Very interrested in the sketches.
I discarded sleeve bearings because I thought there would be too much friction/resistance ( the machine being belt driven, so fast, but I've never seen some used so that's just my imagination at work ), and though of using ceramic linear plain bearings instead, but later on discarded these because I wanted something that can be made using contrapteur parts ( u element ).
If my u-element based design can't work, I think I'll go with the ceramic plain bearings, but I'm still interested on seeing how the sleeve bearings work.

About stiffness, that's the concern that made me do a open-at-bottom design, so that the rod can be supported all the way along/bellow ( with a t-angle, like in : http://www.mcmaster.com/#linear-shafts/=986160 ), so that it can't bend. I even thought because it is fully supported, I can use cheap copper tubing ( which is very smooth ) as the rail. But I'm not sure about that.

I did read the metric version thread/pages a long time ago, that's what had me end up with contrapteur.
About how close to the original it will be, I'd like it to be just a sized-down-to-openstructurse version, but that still sticks to every contraptor idea ( like http://www.contraptor.org/component-design ). If I have the time and I don't run into too much technical difficulties, the goals are to redo every subset, redo every example, redo the mini-cnc, and do my big-cnc, and document all of that somewhere.

Just sent you the sketch with sleeve bearings; not sure about the friction there as I haven't actually tried them. I think it will depend on alignment and quality of smooth rod.

U-carriage makes sense though since you're talking about supported rail. One thing to note is that the rail material must have sufficient hardness because continuous rolling action of the bearings will chew away softer material (aluminum, copper). For 2 lower bearings, their edge is in contact with the rail - this will quickly create grooves even in (unhardened) steel. Hardened rods in that link are pretty expensive, especially pre-drilled ones. I usually try to estimate the attractiveness of a design by cost per linear ft, and there is some threshold since the assembly may become almost as expensive as industrial grade LM rails+bearings (like THK). If there is a way to position two lower bearings at an angle, then I think plain steel tube might work.

Z

About the Z axis, it really needs the same kind of accuracy as the Mini-cnc has, so servo won't work.

XY

On XY, I need also sub-millimetric precision, so 1mm precision won't work.
But I can give up speed, I don't need the 300 ipm the XY plotter has, 60 ( or maybe at worse 40 ) ipm when drawing would be fine, I need the speed more for non-precise operations like moving from a shape to another and refilling paint.
The question is, can I get ( even if it takes time to adjust everything ) the same amount of precision as the mini-cnc ( no visible errors when drawing with a .05mm pen ) using belt drives, sliding elements, and with a 1200mm course ?

If not, I'll need to change the design :(

ACME

The speed you get here would be fine, so maybe ACME rods is the solution, but they are expensive ( is this cheaper than mcmaster or am I missing something ? ), so it would be good to have a belt drive design.

Plotter

Also here :

You seem to get a good precision with just half stepping ( I'll have 1/16, probably later 1/200 ) and at a very fast speed.
And here you get a precision just a bit under what I need.

That's what gets me thinking I can adapt the design to my needs.

I'm kindof dependent on examples I can find and on knowledge I can borrow for this part of the design so any help is welcome.

Preview

Started working on a draft in sketchup :

Also, Sketchup File ( click "Télécharger ce fichier" ).

Short answer: I'm skeptical about achieving sub-mm accuracy.

Long answer:
There are several things affecting accuracy:

One question is sufficient frame stiffness.

Assume the weight of the moving gantry to be 1 kg, or 2 lbs, to simplify the calculations. When positioned in the middle of the ~1200mm X rails, the max deflection of each X rail should be:

(1)

W is the weight, 1 lb on each rail (simplifying)
l is the length, ~48 in
E is modulus of elasticity, 10,000,000 psi for aluminum
I is area moment of inertia, 0.021 in^4 for Contraptor angle (will be even less for 20 mm angle)

This calculates to ~0.01 in, which is ~0.25 mm. This is probably OK for drawing. However, milling is likely to be out of the question as the milling loads are higher. Accelerations will also load the frame, depending on how fast and how much mass you're accelerating. If you decide to build Y on moving Z, the weight (and deflection) will be even more.

