|
[this post got kinda dumb as i wrote it but bear with me] For PC : CNC to be as clean a comparison as that I think you'd need some very interesting developments in manufacturing. Navigating with commands had a very strong impetus for being GUIed out of existence, but playing with the gcode numbers isn't going anywhere anytime soon even if you make it as glossy and painless as possible. I think to really abstract the code itself, machines need to be a lot smarter than they currently are. Most everything that I've heard of is dumb and blind to the work itself outside of specific areas like stock probing and tool length setting, and making the end result match the CAD is generally up to the operator and those numbers. This is basically spec fiction for all the knowledge and expertise I'm wielding here, but a magnitude better sensing could probably let the machine do a lot of that finicky stuff. Imagine high-res laser imaging, something adapted from self-driving car developments maybe, producing a 1:1 digital model of the program being run instead of just its imagined version of what's happening. Deviations from the CAD model are observed milliseconds after they happen, maybe even adjusted for on the fly before tolerances are exceeded, and if a part is junked for a clear reason the program could be tweaked automatically for the operator. Torque, cutting forces and chatter already get sensed by machines, even automatically adjusted for in some cases judging by stuff I've seen in this n the metalworking thread about chatter compensation. Give em better eyes, ears and hands and the code will get farther and farther away, but without that I think some pretty big barriers exist to doing more than putting a veneer over it. e: that said, all that sounds like a particularly expensive and difficult way to cut labour costs and increase productivity, that being the only reason anything like this happens and all. I know iirc Volkswagen was doing a trial balloon thing with the only non-dumb Google Glass application I've heard of, putting headsets on assembly line/warehouse workers so they could watch a screen and do something needing their near-full attention simultaneously. I can see stuff like that showing up real soon to significantly increase the number of machines each operator can run, for example. Ambrose Burnside fucked around with this message at 01:11 on Dec 18, 2017 |
#
?
Dec 18, 2017 01:04
|
|
|
|
| # ? May 17, 2024 08:29 |
|
|
You design parts for manufacturability, once you can manufacture a more complex part with less fixturing, then you'll do it to save setup and increase accuracy. People are doing 5 axis not because of crazy contours (some are) but because it can save a setup. The programs it spits out are not the sort you can troubleshoot at the machine. So you get a downward spiral where your operators need less knowledge to run the part while the software gets more complex. It's certainly much easier to use CAM than it was 10 years ago. Plus you get better results. At a certain point you'll solid model a part, upload it to the Kennametal cloud, pick your machine, and it'll spit out the most optimized code. But the dude on the shop floor needs to know even less. I toured a factory a few weeks ago that put raw motor castings in one end and the other end spit out fully assembled engines without a single person touching it. The machining centers were all linked with ceiling mounted Fanuc robots who shuttled the pallets from one 5 axis machine to the next. It was incredibly impressive to watch it all work. The level of expertise in the field right now is mind blowing, and at the same time distilled into fewer and fewer people than ever before. This isn't isolated to that one shop, a few other manufacturers I visited were dropping the customized production equipment and installing some truly monster 5 axis machines. At the end of the day it'll mean you can make cooler poo poo, even on a hobby cnc, with less knowledge of what's happening under the hood.
|
|
#
?
