Hacker News new | past | comments | ask | show | jobs | submit login
Maslow CNC (maslowcnc.com)
397 points by MaslowCNC on Oct 14, 2016 | hide | past | favorite | 101 comments



I've built several of the 2D plotters (two stepper motors, some string, some math, and blam you have you have a plotter. Most fun was sticking them on chalk boards and then using chalk in them to put some really amazing drawings up on the chalk board.

That experience has suggested that this project is not going to be as successful as they would hope. That is because you can't depend on gravity to be stronger than the angular momentum of a router bit hitting a knot in the wood. Specifically when you're routering away and you hit material that pushes back on the router, the router may not move in a gravity planned sort of way. I've seen that in the plotters when the drawing device catches on the material that its drawing on. You get a line that goes horizontal when it should go down.

These sorts of things are pretty easy to pass off when you just hang a new sheet of paper but if it means throwing out a 4x8 sheet of plywood its going to be a bit more painful.

As for discontinuous lines you put a wall sensor on the drawing puck (or the router) and when you need it to move you tell the operator. Who goes over and lifts the pen/router off the paper. The code detects the lift, adjusts to the new position while you lightly hold the puck, and then indicates you should let it back down on the surface. Once the wall sensor activates it goes back to drawing.


Hey ChuckMcM,

I wanted to jump in here and add something. Full disclosure, It's my project.

The machine is designed to print things out of the sort of manufactured sheet materials we use in most construction these days so things like plywood, sheetrock, MDF..etc. Knots don't tend to be an issue because they are only in one ply of the plywood, meaning the force is very consistent even when passing through a knot. We've never seen an issue. Materials like drywall and MDF of course don't have knots at all.


Awesome project btw! I'm glad you've not seen any issues. Re-reading what I wrote, it sounds more pessimistic than it should. I am sure everyone that uses one of these will have lots of successful "prints" and will make many interesting things.


If the problem does occur you could simply make the thing heavier and run a bit slower. Inertia is a big killer for speed but this isn't a production machine.


I wish it were that simple! On a lot of materials, going too slow will cause problems as well. You want your "chip load" to be within a band; too slow, and you'll start burning the wood instead of cutting it.

http://www.pdsspindles.com/engineering-speeds

I, too, do not want to come across negative about this. I'm very excited about it. If it isn't sufficiently strong to handle some types of wood, even just being able to drop a V-cutter into the router and having it very accurately draw out all the cut/guide lines is a win for the target price. If my work flow is "draw accurate cut lines, do a rough automated cut (leaving some extra around the edges), and finish on the scroll saw and belt sander", that's great!


Yes, obviously if you're moving so slow you're not longer cutting then that won't work but that's so slow that I highly doubt it couldn't be improved on.


Setting the rpm of the router is an ordinary setup/configuration step for doing 2d machining. I used to run 2d router systems at sign shops for years. You're always looking to optimise the speed of travel and router rpm to get the best cut qualities - fast travel with high quality cut - too fast can result in chips/burrs, too slow results in burning/melting and damage to bits, etc.

I always had a list of settings to use for various combinations of materials, thicknesses, router bit sizes/types. It was a bit of a black art actually - sometimes the results of changes were counter-intuitive because of specific material properties (heat retention, etc), and that cheat-sheet wound up containing the accumulated knowledge of years of trial and error experimentation. "The only difference between Science and screwing around is writing it down" :)


Such a cheat sheet sounds like the perfect thing for an Open Source project with a (hopefully) active community to curate.


I don't like that the ropes do not meet at a point. For any given pair of lengths, you have not entirely constrained the position of the router bit. Rotation of the hanging cradle/holder will change the position of the bit. Gravity is nice, but it seems like this problem would be overcome if the ropes met at a point coincident with the bit - or as close to that as possible. That and a Z-axis would make this a pretty cool device that I could get behind.


Can it handle printing a guitar body (wood with about an inch of thickness)?


I would say that it depends heavily on the feed rate, expected bit life, and what wood. Soft mahogany, basswood, poplar might be done on a machine like this. I would NOT expect good results from hard, flamed maple, or other heavy, dense woods.

source: have made guitar bodies on a normal CNC router. Material holding is difficult at the best of times.


This looks like an awesome project. Do you have an idea on when Z-axis will be available? As a (single user) data point, 3D motion is a huge benefit -- if it ends up under $1000 with 3D movement I will likely just buy it to play with. Specs on accuracy would be useful, but unless it is really horrible, not likely to affect my buying decision. Good luck!


Would it be possible to tether the machine to four corners instead of two to avoid this potential? It complicates the pulley assembly (probably doubling it), but should make the bit more stable.


