Neat idea for CNC router

HomeBuiltAirplanes.com

Help Support HomeBuiltAirplanes.com:

Aerowerx

Well-Known Member
Joined
Dec 1, 2011
Messages
5,568
Location
Marion, Ohio
This is an interesting idea. If I understand what they are saying, you can use your 4x16 foot build table, and route out your full size ribs, bulkheads, and skin panels!

This will put extra emphasis on building a square, level, and flat guild table. But I do not know how precise it will be running in the 16 foot direction.

It also appears to use 3D printed parts, so you would need a 3D printer first.
 

ScaleBirdsScott

Well-Known Member
Joined
Feb 10, 2015
Messages
1,232
Location
Uncasville, CT
As long as you have two motors for the long axis, and you have decently accurate rack/track for the motors to run against, you can easily get repeatability within any appreciable tolerances for aircraft construction. The rubber belt method is, essentially, a variation on what drives 3D printers. For light duty work, this whole machine is perfectly capable of accurate movement.

As soon as the loads get very high, I forsee that belt slipping, plastic elements breaking, and in general the whole thing shaking and vibrating more than ideal. I would think MDF and plywood are within capabilities and it can be accurate but speeds/feeds would have to be closely monitored. If a belt slips on one side more than the other, that's when the gantry is now racked out of square and now all your cuts loose a lot of accuracy. This is something I have to contend with even on my machine where I ensure squareness against stops before each job, and I periodically will check for any gross errors over the course of the job. 2-3 thousands over 4 feet isn't bad but the machine will run with the gantry 1/8" out of square without complaint yet your parts may well be rubbish as a result. On the other hand, in some ways the resiliency of a belt may help prevent a heavy load from loosing steps as the belt will still rebound and that should put the machine back where it needs to be. But there might still be accuracy problems with coordinated cuts. Also it's far easier/cheaper to replace worn belt than it is to buy and replace metal rack stock; the teeth on the aluminum racks of my router are starting to show significant wear in some places which is visible as the machine stuttering while the pinion stumbles its way through the chewed gear teeth. I havn't measured any appreciable accuracy loss over the whole part, but I worry that in time this will exacerbate and require me to replace my racks.

I do like the general gist of the LowRider type concept tho, it's a good use of space to pack everything in the gantry and a remote power supply because then the table is more or less just a table with maybe some guides mounted. But I would make my gantry purely from at least aluminum components, and I'd do my research about the types of loads that the belt can handle before loosing position
 
Last edited:

pictsidhe

Well-Known Member
Joined
Jul 15, 2014
Messages
7,521
Location
North Carolina
Has anyoe built a CNC where the position is determined by sensors, rather than by the drive being in a certain number of steps from the origin?

I've just ordered a Chinese 4-axis 3040. Hopefully it will be handy for the innumerable small fiddly bits I will need to make.
 

Hephaestus

Well-Known Member
Joined
Jun 25, 2014
Messages
1,752
Location
YMM
The lowrider and MPCNC designs from v1 engineering are pretty well known and have good reputations.

The usual issue is getting your machine dialed in, and understanding that you can/can't do. The lowriders done some pretty impressive 1/8" aluminum cuts over the years (again it's all about feed rates and tool rpm)

Sensor based distance feedback makes sense in some ways, but once you've spent the time dialing in, it becomes a few thousand dollars of upgrade that's pretty useless at the end of the day and another point of failure (that really bites you in the ass). Pretty much why it's not pursued.
 

FritzW

Well-Known Member
Log Member
Joined
Jan 30, 2011
Messages
3,896
Location
Las Cruces, NM
I've started a small Lowrider2 build, small enough to fit on a shelf under my workbench, maybe 2' x 2' or 2' x 3'

I started with ABS but some of the parts, like the vacuum shoe, came out a little wonky. ...a problem with my ABS settings, not the LR2 files I switched to PLA and everything worked fine (the store bought LR2 kit is PLA). I think PLA would hold hold up okay but I'm going to start over with PETG.

20190709_095529_resized (1).jpg
 

pictsidhe

Well-Known Member
Joined
Jul 15, 2014
Messages
7,521
Location
North Carolina
Hephaestus, i was thinking of a few hundred dollars. With position knowledge, you don't need high precision rails etc. They get expensive on large machines.
 

Hephaestus

Well-Known Member
Joined
Jun 25, 2014
Messages
1,752
Location
YMM
You still need the precision rails etc - yeah you have a bit more give.

The problem is the associated hardware and processing requirements, that ends up running up the cost beyond just the sensors. That processing time to compete where you are vs where you think you should be then add the steps etc.

There was a pi project that was tackling it, but they were running into issues around latency etc - and you have to develop a slicer to manage the on the fly adjusting as well so it becomes a massive all encompassing project.

Whereas using your relative known position and typical monitoring is fairly easy and proven.
 

Vigilant1

Well-Known Member
Lifetime Supporter
Joined
Jan 24, 2011
Messages
4,953
Location
US
Has anyoe built a CNC where the position is determined by sensors, rather than by the drive being in a certain number of steps from the origin?
It would sure >seem< like that would be a good way to go.
 

Radicaldude1234

Well-Known Member
Joined
May 30, 2009
Messages
389
Location
Front Range, Colorado
Has anyoe built a CNC where the position is determined by sensors, rather than by the drive being in a certain number of steps from the origin?
Yup, they're fairly common too and they're called servo driven CNC machines, as opposed to stepper driven ones. Servo motors have an encoder that indicates position. You can technically turn any electric motor into adding an encoder and connecting it to a controller (such as an Arduino), thereby completing a control/feedback loop.

http://www.truetex.com/servomod.htm
https://www.hobbyist.co.nz/?q=convert-dc-motor-to-servo-using-arduino
 

Hephaestus

Well-Known Member
Joined
Jun 25, 2014
Messages
1,752
Location
YMM
Yup, they're fairly common too and they're called servo driven CNC machines, as opposed to stepper driven ones.
But unless you're on a seriously solid leadscrew with 0 blacklash or flex- you're still not 100% sure where your bit is in relation to home.

