Well, it's complete now. But I used 7i33TA and 5i24. Works great doesnt' need anything.

@@calphis also the ATC?

@@alexandrucosmin90 I haven't finished that part yet but have already inspected it. There is a belted motor that slides the carousel towards spindle, another that rotates it. There are two inductive sensors to indicate position and carousel index. The original main board of relays that activate the motors can be controlled with 5V I/o from a Mesa interface. So should be very simple to wire up.

If you go down the pins of the relay board ribbon connector you can isolate which one turns on each motor

@@calphis do you have also hi/low and rigid tap?

there is nothing unique about these machines that would cause them to be retrofit differently than any other LinuxCNC retrofit really. I recommend using a Mesa servo interface card. There is a gentleman who makes a retrofit hardware that takes care of a lot of the legwork. If you have never built a LinuxCNC machine don't view it as a trivial project though. Fair bit of work. forum.linuxcnc.org/39-pncconf/46987-gerber-408-retrofit?start=20#294504

in this video it was GT3 with NEMA23 steppers. but now I'm using NEMA34 and HTD 5M. 10mm pitch screws but would work much better with 20mm pitch

this is not Grbl. It's LinuxCNC

@calphis ahhh, well damn. So its macros

20mm

Thanks. I realized I hadn't posted in so long and had some old video on the camera I had yet to upload. So this is a little older. This also shows the vacuum workholding system. So all the key features of a commercial router system. I now have a 10HP regenerative blower with aluminum bleeder table on the other Gerber Sabre router (commercial unit I didn't build). This one shown actually outperforms the Gerber's by at least a factor of 2x holding power.

Only works for LinuxCNC

Unfortunately I don't think I"ll be continuing work on that project. I got into doing commercial router work and upgraded to a higher end spindle. The cost of conventional ATC is down to about $1,000USD for entry level units, so it just doesn't make much sense to develop something like the PMATC that requires so much effort and support for anyone other than myself who made it. But I'm still happy to pass along the resource files to anyone who wants to pick up the work.

Machine is controlled with LinuxCNC. I used Fusion 360 lathe toolpaths on this machine. Standard turning and profiling strategies. Then back to 3 axis mill strategy to cut flutes or relief.

The X axis appears to have a 1604 screw on this machine. One side of these screws gets a handwheel and the other couples to the motor, so pretty sure they are custom cut to length and machined. if you did not need the handcrank you could just use the machien (even worn) to cut a custom adapter spacer plate to use an off-the-shelf 1604 screw and work in CNC mode only (no manual). Good luck with your project. Have a lot of fun!

@@calphis thank you for your answer. Nice to hear that in case of need i can refurbish the screws on my small lathe. I have a manual with list of parts and exploded views. But detailed drawings of parts are not reported. Congratulation for ypur machine. It is almost sure i will come back to your channel.

Mine was very very high originally. I think in excess of 0.015" if I'm remembering correctly. It's currently under 0.001". I don't have quality metrology equipment nor the rest of thigns dialed in well enough to get a better more accurate measurement, but I suspect i'm somewhere around 0.0003" - 0.0005" now. I assumed my screw would be worn to have that much initial lash. But it turned out to have little to no wear. Screw wear would also be unlikely to be uniform but more loose in the middle and tighter on either ends (unless production cycles had been frequently run in those spots). so that might be a clue. If you do indeed have a worn screw and its not too non-uniform wear, you could try tilting the ballnut to take some more slop out (ie. shim on just one side so as to "bind" against the screw). Not something one would normally want to do , as it lessens contact/bearing surface and results in premature wear. but if it's end of life anyway, no harm done right. Good luck!

@@calphis Wow, 0.015" is a lot worse than mine. From what you're saying I think that replacing the bearings will give me what I need, I'm not doing utra-precision parts right now anyways. Thanks a lot for your response, I really appreciate it!

Thanks! Nice to meet you too. Looking forward to working with you again.

