[SMW Precision]

Project

Adapt a linear scale DRO system for use as a CNC closed loop positioning system. This has been done previously and claims are out there of an appealing 60% increase in machining accuracy versus an open loop encoder driven positioning results.
Scale form used in testing will a GRID measurement technology with added truing or reset points in course of travel. GRID is a line marked media strip using up to a 100,000 marks per inch; a sealed glass scale in this case. Scale resolution is 1um or 0.001mm which matches closely too servo tick count of 40 to 50 millionths of an inch.
Machine form is a Standard RF45 mill/drill with an Industrial Hobby CNC retrofit kit.

Looking for comment or thoughts on project! I have ability too have DRO system modified as needed!

[crazyneosimpson]

I have used a toshiba floor borer with a closed loop scale system, great machine and never needed homeing!
I think it would be a great leap foreward if there were a control software like the mach# to enable it, but I only do hardware!

[pstockley]

I like the idea. What controller software do you plan on integrating this with? I am not too sure it would be possible with Mach due to the timing requirements and communication throughput of the parallel port. However, the Gecko G100 version sounds like a viable candidate.

[SMW Precision]

I like the idea. What controller software do you plan on integrating this with? I am not too sure it would be possible with Mach due to the timing requirements and communication throughput of the parallel port. However, the Gecko G100 version sounds like a viable candidate.

Mach 3 supports linear scale output! Drives are Gecko servo; I going with a standard kit because of time constaints and advantages in doing testing with a known system. I want too minimize the variables as much as possible.
My software question is in later testing stages to be able to compare and measure positional deviation between encoders and scale system. Both systems will be mounted at the same time.

Ken

[pstockley]

If you just want a DRO style read out this would be OK. The problem is that it would be impossible to do any kind of feedback loop. Writing a script is out of the question because they only run every 10th of a second. If you want to maintain accuracy to 10ths you need to have a feedback loop that operates way faster, probably in the range or 10,000 to 100,000 times a second. Having a system that tells you that you have exceeded some kind of tolerance is OK (I know of a board for mach that does this already). The problem is that your part is potentially junk at this point so it doesn't really help much.

[SMW Precision]

If you just want a DRO style read out this would be OK. The problem is that it would be impossible to do any kind of feedback loop. Writing a script is out of the question because they only run every 10th of a second. If you want to maintain accuracy to 10ths you need to have a feedback loop that operates way faster, probably in the range or 10,000 to 100,000 times a second. Having a system that tells you that you have exceeded some kind of tolerance is OK (I know of a board for mach that does this already). The problem is that your part is potentially junk at this point so it doesn't really help much.

No both systems could not be operating the machine at the same time. But recording and comparing measurement data should not be that difficult. In this test linear scales would only be generating a data stream for comparison versus encoder positioning. Thought is to generate data for actual measurement deviation between both system types. Much more informative data than only part generation and comparison.

[gus]

I think older servo systems had an update rate of the 1khz range[think of the sound, 1000hz is the old emergency broadcast system]

[pstockley]

I am not saying that they would be both operating the machine at the same time. Even comparisons you can only do 10 times a second in Mach currently. This is pretty useless unless you just want a sanity check for say missed steps on a stepper system.

[Ron22]

Are you comparing a DRO (Linear Encoder) Closed loop to a Encoder Open loop or Closed loop? Your message said "open loop encoder driven positioning" This sounds more like closed loop.

I have done machine with encoders on the motor for velocity feed back and linear encoder for position with Delta Tau controller.

The improvement in accuracy depends a lot on how the machine was built to begin with. If the machine is built with all quality part and direct drive of the ballscrew or gear rack you will se less of a gain. If the machine has belts or gears before the final drive you will have a larger gain.

Here is an extreme example of what I mean. I just worked on a machine with 3/4 of a motor turn backlash in the gearbox. It sill has good positional accuracy. The machine has very movement when you change directions due to the gearbox but it is still accurate.

I know nothing of the Mach3 and update rate might be an issue.

[SMW Precision]

I am not saying that they would be both operating the machine at the same time. Even comparisons you can only do 10 times a second in Mach currently. This is pretty useless unless you just want a sanity check for say missed steps on a stepper system.

I am going to review Mach 3 manual. Remember this project is only in the starting stages and has a long way to go. Closed loop has some significant possibilities. I expect problems and work resolving them.

[pstockley]

I would talk to Art at Mach. I know he plans on supporting closed loop control for the G100 mach IV. He said that true closed loop isn't possible on Mach3 due to the timing limitations involved. The big problem you have is that most of what you need to do is outside of your capabilities as you cannot modify Mach. Working with Art would seem like the only way to go.

[SMW Precision]

I would talk to Art at Mach. I know he plans on supporting closed loop control for the G100 mach IV. He said that true closed loop isn't possible on Mach3 due to the timing limitations involved. The big problem you have is that most of what you need to do is outside of your capabilities as you cannot modify Mach. Working with Art would seem like the only way to go.

You are correct that seems the place and man to go too! Core element for a closed system that I have is the Linear scale hardware solution. Correct scale build, protection and measurement system needed to make closed loop practical. I also have pricing and production capacity to make it a viable product. It has taken a year and half to get a compatible scale system arranged with needed modifications.
Three elements I need too bring together are Hardware, Software and a result that justifies the product. R & D is both fun and frustrating!

[pstockley]

Sounds like it has a lot of potential. Keep us updated.

[SMW Precision]

Sounds like it has a lot of potential. Keep us updated.

