Its very possible to use magnetic linear quadrature encoders as feedback for servos - most pick and place machines for electronics work this way. The servo controller take the input from one just as it would from a rotary encoder.
Printable View
Magnetic linear quadrature encoders are used in pick and place SMT equipment to remove any errors from the drive system - backlash, thermal expansion, couplers, tolerances on the screws, etc. There's a lot more factors in the motion control downstream of the encoder that can add up to relatively serious errors. In a Haas VMC for example, we see tolerances shifting several thou over the course of the day just due to the 10*c temperature change in the workshop and machine from morning to afternoon.
They easily operate past 2500ipm (high speed assembly line machines move well over a metre a second) and offer better 0.001mm resolution before the quadrature, which gives you 4x that. You can get pretty cheap magnetic linear quadrature encoders on aliexpress and robotics supply stores. To a servo drive, they look exactly the same as a rotary quadrature encoder (which every servo uses) so there is no re-invention of the wheel, this is standard technology that's been on mass produced machines for the past 30+ years. I've plugged a 20 year old Siemens SiPlace 80F3's encoder output into a DC servo drive and it was no problem at all. That machine was running a brushed 100VDC, 30A motor on it's Y axis (long axis the gantry rode on) which only had a tacho output from the motor, no encoder at all on the motor itself. Those machines need 1/4 thou or better accuracy for placing parts reliably (about 12 parts a second) and only have an encoder strip on one side of the axis - newer models need well under 1/50th of a thou accuracy to place 01005 components (0.001x0.0005" components).
Just because they are not common on CNC routers, doesn't mean they haven't been conquered many decades ago in other CNC equipment. There's plenty of CNC mills and lasers running linear encoders on their axes too :) It's a great way to get high tolerance with rack and pinion (thinking Trumpf lasers here)!
I've worked on jobs for NASA, the US Navy & airforce, Cessna, Dillon and several other aerospace/traditionally "high spec" clients and they have been some of my least demanding as far as tolerances and material specs go!
It's been 6-8 yrs since I worked on this, but if memory serves, you can get ~2.4 million counts per revolution out of a rotary quadrature encoder.
I've not done much in the linear world, but have done several projects for a defense contractor needing rotary motion. Using servos (w/ built-in encoders), mounted to a special low-backlash (:: hi-$$) gearbox w/ quad-encoders on the output shaft, we could easily hold rotational accuracy of +/- 0.002 degrees. The spec was +/- 0.010 degrees. At some point, measuring how accurate/repeatable a given system is, becomes more expensive than building it. I could have claimed the 0.002 figure, just don't make me prove it.
Perhaps more interesting and appropriate to the discussion is that the encoders are remarkably affordable. ...What you feed it to, a PLC in my case, may vary wildly in cost. (The gearbox was absolutely stupid-expensive!)
Richard, nice looking machine. May I inquire as to what tolerances your machine can hold? I'm trying to get a good, real world understanding of what is possible with lower budget components.
Richard, I like how your machine is setup to support milling vertical workpieces on the end...great versatility!
There is both a simple and not so simple answer to this. Don’t mean to hijack this thread to talk too much about my machine (I have no pretensions as to it being an “ultimate” machine) but I will do so a bit. My goal when I built it was for it to be accurate to +/- 1/32nd of an inch. Wood moves with humidity, and that is probably about as accurate as I can read the scale on my table saw when making cuts there, so I was trying to be realistic.
On my X and Y axis, I use rack and pinion, 20 pitch. On these axis, it works out that I have around 864 “steps” per inch using my gecko 540’s 10 microteps per full step (non-configurable). So each “microstep” is around .0012 inches, and each full step is .012. If I set a jog increment of .001, and set up a dial indicator to measure movement at the bit, I can actually see each microstep, but about every fifth jog step will not cause any movement. 4 of my “steps” equals .005. These measurements are under no load of course, and it may not be as accurate while cutting. Of course, my pinion gear and the gear rack are also not perfect and there is some slight additional accumulation due to those variations. This is one of the reasons I asked about the feasibility of using linear encoders on the “ultimate” cnc.
Overall, if I cut a cabinet side panel, I am confident I am within my 1/32 goal from nominal dimensions. If I cut several of them, I have found they are very consistent (less than 1/32) variation). If I cut a 2” x 3” pocket and drop a 123 block in, there is probably less than .005 error, but depending on where it is cut, it may be .005 loose to .005 undersized, depending on how the location matches up to full step boundaries, and potentially depending on whether I climb or conventially cut the pocket, and what diameter bit I use. I can cut dovetails that fit tightly. It is a tool that needs to be learned. I am not afraid to use a chisel or handplane to tune a joint if necessary, and sometimes it is. I am a woodworker, not a machinist.
The front table can tilt from vertical to horizontal or anywhere in between, and the spindle will extend past the fixed table about 6 inches. I was actually having some issues with flex of the tilting mdf table and am working on a much stiffer replacement (have a 1 meter length of the 80mm x 160mm extrusion I am going to use).
Richard G....
No worries about the hijack, as I expected it. CNC guys cannot talk about machines without getting sidetracked. Your last explanation of tolerance was refreshingly more realistic than most with a self built machine offer. Thank you. The step resolution number you posted are the reason for my "NASA" comments. Having priced out HIWIN linear encoders for a calibration project, I have an idea of what they cost. And as I expected, you were seeking info on a ten thousand dollar solution to a 500 dollar problem.
Forgetting the cost of the encoders and the control system required to put them to use, the combo of parts you have on the machine are not capable of much better. In reality, most small homebuilt machines will only achieve an accuracy of the half step distance. No matter how much math you recite and how many decimal points you can type. Your machine suffers from a lack of mechanical reduction and too much applied electronic "trickery". I say trickery because we know the drastic drop off in torque when using 1/4, 1/8 etc microsteps, 1/10th microstepping gives good numbers to the uninformed, but also means that on a router with ever changing loads we could seldom see correct positioning. You can't use software to fix hardware problems. Tenth microstepping was invented to keep motors quiet. All I ever found using those settings were bad edges and heat.
If I were you, and without knowing the specifics of your system, I would throw on 5:1 planetary gearboxes or a belt drive on the X and Y, put on drives that have DIP switches to set the current and microstepping and increase the power supply(s) to the max recommended for those drives. And post your results
So far, other than being able to match the cost to value ratio for some of these options, the list is fairly well covered. By definition, if you post an option and then qualify it with a price, you already know you cant afford it.
Forgetting price, what are the features that would make your life in CNC better?
As someone who is relatively new to cnc, I'd appreciate good set of video tutorials for the specific machine and controls.
A questions Gary - I noticed your next 2x3 machine will have a Centroid Acorn controller. Does the Acorn support a closed loop system?
Yes, of Course. All controls that I am aware of support closed loop? Do you have one in mind that you believe does?
I know little about controls, but I'm under the impression that Mach3 doesn't. Unless you're referring to closing the loop in the driver.
Richard, here are the Acorn specs. I am thinking about an Oak upgrade, so i have been doing a lot of reading to try to gain some understanding. Complete specs---> http://www.centroidcnc.com/centroid_...d_diy_cnc.html
Attachment 378198
Now that all the Technical stuff is over, does anyone have something on their wish list that we missed?
I can't remember if anyone mentioned this, but the ability to work on the end of a long part held vertically. For instance to cut tenons.
Also, and this is probably more about software, the ability to scan/map a 3 dimensional surface relatively easily. I have in mind being able to do inlay work on something like a turned bowl or vessel.