The final check before diving into fabrication is to verify that all of the assumptions I made earlier in the design process are still accurate, or at least conservative. If not, I may need to iterate something again (hopefully not leading to any drastic changes!). However, if everything checks out I can move on to make sweet renders and technical drawings of the final design!
Verification
Axis Mass
I eyeballed the mass of the Avid CNC over a year ago to start doing some rough axis sizing. I thought it was an overestimate because I was trying to be conservative, and because the Avid design has a lot of extra thick metal plates. However, after detailing every nut and bolt that goes in the assembly I ended up slightly overweight on some axes but vastly underweight on the spindle.
This mass doesn’t account for non-mechanical parts (like cabling, which can really add up in weight!) and of course doesn’t take into account future upgrades like an ATC or water-cooled spindle.
| Estimate | As Modeled | |
|---|---|---|
| Y: Gantry (kg) | 50 | 57 |
| X: Carriage (kg) | 15 | 16 |
| Z: Spindle (kg) | 20 | 11 |
| Base (kg) | – | 69 |
| Stand (kg) | 34 |
The big offenders in the Y-Axis Gantry assembly were the main beam extrusion, rail, and plates (as expected). However, I hadn’t accounted for the steel screw (2.5kg) or the last-minute addition of an extra box tube on the gantry (5.5kg) which put the assembly slightly overweight.
The X Axis carriage was only slightly overweight, and the categories lined up well with their estimates. I repeated my mistake of not taking the weight of the ball screw (and support bearings) into consideration, but luckily the motor weighed a little less than expected. Hardware weight really adds up!
The Z Axis saved the day by coming in significantly under weight. Every category I estimated (motor, structure, hardware) was about double the final weight. This should make the underestimate of the X and Y axis masses irrelevant because they were both carrying the weight of the Z axis as well.
A future upgrade to a water-cooled or ATC spindle would definitely increase the weight again. For the remaining verification calculations, I will keep the assumed Z Axis mass at 20kg to account for these future upgrades. I will also bump the Y axis mass up to 65kg to account for cables and potentially tubes of water in the future.
Axis Loads
The axis loads (on the ball nut and on the bearings) are primarily impacted by the mass, bearing layout, and stage spacing of each axis. I was able to stay mostly true to the desired bearing layout from much earlier, but the stage spacing evolved over time.
The biggest surprise on the Spindle Assemble was how high up and in the CG was. I had assumed it would be close to the center of the spindle, but the mass of the bearings and ball nut definitely pulled it up and in. This is great news for gantry beam twist! A larger ATC spindle might not lower the CG, but it could bring it further from the bearings.
The biggest surprise on the Carriage Assembly was also how high up the CG was. I had guessed that it might be 75 mm above the bearing centerline, but it was a whopping 190mm above! The weight of the motor really pulled everything up. This is great because it helps prevent twisting in the gantry beam during rapid Y axis accelerations.
I did heavily shrink the axial bearing spacing on the X axis because it has a huge impact on beam length (and total machine footprint). It seemed like it would be OK based on the sensitivity study. The X axis bearing margin dipped slightly negative (-3%) after all of the mass and geometry updates, which I will accept because the driving load case is a 20g crash with a FOS of 2.
The Y Axis didn’t change much, but the CG did end up higher and slightly closer to the back of the gantry than expected. This is bad news for the Y axis bearings, but they had plenty of margin left due to other changes that were in their favor.
Axis Performance
Overall, the improvements in CG location and other geometries offset the impacts of a slightly heavier than expected gantry. With the exception of the one bearing case mentioned above, all of the driving structural cases (bearings, screw whip, screw buckling) have a positive margin.
Aluminum cutting motor torque still has a large positive margin, but wood cutting motor torque at 300IPM is at about -10% in X and -15% in Y. I’m not super worried about this because you can trade those numbers around a little bit with tooling choices (reducing flute count or increasing chip load) and maximum MRR on hardwood is already a fairly nuts point to be running the machine! 250IPM is slow enough to keep all margins positive with a FOS of 2 on motor torque.
Capabilities
It is important to always compare the machine you’ve modeled at the end of the design phase to the machine that you claimed you wanted at the beginning of the design phase. Sometimes adding ‘cool features’ can sidetrack engineers from the real requirements, which might get missed. Over a year ago I wrote down what I wanted to get out of the project, and I think this design managed to stay true to those requirements!
| Capability | Requirement | |
|---|---|---|
| Table Size | 56.5″ W x 48″ D | Easily process full plywood sheets; Clamp down the outside edge of full sheets; Vacuum table upgrade potential |
| Horizontal Travel | 52″ W x 31.25″ D | 48″x24″ working area; Expandable to full 4×8 size in the future |
| Vertical Travel | 8.5″ Clearance; 6.75″ Travel | Access to 6″ table leg on A Axis |
| Speed | 450IPM Rapid; 250IPM Wood; 100 IPM Aluminum | Fast in Wood; Slow in Aluminum |
| Tool Heads | 1.5kW Spindle+ Mounting Points | Fast in Wood, Slow in Aluminum; Vacuum connection, laser head |
| Accuracy | <<0.001″ Deflection; Closed Loop Motors; | 0.005″ locally; 0.010″ globally; |
| Cost | ~$6,000 | Better than outsourcing over ~5 years |
For the categories that can be known ahead of time everything looks great. Maybe a little extra Z axis travel would be nice, but the spindle can be clamped at different heights to adjust the center point of the travel depending on the setup (rotary axis vs. table work).
Accuracy depends on a lot of factors that I don’t have nailed down yet, or just don’t have a hard number for. I’m planning on buying cheap ball screws – will their linearity be terrible? How square/aligned will I be able to get the machine? Can I compensate for these inaccuracies in software? Local accuracy during low-load finish passes should be really good though, so the 0.005″ requirement should be easier to hit, barring any issues with ball nut backlash or something else terrible.
My current best guess at cost is around $6,000, assuming I machine the aluminum plates myself (maybe $7k otherwise). This is still better than the $10k Avid CNC, but since I first dove into this project a few other options have come to my attention. The PrintNC has matured a ton in the last two years and offers a really good value for someone interested in making their own CNC router (although I would want to run the numbers on their bearing choices). The Shapeko HDM is also a great value aluminum-capable router, at only $5k (although it is only just now shipping, and the table size is barely half of what I wanted). I’m still happy with the direction I have chosen and plan to continue on to the build phase.
Sweet Renders
The best part of finishing a design is making some sweet renders to capture the moment! Things will of course go wrong during the build, but right now the design is completely unaffected by the harshness of reality.
Slightly less interesting than admiring cool renders is making technical drawings. At this point all of the dimensions in the model are at their ‘nominal’ size, but when making the drawing you have to consider tolerancing, extra room for clearances, machining steps, assembly steps, and how you can even measure to confirm that you built the thing you designed in the first place!
My plan is to draw and build the machine in stages, starting with the base and stand then making my way up to the spindle. The control electronics and table will go on last. The CAD and analysis for this project are available on github, and the drawings will be making their way onto the repository as they are completed.