Precision mold-making build -- Lancaster, CA, USA

[srange]

I've lurked here for a few months, and have decided to pull the trigger. I'm an aerospace engineer in the Mojave desert, and I'm looking for a tool to build RC aircraft molds and ultralight aircraft parts. I have a few hundred hours operating 3- and 5-axis mills, am fluent in MasterCAM and HyperMILL, but will be picking up welding and control electronics for this project.

I know MM was originally intended for high-throughput, low-precision work, but it looks to me like one of the most rigid frames on the market, with a top-notch community. I think that's as good a starting point as I can hope for, but I expect non-trivial custom work and tweaking to get the performance I'm looking for.

Desired performance:

- Buttery smooth 3-axis profiling
- Roughly 5' x 10' plus a bit (3150 x 1625 mm)
- About 18" Z-stroke (450 mm): likely based on a beefed-up version of smreish's extended z-slide work http://www.mechmate.com/forums/showt...p?t=920&page=3
- Rigid enough to cut epoxy-impregnated MDF full depth at 2mm depth-of-cut, less than 100 ipm feed, with pass-to-pass consistency around 0.0005" (0.013mm)
- Global position repeatability after full-axis traverse less than 1mm (returning to a point -- I realize rails will not be this perfect. I need precision for surface finish, not unphysical dimensional accuracy)
- Sufficient power to hog MDF at all z-depths with 1/2" bit, 50% engagement, 1in cut depth, no rigidity requirement
- Ability to 2-axis cut 1/8" 6061 sheet at top of z-depth
- Jog speed at least 100 ipm (more is always nice)

Current planned components:

- Belt reduction drive, at least 4:1 (more?)
- 2.2 KW spindle
- Bolt-together or welded frame (undecided) with dropped table for extra z-depth
- Probably 2-bar gantry, widened x-rail car, widened y-rail car (for stability with extra z-depth)
- Probably purchase 6' Superior Bearing wheel guides instead of grinding rails -- I assume this will produce a straighter rail than what I could do with a grinder skate

My general plan of attack:

- Lock down a gantry/car configuration, figure out economical rail and steel sizes before dimensions finalized
- Pick a suitable drive system. Order electronics and complete the kitchen table electronics portion
- Build out gantry and cars
- Build table

I expect this will cost about $6K, not counting a welder or control computer/software, and take about 9 months at 15 hours per week to get a cutting, tuned machine.

Please let me know what you think, especially whether you think what I'm asking out of the machine is physically possible, with enough attention to detail and tweaking.

--Sam

[MetalHead]

I think you have a plan! Welcome and we are looking forward to your build.

[KenC]

Welcome, your plan looks good to go.
One of my life biggest bragging right are those grounded rails.
The straightness of the rail is dictated by how well you do your fishing line shimming.

[darren salyer]

Sounds like a great plan.
I guess high throughput, low precision will be relative to the tolerances you desire...

[srange]

I'm starting to plan out my drivetrain. Hoping to order components soon.

Computer: Windows 7x64, Mach3
Computer interface: Ethernet SmoothStepper
BOB: PMDX-126
Driver: G203V
Motors: PK299-F4.5A or KL34H280, bipolar parallel
Drive: 4:1 belt reduction
Rack: DP20
Pinion: 20 tooth
Power supply: 1000-1200W, 56V. KL-5620 or PS-10N56

Voltage approximately set by Mariss+20% (58V), power by 67% power supply max voltage times combined current draw (1000W).

Power supply fed off 120VAC mains power.

I'm trying to pick out well-known, high-quality components in hopes that spending a few extra bucks now will save me from rebuying when the system underperforms and to minimize the part count and wiring complexity (PMDX-126 takes mains power, built to work with ESS).

If I'm doing the math right, with 10 microsteps, I'd get a step size of 0.0063" with the 20-tooth pinion and 4:1 belt reduction, which is basically an order of magnitude off my target. This implies I'll need a custom 2- or 3-stage belt reduction, but this initial configuration should be plenty accurate to build that myself when the time comes, assuming the drag of the drive isn't prohibitive.

2-stage, 16:1 should give a microstep size of 0.0016", maximum feed of 600 ipm (4MHz ESS).
3-stage, 64:1 with 35-tooth pinion (for better smoothness/linearity) should give a microstep size of 0.0007", maximum feed of 270 ipm. My favorite right now.

Some open questions I'm trying to figure out:

- What are the differences between the Oriental and Keling motors? Is one better quality, more linear, more reliable, etc?
- Am I looking at the right type of motors, given that the end use will likely be higher-RPM, lower-torque than typical for MechMates?
- Is bipolar parallel the right wiring configuration?
- Am I calculating voltage and current right? Spec sheets (see Keling KL34H280) specify inductance and current in mH/phase or A/phase. To me bipolar implies each motor has two phases, but all the voltage and power calculations I see on the forum seem to use the base numbers, without doubling.

PS: MetalHead, expect a PM soon about electric kits. Do you sell Kelings?

[danilom]

At those resolutions you will have to relay on rack and pinion precision
and microsteps don't equal to positional accuracy as common stepper motors have a 5% linearity error. Its the size of 1/10 of a step that is the reason behind Mariss making drives not go beyond 1/10 of a step (10 microstep). Microsteps only improve smoothness and resonance in stepper motors.
If you wan't to calculate positional accuracy then count on only 1/2 step as motors tend to lean to only full or 1/2 of step when positioned.
Also steppers tend to have a max rpm around 1000rpm so count that too for rapid feedrate.

There was a nice webinar about steppers linked somewhere on forum if you want to explore further.

Bipolar parallel is the way to go if you want RPM.
your calc are right, the current set on drives is what is in the sheel, bipolar parallel - 6.3Amp

[srange]

Danilo,

Thanks for the 1000rpm rule-of-thumb. I made an error in my original feed rate calculations, missing a multiple of 60 for ips to ipm. Expecting response up to 4MHz seems hopelessly optimistic.

Revised figures:

2-stage 1:16, 20-tooth pinion. Feed = (3.14 in/rev)/(16)*(1000 RPM) = 196 ipm
3-stage 1:64, 35-tooth pinion. Feed = (3.14*35/20 in/rev)/(64)*(1000 RPM) = 86 ipm

I'm currently trying to find data on microstep linearity, but it was my impression that the max step:min step for a high-quality motor/driver combo would be significantly less than 2:1. Pending real data, I'm hoping final precision could be within 150% of theoretical values for a perfect 1:10 division.

[danilom]

Go to STEPPER FAQ - WHY TEN MICROSTEPS

http://www.geckodrive.com/faq.html

Also why you can't count on microstep positional accuracy is that all modern drives morph trough microsteps until 500RPM then they go to fullstep, because if they stayed in microstep they would lose lots of torque, read about this in the MICROSTEP VS. FULLSTEP FAQ

[srange]

Thanks also for that link. I should have been more clear in specifying my goals. I have little need for sub-0.010" precision at high speeds. I'm primarily worried about precision when performing finish profile cuts, and a secondary axis is moving slowly or holding while another axis feeds.

For instance, a common application for me would be cutting the upper surface of an airfoil like this, with the long direction of the airfoil going along the y-axis, and the vertical direction in z.



While cutting at, for instance, 50 ipm in Y, the flat top of the airfoil would be cut with very slow z-movements. If the cutter were to "jump" down 0.006" at a time, the surface would be very ragged, and in some applications, sanding out those steps is not accurate enough for desired performance.

At high speeds, absolutely, the driver would revert to full-step commands, and that would not concern me in the slightest.

[dbinokc]

Are you planning on building RC aircraft where transonic effects are an issue? Or maybe this is for a wind tunnel? I am not an aeronautical engineer, but just recently completed the Intro to Aerodynamics course on Edx .I am also building my MM to help with building my own homebuilt.