Another set of issues is with the sliding elements:

First, you have to make them somewhat loose if you want to use them with the belt drive. Naturally, this will result in a small amount of play that will add to the error. This also means that you can't really put any twisting/rotational load on a single element - such as the Z axis with motor on Y carriage like your sketch shows. A simple pen, maybe with a servo could work, but an entire Z axis would be too heavy and would result in a wobble.

Second, the belt is not entirely stiff and it takes a certain amount of force to overcome static friction to get the sliding element moving, even when they're loose. Imagine a heavy brick on a pavement as a sliding element and a long circular rope on two pulleys attached to either side of the brick. When you rotate one of the pulleys, the brick is moving. But when the direction is reversed, there is some slack in the rope that gets taken up before the brick starts moving in the opposite direction. This is backlash. Some of it is due to the rubber spider from the motor coupling (disk coupling would be the remedy for this). IIRC, on XY plotter (with standard motor coupling) it was ~1/32" (11"x17" work area). Longer belts and heavier stages will exhibit more backlash. Add backlash to the error. Also, I don't think microstepping will help here as the losses are in the transmission mechanism.

You could use screws instead of the belts, but the twisting/rotational load on a single element (Y axis) will still be an issue. Also, each sliding element should be driven individually, otherwise the stage may skew and bind. Driving each element individually would double the length of the screws. The cheap plain threaded rod will give you about 15-20 in/min and I think it's too slow for such a big work area. Fast ACME screws are pretty expensive - the one in the link has 1 start and moves 1/16" per revolution (25% improvement over threaded rod), while the one in the video has 4 starts and moves 1/4" per revolution.

One suggestion may be to make Z stage a moving table on 4 sliding elements + threaded rods (a-la Reprap Darwin). Another is to make gantry the longest axis.

To sum up, I think the Dremel/milling is out of question and I'm still skeptical about sub-millimeter accuracy on XY for drawing (assuming no load on the tool head), but the only way to tell is to try. You could run some tests on 18" long rail of Contraptor, loading the sliding element with the weights and measuring the error. The belt is included :) If you can get imperial angle locally, you could try the same test on a 4 ft length with the longer belt - this would be closer to what you're trying to achieve.

Thanks for all the technical help, it's super usefull !

So I give up on belt drives, it now seems non-realistic for the precision I want.
I'll use ACME rods, it also will make things simpler as I already know the threaded rods drive from the mini-cnc.

From what you say I'm also guessing switching to ACME rods also makes the use of sliding elements less problematic.

For the problem with twisting on Y, I'll change the design to one where no axis has to support Z, that is Z supporting X supporting Y ( or maybe a moving table like in darwin, but I think this could cause problems the darwin has not if I put much weight on the table ), so I will have less weight, so less problems when moving a pen around, and when using a drill. ( I made some tests about the twisting with contraptor angle and sliding element, and it seems that for the drill I intend to use won't cause problems ( it's about 1/3 the weight of the dremel ) if everything is well tight ).

These changes solve the XY precision problems due to belt drives.
So that leaves the problem of frame stiffness.

First I'm changing it to the configuration where there is the least weight on the longer axis ( Z supporting Y ( 800mm ) supporting X ( 1200mm ) ) so Y just has to support the pen/drill, X just has to support Y and the pen/drill, and Z does all the hard work.
Second, I'm guessing that changing the supporting angles to iron angle will improve stiffness ( x2 ? 3x ? ), and I have acces to cheap iron angle. ( or does the increased weight cancel the win with stiffness ? )

It would look something like this ( no motors or rod yet, dark angle is iron ) :

I made a little bit of testing with a 20mm|1.5mm aluminium angle I have ( big-cnc will be 20mm|2mm ) that is 800mm long ( same as the Y axis ).
Putting a 800g can in the middle makes it bend by about half or a third of a millimeter.
Putting the small 12v drill I intend to use on it makes it bend a very little bit, but nothing I could measure ( I'd say 10 times less, not sure ).
I'm not sure how these phenomenons would influx on the precision on the machine : is it just a problem for the Z axis precision, or for all axis ?
Taping a pen to it makes it not bend at all, so no problem with drawing :)
Testing further would require buying empiric angle ( difficult and costy ) or building the machine itself :)

I estimate the 1200mm aluminium angle will be about 220g. That's 110g for each Y stage to support ( plus the tool weight, which is ~0 for pens ).
I'm not sure how much that would be for iron angle, but my current guess is that changing to iron angle will make the structure stiff enough to not have problems. Am I right ?

