That indicates that for machines with long belts (e.g. deltas) the belt stretching can be significant part of corner errors at high printing speeds. Depends on the stepper motor rotor inertia in comparison to the inertia of the rest of the printer moving parts.

I see this on my large printer, I can't run the belts as tight, and with their length (2 meters each), and those long diagonals, it really can create issues. I have to dial down travel speeds, acceleration and deceleration quite a bit to keep it from causing problems, about 1/2 to 2/3rds the Rostock with it's 1.4m belts.

[re: leslieann

3d printing is just motion control with a filament extruder. motion control is a mature problem and the practical solution that people have come up with over the last 30 years is hand programming. sad but true. i personally think that's garbage and that all motion control can be done better faster etc. with smarter compilers; like vvp is saying, the physical parameters of the printer need to be given to the runtime and/or compiler. further, i would insist that a _lot_ more open loop feedback needs to happen.

as far as whether this technology can is beyond the scope of hobbyists, i really don't think so. hobbyist affinity really depends on whether technology is packaged up and black-boxable. it is less about total complexity and more about exposed complexity

that's my 0.02USD anyway.

I'm not saying the problem can't be solved, just not for the money most people are willing to spend on a printer. Particularly one with limited material.

These companies are focused on mass market appeal to regular consumers, which frankly, I think it's a stupid idea.  You can market these to geeks, tinkerers and makers, but to try pushing them into the average home is a complete joke.

I see this on my large printer, I can't run the belts as tight, and with their length (2 meters each), and those long diagonals, it really can create issues. I have to dial down travel speeds, acceleration and deceleration quite a bit to keep it from causing problems, about 1/2 to 2/3rds the Rostock with it's 1.4m belts.

Is it using GT2 belts you mentioned in the past? Do they have glass fibre core?

It looks like most GT2 belts sold use glass fibre core and most T2.5 belts use steel core.

GT2 belt producer claims GT2 should has more precise belt-pulley meshing compared to T2.5. And also GT2 (with glass core) should last longer too. I can agree with them about these too. They recommend it for precise positioning. That maybe be OK but only for short belts (if we are talking bout GT2 with glass core).

Glass core GT2 is not so good for deltas. From this point of view the reprap wiki page on belts is rather misleading by pushing GT2 heavily without even mentioning the problem of low modulus of glass core GT2.

Wolfram alpha claims these young's modulus values:
glass - 69GPa
kevlar - from 70 to 179 GPa (depends on type)
steel - 205 GPa

Looks like steel core is the way to go, maybe kevlar if the belts use the best one (DuPont Kevlar 149).
It would also help to use wider belts than the common 6 mm width.

I did first 3 prints with a fan mounted on the platform:

• Printing on support (i.e. a kind of a bridge) seems to work better. Still, I believe we need an option in cura to slow down first layer on support. I still do not know whether slic3r is better at this. It is important especially when the first layer on support has considerable area, i.e. it is not only a few mm of filament being laid down.
• It looks like corners of big prints peeling off from bed (before the print is finished) is not such a big problem when fan is slowly blowing at the part during the whole printing.

This is all about ABS only. I did  not even try PLA yet.

So it looks like it was a good idea to mount the fan on. Even for ABS only.

nice! that should print nicely with solid HIPs support. i'm still too angry at my printer to bring it up

regardless, more cad drawings and drafters are always welcome here!!

in fact, the big thing i need right now is pieces of 3d printers  skunkworks!

HIPs is a support material that dissolves in pine-sol (d-limonene). you can actually print a solid block that consists of the HIPs material completely encapsulating an ABS part (for support of the abs part) and then dissolve the HIPs material completely so that very complex shapes can be made in the ABS without having to worry about "printing on air"; without support material, one has to set up prints very carefully so that every line of extruded abs has abs below it to support it mechanically.

