Dynomotion

Group: DynoMotion Message: 13654 From: andysontag Date: 7/29/2016
Subject: Geometric Correction Table - Equipment used to create table?

Hi All,


Has anyone tried creating a Geometric Correction Table to correct for machine inaccuracies? If so what equipment did you use? It appears laser interferometers are commonly used although I'm not sure if there are any within the price range of a home gamer. 


I've been noticing some significant errors that can only be caused by my machine's axes not being completely flat and/or straight. It would enormously helpful if I could correct for the machine's inaccuracy using the Geometric Correction Table in KMotionCNC.  


Andy
Group: DynoMotion Message: 13658 From: andysontag Date: 8/1/2016
Subject: Re: Geometric Correction Table - Equipment used to create table?
Hi Tom,

The rails on my machine aren't perfectly straight and thus small (<.010") errors show up when cutting parts. Ideally I could measure where these hardware errors are and account for them in software. At first glance the geometric correction table would be the way to do this, although maybe I'm being too optimistic. If the errors are measured correctly, will the geo correction table allow KMotionCNC to correct for these errors?

After some googling it appears there is a system produced by Optodyne that can be used to perform volumetric calibration. Their MCV-500 and SD-500 when combined can measure "three displacement errors, six straightness errors, and three squareness errors. If the geo correction table can indeed correct for errors, could a system like this be used to perform volumetric calibration on a KFlop based system?

Andy
Group: DynoMotion Message: 13659 From: Hardy Family Date: 8/1/2016
Subject: Re: Geometric Correction Table - Equipment used to create table?
As of 4.32, it appears the geometric correction is tabled for X,Y indices only, not full 3-D (a Z correction is tabled for each XY point, but is assumed to be the same offset for all Z levels - that would work fine for a short-stroke Z axis that was reasonably straight).  I think you could probably modify Kinematics.cpp to perform full volumetric correction, but it would not be trivial.  In 3-D you are talking about measuring a lot of points if the error is random for each point so that you need a full 3-D table lookup.

I would guess that with most machines with nominally orthogonal X,Y and Z axes, you could measure the error of each axis separately, then the overall correction would be the sum of the errors for each axis.  To put it another way, there would be an independent Y and Z error offset function that depends only on X, and the offset functions would be independent of the current Y and Z positions.  Similarly for the other two axis permutations.

If the above assumption is true, then maybe you could lay a reference straight edge along the X axis, then use a touch probe to measure the offsets in the Y and Z directions.  Then do the same for the Y and Z axes.  That would give enough information to correct the entire machine volume.

The assumption may not be true if the machine has complex kinematics (like Stuart platform designs) or if the axes interact (e.g. the weight of the head at different ends of the axis has an effect on the other axes).  In the latter case, I would think that measurements of the straight edge at the 4 extremes of the other 2 axes would be sufficient.  Sounds like an interesting problem, anyway.  Good luck.

Regards,
SJH


On Mon, Aug 1, 2016 at 10:32 AM, andysontag@... [DynoMotion] <DynoMotion@yahoogroups.com> wrote:
 

Hi Tom,


The rails on my machine aren't perfectly straight and thus small (<.010") errors show up when cutting parts. Ideally I could measure where these hardware errors are and account for them in software. At first glance the geometric correction table would be the way to do this, although maybe I'm being too optimistic. If the errors are measured correctly, will the geo correction table allow KMotionCNC to correct for these errors?

After some googling it appears there is a system produced by Optodyne that can be used to perform volumetric calibration. Their MCV-500 and SD-500 when combined can measure "three displacement errors, six straightness errors, and three squareness errors. If the geo correction table can indeed correct for errors, could a system like this be used to perform volumetric calibration on a KFlop based system?

Andy
Group: DynoMotion Message: 13662 From: TKSOFT Date: 8/1/2016
Subject: Re: Geometric Correction Table - Equipment used to create table?
Hi Andy,

What is your working envelope and what level of accuracy do you need?

As SJH pointed out the Geocorrection is basically 2.5D (corrects x,y,z
at one plane at one z height).

Laser Interferometer measurement of everything would be ideal.

Another option much less accurate would be to use a precision flat plate
with a grid of marks or holes.

I've seen Printed Mylar Sheets used. I think Services can print these
with accuracy close to 0.001"

I thought someone was using a tightly stretched wire and some sensor to
obtain straightness.

