Now this is an article I am looking forward to writing! CNC routers fall into that sweet spot of “mechanism+tool+computer” that I find so compelling and that lets builders of all kinds do work that could never be done by hand alone. I have spent a lot of time helping other people think about how to get and use CNC routers and there are no simple answers - but here goes!
To get this out up-front: I have had a problem not buying CNC machinery. To date I have owned about 20 CNC machines of various types. Many have been routers. Most have been either used, small or old. A few are big industrial machines. I have made lots of mistakes and learned a lot along the way - and continue to learn with my own gear and through helping other people with CNC buying, using and fixing problems.
My hope is to help you avoid such a long and bumpy learning curve when it comes to CNC routers!
A CNC router is just a tool - like a tablesaw or a drill. Because it is computer controlled, it lets you do things quickly and easily and with a degree of precision that isn’t realistic any other way. It can open new doors for your business and free you from subcontractors and their schedules. It will become a go-to tool for more and more jobs as you learn how to use it better.
Almost as important - designing things in the computer forces you to think differently about a process. It makes you better. I discovered this as a carpenter (boatbuilder) when I got my first router and felt my focus shift from trying to make a shape, to deciding what shape to make. I have seen it evolve as I started to think about machining as planning and programming instead of carving - and again as I have learned about 3D printing, where execution is even more automatic. Each shift adds another layer of abstraction above manual work, and forces complete decisions earlier in the process. This is annoying at first but once it becomes just “the way” it makes everything easier. And it’s a good learning experience - you can’t micromanage a robot!
WHY A ROUTER?
Assuming you have a type of work you want to do that requires the machining of a relatively soft (foam to aluminum) material in sheets or 3D surfaces, a router is probably the right tool.
If you are cutting only thin sheets of low density material, a laser is worth a look.
If you are machining smaller 3D parts from foam and metal, a machining center may be best.
If you are cutting 2D profiles in metals, a plasma or water-jet may be better.
A router is ideal for 2D machining sheet goods - think production cabinet shop. This is what most of the CNC routers in the world do day-to-day. Most manufacturers sell into this market and have machines optimized for this purpose with high speed, vacuum hold-down, drill banks and sheet loading systems. If your work is limited to 2D machining of sheet goods, you have fewer choices to make and plenty of options.
The composites industry does a lot of 3D shape-making. Most products are designed in software to a high level of detail, and most customers will assume that all but the simplest shapes will be built from machined geometry. Most of the routers I have seen in use in composites shops have been doing a mix of jobs. There’s some sheet cutting, some 3D surfacing and maybe part trimming. In larger operations, there will be more dedicated machinery and more job-specific optimization. This makes sense - you don’t want a million dollar room-sized 5-axis cutting plywood sheets!
DO YOU REALLY NEED ONE?
There are lots of CNC routers out there. Why do you want one? If you are a manufacturer trying to solve a specific set of production problems you will have an easier set of choices. If you are a job shop, you will have to think very hard about what “sweet spot” makes sense as you look at the costs involved and the corresponding opportunities for profit.
It is a good idea to write down a bunch of jobs that would be typical of 80% of the work you would be doing. It is probably smart to optimize for this and subcontract the remaining 20% out - or figure a way to do it sub-optimally in-house. If you are starting out and don’t have workflow or products yet, be very careful about committing to the wrong thing too soon. Try jobbing out the work initially - it will teach you a lot just by listening to the machinists talk about the work. If you talk to shops that have equipment similar to what you are considering for your own facility, you can learn a ton! A big spreadsheet with as much real information as possible is always a good idea, but at some point you’ll have to feel out the risk and make a “gut+data” decision.
Look at what it costs to have somebody else do the machining. Remember you are probably paying for material and labor, and that there is a lot of labor that is not specific to the machining process. Once you have a machine, you will have new commitments in terms of CAD and CAM work and material purchasing and handling. If you are having a large established shop do your work, they may have very nice equipment and it may actually be cheaper to have them produce for you if they have the capacity and can meet your schedules. Ask them what would make it easier for them - and then act on it!
