My CNC Router Build Adventure

[JayneV]

Hi everyone,

I modified the base structure to raise the sides so the rails are up high and the gantry will not need upright supports to achieve the required Z travel. This feature of the Momus and Peter's YaG are what initially attracted me to those machines. This is my take on incorporating that design feature using steel sections I already have. The inside wall would be clad with either plywood or aluminium sheet. I'm a little concerned with distortion from welding, the parts together. I'm assuming the most critical part is the top RHS section that the linear rail is attached to. I could try brazing that section to the upright SHS pieces but don't have a lot of experience with brazing. Another thought is to weld the entire base frame, including the top RHS section, then construct a temporary fence around that top section and pour some self levelling epoxy on top to create a level surface. After the epoxy cures, attach an aluminium flat bar extrusion on top of the epoxy and bolt it down to the RHS. The aluminium flat bar should then present a flat surface for the linear rail to attach to. Does that all sound like a reasonable approach or am I over complicating things?

[peteeng]

Morning Jayne - Welcome to the Shark Tank. 1) The side webs are not needed and if they are used need to be attached the the edge of the section not the middle. In the middle they are attached to air and will either deflect slightly or can vibrate 2) Same thought on the rail. Its mounted in the middle of the section on air and can move and vibrate. Rails need to be mounted on edges or on thicker material then the RHS so you can thread it and to maximise the local stiffness 3) Brazing is very worthwhile to limit distortion. 4) I strongly suggest not to go down the epoxy levelling path. Again its done with mixed success, more issues then its worth and epoxy is 3GPa stiffness so do you want a rail mounted on rubber?

The side webs chew into your footprint and you will run out of geometry by the time you get to our Z axis design to make the Z axis big and stiff. Its usual to a) start at your max footprint and work inward which usually leads to a disappointing working envelope or b) start at the required envelope and get a bigger footprint then you want. So you have to work around all of those things.

Do you MIG or tig weld? and you have an oxy/acy or propane/oxy torch? If you weld the temptation is to weld everything together. This is unnecessary. 25% or 50% welding is good enough for a machine and you only weld the areas that do not contribute to warping, eg you have used equal sized members and the temptation is to weld where the radiused corner edges are (very big gap) this will pull the structure. Only weld or braze were its flat to flat. Use the next size tube down so all faying surfaces are in contact. Preheat the area with the torch to help limit distortion... People use SHS and RHS as its relatively light and stiff and available. For machine design and build consider open sections. A C channel is same bending stiffness (depth to deph), locally thicker and you can mount things closer to its edge.

Steel SHS and RHS is not flat enough or straight enough to bolt rail to directly. The cars will jam... just been through a similiar thing with YaG and its pressed steel parts.

What you have designed is very stiff and will work fine. The issue is to build it so the rails are level and parallel. If you have a solid welding table that is flat it could be built upside down. Tacked up and do small runs spaced out and slowly work it up. Same process with brazing. If welded once complete you could locally bring welds to red with oxy and local stress relive before releasing from the table. Being in a major city look up heat treaters they can stress relieve the entire frame. But you need to be able to then machine the rail lands flat.

I suggest the walls and table be separate boltable parts so you can adjust them individually. Plus if done in one it will be really heavy. I design things now that can be broken down for transport and lifted easily by one or 2 people. Finding lots of muscle or cranes can be an issue... If you build in parts the individual parts can be machined easier then if monolithic as well....Consider heavy aluminium C sections easy to bolt together relatively straight and light... That's my input for the day... Peter

[JayneV]

Good morning! Peter, thank you for the detailed response. I am really enjoying this learning experience.

Good point about the side webs needing to be attached to the edge of the hollow section, however if they are not necessary, then better to leave them out. That would reduce the amount of welding and potential for distortion. I’ll put some more thought into the rail mounting surface to try and come up with a better solution. Epoxy levelling was just an idea which will now be abandoned, I don’t want the rails mounted on a flexible surface. I’d have to do a bit of practice brazing because I’ve never done it for structural integrity before.

I’ve been using option (b) to determine the size of the machine, starting with a desired working envelope and working out. I have some leeway to increase the footprint of the machine. It’s all a juggle to find the most suitable compromise. With the current goal of raising the height of the walls, I am thinking about adding a sub structure to increase the wall thickness to house the electronics. Hmmmm, that may raise problems with the electronics experiencing higher vibrations than they would like. Perhaps rubber shock mounts for the electronics? I’ll think about that later.