I don't change the choice of sliding elements for the stages, as I don't see what other ( documented ) contraptor moving element I could use.

The picture is the current design idea following your advice, I don't see for now what else I could change without replacing contraptor parts with costly parts.
Do you think this would work ?

Thanks !!¹²

While the threaded rods will improve motion accuracy (at the expense of speed and/or $$$), they won't help with frame stiffness. You keep mentioning drill/dremel - while this large structure may work for drawing, the accuracy will rapidly decrease as all kinds of forces are applied to the toolhead. What do you want to cut with it?

Also, don't forget that the ACME nuts for Contraptor are not manufactured - I had to make mine from a Delrin/Acetal bar.

With drawing toolhead, I would likely use the belt drives, trying to improve accuracy as much as possible - by tightening the belts, using disk couplings and limiting acceleration. Mainly because threaded rods would be too slow and ACME screws would still be slow and too expensive. Finally, linear bearings can be used instead of the sliding elements to eliminate the issues with friction - these should help achieve sub-mm accuracy.

If I was building something this size for even light milling, I would probably use 1x3" or even 1.5x3" T-slot. The angle is way too flexible for this.

Regarding steel - yes, in many cases what you would gain in stiffness will be nearly cancelled by the same gain in mass - unless it's a stationary assembly. Also, I haven't tested angle iron with sliding elements at all - and I think there can be issues - such as lack of true 90 deg. angle or lack of smooth finish. So can't really say whether it would work.

Thanks for your help and time !

The machine will be used about 95% for drawing, and 5% to drill soft materials ( like plywood or plastics ) or engraving ( like circuit boards, if possible ).
Whatever I end up with, I will try to drill aluminium ( super slowly ), but if it does not work it's not a problem.

So basically what I understand is that I can use belt drives, if I solve the problem of stiffness and if I don't use sliding elements but something more smooth like the linear bearings.
The problem with linear bearing is that I don't find a metric equivalent to the "flat U" it uses ( only "square U" ).

So I searched for an alternative and I found this :
http://new.chambrelan.com/EN/product/E28-G40/linear/rail/steel/
It's a linear bearing with "ball recirculation". It can support 20kg ( or 10kg with smaller and cheaper stage ) when positioned vertically. ( they even provide sketchup components ).
I asked a quote, it would cost 35 euros for 800mm rail+stage and 45 euros for 1200mm rail+stage, so total 115 euros, which is fine for my budget.
What do you think of those ?

But this still has the problem that it must be fully supported to not bend ( I guess ), so it does not solve the stiffness problem by itself.
I could use vertically positioned flat iron bar with dimensions anywhere from 30x3mm to 30x20mm to support it, the question is : how do I know what dimension will not bend. Also see at the bottom for questions about shapes.
In the formula you gave me I don't see where this ( dimensions of meterial ) is taken into account ( and Wikipedia is not very pedagogic on the subject ), and I don't even know if it has influence ( my instinct tells me it has, aluminium angle being stiffer than aluminium foil ) ( edit : finally found it it's I ).

Any advice on that part ?

Thanks !!!

Edit : after hours of searching I finally found this : http://en.wikipedia.org/wiki/Second_moment_of_area
So that means I can get a better stiffness if I choose a better geometry ( I guess that's why you were suggesting T-slot ), the question is : what geometry is best ? ( T ? I ? box ? more complex ones like T-slot ? ) ( I'd love that square tube be a good solution as it's the one I have the easiest access to ( same shop as the angles ), and it's easier to integrate ).
This says that I is best for one direction of bending, and square tube is best for two directions of bending.
I'll try to calculate this, but my math is rusty.

