Research, Concept, Materials
5 Bears Home Homebrew CNC bench mill

Like many machinists, I tend to focus my efforts on producing the "product", such as an engine, and make use of off the shelf tooling and machines to do so. There is an entirely different aspect of the metalworking hobby which has a huge appeal for many, the toolmaker. Toolmakers are a specialized, very skilled subset of machinists who are concerned with accuracies and precisions which most general machinists only dream of.  We can venture into this realm and get vague ideas of the skills required by producing accessories, jigs, fixtures, and cutters, for our own use.  And we quickly learn that becoming a decent tool and die man requires years if not decades of dedicated effort.

So now I find myself in this role... engaged in the production of a device whose sole reason to exist is to create other objects, hopefully of beauty and occasionally of strangeness.

If the next few pages in this CNC saga seem disjointed, the fault is mine. I decided to begin posting to this site late in the production of Bench Mill Mark 1 (Mark 2 will follow, and correct upon any mistakes made in this phase). Thus, the mill is somewhat in pieces and being reassembled in the same rough order as its production.

To the left is a simplified view of the base of the mill. Front and back plates are evident, with the backplate being 1" taller than the front. The ends of the Y-axis rails can be seen just above the front plate, and the X-axis blocks, with stub rails, is mounted above the Y. The work table (a scavenged, cast-iron XY table) will be mounted to the X axis rails. Not shown is the additional column support members, Z-azis, or spindle. There is no servo motor shown - the 4 tapped holes surrounding the flange will support the servo motor mount. Everything I have drawn (via QuickCAD so far) has made use of abundant .DXF drawings available from the manufacturers like Bosch and THK. It is a real joy to be able to import these 2D models, simply drop them in place, and draw portions of the mill around them.


X-axis ballscrew, bearing block, nut, and nut carrier. THK SHS15 rails visible as well.
My first thought was to buy a small CNC mill off the shelf, and simply put it to work. I quickly discovered that there is a vast range of CNC bench mills out there, which range in price anywhere from $1500 or so to $50,000+ for a tooled Light Machines Benchman XT. Obviously there are differences in these machines; they are primarily due to the precision of the installed components, which ultimately affect both speed and accuracy.

At the bottom of the heap are retrofits of small mills using stepper motors. Please don't misunderstand me, these machines are sweet mills and will do fine work, but are loaded with compromises. I am not going to go into a lot of theory of steppers vs. servos, ballscrews vs ACME, or dovetail ways vs. recirculating ball ways. Each of these represents different portions of the CNC spectrum. At the upper end, we have high-speed spindles, automatic tool changers, zero-backlash ground ballscrews, and recirculating linear ways. Combined with servos, these small mills function like a baby VMC (Vertical Machining Center), with great accuracy and high speed.


X-axis: Table inverted. 2 18" Rails are bolted to the table, and 4 SHS15 blocks form the sliding element. The blocks come in different styles, sizes, load ratings, etc. Best to read the THK online catalogue to familiarize yourself with them.
I could not afford a Benchman XT. I likewise couldn't afford a Minitech Mini-mill/3 or /pro. But perusing these specs, and those of dozens of other models, taught me quite a bit about what was desirable and feasible for my own mill.

Ways: The high-end Minitech mills make use of THK recirculating rails and blocks for the ways. I had never heard of these, and after a bit of research, was amazed at their capability. They consist of hardened steel rails, and riding upon these are extremely precise blocks which have internally recirculating ball bearings. To the left are the rails and blocks which form the X-axis... a set of THK SHS15 rails, and four bearing blocks. In use, the blocks are bolted to a saddle, and with the rails fixed to the table itself, we have now formed a "way" which can handle loads measured in hundreds of pounds (tons for the larger blocks). Amazingly, they are totally rigid on all axes except their intended axis of motiion... the table slides right and left under its own weight with ease, and a gentle push with the hand will slide it back and forth - yet, attempts to "twist" or move the table in any direction except the correct axis is futile. Dovetails are fine, but by gum these are better, as they allow high-speed motion with minimal torque. A drawback - they are expensive. The two rails and four blocks you see are ~$450 to $550 new, and each axis requires two rails, and at least 2 blocks, normally 4 blocks. Ouch! But we'll take care of the costs later.


A pair of HSR25 blocks. You can see the recirculating balls inside the inverted left block. If you think these are ball bearing drawer pulls on steroids, you are wrong, they are dedicated precision devices for just this type of project.
The caveat to the use of linear BB ways for a bench mill is that the leadscrew/bearing assembly must be totally rigid, with zero backlash. If there is backlash in the system, the low force requirements to move the table will cause all sorts of grief when the direction is reversed; in that moment, there is no load in the leadscrew, and the table will flop about like a dead fish under pressure from the cutter.

My initial thought was to refit a Harbor Freight Mini-mill with miniature THK rails. After a month of effort, I was rewarded with a mini-mill fitted with THK RSR12 ways. It looked OK, but the effort was huge, and these tiny blocks weren't going to handle the anticipated loads; larger blocks wouldn't fit. Worst of all, there was almost no space to fit ballscrews, and the stock acme leadscrews were suspect. The project was terminated - the spindle saved. The other parts of the HF mill live in a cardboard box, a testament to my futility. But I learned from it. And isn't that why we do this stuff? Just be sure to hide the evidence if the wife comes snooping.

