Help with making or buying X-Y table/gantry

[cnczoner]

Hi all,

Posted this in linear motion forum, but no bites. Guessing this forum may be better.

Synopsis: I need an X-Y gantry for a very light-duty, but precision application (electronics pick and place system), and wondering if it would be better to buy a "kit" rather than homebrewing it. I expect that these are popular enough to get kits or pre-machined parts nowadays, btw.

Requirements: about 24" X-axis x 18" to 24" Y-axis. and I need about 0.001" accuracy. I'll build the head assembly, which I expect will weigh less than 5 lbs. I need all the mechanical parts, up to the motor mounts, and the rest (electricals/controllers/etc) is easy for me.

Option 1: Buy -- are there basic kits I can pick up to do these easily, or pre-machined end-supports, etc? I found some router kits at Probotix, Lumen Labs, and a couple other places, but they're a bit pricey and come with head assemblies/mounts that I don't need. Any other kits/sources I should look at?

Option 2: Build -- To get this accuracy, how do I decide if I should use linear rails or linear shafts with pillow blocks? If linear shafts, what diameter would I use for a 24" span? Or should it be continuously supported? Should I use ballscrews or belt drives? I expect belts would be lower cost, but is there enough "stretchiness" over this length to not achieve the accuracy? Any links to tutorials on these basics?

BTW, other than McMaster, what other good sources are there for these components?

Thanks,
-Neil.

[Matt Adams]

Neil,

Are you sure that you need .001 accuracy, or do you need .001 repeatable? The difference might be upwards of $750 or more per axis, because you'll need ground screws instead of rolled.

.001 accuracy means that you command the motors to go to any position at any point on the gantry and you'll be within .001 of an inch. As an example, if you command (2.325, 1.750) you'll be within .001 of either of those.

.001 repeatability means you'll command the motors to go to a position, and any time you go back there, you'll be within 0.001. In the above example, with the repeatabiliy, you might be within say 0.004 of the commanded values (e.g. 2.321, 1.753) , but you'll end up at those same positions, +/- 0.001 repeatably.

For pick and place with a fixed number of known stations or positions, you usually want to concern yourself with repeatability, not accuracy, but sometimes you need both. Could you describe the application a little more?

Also, would slides with linear guides, screws and in the $800 for each axis be a price range you're looking at, or is this way off?

Let me know,

Matt

[cnczoner]

I figured I'd start with .001", and see what I can get from there down, but actually I calculated that 0.003" is ideal and I know that even .005" would work for my purposes. I chose "accuracy" rather than repeatability, because I'll be putting a PCB on the board and after indexing/registering the location, the head would be commanded to go to specific positions. Repeatability means more to me on my CNC mill, where I want parts to match, so if one's off by .005, the other being off would be okay to match.... for the most part at least.

Then again, maybe repeatability is all I need if there's a way for me to easily map the error so the software can correct for it. This may be a good way to go.

But you're also assuming screws here. Are belts not be a good option here? And why? Perhaps I can use some "sloppy" drive mechanism and have linear encoders feeding back the accurate position?

The machine would provide a similar functionality to this, but with some differences in how/where parts are held/picked up. However the gantry would operate in a similar manner to pick up parts (vacuum with fine nozzle) from the reel and deposit them on the PCB.

FWIW, I've decided I really need no more than 24" X (width) by 18" Y (front to back). Smallest size is now 20" X by 16" Y.

My budget is a whatever is fair and most cost-effective with all options considered, that still meets the specs. I know that screws would be a big cost, hence my consideration of using belts, if those can somehow meet the specs.

But I'm still unsure of design. I see designs like this that use a single screw to move the gantry and others like this use two screws in parallel to move the gantry. Unsure if a center drive will be as accurate as a dual drive since the center drive may "rack". I've never physically used linear bearings/shafts/pillow blocks/etc, so unsure if they rack or not.

BTW, that last link to the lumenlab micro would be ideal if it had more travel and I also don't need their Z-axis setup. They claim .001" accuracy, btw.

Thanks,
-Neil.

[cnczoner]

Matt,

I got your PM as well. I'm trying to understand the mechanisms and come up with a design before I start purchasing components, so I'll get back to you once I figure out what I need. BTW, one of my frustrations with ballscrews is that I don't have a way to machine the ends, and haven't been able to find a local shop to do these for me. I'm stuck in the middle of a ballscrew conversion on my cnc mill for this reason (still need to convert the X-axis). Is this something you can provide?

