement may be moot if the lathe is not reputable.  The
error might be random.  Maybe the carriage moves like a tuck on a dirt road?

On Sun, Sep 9, 2018 at 6:04 PM Gene Heskett  wrote:

> On Sunday 09 September 2018 18:14:52 Chris Albertson wrote:
>
> > No,  the Software in Merlin is probing the bed using a grid pattern
> > and modeling the bed's actual shape.. Iet to choose how dense is the
> > sampling grid.   I have a capacitive sensor but I'm going to hav e to
> > change to a contact switch because I'm also changing the bad from
> > aluminum to borosilicate glass
> >
> > On a lathe I think the hard part is finding a straight reference for X
> > and Y.  A laser beam would be a VERY straight line
> >
> Agreed there!
>
> > What if there was a 1mm diameter laser beam that shoots through the
> > hole in the spindle?   Then you mount a camera on the tool post and
> > move the tool post.  If the lathe is not 100% perfect the laser spot
> > will move relative to the camera sensor's CCD image sensor. The
> > camera needs to be more like a microscope then wide angle webcam.
> > But you can buy a USB interface microscope now for under $20.  Mount
> > one of those to the tool post.
> >
> > As for the laser, I actually own one that would work and I've actualy
> > used it on my lathe.  It has designed for gun owners.  It is the same
> > size as a .38 round and the idea is you can chamber the laser in the
> > gun and then a green laser beam shines down the barrel and you can use
> > it to align a gunsight.   It is powered by a tiny battery. You can
> > make an adaptor to allow you to "chamber" the round in the spindle
> > hole.
>
> Or it could be chucked in an er40 collet, with the "cartridge" rim pushed
> against the rear of the collet. The problem there is angular since an
> er40 collet has proved to me that it can and will, tilt the equ of 5
> MOA. So the spindle has to be rotating so the beam draws a circle. I
> have a pointer, but its nowhere near straight like the gunsite
> boresighter is expected to be.  As for the camera, I have several that
> might serve if the lens was removed, exposing the ccd to the beam
> directly. Older cameras would be better because the ccd's have shrunk to
> where the beam may even overflow the edges unless it can be tightened up
> beam size wise from a whopping big millimeter.
>
> Its a thought I've explored while drifting off to sleep nights, several
> times. But I've always had a bottleneck shaped "cartridge" in mind, but
> now I'll look for "38" type since they are obviously made.
>
> But right now I just got an order from the missus for a coke float. She's
> dying of COPD, so I always see that famous 2 rule sign about the
> boss. :)
>
> Now, I have just come in from trying to map it again, but I am seeing a
> problem in that the tool location may not equal the centerline of the
> crossfeed, and that can probably effect the correction needed. Humm, but
> thats an absolute carriage position that only changes if I diddle the
> homing switches location. Not terribly likely in the remainder of my
> ride here since I'll be 84 in 3 weeks.
>
> I started out about 1/8" from the er40 collet turning the rod, which is
> about -5.8" from the home switch, touched x off to 0. and logged the
> RAD reading every .2" to about 7" away, at which point I switched to 1"
> steps as I was off the higher wear by then. Finding the tailstock
> off -6.5 thou, which I tried to correct, and found that a cast iron
> bitch to get under a thou, so I need to assume the 6.5 thou error there
> that the tailstock, and scale it down to zero by the time I get back to
> the spindle, then put all those figures into the lincurve, which if I do
> it right, should get me within a thou, at least for the 20" length of
> that A2. And that WILL get it good enough for the girls I go with.
>
> > If you mount some ultra-fine graph paper on the tool post and look
> > really carefully in theory you can see how the bed moves in X and Y
> > vs. in Z. But there are to practical issues
> > 1) It is really hard to see such small movements.   I literally need a
> > microscope to see.   I do have a micrometer grid target.  Youcam buy
> > them in eBay for $13 and yes you need a good microscope to see uM size
> > grid lines.  The lines are laser etched on glass.   $13 is not much
> > It intended use is for measuring "stuff" you see in a biological
> > microscope. "How 

Re: [Emc-users] correct x error from bed wear

[Gene Heskett]

On Monday 10 September 2018 03:52:23 Chris Albertson wrote:

> The lasers are easy to find on Amazon but be warned the laser is not
> perfectly aligned with the housing.But as I said, this does not
> matter if you spin the spindle.
>
> I think a laser is the only option.   If you are measuring the bed
> your reference needs to be at least 10X straighter then the bed. 
> Precsion rods are not good enough but a $20 laser is literally perfect
> (as long as you rotate the spindle.)
>
> With effort you can use the same laser to measure the error with the
> lead screw pitch. Use it as a laser range finder or laser inferometer
>
> But all this measurement may be moot if the lathe is not reputable. 
> The error might be random.  Maybe the carriage moves like a tuck on a
> dirt road?

