PDA

View Full Version : DRO Scales or Linear Encoders



hammerhead74000
03-01-2005, 06:03 PM
I need to generate a calibration table for my Shizouka, and, consequently, need a linear encoder to do this. The problem is the ones I've found are $$$ - it's often as much to buy one encoder as to buy a whole DRO kit! Does anyone know where a good source (besides ebay) would be? I am aware of US Digital, but was hoping to not have to use a mylar scale.

nheng
03-01-2005, 06:06 PM
How about a cheap digital (caliper) scale with RS232 output? What sort of accuracy are you looking for? Den

hammerhead74000
03-01-2005, 06:22 PM
It needs to be 32" long, and needs to have an output that the CNC can read (I'm writing the software, so the exact format is not critical, as long as it's documented - quadrature, serial, etc). It needs a resolution of at least .0005, but more is better. Acuracy should be at least .002/ft. but, of course, I'd like better. http://bbs.homeshopmachinist.net//smile.gif

I haven't found a scale that meets this requirement that's cheap - if you know where one is, please let me in on the secret! http://bbs.homeshopmachinist.net//biggrin.gif

cliff69
03-01-2005, 06:48 PM
cdcotools.com has some cheap scales that are basically the same thing as a digital caliper, they make them up to 40". What control are you putting on your Shizouka, I have one too, I just bought new spindle bearings for it, right at $700 for the 3 bearings.

disclaimer: I have no affiliation with cdco.

[This message has been edited by cliff69 (edited 03-01-2005).]

hammerhead74000
03-01-2005, 07:23 PM
Thanks for the link - I'll look into those.

My machine an AN-S knee mill, from the late 1970's. It came with a "Retro-Tek CNC-88" (yeah... that's what they called it! http://bbs.homeshopmachinist.net//eek.gif ) that I sold the parts out of (some guy down in the San Jose area had another one with the same control, and needed the boards).

The control that I am installing is a custom-built PC based system (it's close to being finished, needs calibration, and I need to hook up the spindle speed control, and make a new cover for the head with the new switches, trackball, & LCD panel in it). The servo drives are Rutex R990H units, and I have a Hitachi SJ-100 vector drive for the main spindle motor (I only have single-phase 220 here, and of course, it's a 3-phase motor). I am writing my own control program, with network integration, various math-derived curves (NURBS, involute, and epitrochidal, to name a few), coordanante space transforms, a proper programming language (I HATE EIA-274 based macro programming - it's doggon ugly ), feature probing, and some other stuff.

Just wondering - which Shizouka do you have? Is it CNC?

cliff69
03-01-2005, 07:31 PM
I have an AN-S also, it used to be a "Spindle Wizard" cnc, I was going to retrofit it, but then I came across a Cincinati bed mill that I am retrofitting instead, so I am rebuilding the Shizouka into a manual mill. BTW, I am using a Delta Tau pc control on mine, I am also going to write my own control software.

[This message has been edited by cliff69 (edited 03-01-2005).]

John Stevenson
03-01-2005, 07:54 PM
Hammerhead.
No email address showing.
Drop me a line
john [at] stevenson-engineers.co.uk

hammerhead74000
03-01-2005, 09:05 PM
Oh, yeah... I have the same handle here as I do on my yahoo mail account - just add
"at yahoo dot com" to email me. (note however, that that account doesn't get checked as often as my other accounts, so mail there might take a few days to respond to, unless I am expecting it) I just don't like entering my email address all over the place - doggon spammers! http://bbs.homeshopmachinist.net//mad.gif

smurph
03-02-2005, 12:38 AM
Have a look at the US Digital linear encoders. They have mylar strips up to 500 cpi that, when combined with the correct encoder head, can give a resolution of .0005 of an inch. Be warned that you can spend hours fabricating this stuff up into a scale. The output is standard quadrature. The repeatability is excelent. I did all 3 axises on my mill with these and it was cheaper than one axis with a glass scale.


------------------
That's my story and I'm sticking to it...

hammerhead74000
03-02-2005, 02:06 AM
smurph - I am very aware of US Digital - I was trying to avoid mylar. How is the accuracy of the setup you have over larger travels (i.e., you try to mill something 24.0000 long, and you get what? 23.998, 24.002, etc)

Forrest Addy
03-02-2005, 02:25 AM
If you gonna write the software set aside some registers for calibration factors. Set a zero, dial off a distance, check with a gage block, note the error, work the math and crank in the correction.

hammerhead74000
03-02-2005, 03:29 AM
Writing now - the windows behind my browser has the source code in it. http://bbs.homeshopmachinist.net//smile.gif

A single configuration variable for each axis won't handle the situation; the errors are non-linear. So, each axis has a table of cannonical position to servo position translation values, for each position on the axis, going both ways (that also takes care of backlash compensation). If I ran the calibration table at full machine resolution (it's table size is configurable in the souce code), that's only 5MB for the x axis, 3MB for y, and less than 1MB for Z - as my smallest CNC (my MaxNC) has 32MB of ram, my largest (the Shizouka), 128; that leaves (worst case) about 20MB for actual job code, but more typically it's about 100MB available on the Shizouka.

I specifically don't want to try to use a set of guage blocks, dial indicators, etc to do this - it gets quite tedious, and is tricky to manage large-scale errors. I want to just set the encoder up on the table, and run the calibrator program, and leave it for an hour as it takes it's measurements.