It is! Though I did upgrade from the version seen here as the machine started getting into production-like loads and the NEMA23 were not cutting it. Upgraded to NEMA34/12Nm about a month after this video, and to a 10mm pitch screw. The "compound nut pulley" part is still printed nylon. About 0.003"/0.1mm runout was the best I could manage what with thermal expansion and printer tuning (even with turning it on the lathe after printing, the gear teeth are not perfectly concentric to the center of rotation). This creates a little vibration and premature wear on the radial bearings. I would also suggest proper thrust bearings at this point if one thinks the machine would see a duty cycle more than hobby or intermittent. About every 40-80 machine hours I find myself tightening the plates or replacing those to take up slop from wear as the RS type bearings are just not quite keeping up with wear resistance under heavy cutting loads (i'm now running a 5hp spindle and pushing material removal rate as much as I can). The NEMA34 weight also causes the taller plate to want to sag under motor weight, and vibrate additionally, so I will recut the housing to be probably 1" Aluminum 6061. FloweringElbow on KZfaq has a similar design that is cast, I just tried for something quicker and cheaper to build (as I don't do casting here). The ShopSabre commercial routers also use this same drive system and you might find some videos of theirs. The current system runs at over 20,300mm/min and I am throttling it to give myself a confidence interval on reliability. I have run it at over 25,400mm/min with this machine's newly casted 400lb/180kg gantry assembly. So I'd say it's up to industrial standards with some considerations. Thanks for checking back. I'll try to post more updated info when I've organized things a bit.

@@calphis thanks for the reply wizard, I see a lot of people saying it doesn’t work or wears to quickly but your solution seems robust at high speed. A new video with the mods you’ve made would be highly appreciated! I have some bearings ordered and going to print the parts soon

Most commercial gantry mills (metalworking) and high end routers (like ShopSabre) rotate the nut instead of the screw. Thomson Ballscrews for example makes some high end rotating/driven nut systems for industry. They also have hybrid systems where both the screw and nut can rotate independently. What I've shown here is just an example of how one could make such a system at low cost for lower duty cycles and a shoestring budget. But industry can and already has far outpaced what conventional screw-turning systems can manage, by orders of magnitude. But there's definitely a gap between entry commercial and hobby grade machines, and those commercial 20-40HP monsters, where designs like these can live and serve well. Oh, one other thing I will note is that I did eventually have some of the nylon threading start to become compromised. Fortunately I formed all 6 holes and only started with 3 being used so I had some buffer to fall back to. This was due to vibration from aggressive cut spindle feedback. Heavy cuts -- 15mm deep and 8kmm/min feed rate in dense, wet wood. Also had some other looseness in linear bearings that was helping these vibrations resonate instead of damp, so maybe not an issue. But some threadlocker at those interfaces might be advised if you expect heavy loads. I figured the nylon itself would act as threadlock as locknuts are often made with nylon. I used Loctite Blue everywhere else but here and resolved by adding it here too.

@@calphis how did you mount the ball screws?

@@jamieclarke321 the new assembly was too big to fit where the version in this video was. so it was moved to the outside edge of the table, and some flanges drilled and welded to the table frame. one side foregoes the use of the original BK/BF## style bearing block and instead just uses a hole. this corresponds to the side that has the bigger block and the conventionally threaded section for the set nut. That nut is used to clamp the bare ballscrew end in this hole without allowing rotation. The other side has oversized mount holes that line up roughly to the screw holes for the BF## block. Those are oversized and allowed to float during initial fit up. Then run the axis down to the static side, and oversized holes for the entire assembly to the gantry allow that part to be tightened in place to establish spacing between gantry and that center hole. then run the axis down to the floating side and tighten those in place. I find tightening in place like this gives much better results than relying precision cutting parts alone. The only tough part about this technique is you do have to precisely cut at least one dimension. for my setup it was the spacer used for the mount between the driven nut assembly and gantry. so that and the center point on one side are fixed, everythign else is floating and tightened in place. wish KZfaq would let me post pic links here.

It's not a file, but a series of files and subroutines that right now that require LinuxCNC and user must be very familiar with LinuxCNC custom macros system to make any real use of it. I'm working on making the files portable to run on other people's linuxCNC machines, and easier to setup, but I'm not done yet. No plans to port to any other controller like Mach3. not sure if it would even work.

This is made with LinuxCNC. Not sure if something like this is possible with Mach3, but with LinuxCNC its quite simple. No need to track the orientation of the spindle. Instead I just put the spanner on a linear guide/rail that both guides it towards the spindle, and on either far end of that travel there is a microswitch. The spanner is pushed forward with an electromagnetic force from a solenoid, so its a steady pressure but wont crash or bind. The socket rotates slowly while the wrench pushes into the spindle. when it reaches the right orientation, the spanner is able to naturally mesh the flats. This allows the spanner/carriage to move forward enough on the linear rails that the switch on that side trips. When it's reverse cycle and the spanner is completely free, the other side's switch trips, and the system knows its safe to move away

in the last video sequence its cuttingn the piece of plastic in question. It's basically a zig-zag with 0.5" stepovers and a 0.125" radius fillet at each step.

nothing that you couldnt do with separate bits, but the goal was to make a single bit to simplify this to a single setup and step.