So far, it seems the wear is pretty well confined to the first 6", by the 
time thats been reached the measured offset doesn't seem to change with 
direction of carriage motion. Backlash of x is in the 2.1 thou range as 
its a ball screw with a slight preload in the thrust bearing, as is the 
much larger Z screw, but the backlash there is around 5 thou due to a 
sloppy thrust bearing. Right at the spindle nose, direction of carriage 
travel causes about a 3 thou diff, but that gets well by 4" out. It 
could be a lot worse, and its rather worse on TLM because the head is 
not well trammed to the bed, the whole head casting has been replaced, 
and I'll probably eventually do this to it. Shoemakers kids, lack of 
round tuits and all that...

[...]

-- 
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 


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[andy pugh]

On Sat, 8 Sep 2018 at 20:23, Gene Heskett  wrote:

> change sign of lincurve y-val, still worse.  Seems like the correction
> is being multiplied by 3 or more.

Can you tabulate the input and output of lincurve (measured by
halmeter at the pins).
If it is outputting numbers that are bigger than the input data then
there is something wrong.

-- 
atp
"A motorcycle is a bicycle with a pandemonium attachment and is
designed for the especial use of mechanical geniuses, daredevils and
lunatics."
— George Fitch, Atlanta Constitution Newspaper, 1916


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Gene Heskett]

On Monday 10 September 2018 06:49:38 andy pugh wrote:

> On Sat, 8 Sep 2018 at 20:23, Gene Heskett  wrote:
> > change sign of lincurve y-val, still worse.  Seems like the
> > correction is being multiplied by 3 or more.
>
> Can you tabulate the input and output of lincurve (measured by
> halmeter at the pins).
> If it is outputting numbers that are bigger than the input data then
> there is something wrong.

I'm not pointing a finger in lincurves direction as what I looked at with 
a halmeter looks good. I may have tabulated diameter previously by 
mistake, so I went out yesterday and using a 0.0001" dial, logged the 
RAD to zero the dial every .2" of travel, so I may have better data now. 
Part of the problem now is that when I got 20" down the A2 rod, the 
*^#^%$ tailstock was off by -0.0065". So now I need to take that data 
and normalize it to a well centered tailstock, in proportion to its 
position relative to the left, start. And I can likely throw away those 
that seem to be essentially a straight line, like from -5.8", the zero 
ref, to -4.1 it needs -.0020, then it goes back to only -.0007 at -3.1"

So that, -0.0007, with the tailstock error factored in which would be 
that -.0065 reduced to what it would be at that distance which is (2.9" 
away from the starting point/20)*-0.0065" = -0.0012675 at that point.
so the first x-val-00 is -5.8, the corresponding y-val-00 is 0.0,
the second x-val-01 is -3.9 (1.9" from start) and its corresponding 
y-val-01 is then -0.0006175, not the -0.002 I logged. x-val-02 then is 
at -3.0" from start or 2.8" out from the anchor point and the 
corresponding y-val-02 I logged at -.0007, becomes -0.00161, yadda 
yadda. And I don't think thats right, I'd better do it again. But 
basicly take the peaks observed, use those for x-val-nn and then add the 
absolute distance/20 to get the amount of that -0.0065" offset to add to 
what I logged at that distance. Something along those lines should get 
me well within a thou, and  by throwing away the ramps between the 
peaks, get it all within the 16 point range of a single lincurve module.  
Thats the plan anyway. :) And if I haven't confused everybody by now, 
I'm sorry. ;-)

Somethings gone sour in my cheap air compressor, its peaking at about 47 
lbs and runs till I turn it off, stinking up the garage with the fog of  
burnt oil, so I guess I'll have to run over to Harbor Freights new store 
and get another. They've got a new 8 gallon that can make 150 psi which 
is 60 higher than I've had this one set at for something over a decade. 
I'll set it at about 80 or 90 so it does a decent job of cleaning up and 
it still can fill up a low tire in a decent time.

Old Toy wheels rust & won't hold air and the mandated nag light on the 
dash makes a nuisance of itself, so you go around it and put about 42 
psi in every tire to put the light out for a week, maybe... PITA. But 
I'm not putting $700 in new cheapest steel wheels on an 12 yo Toy RAV4.

