Can someone help me understand linear scales?

I have a relatively new HMC with Fanuc 30iB and linear scales that is experiencing positioning accuracy errors. I am working with the vendor to fix but I wish I had a better fundamental understanding of how things work.

As received accuracy was .0001 to .0009 depending upon which direction position is approached from, i.e. XY, -XY, -X-Y, X-Y.

We removed the backlash compensation in #1851/1852 and the X error is zero but the Y is roughly .0003.

My questions are;

1. Why does backlash compensation affect position with linear scales. Should not the machine go to exactly where I tell it to regardless of any compensations?

2. When I jog with handle and compensation factors in place, reversing direction using a .0001 pulse causes the machine to move .0005 however the Fanuc DRO would show .0001. I do not understand why the Fanuc DRO is not reading directly from the linear scales. After removing the compensation the jog handle is more accurate however not perfect.

3. This jog handle problem is a big deal because I have found when calibrating the probes using P9801K4M180, I get a cal position error of ~.0005. Therefore I have to manually locate the ring gauge and turn off the M180. Since the DRO using the jog handle does not represent reality I have to use MDI to find the ring center which is a PIA. What is everyones experience calibrating probes and letting Renishaw automatically find the center of the ring using M180 vs manually finding the ring?

PS we did verify the machine is reading the linear scales and not rotary by verifying #1815.1 is set to 1. It still appears to me the DRO is seeing the rotary and not linear but the vendor says different.

This may be a good reference to start from: Linear encoder - Wikipedia

Fairly new isn't new... have you verified it's not a mechanical/mounting issue before changing control parameters? A scale with loose mounting screws isn't any more accurate than the screws or dowels it's installed with, and could definitely do what you're seeing.

I have not checked any mechanical. The vendor told us to remove the compensation parameters and see if positioning improved which it did. The machine was laser aligned after installation and the results were not great (max of .0008) but it was signed off anyways about 8 months ago. Since then we are industrializing new boring programs with relatively tight position tolerances which is why this issues just go noticed.

Badbascom said:

......but I wish I had a better fundamental understanding of how things work......

Click to expand...

I'll try, but this is a very in-depth subject with lots of variables that can cause the issues you are seeing.

Badbascom said:

1. Why does backlash compensation affect position with linear scales. Should not the machine go to exactly where I tell it to regardless of any compensations?

Click to expand...

In a perfect world yes. Your machine is not perfect. In an axis, the ballscrew that moves it is very unlikely to be exactly centered to the load (resistance of the object being moved). Because it is a moving device there are factors like clearance, straightness, stiction that affect the resulting motion. The scale is also unlikely to be centered to the load.

On large machines it is very often easy to see this. No idea what your machine is. Try this. Take your backlash measurement with your indicator contacting a surface as closely as practical to the scale reader head mounting surface. Note your results. Then repeat the test with the indicator mounted as far away from the reader head as practical. Compare the results with the earlier test. IME, they are never the same. Sometimes not very much different, but they are different. This is one of those areas that differentiates a high quality machine from commodity level machines

Badbascom said:

2. When I jog with handle and compensation factors in place, reversing direction using a .0001 pulse causes the machine to move .0005 however the Fanuc DRO would show .0001. I do not understand why the Fanuc DRO is not reading directly from the linear scales. After removing the compensation the jog handle is more accurate however not perfect.

Click to expand...

Taking into account what I wrote above, one can see the possibility for a need of backlash compensation. For this case it would better called something like lost motion compensation. The control adjusts the displayed value by the compensation value.

Also, "just remove backlach compensation" seems like questionable advice. It's gotta be there for a reason. Are you measuring near the edge of travel (easy to read an indicator, but typically less wear) or near the center of travel, where the screws and rails are more likely to be worn? Is there a single backlash parameter per axis, or is there a comp table?

Comatose said:

Also, "just remove backlach compensation" seems like questionable advice. It's gotta be there for a reason. Are you measuring near the edge of travel (easy to read an indicator, but typically less wear) or near the center of travel, where the screws and rails are more likely to be worn? Is there a single backlash parameter per axis, or is there a comp table?