This is just a typical Chinese made 24krpm 4 bearing spindle with a VFD. This is a 65mm diameter body variant. this one happens to be 220V 1.5kW. This approach will work with any spindle capable of low rpm like 100rpm or so.

Huanyang vfd or its many clones will work

Probably not as much as it might seem at first glance, but yeah, a fair bit of work. I was a software engineer in a past life so the programming part was easy. The mechanical parts were mostly already on the shelf, and I incorporated design elements from 3 or 4 homebrew ATCs I liked. Have recently rewritten the software so you can just change variables to match your own table's setup/positioning, and about 75% complete redesigning the parts to be off-the-shelf or very cheap for the 2 custom parts (a few dollars, not a few hundred). Hoping late 2022 it will be ready for the public to adopt on their DIY/hobby machines. It will still be a bit of a timesink to get running, but I will have done 90% of the heavy lifting in terms of design and guidance. The adopter will just need to spend time figuring out how to mount on their machine and tweaking the positions, putting the configuration and scripts in the right place, etc

and thank you so much, and thanks for stopping by :)

Thanks for the update. I'm not certain some of the slop i had to take up wasn't the fault of the garbage bearings I tried, so I'll update the video description to the bearing you mentioned, hopefully save some others a headache

It's LinuxCNC based.

Thanks! Yep, you could also just use a whole piece of flat stock next to the whole tool rack and mill it down in place as the Z sensor. I was tryign to strike a good balance between saving table spce where I'd already used quite a bit on the rack itself. In theory one could semi-permanently attach the tools to its collet and just set tool height once, but I prefer this approach as I can change tools in each pocket without telling the system and it will automaticaly figure it out. It's more about flexiblity, reliability, and automation rather than speed for me. Thanks for stopping by!

Thanks for stopping by! Things are busy around the shop but I will do my best to get this out to you guys as soon as possible.

With encoders already in, you should have a fairly easy time of it. the PNCconf wizard makes it really easy if you have servo amp boards that already have gain/current adjust pots on them. In the wizard you just start running a slow speed, check the position feedback, then adjust the trim pots until the actual measure speed and commanded jog are the same. move speed up a bit higher and repeat until you reach the top speed you plan to use. I'll make another video on this process as well as soon as I'm able. The LinuxCNC folks have done a great job on the core, this wizard, and the servo tuning system. They've even documented it very well on the forums. It's really not nearly as complex as it sounds when described verbally. For some folks like me a video might help clear things up. Thanks for stopping by. And you are welcome to drop by (and anyone else for that matter) to check the machine out if you ever make it out this way

@@calphis you make the internet a better place.

@@jensparrow27 Thanks so much! That's high praise!

@@calphis Thanks, did your machine have inch or metric ball screws?

@@MotorsportDiesel Mine had 5tpi or 0.2" pitch all around

Thanks! I'll do a more detailed walk through soon, with special attention how the Mesa boards are wired, etc. Probably after the tool change is also converted. Might try to incorporate LinuxCNC servo tuning instructions into it as well. There are good written instructions for that on the forums, but I think a video could be helpful to many folks.

Well I don't delude mysef in thinking I'm actually getting 3 tenths positioning in the real world, but the motors commanded angle should be doing their part. In reality the ballscrews have lots of wear at this age (and up to 0.002"" unevenly across travel) and there is considerable backlash, similar to traditional leadscrew. 0.008" on Y axis, 0.024" on X if I remember correctly. Could be some tightening on the ballnuts is needed, and the Lovejoy couplers might also be to blame. I have gotten the backlash compensation values in the neighborhood, within 0.001", but it will take a lot more precise measurement and data entry to home in under that. The actual travel for the X axis can almost squeeze out about 21.7" without having moved the hard end stops. The factory setting soft limit is 20" and I keep it at 21" because the wear is uneven as it has never really been run in those extreme values and would bind and require slower speed out at the ends unless I scrape things evenly, which I may eventually do anyway. From what I can see I think I might be able to get perhaps as much as 23" if I moved the hardstops out a bit more. The Y likewise gives me 16.75" between hard stops and I run it at 16.25" soft limit right now. Z I run at 23" which is probably as high as it will physically allow. The spindle is only about an inch away from table at full depth, without a tool, and at the top it seems to be backing up to the nut. I have also toyed with the idea of adapting a bigger table onto this machine. 16.75" Y would still be usable for a lot of things if I could get 36-40" travel out of X. Maybe find a table from another larger machine and adapt it to fit. There's a lot about this machine that I like that would cause me to retrofit it rather than get a bigger machine. Namely, the weight, frame rigidity, power requirements are low, etc etc. well organized and well thought out design, a real joy to work on. Getting ready to add in an improvised 4th axis from an old mini lathe headstock as well

I revisited this, digging deeper. It turns out the ballscrews themselves have very little wear, under 0.001". The slop was coming from the thrust bearings and issues with the gibs. I was able to clean everythign up and tighten it and get under 0.001" backlash on all but Y. Y needs a new thrust bearing. You can get them for about $40/pair (each axis has two preloaded to take up lash). It needs a precision shim, you can make your own if you keep shim stock on hand. the info on ultimate travel is still somewhat accurate. I could take off the Bellevile washers to get more travel, but that area of the ways seeing less wear means everythign gets very tight out there and I'd have to really slow down the machine or scrape the ways.

@@calphis I’m literally in the same boat! I need to replace the Y thrust bearing….making a clicking noise with every revolution. If you happen to make a video of the repair…I’d be ecstatic! Thank you.

I am in a little rural town about a hour away from Franklin/Nashville, TN. Though if you're not anywhere nearby I'm happy to make another video showing how I did the conversion, what hardware was used, and wiring it up, etc. In fact, I was planning on doing so anyhow. There is only one part I had to make -- the rotary encoder mounting adapter for the drive motor feedback. It's just a disc of aluminum with 3 critical dimensions then drill the 3 mounting holes. It could be lathe turned manually with not much problem. You could probably even 3d print if your machine is dialed in well, it's not under high mechanical stress other than the rpm and the coupling should be done with a flexible coupling anyway due to the high rpm. Fusion 360 design link: a360.co/3HE9aM9 I'll make a rough video soon with just an overview at least but long term I'll make a more detailed walk through of that machine conversion. Thanks for stopping by!

Oh, and this is the encoder I used that fits that design. It's a cheapo one, but it works fine and the accuracy level is more than acceptable for the kind of work I do. Based on the leadscrew pitch on my machine at least, my math shows I can get 0.0003" per pulse. www.amazon.com/gp/product/B07MX1SYXB/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1

@@calphis I actually spent a bunch of time in Shelbyville, South of Nashville. I'm in Dallas. I would love more videos, but once you have the ATC working I really would come visit. I can fly into the nearest municipal airport. I've subscribed to your channel so I'll reach back out as you get further along and then try to come see what it really takes to do the conversion to LinuxCNC.

@@calphis Great work you have done here :) I would love to see a bit more details as at some point my VMC will need a retrofit too for various reasons. Any chance of showing another video with details?

Thanks so much! That's high praise!

thanks!

1. I'm using a 1.5kW 24krpm 65mm body spindle with a Huanyang 220V 1.5kW VFD capable of 400Hz top frequency. ER11. Listed as 0-24,000 rpm, more on that below. 2. Most of these spindles are capable of any rpm up to the 24krpm,, including as low 50 - 60. The sellers tend to list them at 7, 8, or 9k rpm minimum because the torque curve is very poor based on how the VFD supplies power, (just pulsing a fixed current at a proportional duty cycle of the requested rpm) so they may only have 1/3 or less of the rated power at this speed. If your spindle VFD comes preprogrammed with a higher minimum, you CAN reprogram it, just don't try to cut anything below the suggested rpm range. for PMATC purposes we are just using it to thread on the collet nut to pick it up, not tighten it or use any real torque, so it's OK. As always, be very careful about any settings you change on the VFD as you can absolutely burn out the motor if you set something wrong. Under no circumstance should you change the "rated frequency" or "max frequency" settings (usually PD004 and PD005). "rated freq" in particular can burn out the windings in a matter of seconds if its not set to the proper 400Hz setting, as this tells the VFD to supply full amperage at a lower pulse rate, which equates to muliplying the current output basically to a level the motor windings cannot tolerate. 3. Amazon and Ebay both seem to run in the $240-280 range. I recommend the simplicity of air cooled if you don't foresee lots of long hours of heavy production usage (typical hobby or engraving/precision type work). AliExpress sells them cheaper but you will end up with the same amt spent after shipping. These are basically all the same handful of motor makes and VFD for the most part just distributed by different sellers. I'd get the Huanyang VFD (the most common) as there is already Modbus support for spindle control with LinuxCNC for this VFD. Modbus is a serial interface you use with a USB adapter for a few bucks, and 2 wires connected to the VFD to get speed control very precisely and feedback on spindle-at-speed and stuff like that. Thanks for the interest and questions, as I'm sure others will wonder the same things. When the system is ready for public release I'll make a video both showing the newest revision running and likely combine it with a How-To for implementation. You're going to love LinuxCNC. There is a bit of learning curve depending on what you're trying to do, but if you're patient with it and yourself, it's not too bad. There is literally no limit to what you can do with it, and the community is beyond helpful. My 6,000 lb vertical machining center runs LinuxCNC as well. It's totally capable of industrial grade work, perhaps even preferable due to being open. In that regard, it's a great thing to get into, knowing you will never outgrow it.