Regards
TK


On 2016-08-01 10:32, andysontag@... [DynoMotion] wrote:
> Hi Tom,
>
> The rails on my machine aren't perfectly straight and thus small
> (<.010") errors show up when cutting parts. Ideally I could measure
> where these hardware errors are and account for them in software. At
> first glance the geometric correction table would be the way to do
> this, although maybe I'm being too optimistic. If the errors are
> measured correctly, will the geo correction table allow KMotionCNC to
> correct for these errors?
>
> After some googling it appears there is a system produced by Optodyne
> that can be used to perform volumetric calibration. Their MCV-500 [1]
> and SD-500 [2] when combined can measure "three displacement errors,
> six straightness errors, and three squareness errors. If the geo
> correction table can indeed correct for errors, could a system like
> this be used to perform volumetric calibration on a KFlop based
> system?
>
> Andy
Group: DynoMotion Message: 13663 From: Andy Sontag Date: 8/1/2016
Subject: Re: Geometric Correction Table - Equipment used to create table?
Tom,

The working envelope of the machine is approximately 48" x 96 x 6", though the Z will go up to around 8" if the workpiece doesn't need to pass under the gantry. I would be thrilled to achieve +/-.002".  

Unfortunately I have the original CRP Pro Z axis (the one with V-Con bearings) which is about as straight as a politician. Move the Z height a little bit and the cutter will be pushed to a different X, Y location. The movement is visible to the eye, but I'd have to run some tests to find out how much it moves. A full 3D volumetric calibration seems like a very effective way to solve a lot of my issues.

SJH mentioned Kinematics.cpp could potentially be modified to perform full volumetric correction. Is this something I could tackle with some guidance? Measuring error/position will be another ball of wax, but without 3D correction I'll need to rethink/refit the Z axis in order to improve accuracy using 2.5D correction. This might be needed regardless, however full 3D correction is perhaps the only way to ensure accuracy across the entire working volume.

Andy

On Mon, Aug 1, 2016 at 2:52 PM, tk@... [DynoMotion] <DynoMotion@yahoogroups.com> wrote:
 

Hi Andy,

What is your working envelope and what level of accuracy do you need?

As SJH pointed out the Geocorrection is basically 2.5D (corrects x,y,z
at one plane at one z height).

Laser Interferometer measurement of everything would be ideal.

Another option much less accurate would be to use a precision flat plate
with a grid of marks or holes.

I've seen Printed Mylar Sheets used. I think Services can print these
with accuracy close to 0.001"

I thought someone was using a tightly stretched wire and some sensor to
obtain straightness.

Regards
TK

On 2016-08-01 10:32, andysontag@... [DynoMotion] wrote:
> Hi Tom,
>
> The rails on my machine aren't perfectly straight and thus small
> (<.010") errors show up when cutting parts. Ideally I could measure
> where these hardware errors are and account for them in software. At
> first glance the geometric correction table would be the way to do
> this, although maybe I'm being too optimistic. If the errors are
> measured correctly, will the geo correction table allow KMotionCNC to
> correct for these errors?
>
> After some googling it appears there is a system produced by Optodyne
 > that can be used to perform volumetric calibration. Their MCV-500 [1]
> and SD-500 [2] when combined can measure "three displacement errors,
> six straightness errors, and three squareness errors. If the geo
> correction table can indeed correct for errors, could a system like
> this be used to perform volumetric calibration on a KFlop based
> system?
>
> Andy
Group: DynoMotion Message: 13664 From: TKSOFT Date: 8/2/2016
Subject: Re: Geometric Correction Table - Equipment used to create table?
Hi Andy,

Adding the code to read in a 3D Geo calibration table and do the
tri-linear interpolation would be relatively easy depending on your
programming skill level. But that would require a lot of data and I
don't know how you would obtain it. Seems like adding a 1D xy shift as
a function of Z in addition to the 2.5D existing Geocorrection might
work. Do you think the shifting is repeatable over time and there is
sufficient stiffness? Its usually always simpler and better to improve
the raw hardware rather than to compensate in software.