Seen here is an end-view of a Bosch-Rexroth 90mm X 45mm H (Heavy) extrusion, roughly 3.5" X 1.8". VERY stiff, modestly priced. Acceptably straight. Extrusions like this form the support for the mill. The numerous T-slots and tapped holes allow for the mill to be pieced together like a tinker toy, or erector set, allowing a wide range of mounting and adjusting options. They can be ordered online through Bosch.

Structural Support: A bit frustrated at this point, I was browsing eBay one day, drooling over VMC's, when I saw a homemade mill for sale. It used THK ways, rolled ballscrews, steppers, and a Porter-Cable router head for the spindle. The mill made use of beefy structural aluminum sections for support. The price was right, and I bought it! I thought my CNC woes were over, but I was sadly disappointed by the mill. The best I can say about it is that the concept was brilliant; simple, elegant, utilitarian design. But the execution was poor. The rolled ballscrews were not mounted correctly, bad ballscrew bearing design (no thrust preloading), the ways badly aligned. The X-table was an extruded aluminum t-slotted table, probably flat to within .010", no better. The mill would probably engrave nicely, but not do much more. The redeeming factor, though, was the design. It opened my eyes as to what could be done on a limited budget. The mill you will see here on this site is based very closely on the ebay mill, which was reduced to component pieces, and replicated with the best materials and skills that I have. Most of the eBay mill hides in a box with the Harbor Freight mill. I leave the lid closed.

View down the Y-axis to the column, endplate removed. You can see the bearing which supports the end of the Y-axis leadscrew in a block, mounted using the column's T-slots.
Why aluminum? It's not the best material for a machine tool, but given that the ways are ultra-rigid, hardened steel, and the ballscrews likewise, it is an acceptable alternative to cast iron. Certain portions of the mill (Table, spindle mount) will use cast iron to absorb vibrations. The extrusions seen here are far more rigid than they look! They arrive anodized, and cut to the specified length. To replicate this mill with CI sections would quadruple the cost, and are not necessary for the types of work I envision.

Aluminum drawbacks: It will dimensionally vary with temperature. I don't expect this to be a problem, though. It is softer and more prone to damage than iron, but all moving parts on the mill are hardened steel, courtesy of THK. Quite a few of the better commercial bench mills use a cast polymer base, with the THK rails bolted directly to the polymer. So I guess aluminum will work. It does transmit vibration more than CI. I am hoping that keeping the RPM's up (a true HS mill) will keep the low-frequency vibes down.

Two of the Bosch 4590H extrusions form the base of the machine. The ends of the base are 1/2" and 3/4" aluminum plates, which form an open framework for installation of the Y-axis and saddle.

The column will make use of this 9090H section. It will form the core of the column, and be sandwiched with 1" plate aluminum on 2 or 3 sides to form a very stiff upright section. More details will be presented later as the project progresses.

Basically, by this time, I had decided to make use of the structural aluminum for support, THK rails for the sliding ways, and was left with critical decisions to make on the leadscrews.


A ground THK 12mm ballscrew in a shop-made dual angular-contact BB housing.
Lead Screws: Again, without flogging theoretical choices in leadscrews to death, let me say that if you can afford it, a project like this requires ballscrews. The other choice is some type of ACME screw, as is common in 90% of hobby machines we encounter. ACME (or V-thread) screws are fine, but suffer from 2 failings relative to ballscrews - first, they are inefficient. A typical ACME screw will transmit 25% to 35% of the motor's torque into movement... ballscrews are 90%+ efficient! Secondly, without drag-inducing preloaded double nuts on an ACME screw, there is the dreaded backlash phenomenon. Simply put, backlash is the "slop" in the nut, which allows the table to be moved typically 0.005" to 0.025" without turning the handle one bit. When the direction of travel is reversed, the backlash must be considered by the CNC control, and even the best controls cannot entirely eliminate the effects of backlash.

Ballscrews come in two flavors, rolled, and precision ground. A rolled screw is created by dies under pressure, whereas the threads of a ground screw are created by, well, grinding. (Ever notice how the "best" objects are "hardened and ground throughout"? One day, I am going to make an entire complex project made of nothing but "hardened and ground" steel, so I can say, "Yup. Hardened and ground throughout".)

Common to both rolled and ground ballscrews is a nut with recirculating ball bearings. Ballscrews are bulkier than ACME screws, and require precise angular contact bearings for best performance.


The ball nut, mounted on the Y-axis aluminum nut carrier.
Rolled ballscrews suffer from many of the failings of ACME screws, in that there is often a bit of backlash in the nut. This is due to the tolerances in pitch when the screw is rolled. Typical rolled ballscrew backlash numbers are from .003" to .010", not much, but it may as well be 0.050" as far as the accuracy of the backlash compensation is concerned. Normal rolled ballscrew accuracies are 0.004" over 300mm. Not great, but the repeatability will usually be near 0.

Ground ballscrews have none of these failings. Accuracies are usually around 0.0003" (or less!!) per 300mm, backlash 0, quiet, incredibly precise, hugely expensive. Again, cost will be addressed in the next section.

In use, the ballscrews are mounted in some type of housing, which contains either angular contact bearings (best) or a pair of ABEC 5 or better radial BB. The hex nut is tightened and locked, leaving the ballscrew firmly fixed axially. Thus, with the housing also firmly mounted (shown here on the Y-axis endplate), there is zero axial play throughout.

Summary: The project will make use of extensive structural aluminum supports, use THK linear ways, and ground ballscrews throughout. At a minimum, the table will be cast iron, with other areas potentially using CI plate as well.

In the next section, I will describe just how to obtain all these expensive components!