Before any detailed design work can be started there are plenty of big trades to make in the overall design of the router. What will the axis stackup look like? What will the workpiece clamping method be? What will the motion components be? What auxiliary systems will there be? Answering these questions will be critical for guiding the design of the rest of the machine.
Axis Layout
Ignoring some unusual layout options, like H-Bots designs that have limited stiffness and tension line designs with poor control authority in certain areas, a gantry based design is the obvious winner.
For typical gantry routers, the ‘Y’ axis moves relative to the fixed workpieces, while the ‘X’ axis is on a gantry that moves perpendicular to the Y axis. The ‘Z’ axis moves relative to the X axis and perpendicular to both other axes, typically with a fairly short stroke on most routers. For a desired 2 foot x 4 foot work area, the obvious question is which axis to make longer and which to make shorter?
Long X Axis Benefits
The Avid CNC Pro 4’x2′ Router has a long X axis and short Y axis:
- Y Axis can be expanded (potentially to a 4’x8′ full sheet size) without modifying the gantry at all
- Fewer overall linear guides/actuators because the axis with a pair of guides (Y) is shorter
- If placed against a single wall, with the long edge backed against the wall, most of the router space is available a working surface with three edges easily accessible
Long Y Axis Benefits
Most commercial CNC routers this size tend to use this orientation.
- The gantry, which is a weak spot for the overall stiffness of the machine, is shorter and better supported
- It is much easier to drive both Y axis gantry bases with a common actuator
- There is more margin available on gantry width to add extra heads without eating into the desired working area
Decision: Long X Axis
For a hobbyist things like footprint and future adaptability definitely outweigh slight optimizations in stiffness and material usage. This machine will take up a significant amount of the free space in my garage, and I will definitely want to use it as a work table when it isn’t cutting parts. The extra cost of a second motor/drive on the Y axis doesn’t go to waste because it will require extra power to accelerate the gantry assembly, compared to the X axis which only moves the weight of the Z axis assembly. The extra length required on the X axis to fit a laser head is worrisome, and if it becomes too hard to implement a reasonable X axis this decision will need to be revisited.
Linear Guides
The purpose of each axis on a CNC router is to constrain motion to exactly one direction while providing stiffness to counteract linear forces in other directions and twisting forces in any direction. There are a few options available, each with different benefits but also different price points and assembly requirements.
Arranging multiple bearings on each axis can help compensate for the weaknesses of individual bearing blocks while increasing the overall stiffness of the machine, but typically comes with the pain of adjusting and aligning multiple linear features relative to each other.
Linear Rail
Linear rails are what you see on most large industrial machines that need to constrain linear motion. They have very high stiffness in their ‘lateral’ and ‘vertical’ directions, moderate stiffness in ‘pitch’ and ‘yaw’, but limited stiffness in ‘roll’.
- Typically the most expensive option (ground surfaces, ball bearings, large shipping costs and heavy steel parts)
- Due to their extremely high ‘lateral’ and ‘vertical’ stiffness they can be very sensitive to misalignment, leading to premature failures
- The rails typically mount with a single line of bolts at regular intervals running the length of the rail
- A precision shear feature to align the edge of the rail against is usually recommended by the manufacturer to ensure alignment
Linear Shaft
Linear shafts are a good middle ground of price and performance. Usually they are made of a stiff steel shaft mounted on an aluminum T extrusion that adapts the shaft to a flange mount. They have very high stiffness in the ‘vertical’ direction, but less stiffness in the ‘lateral’ direction than a linear rail because of bending in the T extrusion. There is moderate stiffness in the ‘pitch’ and ‘yaw’ axes, but no support at all in the roll direction.
- These assemblies are typically cheaper than linear rails, but must be used in pairs in order to constrain the roll axis
- The unconstrained roll axis and less-stiff lateral axis can actually be a benefit because poorly aligned machines are less likely to damage themselves
- Because the T extrusion is typically not post-machined there are no good alignment features on the mounting surface to use when assembling the guide
- The lower stiffness and more forgiving alignment, compared to linear rails, can make machines with this type of guide less accurate
- When mounting a pair of these guides parallel to each other, using parallel mounting planes does not improve the lateral stiffness, but using orthogonal mounting planes is still not extremely stiff when reacting ‘roll’ forces
V Roller Assembly
V rollers are a very popular choice for budget CNCs. If you have the patience to fine tune the wheel alignment and the willingness to accept a little of slop in some parts along the length of travel then you can save a lot of money on your build. The benefit really starts to become apparent when you use structural extrusions with integrated V roller compatible features.