I remember hearing a story about a Russian tzar who was drawing plans for a railroad. He was trying to draw a straight line by holding a ruler down with his thumb. The writing utensil he used hit his thumb and went around it, creating a bump. The railroad was built perfectly straight except that one bump in the middle of nowhere.

Edit: apparently this was a myth, http://bigthink.com/strange-maps/580-the-legend-of-the-tsars...

It is strange that all the crap I heard as a kid and took for granted for most of my life can suddenly pop into my head and now be verified because of the internet.


Haha, I remember having a similar experience relating to the closing of mysteries as a kid. I remember being really mystified how a toilet and a bunch of home appliances worked. Then one day I had to fix some of them as an adult and thought, "Man this thing used to seem really mysterious and complicated as a boy".


Hey ChuckMcM - curious if you have any build reports of your 2D plotters. I've been very interested in them ever since seeing the Hektor and Victor project but haven't had the time/resources to build one myself (though I'd love to have one sitting at our office drawing out cool line drawings exactly as you described).

http://juerglehni.com/works/viktor


Nice work looks a lot fancier than what I built.

This type of plotter was quite the "thing" after one appeared at the Maker Faire (I believe in 2010), I already had an EggBot[1] from EMSL and the Eggbot board[2] is set up to control two steppers. Its really easy to use, I wrote a simple Perl library for it and put it on github [3]. The "build" was really just strapping two steppers to the corners of a stand up easel board[4], hang a pen in a disk to hold it perpendicular to the page (I believe we used a 2" hole saw to cut out the "puck" and then drilled a hole in the middle for the pen (friction fit with tape for added friction). Two screw eyes at the 10-o-clock and 2-o-clock positions on the puck, a string to a spindle stuck on the left stepper, and a string to a spindle on the right stepper. Then an all night hacking session and boom, drawing pictures :-). Its really straight forward.

[1] http://shop.evilmadscientist.com/productsmenu/766

[2] http://shop.evilmadscientist.com/productsmenu/188

[3] https://github.com/ChuckM/EggBot-Perl

[4]https://www.amazon.com/Post-Self-Stick-Easel-Inches-30-Sheet...


I'm a relatively new owner of an OX CNC (Ooznest kitset). I also researched a bunch of other options before purchasing, including the Shapeoko 2 & 3. I've spent ~$1200USD (incl. shipping literally across the globe), Those in the US could easily come in under $1000USD. The above mentioned CNC routers are really on the low end in terms of minimum required rigidity.

While the feed rate of the machine looks OK, but they do not specify the depth of cut per pass. Combine this with the fact they don't show a part being cut from start to end, even speed up, is telling. I suspect a excessive number of very shallow cuts, leading to long cut times and excessive tool wear. I also suspect if the cut depth is to great you might find out the true meaning of climb cutting.

0.4mm resolution is not that great and the resolution actually changes based on how far you are from each motor!

Mounting material to cut on a vertical surface is going to be a major pain if it's not a quarter sheet of ply or bigger. And once a part is cut loose without tabs the router is liable to kick them out, that's going to be a nasty learning experience for some.

The only great things about this machine are its large cutting area, small foot print and low price. But fair warning, the limitations will start to stack up fast I suspect, this machine would be best suited cutting flat pack furniture designs from full sheets of ply, when delivery time is not an issue. To his credit this does seem to be the market he is targeting.


Hey JaggedNZ,

The OX CNC is awesome.

You are spot on about what we're targeting. We're looking at people who are building big things, houses, sheds, furniture, kayaks, tree houses, geodesic domes, huge signs etc. Basically the kind of things that get built using hand held power tools now. The goal is to make a machine that can download and build those type of things more accurately than a guy who's really good with a jig saw.

As for # of passes, I'll throw some mention of that on the website tomorrow. I usually do 1/8th inch passes so 1/4 inch ply is a 2 pass job, 1/2inch is a 4 pass thing...yada yada


Can you get good results on a single pass if you really slow down the feed rate?


It's not in an obvious place, and for some reason it's not listed on their YouTube channel, but they do have a timelapse video showing the logo from the main video being cut from start to finish. (https://www.youtube.com/watch?v=x7XafvOKoIU) Looks pretty good to me.


Off topic. But I'm currently looking at getting the Shapeoko 3, and I'm wondering why you chose the OX CNC (which I'll admit I haven't really looked at)


You wouldn't happen to be in Dunedin are you?


Very nice job!

An alternative for the router brand: Makita, they offer a number of router bodies that seemingly last forever in both wood and more difficult materials (fibreglass), I've run them for 100's of hours (and probably 1000's) without a problem.