Click to expand...

A linear scale machine should not be using a compensation map unless it is to correct the scale. See this on really big gantry mills but I doubt that is the case here. If it does have a compensation map turned on, that will create issues but not like they are observing.

The lost motion error is the likely explanation, especially if there is some slop in the axis of motion such as loose gibs on a box way machine or worn balls in a linear way machine.

This machine probably has less than 40 hours on it, about 5 hours of actually machining. The linear scale option was chosen after the machine had already gone thru factory testing in Korea, the linear scales were added by the US arm of the manufacturer which I think caused the problem. The vendor is planning a visit so hopefully they will be figuring this out. I just bored a hole and the true position is max of .0004 per CMM which is much better than before with the comp values active.

I guess I should have phrased my questions per the following;

1. Do the Fanuc parameters modify the linear scale positional info to the DRO readout in any way?
2. For those who have linear scales, if you MPG .0001, does your table move exactly .0001?

@Badbascom great question you pose + interesting this only shows up/ get's noticed when higher precision and accuracy boring operations (positionally) are needed.

I need to double check #1851/1852

but I'm wondering pitch error compensation on the ball screws. (laser is one thing but actual pitch error compensation is another)

If the machine is not new and the ball screws were switched out (even on one axis) then the pitch error compensation plots won't match anymore.

I.e. commanded position won't match actual position and then in the feedback loop the machine has to figure out how to correct that move … Does it move in a positive direction or does it have to move in a corrective "Backwards move" which can happen with natural ball screw errors / pitch compensation (+ve and -ve) so normally that machine has to correct for all that before whatever piece of iron stops moving (+ whatever stiction may exist) you can't have the machine servoing back and fourth twitching in position to join up with how the machine "Interprets" how it wants to read the scale. I'm sure there's some special filtering and setting for that , I bet @Vancbiker knows something about that.

Also in the Fanuc scheme there is thermal compensation / guestimates too. So there's thermal compensation without scales and then there's thermal compensation with scales as the thermal expansion of the scales also has to be taken into account and is that compatible with the machine or is there a mis-match ? Does the machine have temperature sensors for the scales and temperature sensors throughout the machine ?

It presents an interesting argument for machines with and without linear scales. (but not rotary direct read scales ;-) ).

Will dig about and put my thinking cap on...

I also remember @PROBE going on about the order of calibration for pitch error compensation versus backlash compensation. I believe he was saying that MAZAk was doing it all wrong. I can probably dig up the thread for that too. But PROBE would be a good goto on that as well.

_____________________

I would have thought that the Renishaw ball bar plots would be good for diagnosing this specific problem.

Comatose said:

Also, "just remove backlach compensation" seems like questionable advice. It's gotta be there for a reason. .....

Click to expand...

I've seen new machines with scales that had backlash comp values set in the .005mm or so range. After the machine has some time on it this is often too much. When setting up at the factory with all new parts (mainly wipers IMHO) there is some amount of stiction that resists movement on reversal. The addition of comp gives the axis drive a little "pop" to accomplish the reversal. After some use things have "broken in" and the axis travels more easily and does not require the "pop" as much. That's when one would often see a need to reduce or zero out the comp.

I wrote the following small program for calibrating with the newer Ren macros.
You move the block delete from 2nd to 1st.or the big brain move might be to remove the block delete entirely. On the floor I dont have two 9801s I rerun the program after changing the words on that line, could be a problem.
You also have to have a functioning 180 orient for this to work.