-- 
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 


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[jeremy youngs]

On Sat, Sep 8, 2018, 10:14 PM Gene Heskett  wrote:

> The A2 is slightly bent so I get about a
> 10 thou wobble in the middle of it, but with the spindle turning, the
> true center is easily seen. I have used the lathes x motor to try and
> straighten it better, but by the time I  get enough bow in it to make a
> diff, the rod shortens and pops the end off the live center.
>


Do you have a torch? Search heat straightening . You heat and cool with
> water and a rag to shrink the metal on one side and straighten, works well
> , it will probably harden the a 2 though. I'm not certain of temp needed
> but on large shaft it doesn't have to be incandescent.
>
> Incidentally ,this method is capable of straightening bent spindles.
>
>
>
>

___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Gregg Eshelman via Emc-users]

I have a LASER that is made to mount in a chuck or collet. It has alignment 
screws and adjustable focus. Aim it at a wall far away and adjust the focus for 
as small a dot as possible.
Make a target to mount on the carriage, chuck and center the body of the LASER 
then use the alignment screws and adjusting the target and rotating the spindle 
until the spot no longer describes a circle when the spindle is rotated.

Move the carriage to the right end as far as possible. Move the cross slide 
intil the dot is on the target. There's your zero. Advance the carriage toward 
the bed until the dot moves off the target. Adjust the carriage until the dot 
is back on target, at least in the X axis. A really good way to map the bed 
would be to have a sensor as the target so the carriage could be moved by the 
control and read a continuous stream of how far off X it gets. Barring that, 
stop every X distance then adjust to re-zero and interpolate slopes between 
adjustment points.
If you put the target on a vertical slide to re-zero Y at each stop, the error 
map (with appropriate math) could compensate for that too. Perfectly straight 
cutting even with a very worn bed, as long as the vertical error doesn't drop 
the cutting tools too low.

On Monday, September 10, 2018, 1:55:36 AM MDT, Chris Albertson 
 wrote:  
 The lasers are easy to find on Amazon but be warned the laser is not
perfectly aligned with the housing.    But as I said, this does not matter
if you spin the spindle.

I think a laser is the only option.  If you are measuring the bed your
reference needs to be at least 10X straighter then the bed.  Precsion rods
are not good enough but a $20 laser is literally perfect (as long as you
rotate the spindle.)

With effort you can use the same laser to measure the error with the lead
screw pitch. Use it as a laser range finder or laser inferometer

But all this measurement may be moot if the lathe is not reputable.  The
error might be random.  Maybe the carriage moves like a tuck on a dirt road?  
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Chris Albertson]

You can actually make good measurements with a larger spot.  This assumes
you are using software and not just eyeball estimates.   Typically you can
do about 1/10th pixels position estimates.With micro-meter scale pixels
this become way-overkill for machine work.  I would not bother to do more
then roughly align the laser.

Yes you need a sensor.  A $10 webcam with the lens replaced by a welding
glass filter would work but you'd need to experiment with different filter
and laser power.

I would not adjust anything r "zero".  between measurements.  leave the
target and laser bolted down solid.  Just record the numbers and when you
are done cal the smallest number "zero"

To get the center point, just use the "centroid" the same math if looking
for the balance point of the laser spot.

You should be able to make a measurement at least once per lead screw
revolution, maybe many times more per revolution.

On Mon, Sep 10, 2018 at 12:09 PM Gregg Eshelman via Emc-users <
emc-users@lists.sourceforge.net> wrote:

> I have a LASER that is made to mount in a chuck or collet. It has
> alignment screws and adjustable focus. Aim it at a wall far away and adjust
> the focus for as small a dot as possible.
> Make a target to mount on the carriage, chuck and center the body of the
> LASER then use the alignment screws and adjusting the target and rotating
> the spindle until the spot no longer describes a circle when the spindle is
> rotated.
>
> Move the carriage to the right end as far as possible. Move the cross
> slide intil the dot is on the target. There's your zero. Advance the
> carriage toward the bed until the dot moves off the target. Adjust the
> carriage until the dot is back on target, at least in the X axis. A really
> good way to map the bed would be to have a sensor as the target so the
> carriage could be moved by the control and read a continuous stream of how
> far off X it gets. Barring that, stop every X distance then adjust to
> re-zero and interpolate slopes between adjustment points.
> If you put the target on a vertical slide to re-zero Y at each stop, the
> error map (with appropriate math) could compensate for that too. Perfectly
> straight cutting even with a very worn bed, as long as the vertical error
> doesn't drop the cutting tools too low.
>
> On Monday, September 10, 2018, 1:55:36 AM MDT, Chris Albertson <
> albertson.ch...@gmail.com> wrote:
>  The lasers are easy to find on Amazon but be warned the laser is not
> perfectly aligned with the housing.But as I said, this does not matter
> if you spin the spindle.
>
> I think a laser is the only option.  If you are measuring the bed your
> reference needs to be at least 10X straighter then the bed.  Precsion rods
> are not good enough but a $20 laser is literally perfect (as long as you
> rotate the spindle.)
>
> With effort you can use the same laser to measure the error with the lead
> screw pitch. Use it as a laser range finder or laser inferometer
>
> But all this measurement may be moot if the lathe is not reputable.  The
> error might be random.  Maybe the carriage moves like a tuck on a dirt
> road?
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>