And still your Arduino doesn't have the power to keep up with normal feed speeds & adjust gcode on the fly to compensate and adjust.
 

Radicaldude1234

Well-Known Member
Joined
May 30, 2009
Messages
389
Location
Front Range, Colorado
But unless you're on a seriously solid leadscrew with 0 blacklash or flex- you're still not 100% sure where your bit is in relation to home.

And still your Arduino doesn't have the power to keep up with normal feed speeds & adjust gcode on the fly to compensate and adjust.
Was using it as an example, but the Arduino is connected to a separate driver component which powers the motor and only provides the signals to control the motor. There are more specialized controllers for this purpose though.

Servo systems also don't adjust the g-code; the g-code tells the machine where to send the tool and the controller does its best given it's understanding from sensors. It doesn't require as much computing power as you think...

You can also, as seen with linear actuators, use linear potentiometers to ascertain your position. That would still complete the control loop. This gets really expensive fast, though, as very large and accurate sensors are pricey...which is probably why stepper motors are seen as the budget solution to CNC and more widely used. They are also fairly reliable and good ones rarely get out of wack unless you're running them to the limits of their capabilities.

Some of newer steppers, like the ones in my Prusa i3 Mk3 3D printer, do provide feedback when they hit resistance like the limits on the machine. This is fed back back to the controller and gives the printer the ability to autohome.
 
Last edited:

Hephaestus

Well-Known Member
Joined
Jun 25, 2014
Messages
1,752
Location
YMM
Again you're providing another form of motor based positioning, which doesn't really jive with the real world. And the i3s force feedback uses a 32bit processor board. Which yes is becoming affordable but where's the benefit if your belt skips, has some stretch, or you get a minor chatter or flexing of the z. Because the link has points of flex - you're not getting precision accuracy.

My CR10s5 on mgn12 bearings and a Titan aero is 1000x more accurate than my prusa (the mmu2 was pure garbage if anyone was considering it)
 

pictsidhe

Well-Known Member
Joined
Jul 15, 2014
Messages
7,521
Location
North Carolina
Put three rotary encoders with toothed pulleys on fixed points above your CNC contraption. Run single ended belts from your CNC contraption over the encoders and keep them taught with a weight. With processing, you can now determine to high accuracy where you CNC contraption currently is. You feed that information as feedback to the desired trajectory into your servo controller. You'll need to correct for the catenary of the sensor belts and their their stretch, but that's not too hard. An Arduino would be plenty of processing power. Now you can use simple angle iron rails.
 

Hephaestus

Well-Known Member
Joined
Jun 25, 2014
Messages
1,752
Location
YMM
Doesn't make much sense, switching back to angle iron is like moving 15 years back in both printing and routing design. 2020 vslot or tslot has proven to be much better (or 2040 for routers), and is generally cheaper (if you're ordering less than 20' lengths)

The lighter simpler machines are working better due to the lower inertia of the system. Partly why my creality got so heavily modified - adding a significant amount of weight at the extruder resulted in a lot of work to bring the print quality back to acceptable.
 

Vigilant1

Well-Known Member
Lifetime Supporter
Joined
Jan 24, 2011
Messages
4,953
Location
US
Does any system use a sticker with bar coded optical locating on the gantry and a sticker on either side of the bed (to eliminate racking) also read optically? Mylar stickers can be darn accurate, and this eliminates all slop due to drive belts/cable sag/stretching, etc.
Since this would be a very obvious approach, I'm sure it has either been superceded or has flaws.
 

pictsidhe

Well-Known Member
Joined
Jul 15, 2014
Messages
7,521
Location
North Carolina
Doesn't make much sense, switching back to angle iron is like moving 15 years back in both printing and routing design. 2020 vslot or tslot has proven to be much better (or 2040 for routers), and is generally cheaper (if you're ordering less than 20' lengths)

The lighter simpler machines are working better due to the lower inertia of the system. Partly why my creality got so heavily modified - adding a significant amount of weight at the extruder resulted in a lot of work to bring the print quality back to acceptable.
Rails need to be straight and stiff. Making a 20' 2040 rail straight and stiff is going to be very difficult. You will need a lot of mounting points to a stiff structure, perhaps some heavy angle iron? If you are relying on steps to position everything, you are going to have to straighten it up with shims etc.
Whereas is you sense where the thing is in space, you can run directly on the angle iron that you need anyway...
Another option is to measure the imperfections in rails. If you know how they deviate from straight and level, your controller can compensate. Remember that wear is a characteristic of moving machinery, as Scott has mentioned above.

Mylar stickers with high resolutioin probably aren't cheap. While a laser printer can theoretically print one, I'm not sure just how accurate they will be. Could be worth a look, though. Yes, you can run at least some grades of polyester through a laser printer.
Perhaps a rubber belt could be bonded to a rigid surface, you'd probably need to rethink the sprockets used, though.
 

Hephaestus

Well-Known Member
Joined
Jun 25, 2014
Messages
1,752
Location
YMM
20' is going to be a massive challenge no matter what.

50x100 would still be less than a 2x2x3/16 angle.
https://8020.net/25-5010-black-fb.html

I'd hate to be the one squaring it up though. 4x8 is difficult, 20' seems insane.

Didn't shurline used to run a laser etched encoder along an axis at one point? But then dust/coolant becomes an issue to visibility
 
Top