Beware of steppers when you're driving routers, you may lose steps along the way ruining your workpiece (step loss is loosely related to Murphy's law, it will happen near the end of complex workpieces).

If you build this thing out of metal with a watertable underneath it will be slightly more expensive but you'll be able to use a plasmacutter to cut metal as well as wood by swapping out the head.

Here is what I built:

http://pics.camarades.com/d/11225-2/dscn3995.jpg

(That's the first version, before making a much lighter gantry to improve speed constancy in corners, not very important for woodworking but extremely important when plasma-cutting.)

Some critical observations about 'Maslov':

Hanging plotters suffer from asymmetric curves when it comes to acceleration 'up' or 'down' which can lead to surprises in overshoot/undershoot of the intended toolpath.

Cable plotters tend to have a problem with that anyway and gravity will make this problem worse. If accuracy is important consider making it horizontal.


I helped with building CNC plasma cutter as well: http://www.boot.lv/forums/index.php?/topic/133809-cnc-plazma... (beware of the latvian language but has nice pics).

How did you solve the torch height control? We ended up using "digital torch height controller" - basically measure the arc voltage, compare to fixed preset and move Z axis up-down accordingly. This is a stupidly simple approach but requires complicated lookup tables of voltage vs height vs feedrate vs material and I'm thinking about how to improve it.


That's exactly how I did it, but I built the controller myself.

The arc voltage is directly proportionate to the height above the workpiece so you don't need much in terms of lookups or feedrate, all you want to do is to keep the voltage constant, so if the voltage is too high you need to get closer to the material (beware of strikes!) and if it is too low you need to back off (fast!, sometimes the material is warping towards the torch very rapidly, especially when the material you're cutting out of is thin).

Where in Latvia are you? I go there with some regularity, I'd love to come visit. Mail: jacques@mattheij.com

Tip: you could lose quite a bit of weight on that gantry but turning it into a lattice.

http://pics.camarades.com/v/jacques/renewables/windmill/09_s...

Weight in a plasmacutter gantry is killing because it reduces the speed with which you can change the direction of motion while cornering in detailed work. That in turn causes you to have to slow down which will burn away fine detail so any weightloss on the gantry should be pursued.

Ideally you want constant velocity relative to the workpiece while cutting. This also helps in maintaining precision because the cone shape that the flame has will cause it to cut thinner and thicker depending on the stand-off distance.


IIRC, the arc voltage was proportional to height only at constant velocity. I guess it was also a function of how much material is below the arc at the instant.

I was working on a controller which was supposed to get it's setpoint by measuring voltage during first few seconds of a cut, but the project turned out to be outside of my EE skills at the time (isolating analog signals, high voltage resistive dividers that didn't blow up because I had no low-voltage feedback signal from plasma source etc.)

About the need for rapid accelerations- that too was one of the things we learned by building this machine. IMO it should be stressed more in discussions about plasma cutters.

In this case the blue linear slide actuator was off-the-shelf part and modifying it would require to cover the belt and linear guides inside with something else and was thought to be too much of an effort.


You can usually tap a point in the plasma cutter power unit that will tell you exactly what the arc voltage is, almost any slightly larger plasma cutter power supply will have a point that's convenient.

You'll still need to use an opto-coupler, the way I built mine was the setpoint and comparator lived on one side of the divide, then two input bits, one for 0.1 V below the setpoint or lower, -0.1 to 0.1 was both inputs off and one for 0.1 V or higher, the setpoint was scaled so that it would have a nice range mapped to the inputs of two op-amps doing the comparison. The outputs of those op-amps drove the opto couplers and the software would track the 'Z' stepper up, down or hold depending on those two inputs. Super simple really.


> lose steps

They did say that it has encoders and closed loop control. Perhaps overshoot could be an issue but losing steps shouldn't be.


Encoders allow you to correct for lost steps, but you'll still be off the intended toolpath. And over/undershoot / oscillation are well known artifacts of a cable based system, it all depends on the length of the cables and how elastic they are and how fast you're moving. It's a tricky combination to get right, I've seen one project like this canceled because of repeat accuracy issues but that may not be so important for everyone (signage for instance, or art projects may be quite forgiving).


This is kind of neat, but the rigidity is questionable. It's significant that they only show sped-up video. You don't get to see the actual cutting speed. I suspect they have a really low feed rate because they can't exert significant side loads on the router. Maybe if they had four chains, so it wasn't just gravity holding the thing in place... Wallboard, yes, MDF, probably, plywood, if thin, hardwood, no.