O7900(COMPLETE CAL)
(MUST HAVE GOOD WOF XYZ)
(AND CLOSE PROBE + TLO)
(SWEEP RING GAGE AND STORE)
(USE K0. AND M180. FIRST)
(K0. M180.FOR XY OFFSET)
(RUN AGAIN WITH K4. )
G20
G90G80G40G0
G54X1.Y0.
G43H40Z3.
G65P9832(SWITCH ON M19)
G65P9810Z.393F40.(PROTECTED)
G65P9801B.236K1.Z0.T40.(CAL LENGTH)
G65P9810X0.Y0.F40.(PROTECTED)
G65P9810Z-.25F40.(PROTECTED)
G65P9801K0. M180. B.236D1.75S1.(1st)
/G65P9801K4.B.236D1.75S1.(2nd)
G65P9810Z3.F40.(PROTECTED)
G65P9833(SWITCH OFF)
N35
G91G28Z0.
G90
M30

Thanks everyone, lots of info to wrap my head around. I will ask these questions when the vendor arrives, I do know the machine is new with pre-tensioned ball screws for thermal comp, not sure if additional sensors are added or not but will find out.

Can you say what the machine is? Scales added by the importer adds a whole host of other things that might be mis-configured.

Badbascom said:

I have a relatively new HMC with Fanuc 30iB and linear scales that is experiencing positioning accuracy errors. I am working with the vendor to fix but I wish I had a better fundamental understanding of how things work.

As received accuracy was .0001 to .0009 depending upon which direction position is approached from, i.e. XY, -XY, -X-Y, X-Y.

We removed the backlash compensation in #1851/1852 and the X error is zero but the Y is roughly .0003.

My questions are;

1. Why does backlash compensation affect position with linear scales. Should not the machine go to exactly where I tell it to regardless of any compensations?

2. When I jog with handle and compensation factors in place, reversing direction using a .0001 pulse causes the machine to move .0005 however the Fanuc DRO would show .0001. I do not understand why the Fanuc DRO is not reading directly from the linear scales. After removing the compensation the jog handle is more accurate however not perfect.

3. This jog handle problem is a big deal because I have found when calibrating the probes using P9801K4M180, I get a cal position error of ~.0005. Therefore I have to manually locate the ring gauge and turn off the M180. Since the DRO using the jog handle does not represent reality I have to use MDI to find the ring center which is a PIA. What is everyones experience calibrating probes and letting Renishaw automatically find the center of the ring using M180 vs manually finding the ring?

PS we did verify the machine is reading the linear scales and not rotary by verifying #1815.1 is set to 1. It still appears to me the DRO is seeing the rotary and not linear but the vendor says different.

Click to expand...

First of all general remark: Linear scales do excellent job minimizing the impact of the thermal change of the length of ballscrews on machine movement accuracy. That is all. Excerpt of that, they cause quite a number of of problems, quite frequently omitted by calibrators, users and maintenance staff. This is even more acute when scales are added in field during attempt to improve machine performance.
The scales will not solve inherent mechanical problems.

Prior to scales installation:
1. Perform geometrical alignment of the machine, with special care on leveling.
2. Run Ballbar test with backlash compensation values (parameter 1851/1852) turned to zero. Observe the backlash and lateral play numbers. Excessive lateral play should be minimized, as scales do not compensate this error. Minimize mechanically (mostly gibs alignment). Regarding the backlash, it's mechanical "play" will remain with and without scales. Minimize adjusting the ballnut and thrust bearing preload.

After scales installation ( I assume it was done both mechanically and electronically in accordance with scale and control producers instructions):
1. MANDATORY: Perform pitch error compensation calibration with laser interferometer.
2. Backlash values in parameter 1851 should be set to least significant digit. Do not set zeros, as reversal spikes compensation do not work with such settings.
3. Once more run Ballbar test, observe scale mismatch, lateral play and backlash.

Your reason for the decision to install the scales in not known. Nevertheless, if above mentioned measures are not executed, this action will not solve "old" problems, on the contrary it can cause many "new" once.

Thanks, the vendor is coming in today, I will tell him to perform these taks.

Badbascom said:

Thanks, the vendor is coming in today, I will tell him to perform these taks.

Click to expand...

I wonder if the vendor is going to claim you need to run the machine for at least three hours before things start to join up properly.

Will dig about.

Is the machine a Makino by any chance ?