It is my opinion that the best reason to buy CNC equipment is to control your future. If you can do something yourself, you control the schedule and the information. Customers will come into your factory and see a level of commitment to technology and capital investment and they will know you are serious. You won’t be waiting on (and blaming) subcontractors for your delays. Modern machine shops are just very well organized rooms full of skilled workers and expensive computer-controlled machinery. If you want to be one of those, you will need to own a bunch of stuff and manage it masterfully. If you can get by with just hiring one that’s a great option - and why job shops exist.
You won’t have a successful business by being a slightly cheaper “me too” business, but you can absolutely become a better provider of value for your customers if you can control your workflow more completely and provide confidence-inspiring certainty.
WHAT TO GET
You have lots of choices!
Big or small? New or used? How many axis? Tool changer? Vacuum hold-down? Control? Lease or cash or bank loan? Warranty? One nice machine or two crappy ones?
This picture shows the basic categories of machine used in the composites industry. I have owned all but the monster 5-axis, but I have always wanted one! They’re expensive.
Your choice will probably come down to money. If you are doing this as a hobby then it is easier - you know how much you can spend and you want the best machine for that money. If you are a business, there is the chicken-and-egg money problem:
“If we have [machine X] we could get [job Y] and then we’d make [Z profit] !”
“If we have payments of [$A] then we’ll have to take [job Y] just to keep [machine X].”
If you buy more machine, you will be able to do more with it - more of your own production or in the case of a job shop, a greater quantity and variety of work. Either way, unless you have a very good idea of what you need and the ROI, it is probably better to buy a simpler machine than you would ideally have, and then once you have learned a lot, buy either a second machine to share the load, or upgrade to a more focused and (hopefully) profitable option. Having expensive machine payments is stressful, and feeling like you have to stray from your core business just to feed ”the beast” is not a good long term strategy. You can learn very quickly and having experience with a simple machine will inform your future choices - think about it as an investment in knowledge and a process rather than the purchase of a tool.
CNC controls are the computers that tells each axis of motion and each cutting tool to move with coordinated motion following the “orders” written by the CAM software. The task is divided into these main activities:
decoding the g-code or part program and turning it into motion commands
timing the motion commands so that they result in the right combination of synchronized acceleration, velocity, and position changes for axes and cutting tools.
sending the motion commands to the axis motors and spindle
receiving feedback from the axis motors and correcting errors
continually checking machine status and safety (e-stop, pressure mats, etc.)
There are two main categories of controls: “industrial” or “PC-based.” Industrial controls are purpose built to drive automated machinery. They are one-trick-ponies and usually have limited interfaces and options, but are ideally suited to running machinery. They can be very expensive, often tens of thousands of dollars for a control system. PC controls are just a standard computer with software that issues commands to machine control hardware.
A PC-based control has a standard PC as the user interface - usually Windows or Linux. It uses the PC’s peripheral interfaces (USB, Ethernet, PCI, etc.) to send commands either directly to a breakout board or more commonly to another piece of hardware that handles timing and interface with stepper or servo drives. The advantages usually start with price, as PC controls are less expensive in most cases. It is also usually easier to interface with a PC control if you are not used to operating CNC machinery. There’s lots of storage and a mouse and files in folders - just like home!
Industrial controls are a different animal. Inside there is still a computer - but also a lot of specialized hardware to allow connection to servos and lots of input and output connections for machine sensors and features. Usually these controls have CNC-specific buttons and switches - feed and spindle overrides and buttons for specific functions. Sometimes it’s hard to get your head around at first! Industrial controls are generally found on more expensive machines, and are typically designed to operate for decades without drama - and sometimes they actually do! They are very robust and operate with a much simpler operating system so there’s no typical PC BS - no Windows update deciding to restart at a bad time!