Shape	Picture	Source	Formula
Square tube		calculator
Rectangle		calculator
Angle		page
T beam		page
I beam		page

Note : y is "distance of axis to extreme fiber".
I'm not sure where it is … I'm assuming it is in the center, if I'm wrong please tell me so I can recalculate.

So now applying to real things ( what I can get is in bold ) as candidates for the big-cnc rail supporting elements :

Element	Dimensions	Calculation	I
Square tube	20x2	(20^4−16^4)÷12	7872
Square tube	30x3	(30^4−24^4)÷12	39852
Rectangle horizontal	20x2	(20×2^3)÷12	13,33
Rectangle vertical	20x2	(2×20^3)÷12	1333,33
Rectangle vertical	30x5	(5×30^3)12	11250
Plain square	20x20	(20×20^3)÷12	13333,33
Plain square	30x30	(30×30^3)÷12	67500
Square Angle	20x2	((2*10^3)+(20*(20-10)^3)-((20-2)*((20-10-2)^3)))/3	4261,333333333
Square Angle	30x3	((3*15^3)+(30*(30-15)^3)-((30-3)*((30-15-3)^3)))/3	21573
Estimated Contraptor	25x3	((3*12.5^3)+(25*(25-12.5)^3)-((25-3)*((25-12.5-3)^3)))/3	11941,75
Contraptor ( inches )	1x(1/8)	((0.125*1^3)+(1*(1-0.5)^3)-((1-0.125)*((1-0.5-0.125)^3)))/3	0,0679, does not match your number I must be wrong about y
Contraptor ( inches, y = a )	1x(1/8)	((0.125*1^3)+(1*(1-1)^3)-((1-0.125)*((1-1-0.125)^3)))/3	0,042, still does not match, let's try with y=0
Contraptor ( inches, y = 0 )	1x(1/8)	((0.125*1^3)+(1*(1-0)^3)-((1-0.125)*((1-0-0.125)^3)))/3	0,179, still not 0.021, I'm lost now, I must be doing the math wrong … computing the rest just to compare, not actually use
T	20x2	Same as angle	same kind of results
I	20x2	((20*20^3)-((16^3)*(20-2)))/12	7189,33
I	30x3	((30*30^3)-((24^3)*(30-3)))/12	36396
Chambrelan E28 rail ( simplifying to a block )	28x9	((9x(28^3))/12	16464

So, the results when trying to compute for contraptor tells me I'm wrong somewhere ( so I won't try to calculate deflection until I know what and correct ) , but anyway I can still see some thing from comparing these numbers even if they are incorrect :

• T is same as angle ( same equation ).
• Worst is horizontal flat rectangle.
• Vertical rectangle is not that good either at compared weights.
• Square tube does well compared to plain square, that sticks with what wikipedia says.
• I is about as stiff as square tube, but square tube has 1/3 more weight ( but I can buy square tube easily and 1/3 weight is not that bad ).

All that tells me : best for me is 30x3 square tube ( same as for the sliding elements stages, that's two in one ).
But I need to get the math right to calculate the deflection and see if it would be good enough.

The linear bearings in the link look good, and the price seems reasonable - works out to be ~$15/ft for a single rail + bearing.

IIRC, the mysterious y in the formulas is the coordinate of centroid for a given cross section: http://en.wikipedia.org/wiki/Centroid

To calculate the area moment of inertia, you can use this Sketchup plugin: http://www.alexschreyer.net/projects/centroid-and-area-properties-plugin-for-sketchup/

And yes, the tube offers the best stiffness/weight ratio.

Hi !

The plugin seems broken : it does not place the centroid of a square tube at the meeting point of its two diagonals ( the plugin's author suggests there is something wrong with the algorythm used ).
If that is what you used to get the area moment of inertia for contraptor angle, maybe that's why we get different numbers.

I'll assume the equation I have for the square tube is right.