Thanks for the suggestion of contacting Art I did so! He is interested though it will be about 2 months before he has time to work on Mach 4 for linear scale.
Actually the time line is about the same as mine to bring my IH kit and RF45 Mill/Drill testing system. Hopefully I get the stand started a 3 phase motor and a Vertorless Hatitachi SJ200 VFD installed this week. They were nice enough to include a 300 page manual in 6 pico type so it fit the box.

While linear scale software system gets together I believe I will put out a couple of New DRO kits. For the RF30 to RF45 mill/drill's a 2-axis DRO system with computer interface and connecting software. This will be a complete installation kit with a sub 2 hour installation time. Included in kit will be a direct Spindle attachment Z-axis digital scale 0.0005 resolution. Also a kits for 36 or 40" Lathes.
DRO system will be 0.0002 resolution standard kit with option for high resolution 0.00004 scales. DRO provides 200 SDM & 200 Tool Offset of built in memory. Bolt circles, curves and pockets are a few of the incorporated features. Standard kit cost I am going to try to do for about $700. I will put them up on the Board in about a month when completed.

[mactec54]

Hi Ken
The PC control I have on my machine will run encoder feed back or linear scale my motors have 17bit encoders std that is 32,768x4 p/rev I have mine set at 16bit that is 16,384x4 p/rev and is supper at any speed the control is made by SoftServo and can run at 60m per min cutting check out there web site may be of some help It is also cosed loop and in velocity mode

[Evodyne]

Ken,

A couple of thoughts. First, your linear encoder moves at tool speed, while a motor encoder turns at, you guessed it, motor speed. So if the servo has a 10:1 reduction ahead of the leadscrew, and the screw has a given reduction (say 5 turns per inch) then the motor will make 50 turns to get one inch of travel. If the encoder was 1000 counts per revolution, then you have 50,000 counts from the motor encoder per inch of travel. Given a x4 quadrature encoder input circuit, your electronics would see 200,000 pulses (per inch, for the motor mounted encoder). My point? Your glass scale has 100,000 counts (marks per inch) and a x4 quadrature encoder input circuit would see 400,000 pulses (per inch, glass scale). So only twice as much output as the motor encoder-not so much-at least for the configuration I mention. You can also get a x1 quadrature encoder input circuit and see only 100,000 pulses per inch. The benefit is that you are measuring the tooling and not the motor, so any lash is not seen.

Also, I have heard that a system with slop can experience hunting when using a linear encoder vs. a motor mounted encoder.

Finally, there are IC's that process the quadrature encoder signals and generate the signals needed to clock a chain of up-down synchronous counter chips. How many chips you add determines the the resolution. This can run in real time and not lose data (counts). This is how certain servo drives track encoder signals internally. At any rate, you can then turn the counter data into an analog output or read it into a PC via a couple of methods. Of course now you're into a "roll your own" approach, which you may not want to tackle.

That's my two cents worth!

Evodyne

[SMW Precision]

Ken,


So if the servo has a 10:1 reduction ahead of the leadscrew, and the screw has a given reduction (say 5 turns per inch) then the motor will make 50 turns to get one inch of travel. If the encoder was 1000 counts per revolution, then you have 50,000 counts from the motor encoder per inch of travel. Given a x4 quadrature encoder input circuit, your electronics would see 200,000 pulses (per inch, for the motor mounted encoder). My point? Your glass scale has 100,000 counts (marks per inch) and a x4 quadrature encoder input circuit would see 400,000 pulses (per inch, glass scale). So only twice as much output as the motor encoder-not so much-at least for the configuration I mention. You can also get a x1 quadrature encoder input circuit and see only 100,000 pulses per inch. The benefit is that you are measuring the tooling and not the motor, so any lash is not seen.

Considering numbers given what is your actual repeatable resolution?

Also, I have heard that a system with slop can experience hunting when using a linear encoder vs. a motor mounted encoder.

That is true varing with amount.


Evodyne

There is one core area that cannot be resolved even with the best CNC equipment of the highest cost or precision. With an open loop system moving metal creates friction. Friction or heat affects the demensional size of metal. Core reason that Linear Scales have come into use is ball screws will change in length as temperature varies during operation. You are attempting to use a variable for a fixed measurement.
With our lower end machines there are a number of other issues Linear Scale or closed loop will also resolve.

Ken

[gus]

I have seen linear encoder machines hunt, but it was an old machine with crappy westamp drives and some backlash. Newer drives run 10 times the frequency and don't constantly drift in offset like old ones, so would not tend to hunt, IE the Copley drives I used I rarely touch

[SMW Precision]

Well it has been a couple of weeks from my last post on project. There has been progress and let me thank all users for previous posts. Your information has been helpful.
My Industrial hobbies kit is in shipment and will arrive mid next week. Computer interfaced DRO displays have arrived and Mill/Drill & Lathe kits will be out over the next several weeks.
CNC interface, odd things seem to happen! Roger walked through the door last week and gave me some major design help. Roger is a prototype design engineer who has several decades of experience in CNC system integration.
According to him direct linear scale connection for CNC presents oscillation problems unless ball screw has almost no backlash, there is a far better way to do it! Using linear scales as a secondary system too measure actual results and update system parameter files.
For this data base a 3 dimensional grid is used too represent machine movement field. Each square represents on encoder tick or resolution point; linear scales measure actual machine performance. System adapts to machine or condition changes through out the machining cycle.
Here is the link to Roger’s company website www.nli-ltd.com. Their development of what they call Genetic Software is particularly interesting.
This week I started hunting Probes for eventual integration and I have identified several possible suppliers. Shortly I will start ordering samples!

Ken