STL files are the lowest common denominator for printing, but solidworks files or IGES/STEP files are quite welcome.

technically, none of our printers actually have floating point units , but the slicers/gcode compilers can take as much floating point precision as you can provide when making the integer gcode instructions.

i've yet to play with the newer support renderer in makerware, but i think one can just make a solid block in solidworks and intersect it with the keycap model. the standard support functionality in the slicer tries to minimize the amount of support struts that are used, which provides less accuracy than just laying down a solid block.

this is something i have to play with to get you a solid answer though, for sure. i have some models in the backlog. once the printer is going again, i will have much more to say about supports and slicing. i've been trying to work out a long-term plan for printing and then production at medium scale based on the printed rapid prototypes, and things are still very much in limbo.

the trick with the support block is that you need two separate objects so that the slicer knows what to print in HIPs and what to print in ABS. otherwise that support block looks quite excellent.

also, boredom is just creativity waiting to happen!!

As for as what is printable on dirt cheap single extruder reprap FDM printers. Mkawa's printer is dual extruder so this does not apply to it. Rotate, position or split your parts so that whatever you do is well printable. It is easy to glue a part together from more pieces.

• No horizontal layer in air except the bridges mentioned below. Slicer needs to add support material if some part would be printed "in air". So you can do it with support material. The surface of the part which is printed directly on support will be rough with occasional significant errors (e.g 0.5 dip in the surface etc). This could be better if slicers would be more clever about printing on support and considered them more like bridges.
• Overhangs can be printed easily if the overhang angle (angle from vertical to the side of the part) is below 45°. Overhangs up to 70° can be done but they will not look good (rough surface, occasional burns). These are overhangs without support.
• Arbitrary big overhangs can be done with support. But then there is a pain of removing it. The bottom overhang surface will be quite acceptable if the filament being extruded at the overhang is at least touching (with small part of its width) the layer just below it.
• Tall thin structures are a problem. They will bent when hotend prints on them and the result will be curved. E.g. 15 cm pipe with 1.5 cm diameter would be problem.
• Bridges can be printed up to about 2 cm quite easily. Bridge is when a solid layer is printed in air, but two opposite sides of the solid layer are supported by previous layers. So the filament is hanged between the two support pillars.
• Thin walls must be thicker than nozzle size. Preferably by 20% at least. Commonly used nozzle sizes are from 0.3 to 0.7 mm. Most printers use 0.5 mm. If the thin wall are not horizontally curved they will bent (as tall thin structures.
• The thinnest printable tower has diameter of about 3 mm if it is very short. If it would be tall about 4 mm or more we would get the problem of tall thin part again. It would bent.

yes the strength using single material printing would be a problem, with a support material one can rotate it 90° and increase it that way.

thx for the info vvp,
based on it i modified and reOriented the cap a little to make it printable with your printer aswell.

This is something Leslieann knows well. Thanks for the warning.

a huge number of the printer designs i've seen just use rods that are too small everywhere. this includes the makerbot.

And the original Rostock. It uses Φ 8 mm rods and linear bearings. I'm surprised Johann did not go for Φ 12 mm rods. Based on a local supplier here the rods would be 35% more expensive and linear bearings would be 10% more expensive. Rod error specified as at most +0μm -8μm. Bearings specified as at most +8μm -0μm. It is the same for both 8 and 12 mm versions. 12 mm versions look worth it. Especially when some people put linear guides on their Kossels which are much more expensive than rods and linear bearings.

Still, I did not compute the dynamic forces. But I can do high bound estimation quickly for the Rostock here: 0.25 kg * 4 m/sē = 1 N. Seems enough to bent 0.8m long Φ 8mm rod a bit. My rough experiment with a scale indicates about 0.25 mm displacement. Not good. Looks like this may be a problem.

Edit: I computed how much the rod should bend with some simplifications which should make the result worse than in reality (simple Euler-Bernoulli bending considering two half rods instead of one rod loaded in the middle). I got the result of about 0.4 0.2 mm. That is in congruence with my experimental result - the measurement "by eye" (about 0.25 mm). Since we would like better precision than 0.1 mm, then this really may be a problem.

