Why would you build a CNC?

My last major project was building a boat. It was much the same as building a CNC router. You don't build a boat to save money. You can buy one much more cheaply than building one. The only rationale for building a boat is that you want to build a boat.

As I do my planning for my CNC router, I think about my rationale is for building a CNC router. For me, it is for fun and for a platform for my real goal of working on the coding end of it. Actually, for not that much more money and a lot less effort, I could have just bought one of the Chinese pre-built machines or one of the kits. Something like Zen Toolworks or Buildyourcnc.com would be a lot less hassle. They are not really charging that much for them when you consider the extra shipping for ordering from multiple vendors and the time spent on sourcing materials, let alone the design time. If I factored my design time at my normal pay rate, I would be much further ahead just buying a pre-built machine.

If I were using a CNC router commercially, or even depended on it slightly, I would seriously lean toward one of the commercial products. At the very least, I would use standard drivers, motion components, and motors. You really can't afford to have to tinker with stuff if you are doing it as a business. Readily available replacement parts beat cheap replacement parts any time. Just having a machine sit there and not be working is a losing proposition. It makes jobs late. It loses you customers.

My initial designs for my CNC router will probably be a failure. It likely will take several versions until I get something that I am happy with. I am building contrary to what have been proven successful designs. I am purposely building an low powered machine. For a few more dollars, I could have bought more powerful steppers. I want to see how well I can move things around without brute force. The other thing that I am focusing on is building a rigid machine that is light enough to be portable and inherently stable enough to move around and not have to re-level or recalibrate. If I can get these two things pinned down, I will be happy.

Another Gantry Design

Another quick sketch for a gantry design would be a large double beam with the router running between them.. If you went with a 3" square 1/8" thick 6063 AL tube, it would be less than  16 pounds for a 4' gantry (1.6 lbs a foot). For a little bit more you could go with 4" square tubing at 2.2 lbs a foot.  Go crazy, 6" square tube is only 3.5 lbs a foot. How on earth can you justify gantries that weigh over a hundred pounds? This is still using fairly conservative and easy to work materials. Go up to thin wall carbon steel with an engineered section and you can get a lot lighter than this.  It still leaves plenty of room for a robust z axis and nothing is cantilevered out adding all kinds of funky forces into the mix.

This is not getting into really sophisticated construction techniques. It is still stuff you can bolt together. It is just looking at the properties of the materials and not just throwing huge chunks of solid materials at the problem in hopes that it will make it rigid. Spread out the footprint of the bearings to add stability. Don't cantilever things out that create big torsional forces. Use large cross section thin tubing. The idea is to reduce weight and add strength.

  

Exploring One Type of Gantry

This is probably not how I will build mine at this time, but one idea I have for the gantry would be like this. Just a rough sketch, but this would probably be one of the best configurations for a gantry. The tool is balanced in the center of gravity instead of cantilevered out to one side. The rails are covered to protect them from chips and dust. The wiring could be contained inside the gantry to keep it safe. Wide stance to keep the lever action on the gantry to a minimum. I added an arch to the design to increase support as you get away from the supported ends.

From the side, you can see where the spindle would ride in this system. Yes, I realize that this looks pretty massive but the structure could be built out of pretty light sheet material. The idea is to increase the section to increase the strength instead of just making the pieces more massive. This expands the distance between the linear motion bearings to stabilize it. It makes it where you have room to build a lift for the z-axis much like the nice router lifts they use in router tables. Making it bigger doesn't increase the mass as much or interfere with the cutting area. Again, the whole idea is to increase stiffness and reduce mass. All the mechanicals could be contained within the gantry and not leave wires hanging out everywhere to catch on things.  It would also give the router a finished look instead of just bolted together scrap. If you wanted to build something stronger, you can go to sheet metal and welding. Something lighter could be made from composites.

So, what is it good for?

My last post might have come off as a bit negative but not intentionally so I wanted to discuss it in a bit more detail. There are many things CNC routers are good for. They are not just the magic wand to set up your own miniature factory. I mentioned in the last post that CNC is not good for production work and that templates are faster and more accurate. It is effectively used to make templates, molds for injection molding, patterns for casting and the like where the production speed is not that critical as you are only making a small number of items. The real use is for one-off items where it is not cost effective to tool up for a high volume production run or to be the tool to create the tools for a production run.