A big problem is that this thing gets its Z position by riding on the material it is cutting. So you have to cut things out of fresh big sheets. If you cut away too much, work near the edges, or use smaller sheet stock, there's going to be trouble. For decorative scrollwork in thin stock, though, it could be useful. Expect stuff from this machine to appear on Etsy shortly.

I've used ShopBots for similar work. Those have enough rigidity for wood, even hardwood. The limitation on cutting speed is damage to the router bit; the machine has plenty of feed power.


My own drivetrain for something like this was simply a small gear on the shaft of a stepper running on a long piece of gear rod. Accuracy was more than acceptable and you could reach very high speeds with the appropriate acceleration curves (very important too to minimize step-loss chances).



This kind of work gets me so excited. Thank you for leading the way, @BarSmith! Open source industrial design communities such as OpenBricks [1] and OpenDesk [2] have emerged to fuel the imaginations of users of Maslow CNC mills.

For instance, one could now mill a vertical gardening wall-mounted planter, with community help, by joining this project: http://www.openbricks.io/app#!/project/566af15e584123b4502f4...

It is a right of passage for a homeowner who DIYs to create a workbench. I haven't created mine yet but have investigated approaches. One novel approach is a torsion box-based workbench. The painstaking process one must undergo to create the lattices could be replaced by a fitting together CNC milled, continuous pieces. So cool.

I'm so ready to join this movement..

[1] http://www.openbricks.io/ [2] https://www.opendesk.cc/


This seems like a cool project but the non-commercial license in the CAD files [1] is a no-go for me to fund their Kickstarter. It does look like the PCBs [2] and firmware [3] are free/libre licensed.

[1] https://github.com/MaslowCNC/CAD/blob/master/License

[2] https://github.com/MaslowCNC/PCBs/blob/master/LICENSE

[3] https://github.com/MaslowCNC/Firmware/blob/master/LICENSE


Why? They put a lot of work into this and are giving the plans away. They just don't want you to compete with them. It seems fair to me.


We all work hard on the things we make. But many of us open source those things so that others can make them better. When we restrict commercial re-use, we're hobbling the project in certain ways. What if someone wants to help make it better, but they want to earn money from their improvements? You can still license the code with attribution and share alike. That way people know you designed it, and you get to use the improvements other people make.

It's okay that they choose a non commercial license. But it may make me less interested in supporting it.

But I'll check out their PCBs and see if I can help make any improvements. I like this project.


This may be very well received in the model railroad community. When building a layout, one of the techniques (called cookie cutter) for creating the subroadbed is to cut out shapes from a sheet of 4 by 8 plywood similar to this gadget. See this for example: https://www.youtube.com/watch?v=iqcUKofeElc



An interesting alternative to a Glowforge. This machine seems less of a "tool for makers" of small things though, and rather an autonomous cutting machine for big projects. It would be silly to use this to cut out a leather wallet, for example, but totally reasonable to build a chair. Neat!

p.s. oh man, make a chair design, and then just set these babies to run day in/out. that could be an fun factory!


There's more to CNC than a hip startup like glowforge. (In the interests of disclosure, it should be noted that Series A investors for Glowforge include Bre Pettis and Jenny Lawton. Given their business decisions with Makerbot, I'd consider them 'tainted' to this market).

This device is more akin to a regular 4x8 cnc router, like ShopBot, or an off the shelf Chinese CNC machine. The innovation here is an experiment with the drawbot mechanics popularized around 2010, but with a much heavier toolhead. If it works, awesome - it's an order of magnitude cheaper than a shopbot. Even if it's not ready for production, it's an interesting enough idea for a few shops, garages, or hackerspaces to replicate the project.

This is really an awesome idea, and comparing it to a startup with questionable investors does it a disservice.


2 more things:

1. this isn't really printing, its subtracting. I think its important to make this distinction.

2. feed rate is REALLY important with getting good tool life as well as good cut quality. Relying on gravity to control a 1-2 HP router is not going to counter the cut forces well except at seriously sub-optimal cut rates.

e.g. A carbide 2 flute upcut end mill with .5" diameter cutting plywood works nicely with a cut depth of about .5" and a feed rate of somewhere around 200-300 inches per minute. Even that may not yield the optimal chip load to keep the bit cool. That assumes a router speed of ~20k RPMs and 3.25 peak HP (AC motor @ 110v).


"Relying on gravity"

I have never built one of these, I wonder why they do that.

I would think 3 strong steppers would be capable of some pretty amazing forces and would be quite well constrained. Or perhaps 4 steppers or more.

