Scale vs encoder feedback. How accurate and how do you see your backlash?

Guess I could go crawl around a few machines but this seems easier. At what point or travel range do scales start to be the shnit? How much more accurate are we talking? No doubt if I was going to be probing a part in a machine and trusting it, scales would be mandatory. I am not sure how thermal changes and such are accounted though.

Also, I am used to being able to count encoder pulses or some simple way to check backlash. With scales, are the encoders still there? I would think you still need the encoders for servo feedback performance but then all the motion feedback would come from the scales? Is there a simple way I am not seeing to check backlash with scales?

backlash

huleo said:

Guess I could go crawl around a few machines but this seems easier. At what point or travel range do scales start to be the shnit? How much more accurate are we talking? No doubt if I was going to be probing a part in a machine and trusting it, scales would be mandatory. I am not sure how thermal changes and such are accounted though.

Also, I am used to being able to count encoder pulses or some simple way to check backlash. With scales, are the encoders still there? I would think you still need the encoders for servo feedback performance but then all the motion feedback would come from the scales? Is there a simple way I am not seeing to check backlash with scales?

Click to expand...

.
i check backlash with a dial indicator go 0.500 forward then 0.500 back with the machine and see if indicator agrees. you need to check throughout travel range as backlash can very from the ends to the center of most used travel. usually backlash compensation is only one amount although it can vary throughout travel range
.
with encoders at stepper motors you rely on ball screws having perfectly even pitch although some machines can use a compensation table or values found with a laser interferometer to measure and compensate for ball screw pitch errors.
.
all digital readouts should also be checked with a long gage block and dial indicator so when you move say 6.0000 the digital readout should say the same or the compensation factor can be changed usually by special mode that operator tells DRO what it moved and it compensates for calibration error
.
just using stepper motors that can be given pulses but not actually move under resistance then with no feed back your loosing count will give errors.
.
if you circular mill a circle then check it will a indicator you can measure out of roundness or errors

My old Deckel has scales but no encoders. Instead it has brush tachometers on the servos for the driver feedback using velocity instead of position. I have no idea how or when the control does the handoff from the tacho-derived position to the scale position.

Almost all of the current systems use sophisticated ball screw error mapping to compensate. There is also a backlash compensation parameter that also compensates for the backlash. The longer the ball screw the greater the issues. This is why it is important to get the machine to operating temperature before starting to make real parts.

Some machine tool builders have also used temperature controlled ball screws to maintain a set temperature for increased accuracy. Others have used temperature compensation to increase accuracy.

The systems that use encoders and scales get to be finicky about motion control when switching on the fly between modes. Not that this is a problem, it is just an extra complication to the builder.

With the current technology, if the ball screw system is well maintained, it is about equal with linear scales in accuracy. If not maintained then the scales are best.

On your last question, this is more a function of the servo amplifier and servo motor. There are different ways of monitoring motor rotation: encoder, tachometer, resolver,magnetic flux, etc. Each system has its own particular method internally. If you are trying to check this out yourself, just put an indicator in the spindle and move the tale back and forth slightly in the axis that you want to check. Compare the axis display difference to the indicator. The difference between the two is your compensated backlash. If you want to measure absolute backlash then you would need to turn off the backlash compensation first before taking the readings.

See, that is the issue with scales. The scales provide feedback to the control for motion. So you want to move 1.000", the control will move the axis until the SCALE sees 1.000", so if there is backlash in the system, you might not see it! I am just curious how you go about testing the system for this?

I have learned that some people sort of disregard backlash and just "comp it out". It is actually a huge deal for machine stiffness those tenths of slop can cause the axis the vibrate thus reduce tooling life, etc.

I guess if a servo does not have an encoder, this might get pretty tricky to figure out! I can't imagine mag flux being used for this though as servos typically need the best holding power possible at 0 rpm and mag flux cannot be monitored without some rotation.

huleo said:

I can't imagine mag flux being used for this though as servos typically need the best holding power possible at 0 rpm and mag flux cannot be monitored without some rotation.

Click to expand...

Not true. You can monitor magnetic flux in a motor without rotation. This is how direct torque control works. VFDs and servo amplifiers can have very sophisticated algorithms in their control loop to do this. This gets to be very complicated engineering wise but allows us to do things today that were impossible several years ago.