Dec 18, 2017 02:02
|
|
|
This is going to get really nerdy and is only tangentially related to the thread. Sorry about that. Yooper's basically got this right: what all this does is transfer the intelligence further back in the chain. But it's nothing new. First the knowledge left the shop floor and went to the programmers when CNC got installed, then it went to the post-processor once CAM got good, now it's going to the CAM providers since they're making the post-processors. Even setting up that automated production line you're talking about? They probably hired another company to do it. There aren't many companies that have the ability to do that in house, they've transferred that intelligence to the distributors, robot, and machine tool companies that do turnkeys and integration work. I work on the machining-side of that sort of turnkey work. Our entire job is handling all of that process intelligence and handing the customer a black box solution. What I see is that there's a lot of processes and automated tools that are common-place in the sort of high-volume production work that I do that are getting simplified and more accessible for job-shops with smaller production runs, allowing sources of error to get removed and further out-sourcing process knowledge. It's not going to take too long for some of it to get to the point where it's accessible for hobby machines. A lot of them are the sorts of things Ambrose is talking about, albeit simpler. A few examples: * Obviously, CAM is getting better, vastly better. With more advanced simulation software like Vericut, you can actually be pretty drat sure that the code will work without needing to prove it out manually. Obviously doesn't account for tool length and part setup, but that stuff's getting automated too. It can already be pretty close to a black-box for simple programs. I do a ton of manual edits and optimization, but my job is pretty specialized, most people are going to be just fine with the CAM output provided they've got a good post-processor (which nowadays is just provided by the CAM company or distributor.) * Tool selection is getting more automated. How we do tool selection is a bit different: we just tell the tool vendors to give us the best thing they've got, they figure out the design, make whatever custom geometries we need, etc. Now job shops can largely do the same thing. Open up Kennametal's NOVO, or whatever Sandvik call their system, tell it what you're doing, and they'll recommend a tool. It can even handle customs. No need to get a whole bunch in house to do the testing or even hunt through catalogs, the manufacturer does that for you. Again, it's transferring the process intelligence back in the supply chain. * Machine setup is getting automated and checks are being added. Lots of machines can check tool lengths prior to actually running the tools (some of them can even do it while the tools are in the magazine so it doesn't impact cycle time.) That avoids crashes by making sure that tools are set up to the correct length. You could do that on even hobby machines pretty easily, just add a whisker switch or something, doesn't need to be a full-on tool probe. Coupled with automated tool-setup, and communicating the offsets to the machines via network or RFID chip, that variable can be largely eliminated. You can have a nice optical presetter that will refuse to let you set a tool up to the wrong length for... maybe $10k these days? Also, if you don't have it in your post processor already: output code comparing the tool offset to the nominal length, something like IF[ABS[#10001-120.]GT0.5]THEN#3000=1(TOOL 5 LENGTH OFFSET INCORRECT). It will save your bacon some time. * Datron's introduced a camera-assisted setup tool on one of their machines for probing routines. A number of manufacturers are introducing on-machine 3d simulation using actual tool lengths and workpiece models, some of them are even tracking machine motion live during the cuts to predict when there'll be a crash. https://www.youtube.com/watch?v=jzBM_lchKtg * Closed-loop production with automated gaging and offset input is getting cheaper and easier to do, Renishaw's Equator comparator, for example. MasterCAM even has a built-in probing system which can do measurements and make adjustments automatically, no need to mess around with macro code manually. * I haven't used them personally, but we do have tool-deflection sensors that can compensate live for that during a cut, and you can probably implement the same thing with one of the smart force-sensing toolholders you can buy these days. I think this is a more realistic expectation than continuous laser measurement of the part. There are machines right now that will take a picture of your tool for length verification. Use that to get the tool form, then you can reasonably estimate and compensate for deflection continuously from force measurements in the spindle or fixture.
|
|
#
?
Dec 18, 2017 02:48
|
|
|
Ambrose Burnside posted:Oh I know, but knowing enough to troubleshoot your own generated gcode seems like a reasonable basic expectation, and in our case even that's generally not there. Just a heads up you will spend your life saying you have an engineering tech degree when people ask if youre an engineer/what you studied and other engineers will ask. If its not much extra time I highly recommend a regular ol' mech eng degree. Not industrial. Not manufacturing. Mechanical. Or Electrical if youre feeling more ambitious and want to really go in to automation. As for on machine inspection and finish machining with adjusted values, I could do that on a Fanuc robodrill with a probe, a sub $100k machine all in. The tools were set by an optical tool setter that could also spit out 2D drawings of new/custom tools. That said, you'll find environmental conditions like vibration and coolant limit some of your on machine inspection options in mass production. I think Neural networks of ____ measurement and offset on the machine (which is networked) to a CMM measurement of that feature are a trivially easy next step that could both be high tech and increase precision. CarForumPoster fucked around with this message at 14:27 on Dec 19, 2017 |
|
#
?
Dec 19, 2017 14:20
|
|
|
CarForumPoster posted:Just a heads up you will spend your life saying you have an engineering tech degree when people ask if youre an engineer/what you studied and other engineers will ask. If its not much extra time I highly recommend a regular ol' mech eng degree. Not industrial. Not manufacturing. Mechanical. Or Electrical if youre feeling more ambitious and want to really go in to automation. As an electrical engineering technologist, this is exactly right. Don't do what I did.
|
|
#
?
