Sanity check- 6'x10' table

Howdy- I was hoping that someone could do a sanity check on my current plans for a CNC table. I need a machine that will take a 50"x100" sheet of acrylic (normally 3/8", but up to 1" thick) and route out round flanges, square panels, key-holes, slots, etc... My end goal is to let the machine do all the work, and I only glue up the pieces.

Similar to these- Note that they have 1/8"x1/16" grooves for O-rings in them, so they are not just simple through-cuts:

http://www.barraquatic.com/skimmers_files/image005.jpg

I am thinking of following the Data-cut plans (not ordered yet) for a fixed table setup.

Current plan:
6'x10' table made out of 80/20 extruded aluminum.
Approx 9' motion in the X axis, via a single 1" fixed ball-screw (.250 pitch) and a spinning nut setup on the cross-bar under the Gantry.
Gantry rides on one 1" steel shafts using two open ball bearing pillow blocks per side.

Approx 5' motion in the Y axis, via a 1/2" ball-screw (.200 pitch), riding on a total of two 1/2" steel shafts and four ball bearing pillow blocks.

Approx ??? motion in the Z axis, via another 1/2" ball-screw and two 1/2" steel shafts (4 pillow blocks).

Vacuum hold-down setup via a ported table, and a MDF scarf sheet drilled with 1/8" holes.

Dust collection system similar to a shopbot.

Three Gecko drivers

Nema 34 Steeper motors on Z/Y, Nema 23 for Z. Unknown torques at this point.

Ideas that I want to keep in mind for future expansion:
Add another Y/Z head to the back of the gantry, with a mill/drill/tap setup.
Leave room that the heads can reach for a quick-change tool setup.

Major concerns-
- Using a single drive shaft for the X axis is troublesome, but changing to a dual shaft adds at least $500+, and I'm not sure I would notice a difference. Would it be possible to do a dual fixed shaft, dual steppers on the cross bar under the gantry? I've been considering rack and pinion but am afraid there would be too much slop for getting perfect edges on acrylic.
- Sag in the table. I also want to use this extremely expensive table as a glue-up table and need less than .015" sag at any point in the table in order to make it fast and easy to glue up large tanks. This requirement, coupled with the single-screw driven gantry option, requires an extremely large extrusion (3"x6") for the main rails supporting the table in order to get reasonable deflections.

I have a little experience with CNC- I built a R/Theta/Z (polar) CNC prober many years ago for a company that needed to move a oscilloscope probe around a test head. I have extensive experience with AutoCAD, mostly 2D, but am proficient in basic 3D as well. I'm reasonable with a Jet Mill/drill and most shop tools, but I've never done tool paths, or run a CNC router.

Greatly thankful for any suggestions-

Zeph

With a table and travels that big plus the weight needed for rigidity, consider using full sized servo motors. Keep all your ball screws the same pitch and at least 1.25 dia X , 1.0 Dia Y and .75 Dia Z

What Castguy2003 said. Go servo all the way!

I can't think of any way that you would keep the gantry square unless you somehow drive it with the screw down the centerline. Even then, off-center loading ( cutter load on extreme edges) could give you jam-ups and inaccuracy. I think two parallel screws would be the simpler answer, and give the machine much needed rigidity.

I was imagining that you could create a truss table, buying a bunch of that fancy T-slotted aluminum extrusion that you can get from Reid Tool Supply. I'm not sure how thick a top you were planning on using, but chances are that the top itself is not going to be all the flat to start with, and if it was, it will sag anyways.

With a truss setup, you can prestress and flatten your table sheet. You would need an upper and a lower parallel bar for each truss, and link the ends together with a vertical straight link. Then you can apply diagonal bracing between these two beams, to straighten and prestress the table. This might allow you to use a lighter top, since top thickness does next to nothing for vertical bracing anyways.

I've been shying away from the servo setups, since looking at the FlashCut page and noticing that the servo packs were +$2500 over a similar stepper pack. This afternoon I did a little more research, and it seams like I can piece it together for much less than that.

For example http://www.seanet.com/%7Edmauch/ has 600oz-in peak torque servos for $159 each. Would twin 600's on the X, and a single for the Y and Z be "more than adequate"?

However I know even less about servos than I do about steppers (except how they relate to RC airplanes), so have lots to learn there.

Can Mach1 control servos?

I've had problems finding ball screws with the same pitch at all the diameters, I'll have to keep looking. I have not found any local shops that know anything about ball-screws or linear shafts, so have to go off of the MSC and McMaster web pages for the most part.


The truss table is a good idea- I've been considering it too, but it seams like a lot time to get it dialed in.

I was hoping that if I could get the gantry to run true, I could just face the MDF table top to get it flat to the gantry. I expect the table to sag some over time, but it should settle in eventually. I figure I can shim under the main sheets of MDF to bring it back up to near flat if required.


Thanks for your time!