-- 

Chris Albertson
Redondo Beach, California

___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Christopher]

> On 2018-Sep-10, at 18:02, Chris Albertson  > wrote:
> 
> You can actually make good measurements with a larger spot.  This assumes
> you are using software and not just eyeball estimates.   Typically you can
> do about 1/10th pixels position estimates.With micro-meter scale pixels
> this become way-overkill for machine work.  I would not bother to do more
> then roughly align the laser.
> 
> Yes you need a sensor.  A $10 webcam with the lens replaced by a welding
> glass filter would work but you'd need to experiment with different filter
> and laser power.
> 

Alternately you could use a quad optical diode array and read the currents 
through the four diodes:
https://www.mouser.com/datasheet/2/414/OPR5911_RevD-1145186.pdf 


Years ago I built a rig to measure the spacing of dots on a slide using a HeNe 
laser and an optical diode array like that and was able to get sub-micron 
precision doing so. 

> I would not adjust anything r "zero".  between measurements.  leave the
> target and laser bolted down solid.  Just record the numbers and when you
> are done cal the smallest number "zero"
> 
> To get the center point, just use the "centroid" the same math if looking
> for the balance point of the laser spot.
> 
> You should be able to make a measurement at least once per lead screw
> revolution, maybe many times more per revolution.
> 
> On Mon, Sep 10, 2018 at 12:09 PM Gregg Eshelman via Emc-users <
> emc-users@lists.sourceforge.net > 
> wrote:
> 
>> I have a LASER that is made to mount in a chuck or collet. It has
>> alignment screws and adjustable focus. Aim it at a wall far away and adjust
>> the focus for as small a dot as possible.
>> Make a target to mount on the carriage, chuck and center the body of the
>> LASER then use the alignment screws and adjusting the target and rotating
>> the spindle until the spot no longer describes a circle when the spindle is
>> rotated.
>> 
>> Move the carriage to the right end as far as possible. Move the cross
>> slide intil the dot is on the target. There's your zero. Advance the
>> carriage toward the bed until the dot moves off the target. Adjust the
>> carriage until the dot is back on target, at least in the X axis. A really
>> good way to map the bed would be to have a sensor as the target so the
>> carriage could be moved by the control and read a continuous stream of how
>> far off X it gets. Barring that, stop every X distance then adjust to
>> re-zero and interpolate slopes between adjustment points.
>> If you put the target on a vertical slide to re-zero Y at each stop, the
>> error map (with appropriate math) could compensate for that too. Perfectly
>> straight cutting even with a very worn bed, as long as the vertical error
>> doesn't drop the cutting tools too low.
>> 
>>On Monday, September 10, 2018, 1:55:36 AM MDT, Chris Albertson <
>> albertson.ch...@gmail.com > wrote:
>> The lasers are easy to find on Amazon but be warned the laser is not
>> perfectly aligned with the housing.But as I said, this does not matter
>> if you spin the spindle.
>> 
>> I think a laser is the only option.  If you are measuring the bed your
>> reference needs to be at least 10X straighter then the bed.  Precsion rods
>> are not good enough but a $20 laser is literally perfect (as long as you
>> rotate the spindle.)
>> 
>> With effort you can use the same laser to measure the error with the lead
>> screw pitch. Use it as a laser range finder or laser inferometer
>> 
>> But all this measurement may be moot if the lathe is not reputable.  The
>> error might be random.  Maybe the carriage moves like a tuck on a dirt
>> road?
>> ___
>> Emc-users mailing list
>> Emc-users@lists.sourceforge.net 
>> https://lists.sourceforge.net/lists/listinfo/emc-users 
>> 
>> 
> 
> 
> -- 
> 
> Chris Albertson
> Redondo Beach, California
> 
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net 
> https://lists.sourceforge.net/lists/listinfo/emc-users 
> 


