If your goal is long parts, why not just do a standard bench mill CNC conversion? I’ve heard that while they don’t list them on their web site, if you call Precision Matthews they actually have some frames for CNC conversion available, or at least have had in the past, and their tables are a bit longer in X than a lot of sieg variants. In that case you could use TTS in an R8 instead of BT30 for an ATC system. Converting the PM-833TV to ball screws if you could just get the frame parts would not horrible from a price point. it’s a bit under $5K OOTB so given your specs I’d certainly be inclined to call them and ask.
X Axis Travel: 21-3/4” Max
Y Axis Travel: 11” Max
Z Axis Travel (Head): 17-1/2” Max
My initial response to this idea was “heck, no” but I quickly realized that I didn’t know why. It’s a good idea on many fronts!
So, I slept on it and spent time poking at my feelings. It seems that I have a mental image of how much and what type of DIY I want in this build and for some reason using a premade frame isn’t in that image.
That said, I am fully aware that I will regret this decision. Probably several times. But here we are.
The taller the gantry, the wider and thicker it has to be, obviously.
The more than 1M depth gives some room to have a very stiff gantry and still have all the travel you want.
With a gantry, you have to decide how to synchronize the screws on each side. The stiffer the gantry, the worse it will be to rack it. Maybe consider a single motor and timing belt between them? You don’t have to limit yourself to belt loop lengths that are commercially available; I posted a way to join a belt into a loop accurately that is very strong (when I tested it using my hydraulic press, it failed elsewhere, not at the joint).
Those weren’t special, I just found what I was looking for and placed orders that went fine.
Now that I’m back at a keyboard, here’s the belt loop splicing technique I came up with, which was illustrated with that 5mm HTD belt, where I made a nearly 5M long loop:
Your tolerances may be tricky. I’d use large ball screws; maybe 2005? C7 is widely available (random aliexpress shows it at $65 for 1M) and is specified ±50µm/300mm. I see listings for C5 that say that C7 was shipped instead, so it looks like you have to pay attention and make sure if you are going C5 or better ground screws instead of C7 or looser rolled screws that you actually get the higher precision you pay for, at least on aliexpress. But that’s just the screw tolerance. I assume you’ve worked through at least roughly how the tolerances would stack up and what you could accept at each interface to still hold two tenths?
Skipping a step on a stepper motor would be expensive at that scale. You might want a servo instead, or a closed-loop stepper. If I were building something in that price range I’d definitely be using a servo.
Looking forward to seeing initial design ideas once you have the time to do them.
Hats off to you for all the upfront planning. It’s great that you have a lot of experience to go by in making a plan. I would not presume to advise you on this since you seem to be coming at it with a lot of experience, but on the chance that you might learn something from the build I did a few years back I can talk about that.
First, I came to the world of machine tools and CNC knowing very little, so I did some experimentation with basic leadscrews and steppers light guide rails driven by a tinyg controller and an existing router for the spindle. It was an okay setup for learning, but not near stiff enough to do well with metal cutting, so I decided to do something that would be stiff enough to do a decent job on aluminum and light cuts in steel.
Version 2 of my setup meant 16 mm ball-screws and 20 mm square guide rails with some larger steppers and a air cooled 1.5 kW ER20 spindle. Still not nearly as ambitious as your plan, but ultimately, I got something working withabout 30" x 15" of x-y and stiff enough to do a good job on steel and aluminum with an accuracy of about .002". The spindle setup will not ever handle auto tool change, and my design isn’t enclosed, so I have to let the chips fly where they may. https://www.youtube.com/watch?v=zExZ78TMWIo
I posted a brief video if you want to take a look just for ideas. Ultimately, I was able to get a lot of stiffness by bolting the whole thing down to a bench that was already attached to the block wall in my basement and bolting the column to the block wall greatly increased the stiffness of the spindle assembly.
Hey, Walt. Thank you for sharing your build! I’ve used a fair number of machines but this will be my first custom build.
That’s pretty smart, using saw tables that you had on hand. My shop space is too small to keep much stock and I don’t keep even one working table saw!
I’m still thinking through options for the table. I’d like something like a Pierson work holding plate with precision bored and tapped holes so that I can quickly swap fixtures with high repeatability. I also need to make it with a low profile to leave space for the Z travel that I want so it’ll be a balancing act.
The control box sits toward the front on the lower left.
With the total footprint that I have for this and the cutting envelope that I want (dotted lined box above the table) I think that I can leave space on the left side of the table for some sort of ATC rack.