Complicated geometry is one of the best uses for CNC. Once you get past the whole rectilinear world of cabinet making, there are many complicated shapes that are hard to reproduce by manual methods. Smooth curves are not that easy to do by hand without lots of hard won skills. Getting three dimensional curves cut by hand that are smooth and match is not easy. It can be relatively easy with a CNC router or at least consistently.

Intricate cuts like scroll saw fretwork, inlay, and many other intricate things are easier to consistently do by CNC. You can do them by hand but unless you happen to enjoy that type of intricate work, it can be tiring and tedious to do by hand. Also, if you are doing structural work, the smoothness of curves makes a huge difference. When the curves are irregular, they don't distribute stress well. For production of circuit boards, it does not do nearly as well as the volume production methods. It does have the advantage of not needing the nasty chemicals.

CNC's Dirty Little Secret

CNC does have a dirty little secret. It is a lousy production tool.  Everything you make with a CNC machine requires that you write a program, plan a cutting path, test that the machine will really go where you intend it to go, decide cutting parameters based on the cutting tool and material being cut and set up the cutting stock. This is not a high speed process. It is wonderful for prototyping and one off type production, but not for mass production.

It is much faster to just use a pattern cutting bit in a router and do it by hand following a template. The setup is faster, it is just as precise and a whole lot cheaper. Have a set of complex cuts?  Make a set of templates.  If you want to ramp up production to mass produce some pieces, just grab a few routers and some pairs of willing hands and you can multiply productivity to pretty much unlimited production rates.

Many times, it is faster and easier to do things manually. If you want to cut out a simple shape, just measure and mark, clamp down a cutting guide, and have the piece done much more quickly than even booting the computer. You really don't need a program to cut a piece of plywood into a 1' square.

I feel better. There are many techniques available that will beat CNC routing in a production setting. Stamping, casting, injection molding, pattern routing... The list goes on.  It is from this standpoint that I can say that  I really don't care how fast my CNC machine will cut. It doesn't make any difference. Why would I care if I can get "1500 in/min Rapids" or some such nonsense. I am not routing a path from Seattle to Miami. For a one off piece, it doesn't really matter whether it takes 5 minutes or even 2 hours longer to cut something out. If I am using an expensive CNC machine to do what a $2 piece of plywood template could do faster and easier, throwing more money at the CNC machine to make it faster doesn't make it less silly.

Keeping the Pieces Together

What is all the fascination of bolting together CNC machines? It's as if the entire world were afraid of a little glue. Modern adhesives are stronger than the wood that they are joining. Even metal is glued together in modern airplane construction. Come on people, join the 20th Century.  Epoxy and Titebond III are great to glue together wood even if they get wet. Look up what it takes to be rated as Type 1 water resistant. We are talking boiling for 4 hours, baking for 20 hours at 140F, boiling again for 4 hours, and then testing while still wet. Regular PVA glue is great if you want to keep it inside and dry. These glues even hold pieces of MDF together pretty well, if you are still hung up on using that nasty stuff. If you have no metal fasteners, they can't eat your expensive cutters when you accidentally hit them.  Glue doesn't cause galvanic corrosion. Vibration won't loosen glue like it will bolts. Most importantly, glued structures will distribute  stresses instead of focusing them like bolts do. The proper term for the focused stress is a stress riser. Bad stuff. It makes things warp and fail.

Now, don't start thinking that I am totally against metal fasteners. They have their place. Go ahead and screw a motor down with them. Drill oversized holes and lock them together with bolts to make things adjustable. Bolt on a proximity switch. Have a screw clamp to hold the spindle down. Use a bolt for an axle for a bearing. Just don't forget that glues have their place too.