I usually MIG weld but I also have a TIG welder. My MIG is currently set up for gasless welding. I have the regulator and plumbing for gas and just need to get some gas and a roll of wire. The type of gas will depend if I go with MIG or TIG. Is one better than the other for distortion? I haven’t done much TIG welding so will need to do some practice welds to refresh my skills. I don’t have oxy/acy (my brother night, which I could borrow). I have a propane torch that attaches to a regular BBQ gas bottle and I have a MAPP gas bottle and torch.

The larger RHS section on the bottom is 100x50x3. Originally I used 40mm SHS for the upright bits but then changed it to 50mm SHS to match the width of the lower RHS. I should have stuck with the first version. These sections are what I already have hence why I chose to use them.

I have a heavy steel bench with a pretty thick steel top plate. It’s not a proper welding table so I need to check it for flatness. I’m hoping it is flat enough.

This is all very helpful information, thank you.

Jayne

[peteeng]

Hi Jayne - MIG uses less heat then TIG and theoretically will produce less distortion. Most distortion is due to the weld puddle shrinking, steel shrinks 2% by volume when it solidifies, aluminium can be 6%. This shrinkage occurs with all welding processes.

Then there is poor fits which is inevitable with cold rolled sections. If you look at the sides they are either cupped or crowned due to the way its made so really tight fits are difficult to achieve. Here is a suggestion 1) make columns and other parts with flat ends these can be welded. Then machine the ends flat and parallel. Then these can be used to assemble the base. They can be shimmed or use screws to adjust fit and position. Once adjusted for fit they can be epoxied in place. This means the top plate that holds rails and drives can be machined or fitted to correct levels and dims and the assembly stage is cold, no heat affected issues.... One thing you will need is a long engineers level. A builders level is good for roughing but not for final fit... ebay has them much cheaper then local suppliers.

Another thing to look at is TIG brazing if you want a hot solution to the joining... Brazing solves the parent metal shrinkage issue but does not solve poor fits. Fits need to be excellent for an excellent result. Do not discount bolted solutions, many buildings have been up for over 100 years like the Eiffel tower and they are rivetted together. Modern high strength bolts (class 8.8 or 10.9 or better) provide much more friction then a hot set rivet.... and its adjustable... To me there are two camps 1) do what your doing but understand that metal welding introduces all sorts of issues. Then have it heat treated then have it machined to correct all the issues at the end 2) choose a process that is correct at the beginning and do everything possible to maintain that correctness. This usually means machining smaller parts and bolting them together. There are other process like bonding that have opportunities to be explored...Planes for instance are mainly bonded together these days. Then theres rivets like metal plane wings are made of... if your into panel beating all sorts of things can be done....eg your gantry could be 4 pieces of flat steel or AL use AL angle at the corners and rivet/epoxy it together. Can make the base the same way as a practice...

When I first got involved in motorsport they were moving from welded crmo tubular chassis to aluminium honeycomb rivetted and bonded chassis. These were much stiffer and lighter then the spaceframes and easier to do, no welding distortion, no cold setting or cutting and correcting.... Peter

[peteeng]

Hi Jayne - Things like this. You don't need honeycomb but can make sheetmetal monocoque. Then its all cold and easy to adjust as you go. Bolted or rivetted. Was soon after doing this sort of thing we then went to moulded composite parts. Me coming from the marine area this was a natural step. The image is not my work. I did formula 3 stuff in Queensland, I also did a bit of bodybuilding in V8 supercars many many years ago, TIG welding up all the holes in Ford bodies... Peter

[JayneV]

Peter, working on race cars must have been exciting work. I work in the aircraft industry and have seen plenty of examples of sheetmetal monocoque construction. There are some clever designs that are deceptively lightweight and very strong. Using rivets is an interesting idea, it's not a method I've thought about applying to build the machine, but it is worth some consideration.