Dec 19, 2017 18:34
|
|
|
ante posted:As an electrical engineering technologist, this is exactly right. Don't do what I did. I almost did the exact same thing until someone gave me the advice you and I just gave. Engineers and the people that hire them are a snooty bunch, even the nice ones. Their view: the effort you put into your training =the effort you will put in once youre there. There is a hierarchy of degrees that no one quite agrees on but goes roughly like: Business Degrees = Web Developers/"Coders" < Industrial/Manufacturing Engineers < Mech/Elect Engineer Techs < Civil Engineers = Environmental Engineers < B.S. Physics < Mechanical = Aerospace < Electrical = Computer Science < Optics < Chemical < Nuclear Also the average pay scale roughly follows this as well.
|
|
#
?
Dec 19, 2017 18:53
|
|
|
CarForumPoster posted:I almost did the exact same thing until someone gave me the advice you and I just gave. Engineers and the people that hire them are a snooty bunch, even the nice ones. Their view: the effort you put into your training =the effort you will put in once youre there. Not to get too off-track from the extremely good CNC discussion, but I was chem eng and we'd walk around like we were hot poo poo cuz our problem sets involved fugacity and Fick's Law and other nonsense. But 1) plenty of chem engs rode gentleperson's C's all the way to their graduation handshakes (and then usually MBA's), 2) only a kept group of like 100 engineers in a multinational chemical corp do the core engineering, 3) you can open up a random chapter of Perry's and see what a young science chem eng still is, most of our bedrock tools exploit simple properties. "Chemical reactor" sounds fancy except they are usually just big blenders. We get paid more than say, environmental engineers, because destroying the planet is more profitable than saving it. Any one of those disciplines will ask every drop of what you've got, if you let it. Okay, except business degrees. As somebody learning this machining trade in his free time: if an engineer tries to intimidate you, he's covering up for ignorance, this poo poo can be as hard as a kinetics problem when you dig into it.
|
|
#
?
Dec 20, 2017 21:22
|
|
|
Mofabio posted:As somebody learning this machining trade in his free time: if an engineer tries to intimidate you, he's covering up for ignorance, this poo poo can be as hard as a kinetics problem when you dig into it. Yeah, but to get that nice chatter free finish on the turbine disc dovetail tree thing, you employ a black art and years of experience. To solve a kinetics problem, you employ a black art, years of experience, AND a shitload of math. Also chem engineering is a black art that nobody really understands. If you do, or at least make enough correct guesses consistently enough, you get a fat 200k+ a year job and get a huge pile of very similar problems.
|
|
#
?
Dec 21, 2017 00:57
|
|
|
What do you guys think of the future of machine shops that are all manual? Excavator repairs / whatever. Will their always be a market for repairing tools quickly on manual machines, or will they end up just going back to the manufacturer for a swap out part?
|
|
#
?
Dec 21, 2017 01:36
|
|
|
There’ll always be commercially-viable toeholds for manual machining, given the right requirements. Those toeholds are just gonna end up real small. Personal example- I used to work in a prop shop that did stuff for film/TV and stage productions, which generally involved fast-turnaround one-off commissions or small production runs drawing on many manufacturing techniques and in a huge range of mediums. A fair proportion of the jobs we did involved machining at some point, but it rarely got past “mill a slot however deep to accommodate the sword tang with no play” or “turn some tubing down on the lathe until it’s as long as this oddball bit of glass for a prop”. I can think of one or two jobs where CNC would have been great- we had to farm out our engraving work, for instance- but it wouldn’t have added anything or sped up production in the large majority of jobs we took. (fun tangentially-relevant fact- artisanal blacksmithing as a real trade in the West was pretty thoroughly eradicated by industrialization, and yet every WW2 military power fielded large numbers of blacksmiths, the US Army going so far as to have a standardized mass-produced Army-issue portable coal forge. More efficient manufacturing approaches put them out of their old occupational niches, but there was no replacement for their particular abilities when those cheap, consistent replacement parts take weeks to get there and more modern manufacturing options were decidedly non-portable.)
|
|
#
?