___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] correct x error from bed wear

[Gene Heskett]

On Monday 10 September 2018 19:37:32 Christopher wrote:

> > On 2018-Sep-10, at 18:02, Chris Albertson  > > wrote:
> >
> > You can actually make good measurements with a larger spot.  This
> > assumes you are using software and not just eyeball estimates.  
> > Typically you can do about 1/10th pixels position estimates.With
> > micro-meter scale pixels this become way-overkill for machine work. 
> > I would not bother to do more then roughly align the laser.
> >
> > Yes you need a sensor.  A $10 webcam with the lens replaced by a
> > welding glass filter would work but you'd need to experiment with
> > different filter and laser power.
>
> Alternately you could use a quad optical diode array and read the
> currents through the four diodes:
> https://www.mouser.com/datasheet/2/414/OPR5911_RevD-1145186.pdf
> 

This would work, but first find out if camview will run on the pi. It is 
a bit of a cpu pig and frame rates on the pi with its teeny usb2 
bandwidth already being a problem. And it seems to me to need fairly 
accurate laser alignment in order to stay within a diode quad of that 
size. I'll plug in a camera, and see if cheese will run sometime today. 
After I get a new air compressor installed under a workbench that I just 
barely got assembled before the rain started again yesterday, and which 
my back is still complaining about this morning.

Camerawise, the laser will overpower the camera, so I'm thinking I should 
mount the camera at a 90 degree angle to the laser beam, and set a piece 
of glass at a 45 degree angle to steer a fraction of the laser beam to 
the chip in the camera, but that would need a thick glass to separate 
the reflections so that only one surface of the glass would hit the 
camera.

Humm, ISTR the pi has a separate connector for a camera, but it takes a 
special camera to connect to it, and that cable would need to be about 5 
feet long to reach the pi from the lathes carriage. But from the 
available cable lengths, I'd have to setup a 2nd pi and put it all on 
the carriage. Which would make it a lot more insulated from lcnc. So I'd 
need a pi which I have, the camera $4.37(ebay, but unk if lens 
removable, maybe even a pinhole?), a psu I have and some sort of a 
display. $25(or less, ebay) for color tft touch screen.

Hu..

Thinking out loud.

> Years ago I built a rig to measure the spacing of dots on a slide
> using a HeNe laser and an optical diode array like that and was able
> to get sub-micron precision doing so.
>
> > I would not adjust anything r "zero".  between measurements.  leave
> > the target and laser bolted down solid.  Just record the numbers and
> > when you are done cal the smallest number "zero"
> >
> > To get the center point, just use the "centroid" the same math if
> > looking for the balance point of the laser spot.
> >
> > You should be able to make a measurement at least once per lead
> > screw revolution, maybe many times more per revolution.
> >
> > On Mon, Sep 10, 2018 at 12:09 PM Gregg Eshelman via Emc-users <
> >
> > emc-users@lists.sourceforge.net 
> wrote:
> >> I have a LASER that is made to mount in a chuck or collet. It has
> >> alignment screws and adjustable focus. Aim it at a wall far away
> >> and adjust the focus for as small a dot as possible.
> >> Make a target to mount on the carriage, chuck and center the body
> >> of the LASER then use the alignment screws and adjusting the target
> >> and rotating the spindle until the spot no longer describes a
> >> circle when the spindle is rotated.
> >>
> >> Move the carriage to the right end as far as possible. Move the
> >> cross slide intil the dot is on the target. There's your zero.
> >> Advance the carriage toward the bed until the dot moves off the
> >> target. Adjust the carriage until the dot is back on target, at
> >> least in the X axis. A really good way to map the bed would be to
> >> have a sensor as the target so the carriage could be moved by the
> >> control and read a continuous stream of how far off X it gets.
> >> Barring that, stop every X distance then adjust to re-zero and
> >> interpolate slopes between adjustment points.
> >> If you put the target on a vertical slide to re-zero Y at each
> >> stop, the error map (with appropriate math) could compensate for
> >> that too. Perfectly straight cutting even with a very worn bed, as
> >> long as the vertical error doesn't drop the cutting tools too low.
> >>
> >>On Monday, September 10, 2018, 1:55:36 AM MDT, Chris Albertson <
> >> albertson.ch...@gmail.com >
> >> wrote: The lasers are easy to find on Amazon but be warned the
> >> laser is not perfectly aligned with the housing.But as I said,
> >> this does not matter if you spin the spindle.
> >>
> >> I think a laser is the only option.  If you are measuring the bed
> >> your reference