Let's calculate deflection for different shapes, materials and lengths, with 1kg ( 9.8 N ) of weight :

• W is load in N : 9.8
• l is length is meters, variable
• E is modulus of elasticity in N/m^2 , 69000000000 for aluminum, 196000000000 for iron
• I is moment of inertia, in m^4, I calculated it for mm^4, so I need to recalculate it

Shape	Dimension	Moment of inertia	800mm alu	1200mm alu	800mm iron	1200mm iron
Square tube	20x2	((20/1000)^4−(16/1000)^4)÷12 = 7,872×10⁻⁹	( 9.8 * ( 0.8^3 ) ) / ( 48 * 69000000000 * (7,872×10⁻⁹) ) = 1,924511999×10⁻⁴ m = 0,19mm	(9.8*(1.2^3))/(48*69000000000*(7,872×10⁻⁹)) = 6,495227996×10⁻⁴ m = 0,64mm	( 9.8 * ( 0.8^3 ) ) / ( 48 * 196000000000 * (7,872×10⁻⁹) ) = 6,775067751×10⁻⁵ m = 0,067 mm	(9.8*(1.2^3))/(48*196000000000*(7,872×10⁻⁹)) = 2,286585366×10⁻⁴ m = 0,28 mm
Square tube	30x2	((30/1000)^4−(26/1000)^4)÷12 = 2,941866667×10⁻⁸	( 9.8 * ( 0.8^3 ) ) / ( 48 * 69000000000 * (2,941866667×10⁻⁸) ) = 5,149709409×10⁻⁵ m = 0,051 mm	(9.8*(1.2^3))/(48*69000000000*(2,941866667×10⁻⁸)) = 1,738026925×10⁻⁴ m = 0,17 mm	( 9.8 * ( 0.8^3 ) ) / ( 48 * 196000000000 * (2,941866667×10⁻⁸) ) = 1,812907904×10⁻⁵ m = 0,018 mm	(9.8*(1.2^3))/(48*196000000000*(2,941866667×10⁻⁸)) = 6,118564176×10⁻⁵ m = 0,061 mm
Chambrelan rail unsupported	28x9	(((9/1000)*((28/1000)^3))/12 = 1,6464×10⁻⁸	( 9.8 * ( 0.8^3 ) ) / ( 48 * 69000000000 * (1,6464×10⁻⁸ ) ) = 9,201748332×10⁻⁵ m = 0,092 mm	(9.8*(1.2^3))/(48*69000000000*(1,6464×10⁻⁸ )) = 3,105590062×10⁻⁴ = 0,31 mm	( 9.8 * ( 0.8^3 ) ) / ( 48 * 196000000000 * (1,6464×10⁻⁸ ) ) = 3,239390994×10⁻⁵ = 0,032 mm	(9.8*(1.2^3))/(48*196000000000*(1,6464×10⁻⁸ )) = 1,093294461×10⁻⁴ m = 0,109 mm
Square tube	40x3	((40/1000)^4−(34/1000)^4)÷12 = 1,01972×10⁻⁷	( 9.8 * ( 0.8^3 ) ) / ( 48 * 69000000000 * (1,01972×10⁻⁷) ) = 1,48567827×10⁻⁵ m = 0,014 mm	(9.8*(1.2^3))/(48*69000000000*(1,01972×10⁻⁷)) = 5,014164161×10⁻⁵ m = 0,05mm	( 9.8 * ( 0.8^3 ) ) / ( 48 * 196000000000 * (1,01972×10⁻⁷) ) = 5,230193909×10⁻⁶ m = 0,0052 mm	(9.8*(1.2^3))/(48*196000000000*(1,01972×10⁻⁷)) = 1,765190444×10⁻⁵ m = 0,017 mm
Square tube	40x2	((40/1000)^4−(36/1000)^4)÷12 = 7,336533333×10⁻⁸	( 9.8 * ( 0.8^3 ) ) / ( 48 * 69000000000 * (7,336533333×10⁻⁸) ) = 2,064975073×10⁻⁵ m = 0,02 mm	(9.8*(1.2^3))/(48*69000000000*(7,336533333×10⁻⁸)) = 6,969290871×10⁻⁵ m = 0,069mm	( 9.8 * ( 0.8^3 ) ) / ( 48 * 196000000000 * (7,336533333×10⁻⁸) ) = 7,269555104×10⁻⁶ m = 0,0072 mm	(9.8*(1.2^3))/(48*196000000000*(7,336533333×10⁻⁸)) = 2,453474847×10⁻⁵ m = 0,024 mm
Rod 10mm	10	(π×(10÷1000)^4)÷64 = 4,908738521×10⁻¹⁰	(9.8*(0.8^3))/(48*69000000000*(4,908738521×10⁻¹⁰)) = 3,086283449×10⁻³ = 3,08mm		(9.8*(0.8^3))/(48*196000000000*(4,908738521×10⁻¹⁰)) = 1,086497745×10⁻³ = 1,08mm
Rod 16mm	10	(π×(16÷1000)^4)÷64 = 3,216990877×10⁻⁹	(9.8*(0.8^3))/(48*69000000000*(3,216990877×10⁻⁹)) = 4,709294814×10⁻⁴ = 0,47mm		(9.8*(0.8^3))/(48*196000000000*(3,216990877×10⁻⁹)) = 1,657863991×10⁻⁴ = 0,16mm