No, it is a Doosan HM1000.

Badbascom said:

No, it is a Doosan HM1000.

Click to expand...

Just some random observations.

The machine is BIG.

The machine is box(ed) way ALL axes.

Pre tension ball screw pre tensioned both ends.

_______________________________________________________________________________________



Just some things I got from the other side of the fence (DMG Mori/ Mori seiki.).

Allegedly with a box way machine that some cycles are needed to lube up the ways, programmed gyrations to compensate for stiction effects.

This where your minimum sensitivity of movement comes in from a dead stop. Where a very small corrective movement is commanded and the machine will store energy like a spring and then lurch to release that energy. Dynamically in the motion control you want all acc/dec moves to be tuned-in in such a way (taking into account the masses and weight of the parts ) to converge on graceful movements that require the minimum of servoing back and fourth. Servoing from a dead stop is not what you want.

I suspect the machine behaves very differently when its running nearly balls (out) to the wall with heavy parts being shunted hither and yon versus nearly static tests. If Doosan are skillful builders then they would bias everything for near maximum running conditions.

These days on a vertical (not horizontal) that uses box ways that linear rolling element guide ways / bearing are used with many trucks on the Z axis for contouring (like mold work). Not sure if that "fashion" is being used on horizontals these days ?

As far as accurate positional boring goes I feel like saying it's not a jig borer nor a dedicated horizontal boring machine (such a fuddy duddy thing to say right ?) , but on the other hand I'm wondering what DOOSAN would say as to be realistic positional accuracies for bored features..? A lot of good machines will creep by as much as ten micron through a day even under pretty good shop environment.

Again from the Mori Seiki end of the pool the scales they use have temperature sensors in them (plus additional temperature sensors at various points in the machine) [Other builders have other ways to map thermal changes more simply with their machines]. The machine's mehanite(sp) giant castings absorb heat at a slower rate than a scale would. [Thermal mass and surface area to volume ratio.]. Some scales have positive pressure and air fed to them to make sure chips and coolant don't mess things up (another thermal effect / difference than the thick castings). Not all scales will expand at the same rate as giant castings. Not sure what the overall scheme that DOOSAN has on that machine for thermal management ? No cooling of castings or redistribution of heat to even out thermal effects. It's a big machine long long travels for its type. ~ That's why I suspect DOOSAN has things tweaked for a sweet spot three hours in running balls to the wall.

So I'd be really careful not to mess up to many original settings until it can be determined that they are really not working for you. I.e. run a bunch of big parts over a few days or a week (expensive I know).

Also there is myriad of diagnostic capability with the humble ball bar plots (there's quite a smorgasbord of diagnostic capability there) but better to run with a machine really warmed up and worked through.

Depending on the distance traveled, the numbers up top don't look so bad for a 1000mm HMC with box ways. Also, my understanding is the hand wheel is never to be used when assessing accuracy.

Badbascom said:

<snip>

PS we did verify the machine is reading the linear scales and not rotary by verifying #1815.1 is set to 1. It still appears to me the DRO is seeing the rotary and not linear but the vendor says different.

Click to expand...

@Badbascom You should press them for a flow chart of how all the compensations stack up. Demand "Clarity" (if you can) rather than have a small team randomly muddling through. That's a big / major machine / manufactory beastie / behemoth. [It is complicated and have a sense of your conundrum / bewilderment .]

@PROBE super nice what he laid out and yeah don't underestimate pitch error compensation and order of calibrations.

There is also a pitch error compensation offset (on the Fanuc control), I assume that shifts the whole plot or mapped out ball screw small increments as effectively a shifted "waveform" <--- Thinking about ball screws with pre-load both ends on that DOOSAN horizonatl (rather than single ended). If someone tightens those differently then your ball screw plot / tables will be inadvertently shifted a smidge or offset ?

I may be totally wrong about pitch error plot / table and how that gets offset +ve or -ve by messing with double ended pre-load at both ends of a ball screw.

+ no core cooling … no biggy probably.