Dec 21, 2017 02:51
|
|
|
Mudfly posted:What do you guys think of the future of machine shops that are all manual? Excavator repairs / whatever. Will their always be a market for repairing tools quickly on manual machines, or will they end up just going back to the manufacturer for a swap out part? As someone who works in industry, there will always be a place for machine shops. Industrial machinery goes obsolete so much faster than it ever used to. You buy a machine, and manufacturers stop supporting it just a few years later. You have to have parts repaired or replaced while keeping your costs down. Paying someone to retool to make a part isn't always cost efficient.
|
|
#
?
Dec 21, 2017 02:57
|
|
|
Mudfly posted:What do you guys think of the future of machine shops that are all manual? Excavator repairs / whatever. Will their always be a market for repairing tools quickly on manual machines, or will they end up just going back to the manufacturer for a swap out part? It'll just become more and more niche. One example is the OEM's are getting into a ton of repair that would've previously been done by dealers/locals. The OE's can do it cheaper, at a higher quality level, with a warranty, where the previous folks couldn't. The niche usually being it's either so big, it's not cost effective to haul it back to OEMland, or as EvilBeard says, the OE has moved on to greener pastures. One local shop here still uses almost exclusively manual equipment but the owners creed for machining work is "if you can lift it, I won't bid on it." All he does is massive stuff. 80 lbs is about where his work starts all the way up to massive ore boat and mining stuff. His shop is flourishing and he can't hire enough people. We used to have this retired machinist do a lot of piece work for us. He did it basically because he enjoyed it and got bingo money or some poo poo. You'll always be competing against a dude like that unless you find a nice niche.
|
|
#
?
Dec 21, 2017 03:45
|
|
|
Ambrose Burnside posted:There’ll always be commercially-viable toeholds for manual machining, given the right requirements. Those toeholds are just gonna end up real small. Ya know I was thinking the other day how portable 3D printers could be really useful for the military to fix poo poo around camp/FOB that breaks. Blacksmithing for 2017. Could also do things like printing small drone bodies and shipping in a crate the electronics and motors to be installed on site.
|
|
#
?
Dec 23, 2017 05:24
|
|
|
I'm a programmer irl and I was interested in gcode for a minute but its such a garbage language. It's obvious that it was always intended to be generated and not written by humans.
|
|
#
?
Dec 23, 2017 07:28
|
|
|
like I would be super into something like 3d postscript but nope
|
|
#
?
Dec 23, 2017 07:30
|
|
|
oops
|
|
#
?
Dec 23, 2017 08:07
|
|
|
When I took a machining lab in college the instructor told us that learning a bit of gcode is good for troubleshooting* but not really worth the trouble of directly programming in. *like for “why isn’t the coolant starting, oh there’s no M08 at the start” type issues.
|
|
#
?
Dec 23, 2017 19:23
|
|
|
When it comes to cnc machinery porn, could we have a PYF? I'm sort of joking. I saw a Mazak Integrex the other day and I was wondering just how advanced these things get. Is there a best in the biz?
|
|
#
?
Dec 24, 2017 10:13
|
|
|
rotor posted:I'm a programmer irl and I was interested in gcode for a minute but its such a garbage language. It's obvious that it was always intended to be generated and not written by humans. Yeah I'm learning to code on gcode and openscad and realize I'm probably not doing myself any favors... LinuxCNC has o-words though (not sure in FANUC etc does too?) and you can write nice(r) clean(er) lathe gcode that way.
|
|
#
?
Dec 24, 2017 18:18
|
|
|
What do you mean o words?
|
|
#
?
Dec 24, 2017 18:41
|
|
|
Mofabio posted:Yeah I'm learning to code on gcode and openscad and realize I'm probably not doing myself any favors... LinuxCNC has o-words though (not sure in FANUC etc does too?) and you can write nice(r) clean(er) lathe gcode that way. Jesus gently caress, don't do that to yourself, neither of those are real programming languages, not even kinda. If your goal is to learn to code, and you're teaching yourself, I'd strongly recommend python. If your goal is to machine stuff, then learn cad/cam. Fusion is good and free. Openscad is pretty cool, but it's hard to do any substantial projects with it.
|
|
#
?