Which adhesives are best suited? Epoxy and  PVA will come into play the most. I used to like the moisture curing Polyurethane glues (like Gorilla Glue) but that was before I met epoxy. The main difference between these two glues in use is gap filling and open time. If I have a nice clean joint and it is not going to take a long time to put together, I go with PVA. If things don't fit perfectly or I need a lot of time to get things aligned, I go with epoxy. Also, if you are forced into gluing endgrain, epoxy can really be your friend. If you need a little extra support, grab some fiberglass, carbon fiber, or Kevlar. If you need to span bigger gaps, grab some fillers. It is great stuff. Of course someone will point their finger and say epoxy is expensive. Not by a long shot. You want expensive? Go compare epoxy with polyurethane glues on an ounce by ounce basis. Polyurethane glues are much more expensive. Also, they don't fill gaps nearly as well unless you count non-structural bubbles of foam. One thing to note: Epoxy will not be as strong if you clamp it hard and squeeze out too much of the glue. PVA works better with plenty of clamping force. Read the instructions on the bottles even if you are used to using them. I found out some mistakes I used to make with some glues. One example is polyurethane glue. I didn't know for a long time that you should pre-moisten the surfaces with water.

What about comparing the price of glue to the price of bolts? Well, you can look at some of the fastener kits that come with some CNC kits. One kit I know of comes in at $120. Having bought a lot of nuts and bolts in my life, I can say pretty confidently, that in my opinion, they are not marking them up enough to be worth counting them all out and packaging them. They are doing it as a convenience.  

Jump right in!

Hey, don't be bashful. I see from the stats on the blog that people are reading. Let me know what you think so far. I won't get all bent out of shape if you tell me that I am full of it.

I am going to start getting into the individual components next. This blog is really a way to keep me working on the plan. As long as I make a little bit of progress each day, I will finish it. When I built my last boat, that is the approach I took. There were not that many pieces in it. If I went out and cut out just one piece a day, I kept the momentum going and finished it. I know that this will be the same way. It will just take a little longer because I have the planning to do as well. On the boat, I was following someone else's plan, so that was done for me.

The Overall Plan

See, you can make a box that is light and strong.

It is not that hard to build a really strong bed for a router table. I have seen some torsion box builds but they tend to be way too heavy and not much better than just stacking lumber together into a pile. The mess of plywood and clamps you see here is my little OZ Racer I built from plans designed by Michael Storer. Yes, I was too lazy to go out and clear it off to take a picture of the finished boat. Use your imagination. It is painted bright yellow with red accents. It are a perfect example for what I am planning. The boat was made much lighter than what I will make the table. The boat is mostly 1/8" plywood. I will probably go ahead and go with 1/4" ply for the router bed. It is about the same price as the thinner and I don't have to travel as far to pick it up.

Rib Framework

This is the basic layout I am planning for the ribs for the torsion box for the router table bed. It will be 1/4" ply (Well, actually 5.2mm Luan underlayment). It will be skinned on top and bottom with a full sheet. To join, I will glue 3/4 in. gluing blocks to distribute the glue join. The intersections of the ribs will be glued with thickened epoxy. Specifically, it will be System 3 Gel Magic since I have about a half gallon left over from my last project. I will add an extra sheet of 1/2 ply on the top to give a more solid work surface and skirt it with 1/2 in ply to give me a solid surface to attach external clamping and the guide rails.  The holes and slots to connect the egg crate framework look more complicated than they are. With the pieces all stacked, I can cut the holes with a hole saw through all the pieces at one time. Same for the slots. I might forget the outside pieces in the frame and only use the 1/2 skirt. Depends on how I feel when I am cutting them out. It doesn't really add any strength but it will add material to attach the rails.

Table down and folded.

 
This is where this little exercise is leading. I plan on my table to be able to be tilted up and rolled out of the way against the wall, so I don't have to give all my space up to have a full sized table. The bed will be self supporting for the length and the rails will be attached to the box. Now, the first thing you might think will be: Won't the bed sag? Some but not very much if it is built well. There is a lot of strength and I am not putting a lot of force on it. There is one other thing that is going to make it work. I don't really care if it sags some. The gantry is running on the rails attached to the sides of the torsion box. If the bed sags any, the distance from the gantry to the table top is unchanged. The rails move with the bed. It is the relative distance that matters. The box is much more rigid than anything I would ever put on top of the table and the sheet goods to be cut will conform to the table top. I have built a much thinner torsion box for my sewing machine. The machine weighs about 40lbs and is set into a hole on the table which is 1-1/2 thick with 1/8 ply on the bottom and 1/4 ply on the top. After a couple years, it has not sagged appreciably. I am not worried about how this much stronger box will hold the gantry which will be directly over the trusses and a sheet of plywood that will be distributed over the entire surface.