Zeph

Z, 6' x 9' seems awful big to cut what appears to be fairly small parts out of whats basically a 4' x 8' sheet.. Keep in mind you can barely get 5' x 10' mdf for your spoilboard (not that you couldn't seam it) especially in single sheet quantity. I have a 5' x 16', and 99% of all work is in the normal 4'x8' zone. I know the on going accuracy debate, but consider rack and pinion for your long axis with the twin motors similar to a Multi Cam setup. Mike

I was reading at that link provided that stepper drives do have the means to reduce the pulse frequency to simulate accel/decel, so that is good.

Do these ultra economy systems include encoder feedback? Part of what you are paying for in a real servo motor could be an integral mounted encoder.

Does anyone know how successfully stepper drives and motors
can perform circular interpolation? That is something else to check on before sinking money into a stepper drive system because you are going to want to be able to cut nice arcs with it.

Mike- I only gave you part of the story. The rest is that I need to accurately cut full sheets down to various sizes, up to 2'x8', in order to make tanks. A sheet of 1/2" acrylic weighs about 100#, way too much for me to handle on the table saw and jointer. Even with a panel saw to make the cuts, I can't handle that kind of sheet on a jointer which is required to get clean glue joints.

My dream is to heave it up on the table, square it, and hit go.

In order to get a good clean solvent-welded joint, the edge has to be nearly perfect. Customers don't like to pay for tanks that leak. :)

I will reconsider the rack and pinion.. Is that what is on your system?

HFD- The ultra economy systems don't have encoders, but that is an option on a mid-level machine. Thanks for the note about the arcs- I had not thought of that potential problem.


Zeph

Z, My machine is a rack and pinion except for the Z axis.
Before buying this machine I had an old Thermwood 4x8 I got basically DOA. I converted the drive system to Flashcut for the signal generator, Gecko drives, and reused the huge nema 42 stepper motors and power supply. I was and still am very pleased with Flashcut, and their new version of software makes it even better. The Geckos while functioning well, I think where a little light for the amps the large motors pull, and the new owner is struggling with that trying to achieve higher feed rates.
The other thing that maybe of interest to you is that in the x axis after removing one of the two heads is that I couldn't cut a full 8' and needed a longer ball screw. Not wanting to spend that kind of money on a machine I hadn't even ran yet I opted to build my own rack and pinion, with the thought I could always change to a ball screw later. Two 6' pieces of Boston Gear rack (Graingers) mounted to a large piece of angle iron, a stock gear, some timed belts and I was done. The accuracy was great, no discernable backlash, and I spent a fraction of a ball screw. Of course once I was up and running I never looked back and the rack stayed. Mike

MikeA, your first hand experience is valuable input. Can you clue me in with a few more details on:
how stepper motors respond and the feedback system your machine is using:

what kind of setup procedure you use to "tune" them:

whether you can perform circular interpolation:

How did you attain backlash-free? Just run the pinion tight into the rack or how did you do this?

Thanks for the info- I've been afraid to go rack and pinion so far. I don't really know how much accuracy I need is my main problem, just a fear of not having "enough" and loosing time/money. I'll be declining work from customers in order to build this so I obviously want to get it "right" the first time.

I just got a multi-thousand quote back for all the ball screws, pillow blocks, shafts, etc.... so that is coming as a reality check. :(

Can you post what your motor sizes are, and what gearing you used? The Shopbot uses 305 oz-in motors with a 3.6:1 gearbox. I didn't count the number of teeth on the pinion though, nor the pitch of their rack.

What kind of top speeds do you get?

Thanks greatly for your help-

Zeph

Quote:

---

whether you can perform circular interpolation:

---

If you are referring to G02 and G03 yest steppers can handle it. It's actually a function of the software that controls the motors.

All the stepper control software that I know of can do circular interpolation.

Eric

First Hu there is no feedback on most stepper systems (I believe there are some drivers that will except encoder feedback). You are at the mercy of the machine. That being said, the old machine with the drivers it had, I knew what my max speeds where and just didn't push it. In processing a 4x8 sheet of material, most fully nested with some type of pattern, and then rehomed I would be off .003-.005" at most. Most wood or plastic workers can't even measure that, especially over 8'. I believe most any newer drive system supports circular interpolation in all axis', as does the Flashcut system. As far as tuning the system this is not required as in a servo unit, although there is a series of ramp up and down settings, baud rate, and other misc. settings in the software that have to be set. That requires a little bit of experimenting with your paticular machine. As far as the pinion set to the rack I slipped a piece of teflon pipe tape and meshed the gears tight together. I had direct drive on .5" lead ballscrews on the other axis' so I used a timed belt and pulley to acheive basically the same lead or as close as I could get. I originally looked for a Nema 42 gear reducer to try to acheive the right reduction, but they cost as much as a ballscrew and there is more inherent backlash than using a timing belt and pulleys. The rack was a 10 pitch 14.5 degree and I used the smallest toothed gear which I believe was a 10 tooth. Z, as far as your project, I wouldn't be scared of rack and pinion, as there a lot of routers out there using them, new and old. My concern would be the longevity of the less expensive racks and gears, but I guess the cost savings would allow you to replace them periodically if you had to. Mike

I did some tests this weekend. An error of 0.010" or more in the wrong place is easily discernable, and would cause additional steps in the assembly process to fix. Picture two glass panels mating at 90 degrees. Even a tiny difference is easily visible, and in my case, creates a weak point in the chemically welded joint.