I've spec'd out NEMA 42 frame steppers that could quite easily start my car (under the theory that a pony engine need be a tenth size to start a larger engine and my commuter car only has 100 ft-lb torque, so 20 or so ft-lb of torque should easily be adequate to start the engine)

Lets see a KollMorgan KM11 with 3750 in-oz of torque applied to a six inch pulley is 40 pounds of force. That should accelerate a ten pound router with some authority, 4 G or so. I know from experience I can push a traditional router against a jig or fence or table much longer and more comfortably than I can curl 40 lb barbells so 40 peak should be more than adequate.

A stepper of that output retails qty one to end users for like $150 delivered maybe? So figure maybe $300 for three in bulk order shipped by pallet-full?

Over the course of my life the prices of steppers and control electronics doesn't seem to change but the torque seems to go up an order of magnitude per decade for several decades now, so it might be a matter of waiting until 2026 to unleash something like the above.


Steppers are definitely strong enough. The NEMA 42s on my machine are geared down 10:1 and work very well moving a 100ish pound gantry.

The issue with gravity is not really motor strength so much as it is rigidity. There is no z axis rigidity in the setup because it only has a chain. So it can't exert pressure against the material from the axis.

The lack of rigidity is the real problem. As mentioned in other parts of this thread, .01" accuracy is not going to give you great tolerance for pieces where that error can be cumulatively compounded.


Would 1-flute, small diameter endmills help keep the feed down? Your jobs would take a lot longer, but I think there's still room to maneuver even if you have to limit feedrate.


Kind of. Routers have very limited speed control, so this is still somewhat problematic. Also, small diameter bits tend to have shorter cutting lengths and cannot cut as deep. They are also fragile. Finally, costs do not scale well with small bits because they last much less due to poorer heat dissipation and early failure.


Hey everybody! Just wanted to throw out a quick reminder that we're launching on Kickstarter tomorrow, the 25th, at 9 AM PST. Hope to see some of you over there!


It seems like it would be worth shopping with whoever has a "normal" CNC machine nearby, and seeing if you can get your stack of plywood machined for less for a given project. If you make a mistake with DIY, you're potentially out a sheet of plywood, not to mention time and effort.

Your project will involve plenty of time and effort at other stages than the plywood cutting bit.

Still, I'd consider getting one (perhaps on v2 after the Kickstarter has played out, and v1 gotchas figured out). It is cheap enough to get "just for fun."


"We're making the choice not to patent it."

Now that the US is first-to-file, doesn't that mean someone else can patent it and use the patent against them? Isn't the best thing to patent it and give royalty-free licenses?


No, this is a common misunderstanding of the implications of first to file.

Publication or public practice of their work is prior art. It immediately precludes third parties from patenting their work. (And, after a grace period will preclude them as well)

First to file deals with interferences, when two parties try to patent the same work at the same time-- it doesn't change how prior art works.

Of course, it is still useful to file defensively since relevant prior art is much more likely to be discovered when it exists as a patent... if it's not discovered a patent could be incorrectly issued. (The patent would be invalid-- but could still be used to harass people until put out of its misery by a court. :) )


There isn't anything that you could patent here to begin with.


Now that there is proof of prior art this isn't patentable any more. You can't patent the wheel either. However, that doesn't mean that an attempt at patenting it wouldn't pass, but you could get it invalidated.


Looks cool... but I'm worried about repeatability and accuracy. I've seen a lot of similarly-designed 2-motor "drawing robots", and the precision definitely wasn't 0.4mm (as claimed). I imagine this could be difficult given the use of gravity to supply the lateral cutting force?

On a related note: I'm pretty excited for the ShaperTools handheld router: https://shapertools.com/ You get the benefit of big work pieces & CAD-based routing, but in a hand-tool form factor.


Oh, Kickstarter.


Small random quote from the vid:

> More details for nerds below

I'm not knocking the person who said this fairly innocuous statement (I catch myself saying similar things), but it goes to show our unresolved social biases against technical people. Yes, some people care about numbers and details and engineering. Unfortunately, that associates them with a term that implies low social value.

If another country overtakes the US in terms of innovation and engineering, you can be sure that they will be showing greater respect to their technical class of peoples.


The crawlbot is another hobby-friendly compact 4x8' CNC. It's $2000. It saves space by attaching directly to the 4x8 sheet with timing belts and using it as the frame for the long axis. The shorter axis and router are contained in a 5x1x1 ft housing that "crawls" back and forth across the workpiece.

https://printrbot.com/shop/printrbot-crawlbot/


Quick tip about your site: put up some photos, not just videos. I'm on a mobile phone and want to see this thing, and playing video is annoying, and sometimes unfeasible.