The Gantry

The real enemy for the CNC router gantry is mass. It makes things sag. It makes people go with bigger and bigger motors which creates more mass. It puts more strain on the system that makes things move and reduces accuracy.

http://gaboats.com/boats/snowshoe16.html
A 15'6" boat that weighs 32 lbs.

Careful engineering can build in strength without adding mass. For this first build, I am not going this far, but here is a perfect example. Platt Monfort was designing ultra light boats that are tremendously strong for their weight. The boat pictured on the left weighs 32 lbs and has a capacity of 500 lbs. There are examples where you can even get lighter.  Except for a little Kevlar roving and the Dacron fabric, this is made with stuff that you could pick up at Home Depot. The Kevlar roving can be had for $25 for a 300 ft. spool. We are not looking at a major investment. Michael Storer had a little Rushton designed Wee Lassie canoe that he got down to the ridiculous neighborhood of 12 lbs. Even a big 3 hp router weighs less than 20 lbs. Why on earth would you need to build a support that weighs hundreds of pounds? These monsters routers are made to use free hand unless you are using some huge shaper or panel raising bits. Why do you need inches thick aluminum extrusions or big steel beams to hold it up? Many of the big gantries I see on 4' x 8' routers could lift an engine out of a car. Come on. Be serious. We are holding up a 10-20 lb router or smaller. I am planning on using a RotoZip on mine. It weighs in at under 4 lbs. It is pretty simple to get a gantry that can support it without sagging over a 4 ft. span in under 30 lbs without breaking a sweat. With a lot of thinking, you could get much lower.

Let the Mudslinging and Name Calling Commence

Up to now, I have been pretty mild. Now, I am going to take off the gloves and say what I think.

Precision Police bullhorn:
"Put down the Dial Test Indicator and
Step Away from the machine!"

Precision measurement and positioning. Wow. These people are living in a fantasy world. I was reading someone describing their 5' x 10' CNC machine. Their claim was "movement to 0.0005" and better" over the entire span of the machine. OK, let's give them the benefit of the doubt for a second. The first thing that comes to my mind is:

 Why??

Where would you even get cutters that approach that precision to match it? Where would you get a spindle that can come close either?  Are you going to measure runout and pull out the DTI every bit change? Just the heat from the cut would change the diameter of the bit from the beginning of the cut to the end. The heat generated by the cut will propagate through the material  and change the dimensions more than that.

At their estimation, the gantry weighs ~300 lbs. Now at the 150 inch/min that they are claiming for the positioning speed, how much stretch does that put on the lead screw? How much will the spiders in the couplings compress? What about the fasteners? Shoving that much mass around with 1100 oz/in motors and claiming 0.0005 in accuracy is just lunacy.

Now, I am not picking on them specifically but this is just an example of the garbage that is out there.

Material Selection

Really, there is not much of a choice here. Let's look at the main options.

A lot of home made CNC machines are made of MDF. There are things to recommend it:

• It is fairly stable. 
• It is cheap.
• It is pretty uniform.
• The big box stores carry it.

Now, on the downside:

• It's heavy.
• It doesn't hold fasteners well.
• It isn't very strong.
• It will sag over time.
• It starts to deteriorate with humidity.

Well, there are lots of other materials but let's not drag this on too long. Hardwood plywood is a much better material. It is lighter than MDF. Stronger. Relatively cheap. Easy to get. Glues well. Holds fasteners well. Maintains its strength and shape well. Less susceptible to moisture. Finishes well. All in, it is a much better material to use for building anything. Softwood ply would be OK but the grades that you can get easily are crap.

MDF has it's uses. It is wonderful for a sacrificial cutting surface. It works well where there is a dry environment and there is no real structural load. It also makes really nice speaker cabinets. For CNC machines, forget about it.