One posibility mentioned above is a long set of digital calipers, another is US Digital's mylar encoder - if I get some good feedback on the accuracy of the US Digital product from smurph or others, I might try that, despite the fact that it's not a glass scale. However, what I'm still hoping for is a good source for inexpensive glass scale encoders.

smurph
03-02-2005, 09:04 AM
<font face="Verdana, Arial" size="2">Originally posted by hammerhead74000:
smurph - I am very aware of US Digital - I was trying to avoid mylar. How is the accuracy of the setup you have over larger travels (i.e., you try to mill something 24.0000 long, and you get what? 23.998, 24.002, etc)</font>

Well... I haven't done too many scientific tests, but I have found the US Digital scales to be spot on with the dials accross the whole X axis's 34" travel. That's plenty good for me, as the longest thing I usually mill is about 12" or so. I'd say that the setup is well below .002/ft that you require.

Having said that, the amount of accuracy you get out of them will depend on how well you build the scales and/or how they are installed. My first attempt at installing them didn't provide the results I was looking for. I went back and was carful the next time and did a better job of fabricating the mounts. Lo and behold, things started looking better.



------------------
That's my story and I'm sticking to it...

Paul Alciatore
03-02-2005, 12:47 PM
<font face="Verdana, Arial" size="2">Originally posted by hammerhead74000:
smurph - I am very aware of US Digital - I was trying to avoid mylar. How is the accuracy of the setup you have over larger travels (i.e., you try to mill something 24.0000 long, and you get what? 23.998, 24.002, etc)</font>

I also would like to find a source of less expensive scales and have been looking with limited success. I have even considered rotary encoders with some kind of line attached to the moving parts. But the problem is chips. It would have to be very well protected.

Just a thought, but if you are writing your own software, it should be possible to add a correction for any errors you measure in the scales. If the mylar is permanently secured to a steel or other solid base, the errors should stay relatively constant. You could then take a series of readings and determine the type of correction needed. A simple linear factor may be OK or you may have to use several points along the travel and interpolate between them. I have considered this as almost a necessity with some of the hair brained ideas I have had.

Paul A.

hammerhead74000
03-02-2005, 06:35 PM
smurph - Well, that makes me feel better about them. http://bbs.homeshopmachinist.net//smile.gif How did you mount the mylar scale - is it clamped along it's length, or did you suspend it on a small spring, or ???

Paul Alciatore - The trouble isn't the software; it's measuring the error in the first place. I don't have a CMM, or other high-acuracy method of measurement; I'm relying on the encoder to provide this. However, I think I might have a line on a glass scale; if it comes through, the problem might be solved. Also, I don't have to worry about chips, because there won't be any while it's on the machine - the idea is to mount the encoder to an axis, run the calibration program, and then take it off; leaving the measurements on the hard drive of the CNC. When machining, the system relys on the rotary encoders in the servo motors for positioning, not the calibration linear encoder (this is, of course, assuming that the calibration process does not reveal any major repeatability problems in the servo positioning system).

Bruce Griffing
03-02-2005, 07:23 PM
Why not just buy a couple of glass scales from the Singapore supplier on ebay and run your servos from them. Then no calibration would be required.

hammerhead74000
03-02-2005, 07:45 PM
There is a small amount of backlash in the X axis; and I think that it will cause oscillation issues with the servo drive - but it is a concept that I had considered. That said, do you have a link to this seller from Singapore? I've been searching eBay, and not found quite what I've been looking for (perhaps I'm not using the right keywords?). Also, I am aware of the 48" Acu-Rite on eBay that ends tomarrow, and might try to bid on it; we'll see how that goes.

Paul Alciatore
03-02-2005, 09:53 PM
If you are worried about 0.002" over 2 feet, I would wonder if counting on the encoders in the motors AND THE LEADSCREWS to hold that accuracy while in actual use is a good approach. Many factors are going to come into play. Heck, a 7 degree F temperature change will change the length of the machine itself (CI) by that much. Dirt, oil film, backlash, tilting and rotation of the table on the ways under cutting forces, etc., etc.

To my mind, high and repeatable accuracy is best achieved by using a direct reading device to tell the operator or servo system exactly where the table and the WORK is. Ideally, these scales should be on a line that goes through the point where the tool contacts the work, not offset. Going at it from the other end, the motor and leadscrew end, is a solution of only medium accuracy.

Paul A.

hammerhead74000
03-02-2005, 11:01 PM
The 23.998, 24.002 figures were just an example of what I was trying to get a feel for; not examples of the measurement I was expecting. Actually, I was expecting perhaps 4 to 6 times that from the mylar encoders, although smurph says he's getting good results with them.

The reason for trying to minimize the error of the encoders is that I am expecting that the total system error will be the calibration error plus the stretch, thermal, and other mechanical distortions that vary according to machining load, etc. If the mechanical error is (just to grab some round numbers) 50% of the total, and I cut the calibration error in half, that's a 25% increase in accuracy, just from being careful about the selection of calibration encoder (again, round numbers, but it seems to me to be worth pursuing - or, at least, asking about http://bbs.homeshopmachinist.net//smile.gif ).

As far as putting the encoders on the work-piece for super high accuracy, you are absolutely right; but I'm afraid that the servo drives won't like it - I think that the backlash is going to cause oscillation. However, I suppose I could implement a second servo loop, with the PC looking at the linear encoders, and generating correction factors in real time; but that's going to require either another $150 of interface hardware, or making another custom PCB - plus a full 3 axis linear encoder setup. Perhaps at some point in the future, I'll try this, along with thermocouples for thermal expansion compensation; but for right now, I'm going to start with a software-based calibration table, and see how far it takes me.