I don't know if the idea is unique (edit: no, they mention it's based on a known concept of "hanging plotter"), but that's a neat way to utilize gravity to bring down costs. It's much more accurate (.4mm) than I would expect, enough for most hobby work with big sheets.

(Off topic, but a hanging plotter (for drawing) would be rewarding project to build with a kid!)


Errr, .4mm is not very accurate. It's about 5-15x less accurate than basic wood cnc machines, which do 0.001-0.005 (they could easily do better, it's just not necessary for wood). If each part was off by 1/64th, you can't even form a 4 sided box that is guaranteed less than +-1/16, which is a lot.

I'm pretty sure they are also not including the router runout (since it uses a router and not a spindle, add about 0.005-0.010, or, rather, halve their precision again)

Now, don't get me wrong, it's a neat idea.

It may be really cool for construction (though drywall/etc guys are a lot faster than this machine is), but ...

But you actually could not likely use this for real cabinet parts, etc.

To give you some idea:

Their specs:

Encoder Resolution: 8148 steps/rev

Real World Precision: +- 1/64th inch (.4 mm) or better

Max feed rate 48 inches/minute

Specs on my pretty basic 36x36 wood CNC router:

Encoder Resolution: 4 million steps/rev (22bit). Electronically gearable to whatever (they are servo motors)

Real world precision: +-0.001 (1/1000) inch

Max feed rate 1200 inches/minute

Now, obviously, there is a range of costs and other machines between these two machines :)

But building precision CNC routers, even with good closed loop servo motors and ball screws, is not anywhere near as expensive as it used to be.


He said, "It's much more accurate than I would expect." He's referring to how accurate it is based on how it is designed. And I disagree that it's not very accurate, given the use case. 0.4mm or 1/64" is way more accurate than almost anyone can do with traditional power tools.

Sure, other more expensive machines have better precision. But how many people actually need that much? Even plywood is going to experience changes in length due to temperature/humidity greater than 0.001" (0.025mm). Solid wood (which I guess this machine isn't really designed to cut) is going to be even more.


"And I disagree that it's not very accurate, given the use case. "

What use case?

"construction" is mentioned, but plywood guys and drywall guys can make cutouts and curves far faster and more accurately than this machine can cut.

So seriously, what is the real use case other than "fun toy"?

" 0.4mm or 1/64" is way more accurate than almost anyone can do with traditional power tools."

This is just flat out false.

With a simple track saw, i can easily get within 1/128th without even trying.

Hell, i can do that with a circular saw and a straight edge without a track.

You can get to within 0.005 pretty easily if you have bothered to tune your tools at all and don't beat them up.

You can definitely get 0.005 if "traditional power tools" includes routers.

Of course, if you go to hand tools, getting within 0.001 is utterly trivial with good hand planes.

"But how many people actually need that much? "

Wood certainly moves, but 0.005 is about being able to make nice fitting joinery repeatably.

https://woodgears.ca/mortise/accuracy.html

"With a mortise of 0.503", here's how the tenons fit: ...

Realistically, the total range of latitude is about 0.005" for a good fit. And this is using spruce, which is a very forgiving soft wood. For mortise and tenon joints in hardwoods, accuracy is even more critical. "


For a wood-cutting system supported by two chains and gravity, .4mm is surprisingly good tolerance. Sure, it is possible to get far more stricter tolerance with tradiotional power tools (for skilled person) but first you need to somehow draw a line to follow, a challenging task in itself for complex shapes. This tool can do CAD, a big benefit.


"For a wood-cutting system supported by two chains and gravity, .4mm is surprisingly good tolerance."

Remember that this is jut the 'precision'. It says nothing about the repeatability.

I have strong doubts it's repeatable to even that level.

"but first you need to somehow draw a line to follow, a challenging task in itself for complex shapes"

????

It's really just not that hard. In the absolute worst case, you use graph paper and cut it out. Most of the time, you just use a protractor or a bow and cut it nearly perfect anyway. Given the tolerances pretty much restrict this to "construction work", you can get the same result quite easily.

"This tool can do CAD, a big benefit."

I'm also going to assume you've not used much cad/cam software, because for most people, it's infinitely faster for a one off to draw it by hand than it is to draw it in a cad program and generate gcode for it using cam software.

People have enough trouble using sketchup, let alone something real :)

The advantage for most people for CNC is the ability to repeatedly do the same thing.

Unless you invest 100k in automated workflows, you aren't ever going to get to the point where it's simply "draw in cad, click button, have machine cut it". It's just never that simple, from workholding to what have you.

Which is another reason why i think it's neat, but has no real use case.