The Business End

The main cutting tools I will be using will be spiral bits. The most useful will be a square ended, plunge cutting bit like this bit from Amana or one of the ones with either a rounded end (ball mill) or a pointy end (v-groove) depending on what I am doing. There are also bits available for cutting metals and other materials as well as wood cutting bits. I will be using a spindle that will take at most 1/4. in bits but most likely will stick to 1/8 in. cutters. This goes back to the diminishing returns idea. I could put a big 3 HP router on the machine. This would mean a heavier spindle support and more powerful positioning to overcome the inertia. Instead, I am going the other way. On my small machine, I will use my Proxxon rotary tool. It is light, smooth running, and I already have it.  The larger machine will use a RotoZip. It is light, easy to mount, cheap, easily replaced, and again, I already have one. Notice a trend here?

Another reason I am planning on the small cutters is waste. Milling is a horribly wasteful process. Whatever I am cutting, I don't want to spend lots of money for materials to have a lot of it thrown away as cutting waste. I would even go to smaller cutters except that the smaller ones don't cut fast enough, cost more, and don't cut deeply enough. If money were no object, I would probably switch over to some other cutting method such as laser or water jet.  

Mechanical Complexity and Precision

Precise mechanical positioning from 150 B.C.

Precise mechanical movements have been around a while. The Antikythera mechanism is thought to be from around 150-100 B.C. It can safely be assumed that they had been working at this for a while before they achieved this level of complexity.

The precursors to modern CNC devices were the ornamental lathes and rose engines. The bottom picture here is a rose engine. One other thing to note: this is a simple one. They got much more complicated. The power to drive this was the foot of the operator pumping a treadle.  I have been reading the Holtzapffel books on ornamental turning Several generations of them were the big names in ornamental turning from the late 1700s on.   When I ordered reprints of them from Dover Books, I didn't realize how comprehensive they were. Cool stuff. The sad thing is that there have really not been any substantive advances in machining technology since that time until computers were brought in to control stuff. 

Style and appropriate use of materials from 1780.

Compared to something like these, the mechanics of  CNC machines are pretty infantile. Maybe few gears, a timing belt or chain. Pretty basic stuff. When you look at the glossy brochures from the CNC companies, you would think that they are the pinnacle of mechanical invention. 

I Feel the Earth Move Under My Feet

The material selection and structure are a big part of the design process. If you don't get them right, bad things can happen. There is not any difference between a gantry on a CNC router and a bridge. You have a span that you want to cover with a heavy cutter that you are moving back and forth and up and down. Just to complicate things, you are also generating forces with the cutter that are constantly changing direction. Throw in the fact that the cutter is at the end of a shaft that multiplies the forces just to make sure that you are not too complacent.

One thing you can do to make things more rigid is to use more stuff. More material can get you there up to a point but it is a matter of diminishing returns. Things get heavier and you need stronger motors to move them. The stronger motors are heavier. The heavier it gets, the more strain and flexing you get along with more friction and inertia to overcome which makes you need even more massive materials and stronger motors. The proverbial Catch-22.

The real answer is to make things stronger and lighter rather than just making things bigger and heavier. My machine is going to be made of wood. It is strong, easy to cut, virtually immune to fatigue, bonds well with adhesives, cheap, readily available. Dick Newick is quoted as calling it Miracle Fiber W. Wood is stronger by weight than steel and aluminum or fiberglass. Converting wood into plywood fixes the real drawback of wood: Wood is stronger in one direction than the other. In plywood, the grain direction alternates to even it out. One other problem with wood is that it expands and contracts radially with humidity. Sealing out water (or at least really slowing it down) is all it takes to address that problem. Most modern machinery is made of metals but don't let the hype fool you: metals change dimensionally too. Metals expand and contract with heat. They deform under stress. If you go back just a few years, wood was the high tech material. It just doesn't have as good publicist as some of the more modern materials because there is a higher mark-up on other stuff and marketing has spent lots of money trying to convince people to spend their money on more "modern" materials.

Gathering components and making decisions

Before I started documenting this build, I have been making some choices and getting up to speed on the available materials and components. There is a lot of good and helpful information available on the Interwebs and I am not going to go over all of them. Too much and too easy to find. A few quick searches will get you a lot of links to investigate.

I am really working on two machines simultaneously. One will be a small scale machine with a bed size of maybe 1' x 2'. I am leaving some flexibility as I have no real firm size in mind or real use for it other than as a test platform. Just small and easy to deal with. I plan on maybe using it for PCB cutting and small scale stuff. I might also put a plastic filament extruder on it for fun. The second will be a 4' x 8' machine bed that will accept full size sheet goods.