Cheers,
-Neil.

[Matt Adams]

Absolutely. I can provide screws that are machined right at the factory.

All I need is a drawing.

Matt

[pminmo]

The only way to get 1 mil absolute accuracy is with a closed loop system and absolute encoders. Screws aren't that accurate unless you are willing to pay for high quality ground ballscrews. And in reality you don't need 1 mil accuracy for pic n place.

[cnczoner]

Yeah I don't need it all the way to .001". I'm currently leaning towards pullies/belts for the drive, and compensating for the error in SW. Matt is correct, that I need repeatability now (due to the SW correction), and from what I read, belts will give me that due to no backlash. Still need to calculate stretch etc to verify this will do the trick.

[cnczoner]

BTW, I contacted these folks today... Nexen. Looks awesome if it meets their claimed specs, but expensive -- $409 f/each roller pinion, and $420 for a 39" length of rack.

[ahren]

Neil,
For a 24" x 24" machine, you might consider going with regular ACME screws and a linear encoder system -- this way you're directly measuring your linear position rather than relying on accurate translation of rotational movement to linear motion. US Digital sells inexpensive mylar encoders you could use, and fairly inexpensive read heads. I wouldn't recommend this for a router due to dust, but for a pick and place system, it should work just fine.

Edit: Link to US Digital site. http://www.usdigital.com/products/en...al/linear/lin/


Ahren
www.cncrouterparts.com

[Matt Adams]

Originally Posted by cnczoner BTW, I contacted these folks today... Nexen. Looks awesome if it meets their claimed specs, but expensive -- $409 f/each roller pinion, and $420 for a 39" length of rack.

Disclaimer: I represent Nexen Products in Illinois as a distributor.

With that said, the Nexen Roller Pinion System is awesome!

It has some pretty incredible properties compared to a typical rack and pinion or a higher precision ball screw with preloaded nut.

One of the limitations ball screws it that the absolute accuracy of a screw increases linearly with the length. Longer screws will have a wider accuracy band.

However, because of the way the racks are manufactured, the error in non-cumulative. Also, because there is a rolling action, the efficiency is much higher than it would normally be in a rack and pinion setup. The lubrication requirements are limited because the rack is self lubricating with the coating on the rack.

Another neat part is because 2 or 3 teeth of the roller pinion are always engaged, there is no backlash.

To top it off, the speed the Nexen RPS allows is incredible - 425 in/sec!

If you compare this to the cost of screws with similar accuracy grades with preloaded nuts, and the fact you can't get nearly the speeds, I think it has a lot to offer. The gantry manufacturers I've worked with agree!

http://www.nexengroup.com/rps/


Matt

[cnczoner]

Okay, I've been researching and have some updates...

I've come up with my basic/preferred design, though some questions w.r.t. engineering the system still need to be worked out. The base will be a sheet of 1/4" or thicker aluminum. Perhaps 1/4" with some 8020 rails beneath it to twist support.

The Y-axis will be 1/2" linear shafts on two sides, with support blocks on each end, and small support blocks in the middle. I chose blocks over contuinously-supported rails since the cost seems quite a bit more for the latter. (Although that's McMaster's pricing, and I know they're quite expensive). The rear support block on each side will hold a stepper motor. For the linear bearings, I'm leaning towards ball bearings, since I expect this will have a high duty-cycle.

I'll use pullies/belts on both sides. Reason is that from everything I've been researching, pullies/belts will give me very good repeatability, and I can map the error. In addition, those US-Digital linear encoders ahren linked to can be used later (although I've not yet determined if steppers with encoders are easily integrateable). I prefer steppers over servos since I have a bunch of steppers and some stepper-drivers laying around here.

Next, for the X-axis, I think one linear rail would be enough. This will be 24" wide though, so I'll put the linear rail on it's side (so the weight of the head won't make it sag). I've considered dual rails/shafts, but that gets complicated to support -- I've considered a "backing plate" with supports to th rails and some truss-type supports, etc (see lower drawing for thoughts I had on this). All I'll need with the linear rail is a couple right-angle mounts to hold it to the Y-axis linear bearings, and one of those plates can be made to hold a stepper.

Partial drawing...

[cnczoner]

Matt,

Now I'm ready for you. Tell me what components you offer for the basic design I have above. I'll need shafts, supports, bearings, one linear rail, and let's assume about 1" dia pullies (most probably 0.25" I.D.) for now with belts -- 6 pcs. What's the widest belt available in this size, btw?

Cheers,
-Neil.