HTH
Regards
TK

On 2016-08-01 16:16, Andy Sontag andysontag@... [DynoMotion]
wrote:
> Tom,
>
> The working envelope of the machine is approximately 48" x 96 x 6",
> though the Z will go up to around 8" if the workpiece doesn't need to
> pass under the gantry. I would be thrilled to achieve +/-.002".
>
> Unfortunately I have the original CRP Pro Z axis (the one with V-Con
> bearings) which is about as straight as a politician. Move the Z
> height a little bit and the cutter will be pushed to a different X, Y
> location. The movement is visible to the eye, but I'd have to run some
> tests to find out how much it moves. A full 3D volumetric calibration
> seems like a very effective way to solve a lot of my issues.
>
> SJH mentioned Kinematics.cpp could potentially be modified to perform
> full volumetric correction. Is this something I could tackle with some
> guidance? Measuring error/position will be another ball of wax, but
> without 3D correction I'll need to rethink/refit the Z axis in order
> to improve accuracy using 2.5D correction. This might be needed
> regardless, however full 3D correction is perhaps the only way to
> ensure accuracy across the entire working volume.
>
> Andy
>
> On Mon, Aug 1, 2016 at 2:52 PM, tk@... [DynoMotion]
> <DynoMotion@yahoogroups.com> wrote:
>
>> Hi Andy,
>>
>> What is your working envelope and what level of accuracy do you
>> need?
>>
>> As SJH pointed out the Geocorrection is basically 2.5D (corrects
>> x,y,z
>> at one plane at one z height).
>>
>> Laser Interferometer measurement of everything would be ideal.
>>
>> Another option much less accurate would be to use a precision flat
>> plate
>> with a grid of marks or holes.
>>
>> I've seen Printed Mylar Sheets used. I think Services can print
>> these
>> with accuracy close to 0.001"
>>
>> I thought someone was using a tightly stretched wire and some sensor
>> to
>> obtain straightness.
>>
>> Regards
>> TK
>>
Group: DynoMotion Message: 13669 From: Andy Sontag Date: 8/5/2016
Subject: Re: Geometric Correction Table - Equipment used to create table?
Tom,

Good to know that it's possible to enable 3D volumetric calibration. I still haven't heard back from Otodyne regarding calibration equipment suitable for homegamers, but I've taken your advice and will be attempting to improve the hardware's performance. The Z axis is already greatly improved and I'm working with the folks at CRP to hopefully straighten out the X and Y axis. We'll see how it goes.

Andy

On Tue, Aug 2, 2016 at 7:21 AM, tk@... [DynoMotion] <DynoMotion@yahoogroups.com> wrote:
 

Hi Andy,

Adding the code to read in a 3D Geo calibration table and do the
tri-linear interpolation would be relatively easy depending on your
programming skill level. But that would require a lot of data and I
don't know how you would obtain it. Seems like adding a 1D xy shift as
a function of Z in addition to the 2.5D existing Geocorrection might
work. Do you think the shifting is repeatable over time and there is
sufficient stiffness? Its usually always simpler and better to improve
the raw hardware rather than to compensate in software.

HTH
Regards
TK

On 2016-08-01 16:16, Andy Sontag andysontag@... [DynoMotion]
wrote:


> Tom,
>
> The working envelope of the machine is approximately 48" x 96 x 6",
> though the Z will go up to around 8" if the workpiece doesn't need to
> pass under the gantry. I would be thrilled to achieve +/-.002".
>
> Unfortunately I have the original CRP Pro Z axis (the one with V-Con
> bearings) which is about as straight as a politician. Move the Z
> height a little bit and the cutter will be pushed to a different X, Y
> location. The movement is visible to the eye, but I'd have to run some
> tests to find out how much it moves. A full 3D volumetric calibration
> seems like a very effective way to solve a lot of my issues.
>
> SJH mentioned Kinematics.cpp could potentially be modified to perform
> full volumetric correction. Is this something I could tackle with some
> guidance? Measuring error/position will be another ball of wax, but
> without 3D correction I'll need to rethink/refit the Z axis in order
> to improve accuracy using 2.5D correction. This might be needed
> regardless, however full 3D correction is perhaps the only way to
> ensure accuracy across the entire working volume.
>
> Andy
>
> On Mon, Aug 1, 2016 at 2:52 PM, tk@... [DynoMotion]
> <DynoMotion@yahoogroups.com> wrote:
>
>> Hi Andy,
>>
>> What is your working envelope and what level of accuracy do you
>> need?
>>
>> As SJH pointed out the Geocorrection is basically 2.5D (corrects
>> x,y,z
>> at one plane at one z height).
>>
>> Laser Interferometer measurement of everything would be ideal.
>>
>> Another option much less accurate would be to use a precision flat
>> plate
>> with a grid of marks or holes.
>>
>> I've seen Printed Mylar Sheets used. I think Services can print
>> these
>> with accuracy close to 0.001"
>>
>> I thought someone was using a tightly stretched wire and some sensor
>> to
>> obtain straightness.
>>
>> Regards
>> TK
>>