Ok, so 30x2 iron square tube seems very good. I'll try to design a sketchup draft using that.
I still need to calculate how much the finished 1200mm rail weights to compute how much deflection it will cause to the 800mm rails.

But from here it seems that this solves the stiffness issues.
Is that right ? ( if it's not either my math is wrong or I need carbon nanotubes :( )

Thanks !

( PS: I'm away to Paris, so no more questions for 5 days :) )

Edit : First draft using the chambrellan elements and square tube as support ( sketchup file ) :

It looks like the plugin may not work correctly for shapes with inner polygons, but the result for angle is consistent with the numbers I remember from early calculations.

I'm not sure I can answer the accuracy question for milling. You would need to model your system taking milling forces into account. In addition to bending, the beams may also twist, forces will multiply under moment arms (vertical rails, toolhead on X axis), workpiece materials will resist differently.. Milling foam accurately would be a lot easier than milling hard wood, but if you go really really slow you could probably even mill aluminum.

In other words, it's difficult to tell what sort of accuracy you'll get without thorough analysis of many parameters. Or trial and error approach which was one of the ideas behind Contraptor :)

Also, when you have larger area, the first thing you'll want is speed - it's very demoralizing to watch the tool/pen take full minute or two to go from one corner to another :) This means you'll likely need more powerful motors (and drivers, and power supply) to move the heavier axes around at decent speeds without losing steps.

Thanks Vitaly!

About milling, I'm more and more seeing it as a bonus, I'll be happy if it works, but I don't worry about it.
Also I changed the supporting bars to 40x3 ( more 20mm-y, also fits better/perfectly the chambrellan stage ) square tube instead of 30x2, which should give even better results ( calculations gave 1/100mm deflection, but I don't know the math for the other forces, so I'll try and see ).

So I hope I can start ordering things next week, and maybe start drilling the first angles before the holidays :)
But for that I need to finish the sketchup design, and find where to order everything.

Here is where the design is now ( sketchup file ) :

Notes on various things, tell me if I'm doing something wrong somewhere :

• Motor mount.

I did a lot of work on my mini-cnc to try to reduce vibration and noise.
One of the big problems I have noticed, is that not all screws of the motor are attached.
Also, any part of the motor touching the structure transfers the vibration and makes things lots worse.
After trying several things here is what I ended up with :

( you have to imagine the 4 screws, isolated with rubber ).
The motor is separated from the mount by a 1-2mm thick rubber sheet ( bike tires ? foam ? needs trying ).
It puts the center of the motor axis at 30mm from the "floor", that solves the inch-metric problem you can have with the nema motors.