Dec 24, 2017 21:14
|
|
|
There are really cheap servos on alliexpress with good reputations for working. The only thing is they have 2500 ppr encoders, some say these are insufficient, you need 20 bit or 17 bit or something. Is this true? Example servo: https://tinyurl.com/ybrww2t2 I am also stuck on the issue of paying $150-$200 more for servos just so I could get name brands with tuning software.
|
|
#
?
Dec 25, 2017 05:29
|
|
|
CBJamo posted:Jesus gently caress, don't do that to yourself, neither of those are real programming languages, not even kinda. If your goal is to learn to code, and you're teaching yourself, I'd strongly recommend python. If your goal is to machine stuff, then learn cad/cam. Fusion is good and free. Openscad is pretty cool, but it's hard to do any substantial projects with it. Are there any FOSS lathe CAM packages for Linux? The only one I've found is NativeCAM for LinuxCNC, but in practice I usually just end up writing the code, or modifying a pre-written o-word subroutine. There are some poorly-maintained mill packages, last I checked, but not much for lathes. My OpenSCAD stuff... everything is polyhedra so all the triangles are generated directly. It's just a big math project that doesn't really take advantage of any CSG or primitives. I might learn python/C++ just to generate my stl files directly but that's the only reason I'd have to cross paths with a 'real' language. That or kicking the tires in LinuxCNC's C++ when I add a lathe c-axis and make it mill-turn :P.
|
|
#
?
Dec 25, 2017 07:06
|
|
|
There's pycam. I've found it useable for engraving, but very slow to do anything actually 3d with.
|
|
#
?
Dec 25, 2017 07:22
|
|
|
Mofabio posted:Are there any FOSS lathe CAM packages for Linux? The only one I've found is NativeCAM for LinuxCNC, but in practice I usually just end up writing the code, or modifying a pre-written o-word subroutine. There are some poorly-maintained mill packages, last I checked, but not much for lathes. Ah, no, CAD/CAM is one of the last things I keep a windows box around for. All the linux ones I've tried have been hot garbage. Hopefully FreeCAD will become not-poo poo at some point, but the last time I tried it, it was still unusable compared to Fusion or Solidworks. I suppose if *all* you're doing is polyhedral primitives, then you're the exact one and only usecase OpenSCAD is worth a drat for. Learning to program is a pretty drat useful skill, even if it's just writing little python scripts to do little BS jobs for you.
|
|
#
?
Dec 26, 2017 02:54
|
|
|
Anyone know much about the advantages of closed loop steppers, other than the obvious? I posted these questions on cnczone and got no replies. "If I buy an 12nm closed loop stepper, like a leadshine closed loop clone, can I use 1/2 microstepping and not go down to 71% torque, or 38% for 1/4 step? Do they work differently, or is torque simply caused by rotor position? If I buy a closed loop stepper, can I have less of a 'reserve torque' and thus probably go down to a smaller stepper motor? Perhaps a 8Nm where 12nm are commonly used (e.g. an 8Nm motor for the z-axis on an rf-45 clone)? Not related to the above question."
|
|
#
?
Jan 3, 2018 21:53
|
|
|
I'll take a shot at an answer. I think it'd depend on how much torque you're expecting. In an open-loop stepper, if you're on a microstep and your tool is repelled by the workpiece, creating a torque higher than the microstep torque but lower than the holding torque, then it will change position back to the full step position. In a closed-loop, the same thing will happen, but the controller will know and attempt to compensate. However, I'm guessing it won't automatically lower the feed rate (ie change the gcode), so the repelling torque on the motor will remain the same. Maybe it will oscillate position between the full step and microstep, until the programmed position advances to the next full step. It depends on how the control loop is programmed. Holding torque on microsteps is a physical feature, where the fields from the stator holds a position on the rotor between its physical ridges. It is mainly useful to smooth motion, not enhance precision. So to answer your question, torque is a feature of rotor position. If you want higher precision, get a timing belt, or gears. edit: just to correct some things from this post after more research: the torque at first glance is a function of rotor position, but more completely, torque is a function of rotor position relative to the stator field. There is an ideal stator field for every rotor position, where the field 'leads' the rotor slightly, causing motion. But the stator fields pretty much immediately run behind their 'ideal', which is only really achieved when there's no load on the motor. Closed-loop steppers allow the field to more closely match the ideal field. There is still a sine wave of torque between steps, and holding torque is still max torque, but closed-loop lets the half-step minimum torque be higher than in open-loop. Mofabio fucked around with this message at 21:06 on Jan 7, 2018 |
|
#
?