@@calphis Thanks for the answers and encouragement! Especially the golden nuggets about Modbus and such. I've ran HAAS machines in college, but looking to get a steel-cutting machine put together on recent-graduate's budget. Putting most of my budget towards the machine frame/motion system. Being frugal with closed-loop steppers and a cheaper spindle solution for now until I can justify a proper Bt30 ATC spindle for some future light production work. Won't be spending anything though until I've learned enough LinuxCNC to confidently tinker. It's a-ways out for me, but when I iterate on your concept I'll share a link if I think it has any additional functionality. Until then, looking forward to yours! Happy Tinkering!

Thanks! I'm hoping to make a final design pass on this early this year and have the design available for public use.

@@calphis Can't thank you enough for that!

thanks!

This is just a regular "run of the mill" 1.5kW 24krpm 65mm diameter air cooled spindle from China, with a Huanyang 220V VFD. no reason it wouldn't work with a 120V or a 400W for that matter on a 3018. it works with any spindle where you can get rpm control and FWD/REV control.

Yeah it took a few days to put it together and several hours here and there later tweaking the motion to fine tune it and get it working reliably. Still worth it for small work. Kress has a 1kW system that's $1,000 for just the ATC mechanism. you still have to bring the spindle and pneumatic controls and spindle drive. You can get one from China for $900, but again, that's not the actual cost. You end up spendign minimum of $1200 if you already had all the pneumatics and a VFD. Kress makes you buy their toolholders about $25 a piece, and lots of ISO30 toolhodlers is not terribly affordable for a hobbyist or a small shop just starting out. In theory when this design is finished being ported for other to use, they should be able to build it themselvse for about $150, and tool holders run about $8/pc (ER11 collet and nut), and no limit on number of tools

Especially if you use a 1.5kW that can be had iwth the VFD for about $160, it's a very attractive option in terms of price comparison

@@calphis Yea, I've made the mistake of buying the Kress 1050 FME-1 spindle with this weird kress collet standard when building my homemade CNC. Next time, when building one at work for my maintenance workshop, I went with 1050 FME-p which has ER collets. Also Kress was no longer Kress. They changed their name to AMB or maybe were bought out or something. I don't know. The point is ATC is not an option for me. At least not for my homebrew budget.

Thanks! How kind!

I am using the stock spindle with an MT2 Morse Taper 2. The literature says they are also available in ISO30. Not sure if this is is just for the ATC variants or manually changed units as well. It handles a 1/2" roughing end mill pretty well on mild steel, despite the fact that it's just an 800W spindle. On the back of the column the spindle electronics/driver is mounted. If you open the cover and run it with manual dial you can isolate one of the screw terminals as being a variable voltage signal based on speed. This can be used with a standard PWM driver to control spindle from LinuxCNC, Mach3, Grbl, etc. There is also a trim pot used to fine tune, and on mine was able to use this to overdrive the spindle a bit. I think I'm running 3800rpm max. I'm driving with DM556 here (5.6A peak) though I think DM542 (4.2A) should also work fine for double step as they are 3.0A steppers. In this arrangement they handle 80in/min very well and I could go to perhaps 120 or even 150 if I lowered the accel, but I personally favor accel over top speed. If KZfaq will let me paste a DRive link here, here is the documentation for this unit, including the stepper motor sheets. drive.google.com/drive/folders/1rpQgxI9LF2YIiHeDT-i_ZQw04x_TWNXD?usp=sharing

Worth noting that these spindles can be rotated to an index/lock position. You probably want to take some time with a dial early on to make sure it's trammed before cutting. I spent some time chasing phantoms at first because the head was slanted about 2 degrees and I didn't realize. :) You can turn the head to any angle, including 90 degrees and use it as a CNC Lathe with a tool in the vise. there some MT2 chuck adapters out there or put stock directly in a collet. The encoder built in is a little too coarse for threading (and it wouldn't have the torque anyway) but good for turning/facing small diameter parts like pins and such.

thanks!

thanks!