Well, I guess from a machine tool perspective, best always trumps "it works". From what I have seen, sensorless vectoring works but encoder feedback is hard to touch. I realize in large fan installations and such, probably where the motor exists and just need want to update a drive, sensorless could be a really smart upgrade.

huleo said:

....

I have learned that some people sort of disregard backlash and just "comp it out". It is actually a huge deal for machine stiffness those tenths of slop can cause the axis the vibrate thus reduce tooling life, etc.

Click to expand...

And that backlash that is always present in a convention drive forces a scale feedback system to have lower stiffness than a motor encoder feedback system.
The problem is the scale moving while the screw has not, control sees it, reverses, waits until the backlash is taken out and then the scale moves. Now the motor has gone too far, back the other way.
At a tight level most scales also have some backlash as the reader head drags on the seals.

Two Feedback Loops are Better Than One | Archive content from Machine Design

Scales can eliminate some problems and create some of their own. Often scales are mounted on the edge of a table which leads to errors from slight twisting of the table.
A big deal in CMM design is eliminating these errors but CMMs don't have to deal with loads the way a metal cutting machine does.
Bob

When I check backlash or lost motion on a screw that uses a scale I have to disable the scale. I have seen a screw with .010" lost motion that repeats to .0001" because the scale controls this. If the lost motion is greater than a setting parameter then a servo disconnect happens. There is also a backlash comp for machines that don't utilize scales, but I would rather fix the problem rather than set the backlash comp.

Machines with scales aren't generally affected by thermal growth. The reason ball screws are stretched when installed is to help minimize thermal growth.

To answer your question, machines without scales will eventually have lost motion as they wear and will lose accuracy. When you have scales on the slides the machine will still repeat even when there is lost motion on the screws and support bearings.

To disable the scale on a Fanuc, (which is what I work on the most) you turn 1815.1 to a zero. Then you have to set up the flex gear to run off of the encoders in the servo motor so the travel of the slide is correct. I use an interactive spreadsheet that updates parameters 2024, 2084, and 2085 based on other machine parameters to set up the flex gear. It's a handy tool for a dummy like myself. Without it I could never calculate what the flex gear should be. If you want it I can sent a copy to you. Daryl

I have 2 5 axis machines that I fought thermal growth on for almost a year. MFG didn't want to do anything. Added scales to both machines and they hold tolerance amazing now. I can probe the table, run a part, probe the table again and get maybe .0002 difference after an hour or running? Once parts are dialed in they stay there over the full run. From 10 pcs to 200 pcs.

In my experience, machines with scales have always maintained their accuracy and performance better over long term use than a machine without scales. I have also rarely had troubles with the scales used on premium machines (Heidenhein, Kuroda, Sony, etc.). How and where the scale is mounted is important. Well shielded from chips and coolant is a must. Somewhat centered between the ways is best as it minimizes the affect of slide skewing.

As a very simple example who would not want scales on a manual B-port verses the dials alone?
Yet the machine might do nasty things in a climb cut when pushed hard.
You can't fix slop/bad mechanics with a scale.
A scale will nail end positions or unloaded running paths dead on which is where most "machine accuracy" testing occurs.
Unfortunately I don't run "unloaded".
They are a huge plus in a machine design or environment that has a thermal growth problem.
They work well in many cases, will make you process capability a bit worse in others.
Bob

Now, this might seem overly complicated or redundant but wouldn't you say 1000 line optical encoders on the servos AND scales be the way to end all? I mean we could easily then detect lost motion and throw an alarm for "axis not holding position". Axis would go to position on the scale and holding would be deciphered from the encoder. Or pretty much run on the encoders but position check with the scales.

Also, Bob, you are automatically assuming a worn out machine running on scales. We already know there is no feedback system that will mask mechanical issues. I am more or less looking at everyday accuracy between scales and encoders. At what travel distances are scales preferred? How much more accurate can a machine get with the scales? I was not too impressed with the latest Mori HMC with I think a rated repeatability of .0004" when I know of machines that are decades old with 4x the travel and scales and rated at .00006"

Guess one of the reasons for the question is this fun chasing of dimensions on VMCs throughout the day.