Thank you for these detailed responses. They track very well with my experience using a 4x8' CNCrouterparts machine. There are so many things acting against precision that I spent my first year of CNC ownership finding out what was really practical. Inlays on machines with >.005" accuracy don't work. Guitar necks do NOT fit. Puzzle joints don't glue up.

The sources of problems are endless, even in a rigid machine: 1. spindle runout. 2. bit diameter changes (its is labeled at .25", but its really .246 3. climb vs conventional milling yields different characteristics in different materials. 4. bit deflection in certain cutting actions (plunge, tight corners, etc) 5. bit wear during the job. 6. dust removal / chip recutting 7. lost steps 8. dust in the ways of the machine 9. linear motion system misalignment. 10. workpiece slippage/hold down. 11. material not being flat/level (especially with cheap plywood, which often varies in thickness by >.005")(try making accurate box joints with that...)(also, it is never flat, it always has some twist/warp because of the way the plys are cut out of the tree).

This machine seems to be a great thing, and I very well may build one for an outdoor plasma cutter, but, I find the accuracy claims very difficult to believe. All of these sources of precision loss will effect the end result.


You are right that the sweet spot for this tool (toy) is very narrow. Let's say I'm still surprised that you can rope-hang a router like that and get usable results, let alone ~millimeter accuracy :)


Can you talk about where to get the cheaper closed loop stuff? I am using usdigital encoders and geckodrive controllers currently...


ebay most of the time :)

You just have to know what to look for.

So, for example, mitsubishi EOL'd the mr-j2-* series, and released the mr-j4 series.

If you look on ebay, you can usually find mr-j2's and mr-j3's , as new old stock, for about 300-400 for drive + motor (I'm going by the 400 watt ones, so mr-j2-40 and hc-kfs43).

If you are really good at it, you can find the mr-j4's for good prices occasionally.

There really isn't much of a resale market because in the automation world, people don't buy used or eol'd stuff (you aren't going to shove used motor into critical factory machine)

I ended up getting 3 new mr-j4-40a's and associated motors for about 1000 bucks.

You can play the same game with yaskawa, etc motors.


How do I learn which mfgs I should look for?


The easiest way is to see what gets used in wood routers by typical manufacturers.

The short list is pretty much: Mitsubishi Yaskawa Sanyo Panasonic

It is hard to go wrong with any of these.

Unless you are doing something nutso, you won't need more than 400 watt motors (which, being brushless dc servo motors, have enough power to pretty much destroy every other part of your machine)


How would this thing lift up the router to go between spots that aren't connected? Or can't it, and is that the reason the logo's circle was open?

And what would be the precision on this thing? It's a bit suspicious they'd leave that out... Would it be able to work at 1 mm precision?


They say 1/64in/0.4mm precision or better.

I'm sure they can lift it up. This kind of mechanism isn't exactly new - but using it on a router is fairly new. There's a lot of projects with pen plotters built in this general pattern, see http://hackaday.com/2011/11/17/polar-pen-plotter-draws-huge-...


I remember seeing that pen plotter before, guess I wasn't creative enough to put 2 and 2 together, haha.

Even for just 2D, this is really clever for the form-factor/capabilities/price. It's in 'impulse buy' pricing territory for most people (I mean, posting here at least). It takes up basically no space. If the steppers are well-engineered, I can't foresee a failure mode other than just normal tool wear.

I was considering buying a Shaper[1] (which is really cool in it's own right) after seeing Ben from Applied Science mess around it. With the Shaper, "you're only limited by the surface on which you can place those encoding stickers"[2] while with this, you can build out effectively until the elasticity point of your support structure (or, I suppose, the chain has a fatigue failure). 15 thou isn't bad at all, especially if you know where the run-out is going to be (i.e. if it's gearing backlash, you can design your features around that with that in mind / build your software to compensate for it / incorporate active air cylinder brakes / etc). Edit: On second thought, they said the steppers were closed loop. So if it's any worse than 15 thou, shame on them, really.

My only concern is if the slop/run-out goes up as a function of the material you cut. The Shaper has what seems to be using some sort of active delta compensation[3] Aluminium and a standard house-hold router can accurately mill your 6061 aluminum that's good for up to like 30k tensile with it. The collet/chuck/gantry/whatever will just actively re-orient your bit for the micro-tuning. (Though, for the $1500 they want, you can pick up one of the thousands of Bridgeports Series 1 CNC's out there, throw another 500 for your carbide cutting tools + retro-fit kit and you've got a home machine shop and come out way ahead only a few hundred behind with the ability to actually mill something structurally sound like steel.)

Either way, this is super cool - I'll wait for a few reviews, but I think I know where my money's going even if it's limited to wood.