Jan 4, 2018 19:09
|
|
|
Thanks, that's kind of what I expected. Tormach do use 3 phase steppers (1.2 degrees per step) on their mills which is 0.0167mm or 65% of 1 thou (I dont get your foreign system but there you go). My lead equates to 0.025mm / 1 thou per step. Maybe if accuracy is what you crave 3 phase steppers and a ballscrew lead like 5mm might be the go (?)
|
|
#
?
Jan 5, 2018 06:37
|
|
|
The best of all worlds is a gearbox or, simpler, a bunch of $10 GT2/GT3 timing belt pulleys from Misumi. I'm doing a lathe c-axis project this year, for a mill-turn machine, and am reducing a 400-step/rev stepper with a 72 and 8-tooth GT3 belt setup (9:1 reduction) to get 400*9 = 3600 steps/rev on the chuck, 0.1 degree per full step and 0.01 per microstep (using a Gecko G540). Lots of accuracy and lots of torque, and not a lot of money. I think both pulleys, the belt, and the stepper put the project cost at around $100, maybe a lil more if I can't find good scrap angle to make a motor mount. edit: no investigation, no right to speak  I went and researched and perhaps your hunch was correct--quote:
This is a big ad so, grain of salt, but info could be useful: https://www.motioncontroltips.com/torque-better-efficiency-closed-loop-versus-open-loop-stepper-systems/ Also this white paper: http://www.galilmc.com/news/servotrends/white-paper-closed-loop-stepper-motor-performance-gains Mofabio fucked around with this message at 18:54 on Jan 6, 2018 |
|
#
?
Jan 6, 2018 10:47
|
|
|
Well, I priced VFDs and 3 phase motors for my spindle and they came out marginally cheaper than a servo. So I bought a 1.8kw AC servo for $500. Then I thought "why not just buy servos for everything?" Aliexpres responded with a deal on 750w servos. Literally cheaper than every other servo on alliexpress currently, including smaller units. I can re-use these on my bridgeport one day, and put the 1.8kw servo on the knee. The 3 servos are about $100 more expensive than Leadshine steppers, when you factor in the DC power suppies leadshine offer. Buyer's remorse setting in but I think this will work. Can anyone recommend a cheap motion controller for 4 servos? I have seen some on alliexpress. I would like to start with Mach 3, though LinuxCNC might be a smarter option in the long run. Also, should I buy a 5 axis or 6 axis controller if I plan on making a tool changer (which I do). Or can you handle tool changing externally to a program like Mach 3. I expect I'd need a servo or two for the tool changer than another for the automatic draw bar process, for which I have a design in mind that's slightly different to normal power draw bars, and uses a standard threaded adaptor / drawbar.
|
|
#
?
Jan 6, 2018 22:23
|
|
|
Today's carving project between cutting out signs. No nips so I figure it's cool but let me know if it needs to be spoilered: The real purpose for this was to do something with the crazy warped scrap of oak sitting on my pile that was useless until then. I clamped it down at the edge of my table and faced it enough to be flat enough for the tape and superglue, stuck it down, made the new top side extremely flat, then flipped it for the carve. Didn't want to risk it popping loose in the middle of a long 3d op. Kind of a self-imposed problem since I've totally gone away from clamps in the last year and don't currently have any provisions installed for them. I want to do actual machining so bad. It'll probably be 18 months before I live in a place with a garage/shed, and I have to move in 4-5 months, maybe again a few months after that. The immediate future will make a tormach-sized machine too inconvenient due to multiple moves and deployment but maaan do I lust after not-a-router. I was eyeballing stupid mini mills today pretty hard for their portability/packability in the mean time. I don't want to waste cash on a conversion money pit (which will end up being $1500 on top of a $600 machine), but then again I spent more than the machine cost upgrading my originally terrible K40 and that thing is has been a total rock, incredibly useful, and a staple of my shop for like 3 years without complaint. Hmm.
|
|
#
?
