What's new
What's new

What's everyone's preferred method of verifying 5 axis kinematic rotation point?

Not a 5 axis guy, but this looked involved.
Not a lot to that. Looks like regular 3-point center calculations being used to set center point parameters and possibly checking/setting from the tooling ball on the trunnion too? I've used both. Measuring 3-point centers and reference point on trunnion. Seems to me cutting an sample part and measuring gets closer than just regular 3-point measuring.

And I didn't have great results with a reference on trunnion either. Tried that on both tilt axis and rotary axis. Is similarly close to what you get from 3-point cycles, but not good enough. In my personal experience.


The "fancy" moves in the video there is just TCP doing its magic. Can hand-program those moves in a few lines really. Especially with the tooling ball in the holder, following the indicators. Not much to it.

Set tool length of that tooling ball holder to the center of ball. Jog it a position to zero out the indicators, or move indicators to the ball and zero. With TCP on, just command movements to different rotary angles. TCP keeps the center of ball in the center of those indicators.



Now something like the grob with an algorithm making adjustments in the background to compensate for machine geometry inaccuracies, like lack of squareness? That's something.
 
This is just the center of rotations, not volumetric compensation, correct?

Because of the grob video we now have two things going on here.

What I can say for sure is that the 3D Quickset that I used was a single sphere, and not dual spheres like in the Grob video, or a dual sphere artifact like in another DMG Mori video I just watched.

I don't have that machine anymore to go and look, so I'm going to assume yes, COR's based on existing linear axis' comps.

Anecdotally, that machine had 0.1µm scales and 0.0001deg rotaries, and the 3D Quickset worked well enough that I never felt the need to machine an artifact. It's the only machine I have used with that function, so I have nothing to compare it to...
 
And after calibrating with probe or calibrating with test cuts it's nice to do a full volumetric check like in this video. I made an elaborate indicator holder and a simple TCP program to see double check that all the parameters are working together correctly. Same concept for B axis mill-turns
Way out of my league, but that's pretty awesome!
 
What I can say for sure is that the 3D Quickset that I used was a single sphere, and not dual spheres like in the Grob video, or a dual sphere artifact like in another DMG Mori video I just watched.

I don't have that machine anymore to go and look, so I'm going to assume yes, COR's based on existing linear axis' comps.

Anecdotally, that machine had 0.1µm scales and 0.0001deg rotaries, and the 3D Quickset worked well enough that I never felt the need to machine an artifact. It's the only machine I have used with that function, so I have nothing to compare it to...
what machine?
 
The only thing I see wrong with setting parameters with probing spheres vs test cuts is the assumption that XYZ and A(B)C are square, true, perfect, etc. Sure the test cuts will correct the exact circumstantial factors, but everything in between is questionable. The math will never lie, and there are many more compensation parameters in the controls I have ran that don't always apply in every instance. Such as when evoking dynamic work offsets, TPC, euler angles, simple positioning, etc. Some of these codes don't factor all error aspects. 5 axis is a hell of a drug
 
All I'm really trying to say is rule out leveling, squareness, thermals, angular positioning, probe calibration, lunar cycle, et al before calibrating with probe and sphere. Then make on the fly adjustments as needed :willy_nilly:
 
Now something like the grob with an algorithm making adjustments in the background to compensate for machine geometry inaccuracies, like lack of squareness? That's something.

The dirty secret is that every modern 5 axis machine is doing real-time volumetric 3D compensation. How much the machine tool builder exposes to the user is the only question - European builders offer the ability to measure volumetric accuracy and manipulate the tables as a feature. Japanese builders are heavily restricted in doing so by Japanese export regulation (the dreaded METI) - providing the ability for an end-user to "increase" machine accuracy beyond the published specifications, outside of the direct control of the factory, is a big no-no.
 
My first encounter with the missup of centers of rotations problem goes back to 1972, while as young plant engineer in aircraft industry I had to solve this problem on Sundstrand Omnimill 5 axis machining center. The measurement methods were cumbersome (understatement big time), and the results have been transferred to programming guys, who made some changes in APT postprocessor in order to get the correct spindle parh.
During the years I was watching the development of measuring methods.The most important (at least in my eyes) breakthrough was the work of professor Tsutsumi from Tokyo University, who proposed quite unique method of using the ballbar to discover the inaccuracies in combined movements of rotary and linear axes of the machine. This TSUTSUMI BALLBAR TEST video shows his test. Few years later Renishaw came with BALLBAR TRACE SOFTWARE, which visualize the difference between the programmed and real machine path.
 
The dirty secret is that every modern 5 axis machine is doing real-time volumetric 3D compensation. How much the machine tool builder exposes to the user is the only question - European builders offer the ability to measure volumetric accuracy and manipulate the tables as a feature. Japanese builders are heavily restricted in doing so by Japanese export regulation (the dreaded METI) - providing the ability for an end-user to "increase" machine accuracy beyond the published specifications, outside of the direct control of the factory, is a big no-no.