Edit2: And bending resistance depends on 4th power of radius so switching to 12mm rods from 8mm rods may be just enough (about 5 times stronger (from bending point of view)).

Edit3: in italic and strikethrough.

Well, those should be compensated by frame. I just assumed the frame is completely rigid. I do not have enough mental commitment to do a good model. Just some "on the napkin" estimates to get an idea how significant it may be.

That is what I should to next, stiffen the frame so that it is not twisting. Some metal beams and a stick welder is in the garage ... I just need not to be lazy to do some manual work

I actually made an error while integrating, the theoretical value for maximal rod bending should have been 0.2 mm and not 0.4. I'll fix it there too. Still looks about right when compared to my "by eye" experiment.

LOL, looks like our el-cheapo 3dPrinters are designed so they just barely work. And if one wants higher quality either he needs to print really slowly or just replace almost all the parts
Makes sense.

i can't believe you actually sat down and calculated to determine this. it's pretty obvious if you just look at it. a) well, it seems to print something b) what the hell? how could this possibly work?

c) it must be designed to the absolute limits.

i can't believe you actually sat down and calculated to determine this

Yes, the conclusion is fairly obvious
I did it for three reasons:

• simple computations like this are better fun to do than e.g. watching CSI (or most other crap on TV)
• just to check that my "by eye" estimate conforms to the model
• because I wanted to compare rod bending error with belt stretching error

I do not think that the model I used for rod bending is much good though. If I put higher forces in it then the results do not conform to experimental measurements.

OK, back to the belt stretching of the delta here (classical rostock with 6mm wide T2.5 belts with steel core). It is still true to the original except the diagonal rods were changed to carbon rods with MP-Jet ball joints. So it should be about the same for other classical rostocks. The data for stepper motors were taken from internet (I'm not sure especially about stepper motor rotor moment of inertia - I did not have it for the exact type I use so I took it from something which look the same).

The maximum belt stretching error is in the range of 0.21 mm to 0.55 mm. What I mean is: The real printer use will lead to some dynamic (temporary) belt stretching. And the maximum of that belt stretching is in the given range. I ignored pulley, idler and belt weight but I doubt they would change it in any significant way.

Btw, the maximum usable acceleration of a carriage is about 8g. At 15g, the stepper would definitely skip steps instead of moving a carriage :-) At 8g, the side force on the smooth rod is about 15N and that can bent it about 0.8 mm. Just to get you some idea what the very worst situation can be.

OK, back to the belt stretching of the delta here (classical rostock with 6mm wide T2.5 belts with steel core). It is still true to the original except the diagonal rods were changed to carbon rods with MP-Jet ball joints.

{cut}
Btw, the maximum usable acceleration of a carriage is about 8g. At 15g, the stepper would definitely skip steps instead of moving a carriage :-) At 8g, the side force on the smooth rod is about 15N and that can bent it about 0.8 mm. Just to get you some idea what the very worst situation can be.

While I haven't looked for steel core, most Rostocks and Kossels are using fiberglass core GT2 belts (not steel or T2.5). I found a chart showing stretch, but I would have to dig it up. I looked at Kevlar, but it's expensive. T2.5, especially steel would be MUCH stronger than common GT2.

As for your speed calculations...
I don't know what G's equate into mm per second, firmware lists it a bit strange and weights vary from one printer to another, but I do know there is a LOT of people claiming some insanely high speeds on their deltas. I have even heard some recent claims of 500mm per second and I can pretty much tell you they are 100% B.S. partly for exactly what you just found. You can only yank a belt and stepper so fast before it skips, stretches, or whatever.

It's pretty obvious when your printer takes a second and a half to cross a 170mm print bed that you aren't hitting 500mm per second. Even if you account for some acceleration, it shouldn't take that long. One clue many miss is when your slicer says it will take 3 minutes to print and then you start to print and now it says 9 minutes.