These assemblies typically have great ‘lateral’ stiffness (after the slop is taken out), but only moderate ‘vertical’ stiffness because the rollers have trouble countering an off center axial load. There is high ‘yaw’ stiffness (again, after the slop is removed) but low ‘pitch’ and ‘roll’ stiffness because of the off center axial loads put on the rollers.
- Very low cost, especially when implementing the V surface on an existing structural member
- Without grinding the V surface or adding a tensioning mechanism it is extremely hard to remove slop at all points in the travel of the carriage
Decision: Linear Rail
If you want stiffness in every direction, 4 linear bearings across 2 linear rails is the best way to go. However, a similar layout with linear shaft only sacrifices stiffness in one direction, which could be designed around. I will begin the design with linear shaft, but be mindful of the known stiffness limitations and circle back to this decision if the analysis shows that the stiffness is an issue.
Linear Power
Like everything else in the design, stiffness and slop are the main criteria in selecting the power transfer mechanisms. Not surprisingly, options with higher stiffness and less slop come with a larger price tag, but sometime there are other tradeoffs (like efficiency). At the end of the day, you need to turn rotary motion into reversing, accurate, stiff linear motion.
Flexible Tension Drives
Belts, chains, and cables are all different ways of turning rotary motion into linear motion that rely on the principle of bending a tension-only member around a curve. The Maslow router is a good example of a spooling cable based drive mechanism, but belt drives (and the occasional chain drive) are much more common on CNCs. Some architectures, like H-Bots, even require flexible tension members to work.
The one characteristic all of these solutions have in common is that they must be tensioned in order to work. The downside of tensioning a system is that it sets a maximum amount of force that can be reacted before the preload is overcome and the system looses engagement. Between that constraint and the low stiffness available from these options they don’t appear well suited for aluminum milling, although they are excellent low cost options for 3D printing and laser engraving.
Screw Drives
ACME (trapezoidal) screws and ball screws are the two most popular options for CNC screw drives. ACME screws are cheaper because the nut rides directly on the screw, but this comes at the cost of decreased efficiency and increased wear. Ball screws use a layer of ball bearings to space the nut off of the screw which allows for higher speeds and efficiencies, although they are more difficult to preload in order to remove backlash. Balls screws are easier to backdrive, which can be convenient on the X and Y axes but a concern on the Z axis.
The two limiting factors for screw based linear motion are screw whip (speed limiter) and buckling load (force limiter). Screw whip is the effect where slight imbalances of the center of gravity on a screw tend to bow it outward when spun at high speeds due to centripetal motion. Buckling is the tendency for long skinny structures to bow when resisting a compressive load, which can happen on the length of screw between the motor and the nut. Both of these effects are determined proportionally to the square of the length of the screw, which is why it is uncommon to see screws operating very long axes.
Tangent Motion Drives
Rack drives and friction drives both operate on the principle of transferring rotary motion to linear motion at their tangent contact point. Friction drives fundamentally rely on some amount of friction and deflection, which makes the kinematic relationship between the rotary and linear motion hard to calculate, especially over multiple rotations. Rack and pinion setups have deterministic kinematics, but it comes at the cost of backlash between the rotary and linear joints.
Tensioning/spring mechanisms can be used to mitigate the backlash but that complexity comes at a cost. Additionally, the mechanical advantage of a rack and pinion drive can be lower than a screw or belt drive, driving the need for an additional gear reduction in the system. As much as a I love the Avid CNC, the rack drive appears to drive a lot of cost and complexity into their system.
Decision: Ball Screw Drive
Given the low stiffness of tension drives and the high backlash of uncompensated rack drives, screw drives seem like the best option for a stiff, low backlash machine. While other machines manage to use ball screws on 3-4 foot axes, screw whip and buckling loads may end up driving a beefier component than would otherwise be necessary for the longest axis on my machine.
Summary
Early architecture decisions should not be irreversible decisions, but instead serve as a guide for the next phase in the design process. Most of the decisions are cost vs. stiffness trades, so when the pencils are sharpened and the first round of analysis is run they should all be revisited. After all, making one or two aspects of the machine stiff enough for steel cutting while others can barely process aluminum is a complete waste. The goal is to trade away just enough stiffness and performance at each decision point to end up with an affordable machine without any overbuilt subsystems.
Making the X axis the “long” axis to save on motion component cost depends on the stiffness of the beam being high enough. Choosing linear rails over shafts only makes sense if the stiffness of the shafts would be problematically low. Using ball screws instead of racks will not work out if the screw length is too long. These are all engineering hunches that need to be confirmed through analysis.