If I cut out a 2'x8' piece (front panel), then a 2'x2' piece that mates up to it (end panel), a difference of more than ten thousands in width means the joint needs time consuming hand work, one on the main things I am striving to eliminate. With experience, I might be able to tell where the table is "off" at and adjust my code to fix it, but since I build full-custom equipment, it is rare that two tanks are the same size.

Can a rack and pinion system be made ridged enough such that it can't get jerked around by a typical 1/2" bit in acrylic?

You know, if people would just accept fish tanks made out of solid wood this would be a whole lot easier. ;)

Thanks all-

Zeph

Zeph, if you made them out of wood, would'nt be a barrel?
Can a rack and pinion system be made rigid enough? This is some what of a loaded question. The rigidity of the machine will be a by product of how it is consructed. The afore mentioned Thermwood weighed in at 12,000 lbs., used 1.5" and 2" Thompson linear bearings. The Multi Cam I have now weighs in at somewhere around 6000 lbs. The rack and pinion I feel are only a small part of the equation. At 5x10 like you want, the frame and guides are what will most likely determine the rigidity and accuracy. Where that threshold lies of what's needed to cut 1/2" acrylic smoothly, and day to day I am not 100% sure?
I sense your eagerness to attempt your own machine, but if these tanks are a profit earning thing for you, could you look to perhaps outsourcing your parts for now and then look to buy or build the right machine later? Mike.

Mike- Good points- It is not cost effective to out-source the custom parts. I'd pay too much to someone to generate the code on each unique part to make any money at it. If I found another hobbyist that was local, it might be do-able, but the big CNC houses in town are too expensive.

There are some standard parts that I could out-source, but the quote I got back was $10 each for the 8" flange shown above. Considering I use 13 in one product that sells for $250 (due to marketing constraints), it is not cost effective either.

I agree that the rack and pinion are a part of the equation, but all the slop does add up. I am working on the assumption that if I do the best I can reasonably do at each step, the over all slop will be adequate when I get done. That's a lot of assumption, but I know of no other way to proceed.


My goal is to make a "better" machine than a shop-bot, for similar or less money. Their total accuracy is rated at 0.015", which is larger than I desire for my machine, so I'm looking at where I can get the most improvement for the least money. At this point, I'm thinking of a linear shaft/ball bearing setup to carry the weight, and am waffling on the Rack or ball-screw for the drive.


Thanks for the help-

Zeph

What happens is machines loose the accuracy when they change directions. If you make a rectangular part, I doubt the straight lines will vary by .015, but if you make two rectangles that are 10" x 15", the size may vary by .015"

My machine built for low dollars only varies by .0055", and I took that out in the software. It's only 12x24" mind you. Measuring several identical parts shows more like +/- .002".

I would go with acme rod and antibacklash nuts. I would turn the nut and not the screw on a table that size. I would probably go with 3/4" OD rod. There will be no backlash when the machine is new. Changing the nut will be easy if some develops that you can't live with. The plastic/delrin antibacklash nuts are pretty durable.

Just my 2 cents.

Eric

Zephrant, another possibility, although it could be expensive, is to mount scales on the machine for feedback, instead of spending the money on encoders for your servos.

You might only need a scale with .001" accuracy which is not a premium scale in today's technology. There is a Newall brand scale, based on a sensor which reads position from a tube filled with ball bearings. You would need to contact them to find out if you can get a suitable digital output from the scale unit to interface into your motion controller.

I did this with a Sony magnescale unit on a lathe, but the scale was only 8" long, not 10 feet. The conversion box cost $400, plus the $1500 (Canadian) for the scale. The trouble is the Magnescale is likely overkill for your application, as it is capable if incredible precision. Anyways, using a scale for feedback eliminates the positioning errors due to backlash, although a loose fitting screw makes a very poor "clamp" when the servo on that axis is supposed to hold a given position while the other axis is busy.

Thanks guys- I've been out of town for a few weeks, just got back and am in the scheming stage again.

BM- Thanks for the info- One of my requirements is to cut out four pieces of acrylic that are 4x7", and have them all come out the same size +/- 0.005" or better. If one pieces is 0.010" larger than the other, the glue joint gets marginal. Much more than that and the box is not water tight without additional work.

HFD- I have seen scales before I think, magnetic ones, but was scared off by the prices. I like the absolute feedback they present, I'm just not sure I can afford them.


Thanks for the ideas-

Zeph