Jan 8, 2018 06:22
|
|
|
^^ I can't quite see what that is could you make it bigger. On converting an existing mill to cnc.... I'm converting an RF-45 clone and I disassembled it and measured the ways on a surface plate. Not good. So far I've just been scraping the column but there is significant twist and other areas where it almost looks like they went at it with an angle grinder. I'm rescraping the whole thing before I cnc it, but even then I think I'll need some way of nod in the column due to the head weight. Looks like a bit of a model fault, the column needs more beef. So does the g0704 AFAIK, the other likely contender for cnc - I believe it's known for its lack of column rigidity. So both the most often converted mills seem to have issues, and coming out of Chinese factories unless your importers QA is quite good the mills themselves won't be great. OTOH, I believe Australia where I am gets the worst of it, if their distribution rumours are true - make 2500 in China, send the best 500 to Germany, the next 1500 to the US, and pepper the rest out to be rebranded in other places. Tormach seem to start with a better machine and also hit the nail on the head by putting quality where it can be used. Their engineered choice of ballscrew grades, steppers and other components seems sensible. I think other manufacturers might have better mills for the $$ though. On weight, my RF-45 is unfathomably heavy together (340kgs + stand?) but a cinch to move once apart. I wouldnt be too dissuaded by the entire Tormach weight - look at the individual component weights. Also , the parts aren't like moving a fridge. Tie it to a trolley with rubber wheels and you'll be fine. Then there's time researching everything and building it. Glad I have another mill. I think at the end I'll get a Tormach like product but I wouldnt do it if I lived in the US and could buy them at the prices you can. So yes... I think you are correct in buying a purpose built cnc, but could perhaps spoil yourself and get it now. edit: Perhaps a sherline would suit you, and compliment a tormach later? Also I think I saw a kickstarter for a BT30 Sherline head with ATC spindle, so you could perhaps build up your tools for a Tormach if they support BT30 with their ATCs? Mudfly fucked around with this message at 12:00 on Jan 8, 2018 |
|
#
?
Jan 8, 2018 11:56
|
|
|
I was just looking at the pre-CNC’d Chinese mini mills (seig kx3?) at LittleMachineShop like a month ago, but it looks like they’ve wiped any evidence of them and become a Tormach seller.
eddiewalker fucked around with this message at 15:33 on Jan 8, 2018 |
|
#
?
Jan 8, 2018 15:26
|
|
|
Mudfly posted:^^ I can't quite see what that is could you make it bigger. https://i.imgur.com/2hK5Yb9.jpg There is a nice CNC'd HiTorque 3990 near me for $2.2k which would be just about perfect as far as single-person-portable goes. Considering it. I know the bigger-than-benchtop machines are easily portable once broken down but I do all my moving on my own, and with a couple moves in the near future I just don't want to deal with trying to get a 1000lb mill delivered, moved into current basement, moved out into my trailer in may, moved across country, moved into garage, moved into storage, moved out of storage, etc in the next yearish. So the responsible choice is just to wait, but maybe a baby mill could be a nice stopgap. In other news my router is slowly moving in the direction of paying for itself due to etsy work. I gotta pump those numbers up, though; had no bites the first month, then a handful of sales in the same 2 weeks, then nothing for a month, then one lone sale. Now it's been a few weeks of drought. If I make like 30 more signs it'll be free, should only take a few years at this pace. 
|
|
#
?
Jan 9, 2018 01:37
|
|
|
The mini-mills are good desktop machines, and as long as you don't need the speed, accuracy, or reliability of a bigger, nicer machine, they're pretty alright for the cash. A Tormach 440 with enough goodies to get you started will probably run you about $7000 by the time you're done buying crap you want/need. My 1100 ended up being about 20k by the time I was done picked up all the poo poo I wanted/needed for it. Worth it though, anything you want to cut it can cut beautifully, support it top notch, and the controller is so much nicer than Mach3 you would poo poo yourself if you saw them side by side. Fancy machine tools are entirely buy once cry once. And a 440 is just small enough that you could conceivably move it yourself with minimum external rigging or extra tools.
|
|
#
?