• Motors

I'll build a first version using the same motor/driver configuration as the mini-cnc, and see how fast/precise I can get it, just to see.
Then I'll change the motors/drivers to something bigger like this :
http://cgi.ebay.com/CNC-NEMA23-Stepper-Motor-Driver-2M542-4-2A-Controller-/170534793478?pt=LH_DefaultDomain_0&hash=item27b4aa7106#ht_7181wt_1141
I have a 7.8A version of the same design, I never used it at more than it's 4A setting ( works very nice, smooth sound, no vibration, high torque ), so I'll get 3 of the 4.2A version and corresponding stepper motors, probably something like this :
http://www.alltronics.com/cgi-bin/item/28M027/55/Lin-Engineering-5618L-52PD-Bipolar-stepper-motor
The brain will still be a sanguino.

• Pulleys

I'll need 10 of these ( 4Z 4Y 2X )
I'm looking for the same thing as contraptor ( XL 10 teeth ), as all the metric sized belts I found were way more expansive, so empiric belt, so empiric pulleys.
I googled for a long time and found 3 vendors that were not too expansive : 

1. mcmaster 7,40$ each
2. econobelt 5,14 each, but with no screws for fixation so I must learn to do that myself.
3. sdpsi 5,17$ each, not very verbose on what I'll get in the end, like screws etc …

I prefer the last two of course for the price, but mostly because they also sell belt clamps, wich I need and did not find at mcmaster. But mcmaster has the nice screws for fixation …

• Belts

I need : (1,2×2)+(0,8×4)+(1,2×2)+(0,8×2) = 9,6 = 10 meters of belt.
Prices in the 3 stores : 

1. mcmaster 2,50$/foot, 10 meters is 32.8 feet, that's 82$
2. sdpsi 30$ for 100 inches, 10 meters is 393 inches, so about 120$.
3. econobelt 26$ for 5 meters, so 52$ for 10 meters.

So I think I'll go to econobelt, but I'm not sure I understand all the parameters correctly so mcmaster is maybe where I have the less chances of error.

• Belt clamps.

About 5$ at sdpsi and econobelt, I'm not sure how I could DIMyself.
I think I can use belt clamps even for the Z axis darwin-like system ( the clamp must not touch the pulleys, so I must move the Z axis with no longer than about 1100mm, that is about 55 pulley turns, that is 55mm of freedom for the Z axis, which is enough ).

That's all for tonight, thanks for reading.

Ok so no "NOOOO DONT DO THAT!!", I'll be presuming there was no big mistake in the message above.

Now below is the 0.9 version of the design, it's as far as I can go with my knowledge, if it was just me, it would be what I would build, and see if it works.
But please please please tell me if something is done wrong !

Picture !

And sketchup file.

What is still missing :
- Bolts and nuts ( booooooring, won't add in sketchup )
- Z motor needs better fixation.
- X and Y motors need support bellow when upgrading to bigger ones.
- Motors will have double shafts, so I'll put flywheels/dampers there.
- Fixation for the ball bearings ( not sure I'll do it the way contraptor does, I have a few ideas I need to try ).

The machine will be mounted with bolts/nuts to a 1200mm/800mm table, the dimensions are made to fit that. ( just starting to realize how big it is, I need a bigger house ).

Tomorrow I'll be starting a bill of materials.

The ball bearings will be 6x12x4 ( fits the angles size even on the inside ).

I think the angles will have 6mm holes instead of 5mm, I chose 5mm randomly in the beginning, but 6mm fits better the threaded rods, pulleys, ball bearings, etc … and gives the possibility to use 5mm nuts where freedom is needed.

This kind of thing looks so much easier in the Iron Man movie … but it feels as cool :)

Ok, so starting the BOM ( don't miss the message above ) :

Underlined, what has been ordered.