The primary reason for question was actually taking a look at CMM type in machine probing work if a machine has scales? I am hoping with scales, it is harder for a customer to say "what if the machine is not right?" I mean, a CMM could have a bad day too but a CMM would operate just like a machine tool with scales!

I have not had a job that made sense to probe in the machine yet but working that way. One thing that stands out and probably thought of is concentricity and repeatability of the tool change. A CMM is dialed in and that probe stays where it is. The machine tool has to stick that tool back in.

Now, would I trust the probing for some +/-.0001" feature? probably not! but if you have +/-.005", I would think a probe could be well inside that? Hell, our beloved digital height gauge with scale repeats all day long within .0002".

We use gauge blocks as our standards for the shop and hold certs on that stuff and work from them. If we will be testing something about 6", we will test on a 6" standard to verify. We had a big customer in a while back that was not enthused but also could not find a reason not to like it.

So I guess I am looking at scales from the perspective of how much more accurate or repeatable my parts will be, and how well probing can work in machine with them?

I think you are making an assumption that all scales are created equal. Depending on the exact type and model there is significant variation in actual performance. All of them have limits on accuracy and temperature control is vital for consistency.

Take Sony's Magna-Scale, a very good system that is consistently accurate however you must understand what the scale is actually using for its metric. In this cause it is a wire with a magnetic signature that gives us the read out. The stretch and or tension on the wire changes the actual indicated distance. These need to be mapped when commissioned and checked regularly. Temperature control is crucial.

Glass scales are best temperature wise but have a limited practical length. Again these should be mapped for best accuracy.

If accuracy is of utmost importance than you need a laser interferometer.

I don't see why machine probing would not work for you. The big issue is to know what all of the potential error is and where it is at. In essence you need to know your machine and or measuring device. The complete process on how you take the measurements is the key. Tool change issues should not be a problem if you have a ball bar standard mounted to the machine table and calibrate the probe prior to using it after the tool change.

I think your real question goes back to precision vs. accuracy. No system is accurate unless it is used properly and in the context of what its total achievable accuracy is.

I see a lot of precision measurements being thrown around in industry and yet when you analyze how they made there measurement, the actual measurements accuracy is at least on the order of a magnitude less then what they think it is.

Just remember, when using a mic, ignore that clutch thimble, you want it good and tight!!!!

Was going to have some brake rotors turned at Oreilly. I started biting my nails when I watched the tool do this... Then he could not read it. I actually already tested them and recorded the value right on the face but no.... he needed to check....

Agreed, anything tighter than say a tenth and we are talking TIGHT temperature controls! Even at that, we have parts that need to chill for a minute out of the machine to get in tolerance. Kiss my ass on grinding ABEC9 AC bearings.....

If glass linear scales are mounted right, the almost temp comp themselves. The glass is chosen to have a thermal coefficient matching iron.

A real world example of scales vs. encoders: my VMC was available with scales as an option. The guaranteed absolute accuracy over the envelope without scales is 0.1 mm, with scales it is 0.01 mm. A recent laser check on my 14 year old scale equipped machine showed it to be well within the as new spec. The accuracy is ensured by the scales, which don't wear.

Usually the way a control is set up with scales is the servo velocity loop uses the encoders and the position loop uses the scales.

swarf_rat said:

If glass linear scales are mounted right, the almost temp comp themselves. The glass is chosen to have a thermal coefficient matching iron.

A real world example of scales vs. encoders: my VMC was available with scales as an option. The guaranteed absolute accuracy over the envelope without scales is 0.1 mm, with scales it is 0.01 mm. A recent laser check on my 14 year old scale equipped machine showed it to be well within the as new spec. The accuracy is ensured by the scales, which don't wear.


Usually the way a control is set up with scales is the servo velocity loop uses the encoders and the position loop uses the scales.

Click to expand...

Ah! Orders of magnitude do count in places beyond paychecks. :-)
I see you've been keeping up with theory. Early on I had just glass scales ( 5 um ) on my mill. Worked but hard to tune. Now I have 2500 cpr encoders, at .25 pitch ballscrews that give me 40000 c/in. Tuning is infinitely better. Next step is to hook the glass scales back up. :-) Rule of thumb for control is to have your resolution 5X your intended accuracy. So I'll only be good to about 0.001". Probably adequate for a machine that went in service about 2 yrs after the Berlin wall went up.
Stay healthy, stay safe.