Jan 9, 2018 06:05
|
|
|
Can someone recommend me a motion controller and software? I am fitting 3x750w servos and 1x1.8kw servo (spinde) to an RF-45. I'd later like to use two or more steppers to do a very basic tool changer. Maybe 5 axis wayyy down the track but not sure I need to think that far ahead now. Are there budget motion controllers out there that work pretty well? There are pretty drat good Chinese steppers and servo clones out there, I was hoping there might be good MC clones. I like the look of LinuxCNC but is it really hard for a beginner? A pain to get up and running?Mach 3 looks nice but costs a bit - and is perhaps too simple, though I won't find it that way for a while. I think I could get an old PC to sit beside my mill permanently pretty cheaply. Do I need to get 6 axis motion controller + software for my xyz axis, spindle and two steppers? Is every single motor an axis?
|
|
#
?
Jan 9, 2018 11:58
|
|
|
eddiewalker posted:I was just looking at the pre-CNC’d Chinese mini mills (seig kx3?) at LittleMachineShop like a month ago, but it looks like they’ve wiped any evidence of them and become a Tormach seller. I remember seeing those; it looks like you're right and they've disappeared off the website. Interesting. At the same time DRO Pros has started selling some kind of barebones CNC mill made by Weiss where it looks like you add your own stepper motors and controller? I'm not that into CNC and haven't compared prices but it looks like that be a good alternative option for people who are looking to buy a manual machine specifically to convert it.
|
|
#
?
Jan 10, 2018 02:52
|
|
|
Mudfly posted:Can someone recommend me a motion controller and software? LinuxCNC/Pathpilot as the machine controller, with a Mesa card and breakout board for the controller and driver end. You can poke around on the linuxCNC website for the details and example builds and configs from people with something similar, and pathpilot is basically a tormach specific version of linux CNC with a much nicer frontend on it. People have jailbroke it and got it working on old rear end big-boy CNC machines that had big 2kw servos on them, and a motion controller from the 1980s. Fair warning, it is a LOT of god damned work, and in the end will be exactly as good as you'd pessimistically expect it to be. I know when I tried doing something similar it ended up belonging in the DIY electronics version of the crappy construction tales thread.
|
|
#
?
Jan 10, 2018 08:40
|
|
|
Methylethylaldehyde posted:LinuxCNC/Pathpilot as the machine controller, with a Mesa card and breakout board for the controller and driver end. I would like to use my glass scales on the mill as another form of feedback, and have read that the servos should be in velocity set mode rather than step dir mode for this process. Can all the MESA boards do this, or do I want those that can only do analog? Bit confused as to analog vs digital. Also, can you (or anyone) recommend a motion control basics document/book/web page series to learn about it? I don't even really know what step direction is, let alone the myriad of terms used when trying to work out which motion controller to buy from the linuxcnc site. Like, this description - http://store.mesanet.com/index.php?route=product/product&product_id=290 DO I want an ethernet interface board? edit: is linuxCNC as much of a nightmare as people make out to set up? Can't be that hard.... I will go for linuxcnc or kflop I think. Leaning towards LinuxCNC. Not sure on ethernet card vs others. Mudfly fucked around with this message at 13:06 on Jan 11, 2018 |
|
#
?
Jan 10, 2018 20:05
|
|
|
|
| # ? May 17, 2024 08:29 |
|
|
Mudfly posted:I would like to use my glass scales on the mill as another form of feedback, and have read that the servos should be in velocity set mode rather than step dir mode for this process. Can all the MESA boards do this, or do I want those that can only do analog? Bit confused as to analog vs digital. The basic process is as such: List all your inputs and outputs: Axis motion, limit switches, any extra feedback systems you have such as your glass scales. What kind of inputs and outputs are they? Quadrature encoded? 0-10v analog? What kind of current? How are you going to drive the motion? Most servo drives will do analog or digital step+dir, so you can easily enough wire them as if they were a stepper. Once you have a complete list of all the different ways the machine needs to talk to the computer and the computer needs to talk to the machine, you can pick out the Mesa card and breakout board for it. The breakout board generally gives you all the actual wire-attach points for the machine, so you can wire up the limit switches, servo drives, feedback systems, and misc poo poo like spindle VFD. Then in LinuxCNC you have to define what all the inputs mean and how the computer is supposed to handle them. Each input and output is a pin, which you define and link to different things. This is probably the most annoying part, but if you're good at cribbing code, you can cobble together a working system very quickly just by looking at other people's configs and the wiki. This should work most likely
|
|
#
?
Jan 11, 2018 23:12
|
|