Category	Item	Minimum detail	Minimum total	Minimum price	Actual order total	Actual order price	Comment
Metals	20x20x2 aluminium angle	6x angle-20mm, 18x angle-40mm, 1x angle-60mm, 10x angle-80mm, 4x angle-200mm, 1x angle-1370mm, 2x angle-1280mm, 2x angle-860mm	(6×20)+(18×40)+60+(10×80)+(4×200)+1370+(2×1280)+(2×860) = 8150 mm	8,3293 €	30000mm	30,66 €	Basic angle
Metals	40x3 iron square tube	1x 910mm, 1x 890mm, 1x 1300mm, 4x 80mm	910+890+1300+(4×80) = 3420 mm	14,80 €	6000mm	20,85 €	Supporting elements
Metals	30x30x2 aluminium square tube	4x 40mm	160 mm	0,48 €	2000mm	6,10 €	Sliding elements
Metals	60x2 iron square tube	2x 60mm	120mm	0,43 €	1000mm	3,63 €	For motor mounts, cut in half
Plastics	UHMW-PE channel	4x 20mm	80mm, minimum 1 foot	2,63 $ = 1,98 €	5 feet	13,15 $ = 9,93 €	1/8 inches = 3.175 mm, we need 3mm min
Mechanics	10 teeth XL pulley, 6mm bore	4Z 4Y 2X	10 units	50.20 $ = 37.94 €	15 units	75.3 $ = 56.91 €	Cant find 6mm bores with screws … MUST NOT forget, stepper pulleys have 8mm bore, wait no, they have 6,35mm bore … ?! must check with seller
Mechanics	XL 3/8" belt	Z ( 1260x2 + 820x2 + 200 waste = 4360mm ) + 2x Y ( 880x2 + 100 waste = 1860 mm ) + X ( 1200x2 + 100 waste = 2500mm )	1x 1064 teeth + 2x 394 teeth + 1x 516 teeth	26.47$ + 2x10.81$ + 13.01$ = 61,1$ = 46,18 €	+2x 1064 teeth	+ 2x26.47$ = +52,94 = 114,04$ = 86,19 €
Mechanics	XL 3/8" belt clamp	1+1+1	3	5.33$x3 = 15.99$ = 12.08 €	6	24,16 €
Mechanics	6x12x4 ball bearings	7Z + 6Y + 2X	15	0,54*15 = 8,1$ = 6,12 €	100	54,95$ = 41,56 €	Minimum order 100
Mechanics	Chambrellan rails E28+G40	2x800mm + 1x1200mm		149,26€		149,26€	Will have to cut a bit on all rails, price incl. shipping
Partial total				277,59 € = 370,19$		429,25 € = 570,77$	That's without shipping for all except chambrellan

Update : 

• Just ordered the chambrellan rails … I think the shipping is free to france … no going back now !
• Ordered the 100 ball bearings, shipping 10$ for france
• Just asked quotes from several local machine shops about milling/cutting the angle, we'll see how that turns out …

Just got a quote for the electronics/motors parts.

It'll be 390$ inc. shipping for 3 double shaft 425oz/in NEMA23 motors, 3 4.5A 256steps motor drivers, and a 400W 36V power supply.

I may be adding soon diagonal bars ( like on the mini-cnc ) to the design to improve stability, seems after reflection like it is needed.

Update:

• Got a quote for XL 3/8" belt : 41$ for 30 meters of belt, way cheaper than econobelt ( wich is more like 150$ for 30 meters, mcmaster is more like 240$ for 30 meters ) ( still need to count VAT, customs and shipping, waiting on a quote for shipping ).

Agrr, it's not very convenient to have to scroll all the way up in order to reply to the thread..

Anyhow, I still do have some concerns. For one, I think steel tube is overkill, harder to work with and weighs a lot more - I would go with AL.

Another thing is I'm still not convinced that the frame is sufficiently stiff. Imagine four legged table up side down and try to move the legs - they're up in the air and fairly flexible.

Also, if you're going for 4cm beams, why not use them throughout the entire frame?

Yes there are no holes I will have to make them myself.
There was an option for holes at order, but it was too expensive.

Also, they are bit noisier than what I was imagining, but still it's quiet enough.

I think you can make it work even with closed belt :
You never have belt all around the pulley, only on 1/4 or 1/2 of the pulley, so you just put the belt around the pulley, the pulley on the rod, and then theoretically you still have access to the screws to fix it in place.
It may put you in a complicated pose for screwing it, but it is still possible in most situations, no ?

Yes, possible, just might be a bit inconvenient with L shaped hex key

Although.. it may only need quarter of a turn to tighten, so no big deal.