[peteeng]

You'll need a few Clecos and since these things don't need strength you can use a soft grade of aluminium or thin steel and you can make curves.... Peter

[routalot]

If I may put in a word for LinuxCNC I would recommend giving it a try.It won't cost anything to download an .iso an run it as a live install to see what it looks like.I use it on my hobby machine and it has been utterly reliable.I built my tiny machine as an exercise in making a machine that worked at a scale for modelmaking and took a few shortcuts in order to get cutting.I bought a very cheap Windows box and installed LinuxCNC easily.Configuring the drives wasn't too hard as I use the parallel port for this machine.There may be a future machine with a more up to date solution but the principle of all PC driven machines is the same-you need to tell the PC how many steps it will receive to move one unit of measurement.I know absolutely nothing about anything Apple but will raise the topic of how you might generate toolpaths for the machine once it is complete.You probably know that you will need to find a CAM system that will work with your chosen controller and the choice of post processor might guide you.

[JayneV]

You'll need a few Clecos and since these things don't need strength you can use a soft grade of aluminium or thin steel and you can make curves.... Peter

On a side note, I was seriously considering building a plane some years ago. I spent quite a lot of time researching different kits and plans and even visited Vans Aircraft factory in Portland, Oregon, USA and later ordered a set of preview plans from them to study. Around that time, the Australian dollar had overtaken the US dollar by a good amount and should have ordered a kit then, but the timing wasn't right, then life happened and the plane never got built. It's still a distant dream. Had I proceeded with that project, I would have become somewhat of an expert at riveting and placing Clecos in and out of holes.....there were literally thousands of rivets to set.

[JayneV]

If I may put in a word for LinuxCNC I would recommend giving it a try.It won't cost anything to download an .iso an run it as a live install to see what it looks like.I use it on my hobby machine and it has been utterly reliable.I built my tiny machine as an exercise in making a machine that worked at a scale for modelmaking and took a few shortcuts in order to get cutting.I bought a very cheap Windows box and installed LinuxCNC easily.Configuring the drives wasn't too hard as I use the parallel port for this machine.There may be a future machine with a more up to date solution but the principle of all PC driven machines is the same-you need to tell the PC how many steps it will receive to move one unit of measurement.I know absolutely nothing about anything Apple but will raise the topic of how you might generate toolpaths for the machine once it is complete.You probably know that you will need to find a CAM system that will work with your chosen controller and the choice of post processor might guide you.

I don't recall if I already mentioned that I have an old Windows laptop which has been sitting on a shelf unused for years. From memory, it has a parallel port but I think I read somewhere that a laptop parallel port isn't suitable to run a CNC? At the very least, that computer should be good enough to play around with the software.

My brother sent me this link https://www.thingiverse.com/thing:4961410 last week, It's for a 3D printed mini milling machine to learn the gcode and software control side of things. I may build that as a test platform for LinuxCNC so I can see it in operation. Shouldn't take too long to print.

For CAM I'm planning to use Fusion 360. I don't yet know specifics about which controllers do and don't work wth Fusion.

[jckstrthmghty]

I don't recall if I already mentioned that I have an old Windows laptop which has been sitting on a shelf unused for years. From memory, it has a parallel port but I think I read somewhere that a laptop parallel port isn't suitable to run a CNC? At the very least, that computer should be good enough to play around with the software.

I have a very old eurocom latop, dual core pentium something, running mach3 directly off the paralell port. Not the smoothest but runs reliably. Just have to turn everything else off, firewall, antivirus, screensavers, updates, etc... I've since bought an ethernet smoothstepper and moved to a better computer but the old laptop was viable. I definitely recommend an external controller like the smoothstepper but for learning/testing phase your old laptop should be adequate.

[JayneV]

I have a very old eurocom latop, dual core pentium something, running mach3 directly off the paralell port. Not the smoothest but runs reliably. Just have to turn everything else off, firewall, antivirus, screensavers, updates, etc... I've since bought an ethernet smoothstepper and moved to a better computer but the old laptop was viable. I definitely recommend an external controller like the smoothstepper but for learning/testing phase your old laptop should be adequate.

Thank you, that is good to know an old computer would work. Hopefully, it will still turn on. I think it had some hardware problem when I stopped using it, but it might only have been the wifi that stopped working which won't prevent trying out some controller software.