[1] https://youtu.be/Kg2HOk1XHsA [2] And those stickers smell of vendor lock-in. It's got the opposite 'feel' to it than this product. [3] https://en.wikipedia.org/wiki/Delta_robot


When can I get my hands on this?!

I was just looking for something like this yesterday! I pre-ordered a Shaper Origin (https://shapertools.com), which ships in September 2017, but I need something to use now.


This is interesting in that it is somewhat similar to the "delta" style 3d printers in concept, just in 2-axis instead of 3.


This is a 2-axis router it looks like; though if you used the recommended router, you could adjust the depth between runs.


The third axis is always the killer on any machine tool. The first two axes run in a shared plane, so it's easy to make it rigid. The third axis has to be supported by some sort of cantilever or frame, and stay out of the way of the workpiece, and thus is more complex. Also, if each axis has to carry the weight of the next axis, the overall design has to be stouter. This is the reason why the third axis is often separated from the first two, as in a typical knee or gantry mill.


Most of the work done on CNC routers is 2-axis anyway. Or 2-axis with just a couple z-heights, which can be accommodated by changing the router depth by hand. Ultimately it's made to cut up plywood, not sculpt things. Like a sized-up version of small scale laser cutters.

Once you accept the constraints of a 2-axis CNC router, you can design in a way that maximizes it and build incredible things. See https://www.opendesk.cc/ for examples of a lot of furniture that can be built this way.


Okay, that makes sense; this is for replacing a woodworker with a jigsaw, not for fine work. Seems like a really good solution in this space (though I'm biased, since I suck with a jigsaw).


The project in this video on their Inspiration page is pretty neat:

https://www.youtube.com/watch?v=gij49fF-4X0


True. Also, in the FAQ they say a z-axis is in the works.


Nice concept, and I love their philosophy about openness. I'm interested to see how well it works!


Looks brilliant - if I could make one suggestion for the video - more examples of what it can do.


plywood is peachy but aluminum is awesome. have a robot glue a line of thermite to horizontal sheet metal. Light a match and you have your part.

(Tolerances may not be exact).


You can't light thermite with a match (source: tried).


sparklers apparently work


I like the low cost idea and approach. rly neat!


I'm no expert in hardware and software that is developed close to hardware. Can someone explain to me where the achievement is, before the Kickstarter even starts?


Kickerstarter doesn't require any achievement, innovation or novelty - success there depends only on your ratio of marketing folks to engineering folks (the higher the better). Oh, you also need a bunch of bay area hipsters saying how this product transformed their life.


How does this do splines?


The chains are pretty neat! Any idea on backlash?


Author made a comment in the last video; the weight of the cutter keeps the motors under constant tension one way, so there is no play caused by changing direction.


Mostly true but not quite. The forces on the chains are actually higher when the chains are "shorter" and the router is higher up. The farther from vertical the chains are, the more horizontal tension there is on top of the vertical component.

The same basic principle shows up in rigging - http://www.fdlake.com/rig-slng.html - the illustration where the straps are at 90 degrees is rated higher than 45 degrees.


Also there is slack in any gear/chain train when switching directions as they never fit perfectly. Same with gears. Even tiny amounts will throw you past .4 mm quickly


I think by keeping the object to be cut out in the centre, having small (relatively) size would experience lower slack, hence reduced deviation from the said accuracy.


Actually that kind of backlash isn't a problem in a machine built like this. The tension of the chain against the sprocket is always pulling in the same direction around the sprocket. Even when the sprocket reverses direction, the chain is pulling the same way and staying tight to the same side of the teeth it was on.


That would be correct until you figure in momentum as long as you stay under that threshold your hypothesis holds, if you get the device moving or swaying it won't. Of course this can be compensated for in software. So Point taken.


His explanation is pretty good that the weights always pull the chain in one direction, which eliminates most of the backlash. I think he mentions that the remaining backlash is taken out in software.

There could be some error due to rotation of the entire router as it encounters friction, but it should be manageable, and whatever error remains is basically how accurate you can get.


In the video where he explains that rotation, to me it doesn't look like the chains come down mounted in the same axis as the router bit to form a V, just a little higher. I just wondered if that by its nature is imparting a moment on the router.

I wonder if you could mount the router on a bearing, have each chain attach to a free side of the bearing so each side can rotate around the router. Then just figure some way to lock the routers torque that isn't the x/y control mechanism. Perhaps 1 chain pulling directly down.


The title should be changed as this product is not currently available for purchase.


The sources are available...so you are free to start building yours


Which is great!

I only mentioned it because the original title read something along the lines of "4 x 8 foot CNC available for purchase for under $500".




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: