[jsheerin]

Jumping on the bandwagon of 5 axis router builds, this is a build log for my second cnc wood router or possibly a build log of rebuilding my first router. I've been thinking about these issues for at least 6 months now and this first post summarizes a lot of this thinking. I imagine this process will be a long one, and I'll continue to update this thread as I make progress.

For some background, I designed and built a 4'x4'x6" router based on 80/20 extrusions and Ahren's router parts (cncrouterparts.com) about a year ago, finishing in early 2009. I learned a lot of stuff and used it to make quite a few parts but eventually I got tired of several issues with it:
- Adjustability of the z axis bearings - they're a pain to adjust, and they come loose more than I'd like.
- Dirt build up on the flat steel rails - I did not incorporate any dust shielding in my design. Since the rails are flat, the bearings tend to pack down dust that rests on the rails. On my z axis, this eventually led to the stepper motor occasionally stalling. The dust sticks to the steel rails even when vertical (and even using dust collection). I also put the x axis rails below the work surface which leads to them accumulating more dust than they otherwise probably would.
- The plastic anti-backlash nut on the z axis wearing out (although I've since learned I could probably grease this and it would work better).
- The router (Porter Cable 892) being a piece of junk. I took it in for repair, and they found nothing wrong. However it bogs down cutting with sharp bits in MDF at any kind of reasonable depth of cut and draws enough current to trip 15A circuit breakers. I eventually settled on running at 0.125" DOC at 120ipm. I swapped in a 15 year old Dewalt 1.5 HP with the same cutter and it performed vastly better. I think I got a dud router from Porter Cable, but there's not much I can do about it. So it's getting replaced.
- Limited speed due to motor and screw choice. I used 495oz*in motors from Keling with 1/2-10 5 start precision acme screws and a Gecko G540 with a 48V power supply. These motors really need to be geared down more than what is provided by these screws to give a good force vs speed curve, but to do that I would have to add timing belt reductions on the motors. This would be possible, but I figured I'd go for bigger and better at the same time. This makes everything run slow (which is not the worst thing in the world for me as I'm half hobby, half production usage but nothing on a full time basis), but a bigger issue is that I can't get good chip loads for my cutters. This leads to burning of the material and dulling of the cutter on some cuts.
- Lack of closed loop control. I've had several cases where parts were ruined due to the steppers missing steps and the controller (Mach3) not knowing about it. I believe I solved all these issues since then (bad parallel port in my computer and binding issues on the z axis as previously mentioned), but in the future I'd like to have a better chance to save the parts should something go wrong.

In the middle of 2009, I also bought a large number of VFD's (variable frequency drives for running 3 phase motors) and THK linear bearing carriages for almost no money at an industrial auction. I've also recently been learning how to weld.

From all of the above, I decided to upgrade / rebuild my router with the following changes:
- THK linear bearings for all axes
- Covers for all axes (as much as possible)
- X rails above the work surface if a gantry to reduce the moment on the bearings
- High speed 3 phase spindle with VFD
- Possibly dual Z axis slides, one to carry a low speed drilling / milling head and one to carry the high speed spindle
- Ball screws instead of acme screws
- Rack and pinion drive on the X and Y axes to eliminate screw whipping at high speed and provide a better gear ratio
- Servo motors with appropriate gear reduction to give high speed, high acceleration performance (target of 700ipm cutting speed capability and 1000+ ipm rapids with around 1g acceleration)

I also did some analysis of what slows my construction process down the most on the products I have been making. I determined it was sanding out uneven spots on my finished parts (I cut out slices from sheet goods like plywood or MDF and glue them together to form the finished product). There were many other smaller issues that this machine design would attempt to eliminate, but by far the biggest issue was the sanding. One way to reduce that would be to surface the exterior of the cabinet in one operation after it was assembled. This would require a 5 axis machine. Initially I decided not to build a 5 axis machine due to the cost of 5 axis cam software (~$13k for Mastercam for example). My extremely part time side business currently won't support a purchase like that. However, I eventually decided that even some manual programming to rotate the head and machine in different planes would make the 5 axis machine worthwhile, and eventually it would probably make enough money to pay for itself and a 5 axis cam program. So I decided to design and build a 5 axis machine knowing that I won't fully utilize it until some time in the future. Plus it's just cool.

In addition, I plan on starting off this project by quantifying cutting forces through some experimentation. This will give me a design target to shoot for instead of just picking a force that I think will be required to cut. However I did get a bit ahead of myself and already have these parts available for this machine:

- 3.7kW (5HP) water cooled high speed spindle from Chai (linearmotionbearings2008 on ebay)
- Mitsubishi 7.5kW VFD to run the spindle from previously mentioned auction using 240V single phase power (derating the VFD by half per Mitsubishi tech support for running off of single phase power)
- parts to make a water cooling setup for the spindle, mostly from www.crazypc.com
- 20mm x 5mm pitch ball screws with anti backlash nuts and 15mm end blocks from Chai
- X2 milling head spindle from Little Machine Shop: http://www.littlemachineshop.com/pro...ory=-269978449
with a set of R8 collets and a drill chuck
- THK HSR25 bearings and rails (bearings from previously mentioned auction, rails from ebay)
- Fanuc 5s AC servo motors from ebay (~0.9kW continuous output - 5.9N*m of torque, 53N*m of torque max, 2000rpm max speed)
- Fanuc 10s AC servo motors from ebay (~1.8kW continuous output - 12N*m of torque, 78N*m of torque max, 2000rpm max speed), some including brakes
- 15kVA transformer from ebay for the power supply

I've also been thinking about the C and A axes drive mechanisms. So far it is looking like a worm gear setup with manually adjustable mesh clearance might work. I'll rotate the head to distribute wear evenly around the gear for longer life. See this thread for more info:
http://www.cnczone.com/forums/showthread.php?p=737050
However I also have a Thomson Accutrue 100:1 planetary gear head and 2 harmonic drive 100:1 gear heads that will support the torques I am interested in (over 300N*m at the output). The harmonic drives should have very low backlash, but I will test them and see. The Accutrue should have more backlash, but it might have an acceptable amount. I'll also have to test that and see how it performs.

I am currently working on writing an Excel sheet with calculations for various parts of the router. I'm also thinking about different form factors than the typical moving gantry. If I want more z travel than the typical 6"-12" or so, I will probably move away from the moving gantry. One possibility would be a fixed gantry moving table design. However some quick calculations show that my part height will still be limited to 2' or less depending on the exact height of the ceiling in my garage due to the requirement to get a long tool above the entire part and the extra vertical space that takes. (I am moving in a few months, so this question along with other experimenting will have to wait until then). An option to get more z travel given a fixed ceiling height is a moving y axis design like the Tarus claymills. This would get me up to a 3'-4' part height depending on several factors. However some quick calculations around the THK HSR25 bearing show that some serious consideration would have to be given to minimizing forces on the bearings if I want to get any kind of reasonable life out of them. It looks like I would need bearings on both sides of the y and z axes assemblies unlike what the Tarus machines look like. A sketch I did today of an idea for this style of machine is attached below. It would use all rack and pinion drives. The spindle is a bit out of proportion to the rest of the machine.

I've also been thinking about construction methods for this machine. I plan on using steel for framing members and using some combination of pins, bolts and welding for joining parts together. My primary concern is how to get a flat, plane surface for mounting the linear rails. I don't want to haul a massive frame to a surface grinder. I also don't relish the thought of long hours with a machinist's level and a scraper or grinder and a surface plate. Because of all this, epoxy surface plates were very interesting to me, but after some calculations I decided they were not suited to the task of supporting linear rails due to issues of static deflection under the weight of the moving components, creep (deflection over time), and differing coefficients of thermal expansion between the epoxy and the steel it would be adhered to. However, it would be a good surface to mount linear rails to and mount a surface grinding wheel to. This wheel could then be moved back and forth over the rail mounting surface and the finished rail surface height set relative to an epoxy surface plate poured around this surface. The epoxy surface plate will connect the different sides of the machine together (in whatever configuration that might end up being) in order to allow two planar rail mounting surfaces to be created.

[spoiledbrat]

I love your concept sketch! Looks awesome.
78 Nm? Those sound like some serious servos!

Will the part table depicted rotate, so that you can get to the back of the part? I realize that there are parts that require the tool to be normal to the surface at all times. Dont you want to be able to machine 5 sides of the part, though?

I look forward to seeing some build pics. Good luck!

Rob

[Zygoat]

i read your book. ill be watching. i may have to pm you too when i am finalizing my calculations for my build.


good luck!

[jsheerin]

Thanks!

Yes, I'm thinking that the table will have to rotate. However it might not be a super precision thing - just turn 180 degrees (and maybe stop at 90 as well). I have a planetary gear head with a 100:1 ratio, so I could use that to accomplish this. Maybe put limit switches at either end and just use a simple motor and drive to move it. Then have some kind of mechanical brake or clamp to lock it in place while machining is going on. Or I could just do it manually - unclamp it, rotate the table and reclamp it. I could have tapered pins for alignment. One big issue with this will be how compact I can make this assembly. Any height of the table above the floor directly cuts into the height of the parts I can cut. Since this will be in a garage with an 8' or 9' ceiling, this will be a major concern. If the table did not have to rotate, it could sit directly on the floor and maybe just be a couple of sacrificial pieces of MDF that I would surface.

[Zygoat]

have you considered turning the table sideways. so that it will rotate more like a normal 4th axis or like a lathe? i imagine that your material is much taller than it is wide. this would allow you to handle material that is 8' WIDE and as "tall" as your garage is long. (probably longer than 8') just an idea since you sound like you are wanting to work some big material. and this way the long axis does not have to fight gravity so you should be able to move faster, which is a big plus on the longest axis.

[jsheerin]

No, most of my parts are wider than they are tall and also typically need to be machined on 5 sides. For example, see a few of the parts I've made on my current router and one I've been designing below. Most of these would benefit from some form of 4 axis machining with an indexed 5th or even 2 and 3 axis machining with other indexed axes. However they're all pretty heavy and would not be easy to support and turn on a horizontal 4th axis. The last one I would probably lay down in a fixture to machine the front, then flip over and machine the back. Although if I got 4' of z travel, I suppose I could do it standing up...

[Zygoat]

ok. just thought i would through it out there just in case you had not thought of it.

[jsheerin]

I appreciate it. I had not specifically thought out that configuration, but I think that's because most of the parts I make would be difficult to mount that way and might not support their own weight when mounted horizontally. I suppose if I got a part that would be a lot easier to make that way, I could always make another (6th?) axis and use that without one of the other axes running.

[jsheerin]

I've been doing a bit more thinking, and now what I'm leaning towards is a column mill as sketched above, but with an additional gantry for processing sheet goods. One or the other would park out of the way at the end of the machine when the other one was in use. I'd only have electronics for one set of motors at a time. A sketch is below, although not really to scale - it would be smaller relative to the column. This was motivated by my continued need to process sheet goods quickly, and I was worried that I'd be making some compromises in the column design if I wanted to be able to cut sheet goods in a reasonable manner.

I'm also thinking about using my minimill head only on the column. This would be smaller and lighter than a high speed spindle. I could initially run it from a 700W Yaskawa AC servo and AMC drive I already have. I could set this up to control speed using Mach3 I think, as the AMC drive is a +/-10V analog input, so I could just send it 0 to +10V for forward. I'm using this setup currently but with a 300W motor on a handle held spindle sander I made. A pic of that is below, and more are here:
http://ldsg.snippets.org/HORNS/images/100413_sander/

I'm thinking about mounting an R8 to ER40 adaptor in the minimill spindle so I can run 1" ball end mills. This would be for taking light 3D (or 5D I suppose) cuts to finish a part. Or I could go with an ER32 collet extension in a 1" (pricey) R8 collet with a 3/4" shank EM. Something along those lines to get some reach away from the spindle.

[jsheerin]

I've been thinking about writing my own 5 axis CAM software. It doesn't seem like it would be that hard for simple operations. Basically I'd have to read in an STL file, sort the facets by whatever direction I wanted to step over in (Z, for example), interpolate to find all the points at a given Z level, put them in order to form the cutting path, and then calculate the xyzca values that would position the tool correctly at each point. For a straight endmill, the math to do that is pretty simple. I wrote it out the other day. The programming should also be fairly straight forward (not that I'm an amazing programmer or anything so I'm not saying I'll have this finished tomorrow). The parts that would start to make things complicated would be graphics display and code for keeping the machine or cutter from running into the part, keeping a constant tool speed across the surface, etc. However I've flow charted out how I'd do all that as well and it doesn't seem impossible - just involved. This would all just be to cut with the tool normal to the part surface. I'm sure I could figure out side milling as well, but I haven't really thought about that one yet.

Anyway, this gives me hope that I could put together a program that would give me simultaneous 4 and 5 axis gcode for some of the types of shapes I want to cut using software solutions that I already have (cad to generate stl files).

[Miguel-]

Look at heekscad / heekscnc. It could be a good start -it is currently quite active open source project and there is a possibility to write plugins for g-code creation etc. It can open STEP, IGES and STL so it has good import possibilities. You can find it here:

HeeksCAD
HeeksCNC
And here's a 3d-cnc plugin written for heeksCNC: OpenCamLib

[jsheerin]

Thanks - looks interesting.

[jsheerin]

I've been playing around with my 5-axis cam idea since I am currently waiting on moving before doing any more design work on my router (new garage - woohoo!). So far I've been able to read in an stl file, calculate profiles of the part at various z heights, and calculate offset profiles from these profiles. See attached screenshots of a test stl part and plotted profiles (from excel) from various z heights.

I've got a lot of the code written to generate 5 axis motion from these profiles and then convert it into g code and output it, but I'm currently rethinking how I calculate the angular machine positions. So far what I've done would only work on convex shapes, but I think I have it written so support for ball endmills would be pretty easy to add.

[jsheerin]

My 5 axis program now spits out gcode that is reasonable. The primary issue at the moment is getting the start point in the right place and then having the program keep track of the angle the C axis is at so that it doesn't unwind part way through the cut on a round part. However it looks like the x and y gcode values bounce back and forth a bit, like the cutting head is moving forward and backward around the part. I'm wondering if my method of interpolating points and how I calculated the normals isn't working and that error is being magnified by a long tool length. More things to play with... However, it is nice that the general frame work is functional. I'm using a ~700kB stl file for input and getting gcode less than a minute later even with the brute force approach I'm using. I've also started looking at some display stuff.

[FandZ]

Don't mind me, just subscribing for future updates.

[jsheerin]

If you go to the top of the page and click on 'thread tools', you can subscribe to a thread without posting in it.

[FandZ]

If you go to the top of the page and click on 'thread tools', you can subscribe to a thread without posting in it.

Duh.... but I wanted you to know that I'm subscribing to your thread because I think your project is just that cool :P

[jsheerin]

I did this a while ago, but I figured I'd post some pictures of it. I was playing with one of the Fanuc AC servo motors I bought for this just to verify that the encoder was an incremental type and to verify that the motor would probably work. On the advice of Al the Man, I hooked up three 10k resistors, one from each motor phase to a common ground. Then I looked at the output of each motor phase relative to the common ground using my scope while driving the servo motor with another electric motor. I got a sine wave from each winding of the same magnitude, so this one should be okay. I also figured out the encoder pinout, that it is a differential incremental quadrature encoder, and that it has an index pulse. So I should be able to use this with a Granite Devices drive which was my plan. Another zone member has had success running the Fanuc AC servos with these drives as well, so I'm looking forward to getting that all set up. Anyway, here are some pics of testing. The last one is my setup for testing the encoder. An adjustable power supply set at 5V and supply about 0.2A is sittign on the stool. This is a 5S motor which is currently slated for driving a rotary axis. I also have 10S motors to drive some of the larger bits.

[Boltz]

Did anybody else smile when they saw the picture with the washing machine, the angle grinder, and the oscilloscope?

I'm pulling up a chair for this one...

-Jim Hart

[jsheerin]

Yeah, my basement is a bit cramped... ;> There is a metal lathe and a welder just to the right of the welding bench in all those pictures. I'm moving in a month and will be able to spread out a bit. I'm waiting to do any serious design on this machine until then - I need to measure my available space so my z axis doesn't end up poking through the floor of the living room (which is above the garage). That's part of why I'm considering the Tarus style machine - I'm vertically constrained.