The control of each will be drastically different. The small one is going to be controlled by an Arduino Mega with EasyDriver 4.2 stepper controllers and some small steppers I picked up at Sparkfun ( http://www.sparkfun.com ) Why? Because I already had the parts. If I didn't, I would have grabbed some of the pre-made Sanguino stuff that RepRap and Makerbot use. It would be less hassle to use stuff that was already integrated.

The bigger one will be computer driven from a parallel port using parts I have ordered on Ebay from Hong Kong. It should be a month or so until they actually get here. Since it was only $260 including shipping for a 4-axis set including 4 steppers, controller board and power supply, I am happy with it in theory. What's not to like? The steppers are 1 NM (about 140 oz/inch) stepper motors which should be more than powerful enough for what I plan to do with them. Why so small? That goes to the whole idea of my building ideas. If things are designed well and you are making reasonable cuts, it shouldn't take a lot of power. This is going to be a hobby machine and not some production monster that will be going nonstop for 2 shifts a day.

The linear bearings will be cheap. I really can't justify going beyond stuff based on standard 608 skate bearings. They are high quality bearings that are dirt cheap and readily available. I really can't see spending lots of money for the fancy recirculating bearings and such. Life is full of compromises. In this case, the price is going to be the #1 priority in the decision making process.

The last design choice that has already been made is for the actuator mechanism. I thought about this for a long time. I considered plain threaded rod lead screws, acme threaded lead screws, ball threaded lead screws, rack and pinions, timing belts, and timing chains. The choice really was super easy in this case. I went to SurplusCenter.com and found the #25 (1/4 in pitch) chain for 60 cents a foot in a 100' spool. The sprockets came in at $1.75 a piece. Pretty hard to beat that short of getting it for free. Is it ideal? Well there are a lot of factors to consider but the price was so good on this that it outweighs everything else. It should also have very little backlash issues to deal with.

Software. Well, I really don't know yet. The only CNC experience I have had so far is to play with a little nearly antique Emco Unimat-PC lathe and a Roland CAMM3 that I brought back from the dead. I have never had a real CNC machine and I will probably start with EMC2. I am also following the development over on Phlatboyz.com for a plugin for SketchUp! that generates G-Code. Looks like it is coming along really well and will be working long before I have a machine to use it on. We shall see. If I don't like it, I am also looking at the USB controller and software combination that they have over on http://www.planet-cnc.com/. It looks like it could be a really nice solution as well. I will try the free stuff first.

Most of the parts will be off the shelf. I know I will probably have to fabricate some things. For that I have a little Harbor Freight 7x10 lathe with a milling attachment. Nothing fancy but it should be good enough for what I need. Beyond that I will just be using basic hand tools and a few portable power tools. If I get in over my head, I do have access to a fully outfitted machine shop and a pretty cool array of rapid prototyping equipment at work so I am not worried about painting myself into a corner. I also have access to a whole building full of engineers to pester if I have anything I can't figure out for myself. Standing on the shoulders of giants and all that....

Bringing CNC Router Tables Out of the Dark Ages

The modern design of CNC routers is like looking back in time to the Stone Age. Take a look at the current designs. Out of all the possibilities for structural design, they have generally selected the absolute worst possible choices of putting together a structure short of just stacking the pieces together and wrapping them with bailing wire and duck tape.

What I am in the process of doing is to design a CNC router for myself that brings in at least a couple of the improvements in engineering that have come about in the last few thousand years. I know that it will not be perfect but, I do know that it will be better than what is currently available. It would be hard to make it worse.

As I develop the design, I will take it in the same order as the building. At present I am going to build two units. One will be a small scale model before I progress to the full size machine. I am starting this a bit late as some of the design is already set which I will cover shortly. The main constraints I am under are budgetary and availability of materials. I am building this for fun and I don't plan on just throwing huge amounts of money at it. I am going to use features that fulfill their purpose and not just do something because all the other designs do it.

At present, I am going to focus on building a machine that will use a router spindle but I am leaving open in the possibility of 3D printing with ABS or other plastics. That will probably be the better use for the first prototype. I really don't need two machines.