What is modern? What defines volumetric compensation? Sounds like we are getting into murky waters here. I have no proof one way or another but my logic on this is as follows. I'm skeptical this is the case because as soon as the machine moves, the volumetric compensation would need to be "re-compensated". There is never a way to get the machine sitting exactly how it was in the factory. Really don't see any need for going beyond thermal comps, mechanical squareness, and an accurate way to pickup and adjust centers of rotation unless the end goal is chasing tenths across the entire envelope. If that is the goal, it has to be something that can by adjusted in the field because of the changing conditions, otherwise, what's the point. I'm out of my depth at this point, pure speculation. Would like to see evidence otherwise. Anything in Fanuc documentation describing this maybe I missed over the years? Seems like every Fanuc control needs an added RISC board for anything remotely complex (like NURBS), can't imagine what the 1980s technology would need for advanced 3d volumetric compensation.
 
Thinking about this more, if anything has to be serviced... not even from a crash but lets say a seal fails somewhere and the trunnion has to come apart. It never goes back together PERFECTLY, makes non-adjustable volumetric compensation useless, and perhaps worse than none at all. Seems very non-Japanese. Unless what's being implied here is that volumetric compensation is being used to makeup egregious non-squareness, which would be measurable with handle jog, or detectable seeing the other axis moving while only jogging one, Which I know doesn't happen on any machine I've ever used. Just trying to conceptualize this new territory.
 
The math will never lie, and there are many more compensation parameters in the controls I have ran that don't always apply in every instance. Such as when evoking dynamic work offsets, TPC, euler angles, simple positioning, etc. Some of these codes don't factor all error aspects. 5 axis is a hell of a drug
What are the inputs besides kinematic arrangement, x y z offset (4 geometric) , angle of head (if non square), and angle comps (Interpolated table or similar for a list of abc angles)? Honestly asking as I’ve heard people talk about other settings but never seen any documentation on ones besides these and never came across them in any controls. Could you post a pic of these settings and what they are referring to?
 
The further away from axis rotation centerline the checks are are done the better you will will get.
Let us just take the first.
3 points as the ball center make for an arc at some center.
5 or 6 (24 probe touches for a 6) gives a varying center. Best fit is used here.
Now, and we are just using one axis comes how straight to XYZ to the axis rotation axis.
A simple 4. Near the collet or table the same as 6 inches out?
Imagine a 4th axis mounted 0.5 degrees twisted off. How to comp that? Can your controls or software handle that?
Now add 5th, 6th or more.
One comps or tells the CAM rotation centers in the machine's XYZ space.
But none of these rotaries are perfectly true to to the world so a one time 3 position check (12 touches) will not work.
This just a 4th. Add 5,6,8 :nutter:
Each rotary has errors on the rotate (that is easy) but also a twist/tilt to the world (that no so easy to comp in most cncs).
 
Thinking about this more, if anything has to be serviced... not even from a crash but lets say a seal fails somewhere and the trunnion has to come apart. It never goes back together PERFECTLY, makes non-adjustable volumetric compensation useless, and perhaps worse than none at all. Seems very non-Japanese. Unless what's being implied here is that volumetric compensation is being used to makeup egregious non-squareness, which would be measurable with handle jog, or detectable seeing the other axis moving while only jogging one, Which I know doesn't happen on any machine I've ever used. Just trying to conceptualize this new territory.

I didn't say these systems (on Japanese machines) were non-adjustable. I said the *end user* does not have access to them. Fanuc and the MTB continues to have access to these parameters and can change them on behalf of a customer when doing service that might require it.

Two examples I can give because I've been directly involved; on a new Speedio (D00 controller machines), you cannot access the pitch error comp tables, and neither can Yamazen. If those need to be changed, we can bring in a laser, generate the numbers, and Brother USA will return us an appropriate and complete comp table (with perhaps a couple of rounds of adjustments for good measure).

Our 5 axis machines just introduced a sort of volumetric comp the user has access to called "Table Weight Compensation." On a parameter page, you can set the machine to make A and C axis motions with your work holding and part installed; it will use servo data to estimate the moment arm of everything on the table and generate a compensation matrix. It also tweaks the servo jerk with the given results. You can (optionally) manually enter measured data points into this table, and the control will interpolate between them in the background. The control has had a similar function for XYZ axis motion since launch.

Is it full volumetric sub-micron compensation with weighted interpolation between points? No, but we are very clear to users that we aren't selling Kerns; our machines are for the other 99% of 5 axis work where a machine designed to hold +/- 2 microns over the volume is obscene capital expenditure overkill.
 
I didn't say these systems (on Japanese machines) were non-adjustable. I said the *end user* does not have access to them. Fanuc and the MTB continues to have access to these parameters and can change them on behalf of a customer when doing service that might require it.