Don't even get me started on those claiming 300mm per second on a Kossel Mini with magnetic ball joints. The ball bearings add a LOT of weight to the moving assembly, probably doubling it. I barely trusted them at 150mm per second on my Rostock.

_everything_ causes problems, as we all know. we have all had our share of incredible frustration. now that i have a more global view of the FDM printer market, i can laugh about it (even 60 grand printers are incredibly frustrating), but my response to this is to sit down and engineer stuff. after working in research for so long, i don't really expect anything to work, ever.

Tell me about it. LOL
I would hate to do design for a living, not that I need worry, I would be fired for taking too long and far too many mistakes. 

Revision #350.... Oops... Revision #351... Oh darn...

What I have written down are absolute physicaly possible maximums for my machine (if I would allow it to run in SW to the very limits) ... and I'm not sure even they are achievable with my 24V power source (but they would be achievable with 48V power source). The reason I'm not sure here is because of the static moment charts for my steppers are missing or incomplete for 24V power source. I considered only 1/3 of the maximum static torque of my steppers but may be it should have been 1/5.

The point is when you actually want a high acceleration to lead to some error you must be able to sustain the acceleration long enough to actually move the carriages for a distance bigger than the error they should lead to and that means you must be able to run steppers at high enough speed with some torque still available. With high acceleration speed ramps up really quickly. The motor torque goes down with the speed because of wiring inductance. But to sustain the acceleration printer needs the torque.

Anyway I rerun my notebook with only 1/5 of the maximum torque and got maximum acceleration of 5.6g and maximum belt elongation of 0.3mm and maximum radial force on carriage of about 10N (and that should lead to rod bend of about 0.4mm - this is just "by eye" measurement - I'm lazy to pull the rods out of the printer and do it precisely).  For these to be achievable the maximum printer speed limit can stay at probably common 0.3 m/s
but the maximum acceleration would need to be set to uncommon 56000 mm/sē (most people probably use only 10000 mm/sē or 9000 mm/sē, which was (and probably is) the Marlin default) or it can be somewhat smaller but then jerk must be non-zero.

Most people probably use the default acceleration limit which is about 10e3 mm/sē = 10 m/s^2 ≅ 1g. But they also will probably have non-zero jerk (probably about 20 mm/s).
If they would use also zero jerk (which probably nobody has set to 0 but lets assume it for a while) then that would mean that their maximum belt stretch error for T2.5 steel core belts is about 0.07 mm. 20 mm/s jerk will probably increase it only by about 0.005 mm (just an educated guess).

So really, if people run at low acceleration limits  of about 1g and low jerk (at most few tens mm/s) then they can assume their belt stretch just below 0.1 mm. That is if they use steel core T2.5 belts. If they use glass core GT2 belts then the maximum error is about 3 times worse (i.e. 0.3mm). The rod bend error will be around 0.1 mm.

For these numbers to be valid speed limit may stay rather low at about 300 mm/s. If they use some insane high jerk (but which still does not lead to stepper motor skipping) then that may increase their maximum error by at most half step (i.e. about 0.1 mm). This all assumes their frame is rigid and no resonance happens.

Now, do I typically hit these errors. Hah, not at all because most of the time:

• my maximum acceleration is 0.4g; limits belt stretching to 0.02mm at most (about 1 micro-step)
• my printing moves are limited to 60 mm/s (I still did not improve the extruder), which is effectively the same as maximum acceleration limit of about 1.8g (from the point of view how it can influence belt stretching); limits belt stretching to 0.1mm at most
• although I typically use non-printing moves of 200 mm/s these are mostly initiated with retract and ended with retract reverse, since these take about 0.1s that is plenty enough for position to settle to the precise spot

So if I would improve the printer then the order of improvements should be:

• stiffen the frame (slight vibration of the top plate can be seen while printing), I doubt the amplitude is more than 0.1mm, I could measure it with accelerometer in my mobile - easy and simple, but I'm probably lazy
• check whether there is a play between belt and pulley, people say a lot of pulleys are not precise enough
• and I should do something with the extruder so that I can print more quickly

As for as people claiming some insane speeds. Well maybe they citied their slicer limit and forgot about lower limit in firmware :-)

your stepper moment depends on the driver and motor, as the driver does not pass off vcc to the stepper. the driver is buck charging down to a lower voltage and high current. if you buy matched drivers and steppers, the driver datasheet will give you the holding torques, but the vcc voltage of the driver doesn't really matter, since it's buck charging whatever it gets down to the stepper voltage. the advantage of giving a driver higher voltage is generally that the power supply will run more efficiently, as it doesn't have to push out nearly as much current.

It's pretty obvious when your printer takes a second and a half to cross a 170mm print bed that you aren't hitting 500mm per second. Even if you account for some acceleration, it shouldn't take that long.

500 mm/s is hight speed. If they use the common 1g acceleration then (even with 0 jerk) crossing 170 mm bed should take 0.34 s. That is really quick. Although 1g (10000 mm/sē) or 0.9g (9000 mm/sē) is the default acceleration in Marlin. It is rather high. E.g. it takes only 12.5 mm (and 0.05 s) to get to 500 mm/s with 1g.
If they use non-zero jerk then it should be even quicker.

Notice that if I lower the acceleration to only 0.5g (almost everybody has it at this or higher value) then the time to cross the bed increases only to 0.44 s. Still significantly less than a second.

If they cross 0.17m bed in 1.5 s then that would mean that their acceleration is limited to 0.03g (300 mm/sē) and the maximum speed they would reach while trying to cross it would be only 225 mm/s (far from 500 mm/s). Assuming 0 jerk. I doubt anybody uses so low acceleration.

Don't even get me started on those claiming 300mm per second on a Kossel Mini with magnetic ball joints. The ball bearings add a LOT of weight to the moving assembly, probably doubling it. I barely trusted them at 150mm per second on my Rostock.

They probably do not really know what speed they are actually running. I'm very satisfied with my ball joints. No way I would switch them for magnetic joints (especially after what you reported about them). One ball joint weights only 5g (even with the associated screw/nut to platform/carriage and the screw to the carbon rod). One finished rod with two ball joints and the nuts to connect them to platform/carriages weights only 15 g. It is stiff, strong and does not have any play I can notice. No way I could achieve this with magnetic joints.

here's a pretty detailed treatise from STM: http://www.st.com/st-web-ui/static/active/jp/resource/technical/document/application_note/CD00003774.pdf

start at page 8. the basic idea is torque is _only_ a function of stator topology and current. you're better off buck charging, even from a 0 step frequency. buck charging using BJTs or IGBTs is basically a V/I operation, so a higher voltage will be more efficient, especially, as you said, when the stators are starting from 0 spm and there's no inertia to lower coil resistance.

to make things more concrete, here are some of applied motion's torque curves for their matched driver/motor pairs:

here's a very small motor with Vcc at different voltages http://www.applied-motion.com/sites/default/files/5014-842_STR2.jpg

here's a very large motor with Vcc at different voltages http://www.applied-motion.com/sites/default/files/HT23-601_STR2.jpg

you see repeatedly that higher voltages give you better low frequency behavior, when the resistance of the windings is high, but that as rotational velocity increases, you're just kind of freewheeling and it doesn't really matter how much voltage you give the driver.

if the STM document doesn't have any schematics, it's easier to just visualize it. a stepper motor, like pretty much any other rotary motor, is just a bunch of coils that you need to energize in some sequence in order to move a magnet. so you put a bunch of BJTs in front of the coils and you switch them according to the sequence you need to fire. these BJTs are going to have to split either current or voltage, and your goal is lots of current, so you have them split voltage. that's pretty much it. there are a lot of tricks to minimize noise and weird resonances, but that's the basic idea.

yah, they mean microstep. the str units work in units of microsteps. they interpolate when given larger steps. to get the full step graph though you just make it a straight line

hah!

as far as avoiding the initial torque drop, afaic, change the motor, not the driver. lower impedance windings are really key. this is one reason why servos are much preferred to steppers whenever possible. in motion control, without a LOT of smarts on the control side, it's almost impossible to use servos, but that's the ideal.