It is hard to say what you should choose - typically a machine will come with a control that is aligned with the price point and level of performance. I would choose an industrial control for a larger machine because it is safer and generally more able to provide a complete operating experience. Things like feed and spindle override knobs (instead of sliders on a screen) and real buttons for cycle start and feed-hold are easier to use and give better intuitive connection with the machine. File handling on older industrial controls can be a hassle - drip feeding with DNC when there's not enough onboard storage for a large program - but once you’ve sorted it out it isn’t a big deal.
It is good to try to find a control that uses a relatively standard kind of g-code. This will be nice as you or your operators and programmers get used to looking at code and modifying it - and will make it easier to deal with post-processors. You might not hand write any g-code (why would you?) but you’ll become good at looking at it and knowing what going on roughly and sorting out any problems. Standard g-code is easy to read and the internet is full of discussions about all manner of problems. I guess this is an example of the “don’t get/do anything too weird unless you like adventure” model!
The spindle is the “teeth” of the machine. Your spindle could be anything from a 1hp trim router to a high precision tool changing water cooled 25hp monster that costs as much as a nice car. Something in the middle is probably going to be good.
Unless you have a very small machine, a standard wood router is going to be disappointing. Yes, the price can’t be beat - but they are noisy and weak and probably not a good choice for commercial work.
Electrospindles (I’ll just call them spindles…) are precision high-speed electric motors that combined with heavy duty bearings and some tool clamping mechanism, allow for lots of spinning under a variety of load conditions. The big price jump happens when you go from a wrench (manual, with wrenches) tool change to auto tool change (ATC). With an automatic tool changer, your machine can run a program unattended that uses a bunch of different tools. This is awesome! The added benefit is that the length of each tool is consistent from tool change to tool change. With a manual tool change spindle, you have to re-measure tools every time you change them unless you have a clever system - and clever systems are easy to mess up!
Wrench or manual tool change spindles cost from $1000-$5000 roughly. They are usually very robust and can run forever with very little maintenance. Typically there is a threaded collet and nut assembly that allows you to clamp a tool in the spindle. Most wood router spindles use ER collets. To change the tool, you use two wrenches and open the collet nut, remove the collet if you need to change to a different size, snap the new collet in and loosely screw it onto the spindle. Then you put the new tool gently into the collet and tighten manually until it won’t fall out - then you use the two wrenches again to snug it up tight. You’ll have to re-measure the new tool and run the program. It works great, but after the 20th time in 3 hours you’re going to wish you had an auto tool changer!
Auto tool change (ATC) spindles use a “taper” and a draw-bar mechanism which grabs a feature in a toolholder and pulls it tightly into line with the spindles shaft. The two main types you’ll find in wood routers are ISO-30 and HSK-63 (usually with an F). HSK spindle tapers are a little higher-end and offer some benefits. I can’t quantify the benefits, but have had both kinds and HSK seems like a more “positive” grip - but not so as you’d notice in day to day use. Some really beefy machines (Haas GR series routers for example) use CAT-40 spindles which are bigger and more metal-oriented.
ATC spindles require tool-holders for each tool. These are precision adapters that clamp the tool on one end and interface with the spindle nose on the other. They cost from $50 (cheap or used) to $400+ each. Toolholders will have a system for clamping the tool, often an ER collet or a heat-shrink or hydraulic clamping system. ER collets are fine for most anything wood or foam related and there is a whole world of high end tool holding stuff if you need to go there.
A machine with an ATC will have some type of tool rack or carousel for storing all the tools in their tool-holders. It is programmed to find a certain tool number in a given spot. When your code asks for a new tool, it will go drop off the current one and grab the new one. Your tools will be different lengths from the nose of the spindle, so you will have to tell your control (often this is partially automated) how long each tool is relative to a reference. This makes it possible to use many tools in any given program, and also to change out tools or adjust the amount a tool sticks out of its holder and not re-post the code from the CAM software. This is especially useful with a 5-axis machine and is one reason to avoid machines with a “double-ended spindle” or without a tool-changer if you plan on using multiple tools per part. A double ended spindle has only two tools at a time, and they are changed with a wrench - so the tools need to be remeasured and the program re-posted each time the physical distance from the tool tip to the point where the rotary axes intersect of the rotary axis - called the “pivot length.” Double-ended spindles really shine for part trimming in a production environment where two tools are plenty and the time to switch tools is just seconds.
Spindles need to be cooled because all that spinning makes a lot of heat! The most common method is a fan - either shaft mounded or electric - that blows air through the spindle housing. Other options are compressed air and water. Water cooled spindles tend to be either very cheap or very expensive. I would suggest you avoid the cheap ones. A fan cooled spindle is fine for almost all situations. Very high-end machines will use liquid cooled spindles to control thermal expansion and because they are more compact than externally mounted fans.
You’re going to have to hold down the things you will be cutting. This is called “fixturing” and is a major component of the art of machining.
There are a few different methods commonly used but the basic goal is always the same - don’t let the workpiece move, and don’t spend any more effort to do this than necessary!
The simplest option is often to use screws or clamps to hold material down. For extended 3D machining of blocks of material, strap clamps like you’d use on a Bridgeport mill are awesome - just don’t hit them! Many machines come with an aluminum t-slot table (Techno, AXYZ, etc) which can do vacuum or t-slot hold down. Other machines come with threaded inserts fit to the table in a grid. My first few machines had two sheets of MDF fastened to the frame of the machine, and I just held stuff down with square-drive screws - only hit a few dozen! The key here was to fasten the first sheet with countersunk bolts and then glue the second sheet on top so it was fastener-free and could be machined down to a flat surface.
For sheet-goods, the industry standard method of fixturing is vacuum hold-down. This is when a large (or huge!) vacuum pump sucks the workpiece down to the table and holds it there with vacuum. Vacuum fixturing is often a factory option on mid to high end machines. If you are cutting sheets of material, chances are vacuum hold-down with be the best option. Pod and rail machines use vacuum to hold smaller pieces so the edges are available for machining and hole drilling. This is a great setup for small parts and can be added in modular fashion to machines that have just a flat table. Vacuum is expensive though - running a huge vacuum pump - and if things go wrong it is easy to have a part shift on the table - resulting in junk!
My rule of thumb is that if a cutting job is going to take longer than an hour, you should fixture it with clamps or fasteners - but that’s just me.
There is an infinite variety of ingenious work-holding options but most setups will be pretty standard. Find a process (or a few) that is fast and accurate and work on streamlining it. Mastery of the process you choose is often more important than the choosing!
NEW vs. USED
The only things you get when you buy a new machine are the knowledge that nobody before you screwed it up and that you have somebody to call when it stops working right. These are valuable considerations! Even if you are an electrical wizard and gifted computer whisperer, these machines can be complicated and it is often not obvious what is messed up. Parts are hard to come by and down-time is expensive.
But used machines can be so cheap! This is what has always got me into used machines - the promise of a lot of bang for the buck. For the most part it has been a good strategy. I have also had broken machines sit for years because I was too busy to fix them - and this isn’t so good. It comes down to knowing yourself and your tolerance for risk - and your ability to sort out what is broken and fix it.
Generally the calculus comes down to the value the machine will provide vs. it’s cost, and the relative cost of down-time. If you are supporting a factory assembly process where a down machine stops production then by all means get a new one. If you are finally starting your own business but spent years running and fixing the boss’s CNC machinery and can take on maintenance and repairs - your cash will go way further on a used machine. The internet is a vast source of information and collective experience - so many times somebody else has been where you are and (bless ‘em!) wrote about it online. It won’t always be a walk in the park but given how much money you will save compared to a new machine, there is a compelling argument for taking the chance!
Keep in mind that machines have gotten consistently better over time. Controls are faster and more reliable and servo systems are more powerful and accurate. A new mid-range machine will have the same specs as a decade-old high-end machine, but it probably won’t be as heavy. You may find that all told you can get just as much value from a new cheaper machine and avoid the risks of buying used.
I am more of a proponent of the used machine route than perhaps I should be - but there are some good reasons. In situations where cash is tight and the machine isn’t the core of a business - or where a used machine is the only affordable option I think it is better to just get something and use it. Say you want a 5-axis machine to trim your parts, but the cheapest new one is $150k - that’s a lot! You can find a 15-20 year old 5-axis for 1/4 that much. The metal parts are pretty much the same and it works. One of my most productive machines was a 1998 5-Axis Motionmaster that was 15 years old when I bought it. It was old but did tons of work I could never have done without 3’ of z-travel, and rarely needed repairs. The only down side is that the computer bits are so old that your CRT (what’s that?) could go any day and those control cards are getting rare. You may get a few more years out of it, but someday you’ll have to retrofit the control. It may just be better to get the machine, use it, make money with it and learn than to wait until you can afford the fancy new one.
Old machines are sometimes a little worn out. When you are looking over a machine, be sure to check it for sloppy-ness in the movement and see what is causing it. Most of the time there are ways to adjust the slop out of an axis once you measure it. With the machine on and the servos or stepper enabled, grab onto the various moving parts (but not the spindle!) and push on it. It may move a little. It is a great idea to get a dial indicator and set it up so you can quantify the movement you feel in each axis. Most often this movement is fixable. What you want to look out for are situations where you push things back and forth and feel a clunk-clunk at each end and some measurable play. Here are some things that might be wrong:
Gear pinion may not be seated all the way into gear rack. Most machines allow this to be adjusted so see if it is just loose. Worst case, gears can be replaced.
The bearings at each end of a ball-screw often apply some tension to the screw, and this can be adjusted. If the screw itself is sliding back and forth as you apply pressure, this may be an issue. This can usually be fixed by tightening things up with some wrenches - or adjusted some other relatively simple way.
Ball-nuts (that move along the ball-screw) can wear out, bind up or lose their balls. If the screw stays fixed but the ball-nut bumps back and forth this may be a problem. Ball-nuts can be replaced or repacked with slightly larger balls.
Attachment of axis drive motors to the rack or balls-crew is often through a gearbox or belt. Key-way and setscrews in these can be a source of slop, as can work or loose belts themselves.
Parts availability is a big concern for older machines. Some manufacturers used common off the shelf parts - and this is great. Others made their own or mixed and matched and this can be a real challenge if things break. Ebay can be a great resource as long as you can be sure that your are buying the exact same component that is in your machine. There are a number of variants of common servos and drives and a single letter or number on the tag can be the difference between success and expensive smoke.
It is never reasonable to assume that just because a manufacturer is still in business, that they will support an older machine or have parts. Several common machines (Multicam, AXYZ, probably others…) have machines that aren’t that old for which the main control cards are no longer available. The rest of the machine may be fine but if that manufacturer -specific board goes, you’re looking at $20-30k to retrofit to a new supported control. This is fair enough - machine manufacturers can’t support things if they can’t get parts from their suppliers, but it is worth a call to the manufacturer to see what the real deal is on parts and support before you negotiate a price. Some manufacturers will help you out for free even if you aren’t the original owner and others will want you to pay for a support program before you can even get a technician on the phone. Just make sure you know what you’re signing up for!
Program transfer from new PCs to old CNC controls is a common problem when using old machines. Usually this is done with DNC (Direct Numerical Control) software over RS-232 - which is a standard for serial communications. In 1996, PCs came with serial ports - and your dial up modem might have had 56,000 baud rate for data transfer. Now its actually hard to get a computer with a serial port and DNC software sometimes requires DOS or some older version of Windows. The problem is compounded by the fact that big programs use a lot of memory, and older CNC controls didn’t have much. You can’t fit too much g-code in 128k of memory! The solution is to “drip feed” programs over the DNC link instead of fully uploading them into the controls memory. This is great once you get it sorted but it leaves things open to problems because the computer has to keep streaming data over the RS232 line as long as the program is running. Its a bummer if you’re 4 hours into a big 3D surface and the DNC stops dripping!
If you are going to buy a used machine it is also REALLY important to copy the control parameters to a backup before the machine is disconnected from power. Parameters are the long list of numbers that tell a generic CNC control (Fanuc, OSAI, Fagor, etc…) how to interface with the specific machine configuration. This includes useful stuff like how many axes there are and how much a servo has to turn to make the X-axis move 1mm - and how to find tool #6, and how to open the spindle tool clamp. Important stuff and your machine will be a total brick if those numbers get lost. Most controls can backup their parameters over DNC, Ethernet or USB. Don’t risk it - machinist forums are full of people begging for parameter backups to some obscure machine and too often this sends an otherwise usable machine to scrap.
REPAIRS and RETROFITS
If (and when) you do have a major breakdown with an old machine, there are plenty of good repair and retrofit options. Even if the original maker of your machine is no longer in business (like Fadal, Motionmaster, etc…) you’ll usually still be fine because many of the people who were specialists are still selling that experience - somewhere else. Many independent repair technicians are available and if your machine is common there will be parts and knowledge readily available. Retrofit controls like Centroid and Machmotion are expensive seeming but effectively bring a 20 year old machine up to modern status. For similar money, the newer version of the machine’s original control can often be installed. Many manufacturers will supply upgrade paths for machines that are too old to support.
Retrofits are easier on simple machines. You may even be able to do it yourself with something like Linuxcnc, Masso, Wincnc or Centroid’s Acorn. I wouldn’t recommend trying to retrofit large complex machines yourself unless you feel very confident. 480v power is scary and the parts are expensive and one wrong move can cost thousands of dollars. You can expect to pay a qualified technician from about $10,000 to $50,000 to retrofit a machine depending on the costs of the parts and the level of complexity. Putting a Centroid Allin1DC control in a 3-axis router and re-using the old servos would be on the $10,000 end of the range. Retrofitting a 5-axis machine like my old Motionmaster with a new Fagor control would cost maybe $25,000. And one time I discussed the cost to retrofit a 5-axis CMS machine and the repair company was talking north of $50,000 for a new control and all the work to integrate the complexities of that machine. Some of this is labor and it would take days or even weeks to do, but the parts themselves are very expensive. At least half of each of my estimates is spent on hardware.
I think there are two general situations where a retrofit makes a lot of sense. The first is for cheap machines, or older simple machines where the owner can do the retrofit themselves. The other time is with large or complex machines where there is a lot of value in the machine frame, mechanical parts, structure, etc. and it would be quite useful in working condition. People retrofit big 5-axis machines all the time but it makes less sense to retrofit a simple panel processing machine, for which there are lots of substitutes available in working condition. Not too long ago I parted out a 3-axis router because a key control card blew and replacements weren’t available - it just didn’t make sense to fix it.
A CNC router is a big tool and can create a lot of value for you and your customers. It can also be an expensive item and one that may rapidly become integral to your production process. Here is my best advice on things to consider when purchasing one:
Learn as much as possible - talk to people who run machines and sell machines. Find a few people who can help you out. You may have to pay for this and that’s ok.
Look very hard at what you need to do with it, what you hope to do with it, and what you could end up doing with it that you haven’t thought of yet. Look at what other people are doing. Make a list and put numbers in boxes.
Be realistic about the value it will bring and align the machine you buy with the value you hope to create with it. Again, numbers.
Understand your tolerance for risk. Make that part of your calculation.
Be realistic about the costs involved - its not just the machine but tooling, CAM software, rigging and power.
When in doubt, start small. You can always trade up and you’ll have gained hands-on experience.
Mostly, treat the process as a chance to learn. You may be shopping for your side gig, your business or your boss - the knowledge you gain is yours so enjoy the process and ask lots of questions!
This article contains information that reflects my opinions - I make no promises about its usefulness! It may contain mistakes (please let me know if you find some!) and will include prejudices based on my limited experience. If you disagree with anything here, please get in touch. This is not just for me to share what I know, but to learn from others. I will gladly insert additional information and differing opinions so readers are more aware of the diversity of “right” answers!