[srange]

As it happens, I am considering building a dynamic soaring ship, which reach top speeds between Mach .7 and .8. But there are numerous airfoil types sensitive enough to require this sort of insane precision. Laminar flow airfoils on gliders are widely believed to require contour accuracy / waviness below 0.005" to achieve design performance. To reach that in a final product, after mold-making, layup, trim, assembly, surface prep, and painting, I'd like to get below 0.001" for the initial plug on such projects.

[dbinokc]

I have seen videos of dynamic soaring with rc gliders on youtube. Those gliders do get to insane speeds.

[ger21]

I think you're going to want to be cutting at around 300+ipm. 3D contouring at high precision typically uses toolpaths with very small stepovers to get the smoothest finish possible. On large parts at 100-200ipm, this could turn into a multiple day cutting job.

When using steppers, you can't get high resolution and high rpm's together. Steppers lose torque as rpm's increase. So typically, to go fast, you spin the motor slower, where it has more torque. If you want more resolution, you spin it faster so you have more steps/inch.

- Am I looking at the right type of motors, given that the end use will likely be higher-RPM, lower-torque than typical for MechMates?

Keep in mind that generally, the highest torque requirements are to accelerate the gantry to a given speed. Even though your cutting loads may be light, you're torque requirements may not be much less than that of other Mechmate users.
If you're seriously considering a 64:1 reduction, then smaller motors capable of higher rpm's might be a better choice.

When you want to go fast and have high resolution, you get into servo territory.
Leadshine makes some affordable AC servos that look promising, and I plan on using them on a non mechmate router I'm slowly working on. They run about $450-$500 for a 400w servo and drive.

[srange]

How embarrassing. I must stop spouting off numbers without double checking my calculations. I believe I've been significantly overstating the size of a microstep.

These are the formulas I'm now working off of:

Microstep size = Pi * (pinion diameter) * (output/input belt reduction ratio) / (microsteps per revolution)
Feed= Pi * (pinion diameter) * (output/input belt reduction ratio) * (RPM)

I want to stick with steppers, as a servo-driven machine of this type is electrically a very different beast than a MechMate. With these calculations, I think I can find a sensible stepper configuration for this application.

Admittedly, 64:1 is very excessive. But 1000 RPM also seems a very conservative upper bound on speed. Mike on Richards on talkshopbot shared his experiences with the PK295-F4.5A, suggesting it can "easily run at 1,500 RPM at 35V." Based on its bipolar parallel inductance of 1.5 mH, that is operating at 89% of Mariss voltage. Assuming the 299 properties scale with that of the 296, and that max speed increases more or less directly with voltage, the PK299-F4.5A should be able to function reasonably at 2000 RPM. Oriental Motor's spec sheet show its torque tapering slowly through 2000 to at least 3000 RPM, powered at 48VDC, so this seems like a still conservative speed.

Consider this:
PK299-F4.5A, driven at 56VDC, turning at 2000 RPM
2-stage belt reduction 9:1, 35-tooth DP 20 pinion (very similar to 4:1 with 20-tooth, but smoother running)
Microstep size = (3.14*35/20 in/rev)/(2000 microstep/rev)*(1/9 belt ratio) = 0.003", maybe 0.005" after bad motor non-linearity
Feed = (3.14*35/20 in/rev)/(9)*(2000 RPM) = 1200 ipm feed "comfortably" (under light finishing load)

That is more than adequate for my needs, and with that belt reduction, hopefully my machine would still have plenty of torque for acceleration. If anything, I feel that speed is probably excessive, and I might be more inclined towards a 1:12 belt drive, or similar, but such things can be achieved just by switching out a belt and gear.

If I wake up tomorrow and still believe those calculations, I think I'm pretty well satisfied that this gearing arrangement will work.