Well, it looks like DOMO cannot accept a Cherry MX switch into that nice middle hole. If so what it is good for?
This is what I would like to print ... eventually. It can accept 40 switches. Not only one, not only after the teeth are filed away

On the other side, DOMO has a decisive advantage. It can be printed now. My part is far from ready to be loaded to a slicer.

I'm sorry because I did not take the time to read the 600+ messages of this thread and I'm afraid my question may be off topic...

A friend did print my Ergodox case using SLS 3D printing with white polyamide powder.

The case is really nice. The only point is the granularity/porosity of the surface which will make it get dirty quite quickly and impossible to clean up.

I got told to spray the case with Sikkens Spotprimer then spray with any color I want.

I built the Ergodox because:
- I found the project really nice
- I'm not 100% satisfied by my TEK
- I'm a geek (at least this is what the other say)
- I'm willing to reduce some elbow/shoulder pain

I would like to avoid exchanging elbow RSI with some finger skin allergy due to the "any color I want" I would have sprayed on my everyday keyboard.

Any suggestion about which color or kind/type of color I should select for that usage ?

Thanks.

for a porous surface i'd start with a layer of white artist's gesso. it's a very white acrylic primer that can be bought from any art store for almost nothing and loves porous surfaces. just brush it on.

once the gesso is on, you can go crazy with any kind of paint you want. seal with a layer of acrylic clearcoat. i use a golden glossy UVLS clearcoat (uv protection layer) because it's water based and yet very hydrophobic once it dries. it is also made to allow even oil paints to dry underneath it. however, once it's on, it's on. you can't paint over this clearcoat.

for a porous surface i'd start with a layer of white artist's gesso. it's a very white acrylic primer that can be bought from any art store for almost nothing and loves porous surfaces. just brush it on.

once the gesso is on, you can go crazy with any kind of paint you want. seal with a layer of acrylic clearcoat. i use a golden glossy UVLS clearcoat (uv protection layer) because it's water based and yet very hydrophobic once it dries. it is also made to allow even oil paints to dry underneath it. however, once it's on, it's on. you can't paint over this clearcoat.

Thanks

Hi guys,

I'm considering building a 3d-printer, my primary motivation is vvp's keywell prototype:p
I want to 3d print custom keyboards, like the ergodox but with curved keywells.
I'm aware that building a 3d printer takes a lot of work, that's OK, I like building stuff and have some experience with other kinds of DIY projects.

What kind of kit would you recommend?
I'm currently looking at various prusa i3 models, e.g. the i3v, i3xl or the omega.
What kind of material would you recommend for my purpose?

Yes, the 8x8" bed on the standard i3 is why I was looking at the derivative models.
Massdrop has the i3xl (9x12x7) up currently: https://www.massdrop.com/buy/diy-tech-shop-prusa-i3xl-3d-printer-hardware
Here is the omega model (9x12x7): https://shop.diytechshop.com/index.php?route=product/product&product_id=54

What size do you think is needed?

The flashforge models do look pretty good, but the ones I can find seem to be a smaller print size?
Is this what you are referring to?:
http://www.ebay.com/itm/Flashforge3D-printer-extruder-3d-printing-machine-nozzle-double-extruder-creator-/190877628239?pt=COMP_Printers&hash=item2c7131774f

Having a ready-to-go solution is tempting, but on the other hand I do like to build stuff, and the reprap concept is very nice.
I guess the flashforge is less customizable?