I’m considering Teknik’s NEMA 34 Clearpath Servos that have integrated drivers. This model is driven by the boring old “step and direction” signals so when I need replacements in the future I’m not stuck with one vendor.
I’m a bit annoyed that searches for “servo” on sites like eBay, alibaba, and Amazon turn up results that are almost entirely for closed loop steppers pretending to be servos.
A couple of these power supplies (also from Teknik) would fit into the cabinet and (I think) handle the three initial servos and a potential 4th axis servo in the future.
Those Mesa boards are mostly out of stock, right? I ordered a 7i96s control board and a 7i84 I/O card because they’re in stock and supported by LinuxCNC. I’m not so confident in the longevity of a company that has so little stock, though.
I’m still browsing spindles and VFDs. There seem to be a couple of spindle kits that are rebranded by everyone and their brother so search results are almost exclusively filled with the same kits.
One nice aspect of working on a machine design in public is that people with a variety of experience will offer advice and pointers to references. For example, in this Fediverse thread I’ve received links to helpful PDFs about linear motion guide configuration:
I’ve also received helpful advice about strengthening the section of the frame that handles most of the forces involved with cutting steel. I added space frame (ish) support and am planning to fill that section with epoxy granite.
One other piece of advice that I’ve received from multiple people is to use boring old open loop stepper motors (instead of the expensive Clearpath servos) and use LinuxCNC’s trick of closing the loop with glass scales. Their arguments for this are that it’s less expensive, I won’t drop steps unless I’m misusing the machine, I won’t be locked into a vendor, and directly measuring the position of the gantry and spindle is better than indirect measurement of the motor position.
To get some direct experience with LinuxCNC I’ve ordered a couple of boards from Mesa USA, as well as a motor and driver that was recommended by a machine tool builder in my neighborhood.
Here are my current numbers for target speeds, feeds, and power:
This is my first attempt at calculating power for a new machine design so if any of you have more experience and want to weigh in then I’m open to suggestions.
There seem to be a couple of common spindles like this one from RATTM that have a BT30 taper and ATC but they all seem to be designed for higher speeds like 15k to 24k RPM, presumably for softer metals and plastics. For steel I want more like 600 to 1200 RPM.
There are also a few kits like this and this that can be belted to external motors so perhaps I could use a motor with the right power curve for steel.
I wonder whether it’s too much to hope that the enclosure lower the cutting and motor noise to ~60db, or around the level of a regular conversation. With the air compressor nearby and the fans running, it may be a pointless goal.
I went back and re-read the thread, and saw that you were considering both flood and mist coolant. With it enclosed, is there any reason not to use only flood? Given flood coolant, would the air compressor be needed? If only for a pneumatic ATC, you could easily get away with a quiet compressor, right?
You have more experience than I do, but my manual mill isn’t all that loud and doesn’t typically inhibit conversation, so not sure why an enclosed cutting area would be any louder.
With a slower, higher-torque motor, what fans will be running?
When correctly cutting steel the noise from the tool and part aren’t terribly loud. Ideally, it’s a low pitch thrum. So, I’ll just always nail my speeds, feeds, and fixtures.
For the air compressor, I was assuming that something like the obnoxious 2.5 gallon pancake compressor I currently use (as little as possible) would be necessary. Maybe that’s a bad assumption. I’d like an air gun to blow off the fixture and part before swapping in new stock but I have a little handheld electric blower that could work. If air it is only used by the power drawbar then perhaps I can find a quiet compressor.
I have yet to find a low speed spindle with an ATC-compatible taper for less than $4k, though. My best choice so far is an unpowered head kit driven by a geared-down motor connected by belt. That’ll need some sort of cooling, air or water.
The electronics cabinet will need fans and the steppers make more noise than I’d like when they’re rapiding but maybe I’m incorrectly thinking about the noise levels of the motors that move the heavy beds of the big machines.
My 25G 175PSI compressor is enough to keep up with kool mist, but I don’t think I’d use a pancake for it; it might exceed its duty cycle.
The belt-drive spindle motor on my PM-30MV is passively cooled. For low RPM units I think that’s normal. Here’s the kind of motor that it shipped with (2HP, three phase) not including a VFD):
It’s pretty easy to cut custom v-goove pulleys for whatever range you want to run. I did that for my Z lift; I just ground a lathe tool until it fit nicely in an existing v-groove pulley groove, measured the pitch, and then cut to fit.
The stepper controllers you found say that they are quiet. It looks like you expect to put the steppers inside the enclosure anyway?