[peteeng]

Hi Jayne - The clocking requirements for CAM is very strict, that's why external motion controllers are now used. They are responsible for the step timing not the computer clock. Computer clocks slow down and speed up depending on the threading and application activity. So a slow computer does not matter as the MC does the work... Old PC MC's froze various internal clocks to achieve CAM timing eg Mach3 don't go there. Peter

[JayneV]

Hi Jayne - The clocking requirements for CAM is very strict, that's why external motion controllers are now used. They are responsible for the step timing not the computer clock. Computer clocks slow down and speed up depending on the threading and application activity. So a slow computer does not matter as the MC does the work... Old PC MC's froze various internal clocks to achieve CAM timing eg Mach3 don't go there. Peter

Ahhhh, that makes sense. I never really liked the idea of the steps being controlled by a parallel port. It seemed like obsolete technology. It seems logical to have a dedicated piece of hardware to control the motors reliably. An average personal computer, whether, Windows, Linux, Mac, is a multifunctional device designed to do many different tasks, it's not really intended to accurately control time critical external hardware, although it can be done.

I haven't done much research on the various electronic components yet. Is the gcode streamed to the motion controller from the PC in real time or is the entire code sent to the MC at once and then the motion controller runs independently of the computer? I'm thinking about my 3D printer, where the gcode is generated in slicing software and saved to an SD card and then the card is plugged into the printer to execute the code.

[peteeng]

Hi Jayne - The MC is a dedicated mini computer. It will have some ram in it. The gcode gets streamed to the MC and the MC releases the gcode in strict time with velocity data attached. You will see in the gcode generated by the CAM it only has F code in various spots. The MC looks ahead and figures out if a corner is coming and then slows down the feed velocity as per its settings, then speeds up again if the road is straight. I'm sure if the data stream is big it will be buffered somehow in the CAM or in the MC. Also depending on how you set up the parameters in CAM the MC will steer through a corner within the "road" width set in CAM, like cutting a corner in a car. It does not go through the exact gcode positions unless you set it to exact or precision (forget the actual term) if you do set it to exact, the feed will be extremely slow. Peter

[peteeng]

Hi Jayne - its called exact stop mode or constant velocity mode - download uccnc manual and read from P36 about path tolerances... pdf too big to upload here...Other controllers will be similiar,,, Peter

https://cncdrive.com/UCCNC.html

[mactec54]

Ahhhh, that makes sense. I never really liked the idea of the steps being controlled by a parallel port. It seemed like obsolete technology. It seems logical to have a dedicated piece of hardware to control the motors reliably. An average personal computer, whether, Windows, Linux, Mac, is a multifunctional device designed to do many different tasks, it's not really intended to accurately control time critical external hardware, although it can be done.

I haven't done much research on the various electronic components yet. Is the gcode streamed to the motion controller from the PC in real time or is the entire code sent to the MC at once and then the motion controller runs independently of the computer? I'm thinking about my 3D printer, where the gcode is generated in slicing software and saved to an SD card and then the card is plugged into the printer to execute the code.

If your computer runs then you could get a UC100 ( buy this from the manufacture or there resellers as there are copy's out there that you do not want to buy ) this works with Mach3 and has it's own CPU so the computer can be anything, it only needs a USB port and you can be running with a very basic Breakout Board

[mactec54]

In your case your computer is the controller where the G-Code is sent to a motion controller like the UC100 or direct to the drives through a Breakout Board, having a motion controller in-between the computer and the Breakout Board is the best way to go for Hobby level, and advanced, this take all the load away from the computer

[JayneV]

I understand now. The motion controller takes care of all the critical timing and signals to the motors. It also optimises the gcode to create the actual real life path followed by the tool based on user settings for desired tolerance and speed. The computer is basically the user interface for the machine and keeps the buffer memory topped up for the motion controller to use. So as long as that buffer doesn’t become empty, the motion controller will happily continue to do its thing. This seems like a superior arrangement than one where the computer performs the role of motion controller as well. Does that all sound about right?

Peter, I read a little about path tolerance in the LinuxCNC manual. I’m sure the concept is the same across all controllers. I assume this is how machine limitations are handled when transferring from the ideal CAD/CAM world to the real world, like the acceleration limits of the machine. And different machines would have different solutions to the same problem?

Mactec54, is the UC100 the motion controller hardware used by Mach3 or is it a hardware interface to connect a computer to various brand motion controllers?

Thank you everyone for all of your help. It is all so very helpful to help me have a better understanding of CNC machine construction and their controlling electronics.

Jayne

[peteeng]

Hi Jayne - Think of them as parameters vs solutions. For instance Brevis is my small belt driven router and it can run 15m/min velocity and 1000mm/s/s acceleration. YaG I've cut down to 5000mm/min and 200mm/s/s as it has ballscrews and can't accelerate that fast as the screws have lots of inertia. Still playing with its parameters but its time to make some dust... Peter