Two examples I can give because I've been directly involved; on a new Speedio (D00 controller machines), you cannot access the pitch error comp tables, and neither can Yamazen. If those need to be changed, we can bring in a laser, generate the numbers, and Brother USA will return us an appropriate and complete comp table (with perhaps a couple of rounds of adjustments for good measure).

Our 5 axis machines just introduced a sort of volumetric comp the user has access to called "Table Weight Compensation." On a parameter page, you can set the machine to make A and C axis motions with your work holding and part installed; it will use servo data to estimate the moment arm of everything on the table and generate a compensation matrix. It also tweaks the servo jerk with the given results. You can (optionally) manually enter measured data points into this table, and the control will interpolate between them in the background. The control has had a similar function for XYZ axis motion since launch.

Is it full volumetric sub-micron compensation with weighted interpolation between points? No, but we are very clear to users that we aren't selling Kerns; our machines are for the other 99% of 5 axis work where a machine designed to hold +/- 2 microns over the volume is obscene capital expenditure overkill.
is this all referring specifically to brothers and the latest control?

Also, is the pitch compensation you are referring to for linear motions? I’m guessing you were just using that as an example for a “locked” parameter.

Additional user adjustable (or automatic) parameters for “table weight comp” does sound like an interesting feature. Is there any documentation available on this and how / what it does exactly?

I don’t really play around with a lot of brothers, especially new ones, but still interested in learning more.
 
MAM72-63v

Yes, something I made.

What issues do you envision? Besides of course relying on the probe length remaining constant and booger free? I suppose the one "manual" operation I need to do is regularly check the probe tip is clean.
I will have to read through your post a few times to fully get it but on first glance…

Focusing only on “z”….
it’s important that the tool length setter “agrees” with the spindle probe length. Therefore if you do a face pass at a freshly probed 0 with a freshly probed tool, it shouldn’t cut anything or leave anything. The accuracy of feedback from your test part program relies on the relationship between tool setter and spindle probe. Because of this relationship, it’s important that the tool being used for the test part probes within your accepted tolerance from the last time the test part ran. If it doesn’t, this could mean something is up with the tool or it could mean the setter moved. At that point you’re kinda in trouble. You can see how the accuracy of this relationship has been compromised and you’ll have to get to the bottom of it before moving on, same goes for bogers on the spindle probe.
 
What are the inputs besides kinematic arrangement, x y z offset (4 geometric) , angle of head (if non square), and angle comps (Interpolated table or similar for a list of abc angles)? Honestly asking as I’ve heard people talk about other settings but never seen any documentation on ones besides these and never came across them in any controls. Could you post a pic of these settings and what they are referring to?
You are right, there are only so many things that can be accounted for, I guess I was referring to how and when different controls apply any of them.
My experience with high-end 5 axis was mostly Mazak variaxis and integrex, and I found it to be overly complicated. There are so many sets of parameters for center of rotation, compensation, compensation for compensation lol. Some are very benign like graphic representation. One fun pair of parameters were the L104/L105 rotary axis misalignment on i-700 variaxis. You could actually tweak these and XYZ would move on the fly, not something you should really have to mess with once it's right. There are K,L,M,BA, I, N, S, + parameters which some only apply in certain situations, such as during TCP or dynamic work offsetting, or G68.2 (euler) meaning if you were to bore a hole with TCP vs simple positioning canned cycle those positions could be off from each other due to blind compensation. And then, some parameters would apply immediately, some at machine reboot. And you're trying to troubleshoot while everything is at stable temperature. The shop I was working in had overhead radiant tube heaters 5 feet from the top of the machines blaring in the winter which I could always see thermal creep in mismatched cuts and CMM reports. I drove myself insane chasing it, before deciding to just adjust C.O.R. around dramatic seasonal changes and live life.

My Haas at home is pre-NGC with a pair of rotaries bolted together and I find my own MRZP by probing 3 sphere points around B and C axes. I made my own macros to calculate each into 2 work offsets. This also gives you the axial misalignment which I input into Gibbscam and all compensation is in G-code, meaning strictly inverse feed-time for simultaneous. I routinely remove and re-install this trunnion arrangement and am always amazed how accurate I can get it dialed in by sphere probing given I ensure the B and C are perpendicular to linear travel.

I have been very successful with 5 axis while never really feeling I had a total handle on getting it perfect. No doctoral theses here I tell ya, shit I hate even re-reading my own post
 
Last edited:
Our ibarmia has a cycle to probe the artifact and calculate the MRZP automatically, and i use it once a week or so.
i was wondering how some of you guys verify the alignment? machine a bore from both sides in a square, measure the mismatch in the middle with an indicator? examples would be sweet to see!
What control is on your ibarmia?
 
My Haas at home is pre-NGC with a pair of rotaries bolted together...
This is exactly the same boat I'm in (T5C) except mine is 4+1 (external rotary box). It works for positioning and that's all I need right now. I'm still paying attention for the day when/if the need creeps to simultaneous 5.
 








 
Back
Top