Ger correctly points out that the profiling cuts I'll be doing are time consuming. A typical finishing pass for a large (2m-ish wingspan) RC plane might be 900 in^2 of profiling, with a 3/16" ball mill, with 0.003" stepover. At 200 ipm, that's 24 hours of continuous cutting, or 12 hours at 400 ipm. I'm no stranger to multi-day jobs (I used to do this sort of work, though in smaller pieces, on a 3-axis CNC Trak knee mill, 50 ipm! But the 0.002" z-axis backlash was far more problematic).

Does anyone have experience with unattended running of these machines? It would be nice to have a low-fidelity onboard encoder on all the motors to detect a stalled motor or similar, though I don't know if Mach3 has provisions to monitor such things. Tool break detection would be nice, but might be more of a stretch. Also very importantly, running at a few hundred ipm, will these setups reliably not miss steps for hours at a time?

[danilom]

Considering you can get a 2000 RPM out of steppers on a production machine is not going to happen, even Leadshine stepper / servo drives can't easily push them that hard maybe. You may spin them that fast on a workbench not loaded or in a belt driven laser. You are here in servo territory as Ger21 stated.

You loose torque with speed so look at a RPM / torque for each motor you specified. They state a flat line up to 4000 RPM on about 0.56Nm (80oz/in) but that has to be proved in practice. Look for Tormach using 3 phase leadshines for better responce and linearity.

Monitoring the VFD load you can detect tool break if the load is at idle for long time. It should not loose steps if you use it along its capabilities.

This is the last from me on this topic, considering the motors contribute to all the precision is wrong, you will have to calculate your spindle runout, rack non linearity, precision and rails missaligment, belts strech etc. into all this.

[KenC]

Mach 7~8 Now... this draws my eyeball very hard...
Task, :- 0.001" 0.0254mm which is 25micro meter precision on a plug...
It will be impossible to have MDF or plywood plug to achieve this sort of precision with temperature stability, moisture stability... etc. one careless extra sweep of sand paper (or just news paper) & the whole thing will deem ruin...
You must go for mold grade Alum or even steel.
Generally, lathe gives 0.1mm, milling machine gives 0.01mm
You need a heavy metal CNC milling station or just outsource to mold makers, they will know exactly what to do. surely cheaper than to buy a 6 axis CNC station.

[srange]

Ken,

Sorry for the confusion, but I think you missed a decimal: that is Mach 0.7 to 0.8. Standing record is 498 mph by Spencer Lisenby.

As I mentioned upthread, point-to-point accuracy, say wingtip-to-wingtop is not remotely required. 0.001" is the target profile precision. I look to avoid gouges and steps above the size on the tool surface.

To get a sense for the common practice in this activity, consider the build thread here. The base material is MDF soaked with epoxy resin, so the cutting surface is akin to a plastic. Sanding is still required, though I think it is overdone in common practice, and would readily run the machine at 1/4 the stepover seen here. Careless sanding can definitely ruin work -- one of the main reasons plugs are still used in low-rate RC airplane work is that maintaining a razor sharp mold edge in MDF is near impossible.



Danilo,

Very well. Thanks for your contributions thus far. You are quite correct that I'm not factoring all the components of absolute accuracy. But I'm not setting out to achieve 0.001" absolute accuracy. Z-axis flex, belt stretch, runout are all continuous functions of tool load, and the target scenario is cutting with an extremely uniform tool load, such as a 2mm offset cut from the final tool surface. Stepper motors are a particular concern due to their discrete motion paths, hence my focus on them thus far.

[KenC]

My bad, too much hop-reading. still .7 ~.8 mach is eye pulling.

They way you describe, MM will be fine, you still need to do plenty secondary processing on the plug mostly by hand. I reckon you need high precision measuring gears more than a precision CNC router.

measure, sand, measure sand, measure, sand.... you can even get to within 0.0001" precision if you are patient enuf.