Why Spindle driven Delta's are more precise than Belt Driven

Quite a few Delta's on the market are offered with Timing Belt driven sliders and I wonder if anyone ever has questioned this system.

A spindle driven slider can be much more precise as a timing belt driven one. Especially when printing small parts in the center of the print bed of a Delta printer, precise movements do matter.
Maybe you have noticed that when printing small parts in the print bed center with holes of about 4 mm (0.157"), the holes are not round anymore. (Timing belt driven delta) There is a reason for that behavior and it has nothing to do with your build but more with the quantity of steps per unit the system needs.
In the center all three parallel arms will move much more aggressively per unit as out of center. (At least two of them) (See schematics attached)
For a given belt driven system, for 1 mm movement it takes 5 MOTOR steps (1/16th driver makes it 80 micro steps); 0.1 mm takes 0.5 MOTOR step, what won't happen, it will be skipped till the next full step, due to many variables working together with Murphy.
A spindle system with a spindle pitch of 2 mm will take 100 MOTOR steps for the movement of 1 mm (1/16th driver makes it 1600 micro steps); 0.1 mm takes 10 motor steps and the system will do most probably his job.

See attached schematics and new design picture below

Of course if you wanted to increase your steps per mm you can just use a reduction gearbox and you don't need to redesign the entire machine

[www.robotdigg.com]

The next problem you develop is a loss of maximum speed, and since your running your stepper driver at a higher frequency probably more heat

On the plus side you will get loads of thrust, and could probably run your effector with a direct drive extruder

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

Thread is moving along the spindle axe when carriage is moving along tower. This adds an error to the carriage position.
For belt systems, the typical microstep size is around 0.015 mm and not 0.1 mm as you claim.

He was quoting motor steps, micro steps are fractions of motor steps.

A well floated ball screw shouldn't induce error, but yes its a concern, and if your not familiar with floating ball screws then you could have a problem.

My XL uses 16 tooth drive gears so its actually 0.01mm/step frankly this is more than enough really cant see anyone needing 0.005mm steps, and frankly my printer with its 16 tooth belt drive and 200 steps per rev motors now prints excellent holes even when printing at 100mm/min, may upgrade it to 400 step/rev motors at a later date, but for now I'm more than happy.

Edited 1 time(s). Last edit at 05/20/2016 04:57AM by bgkdavis.

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

... the discussions regarding "belts_vs_spindles" aren't new -- you have to find the best compromise between speed, stiffness and accuracy ... and this is sometimes better for spindles, sometimes for belts.

Another essential point is the used motor -- for costs and building sizes most 3D-printers use NEMA-17 steppers ... but with NEMA-23 (or bigger) motors and much higher torques the usability/accuracy of microstepping gets into account too - I have some micropositioners with NEMA-23 and -36 steppers with 2Nm torque, where I can can use high microstepping rates with pretty good repetability between the fullstep-positions

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

HERCEK, I don't think because of using a spindle I add an error to the carriage position at all and second, I'm talking about motor steps. (Backlash is taken in regard by Machinekit)

Quote
evandene
Quite a few Delta's on the market are offered with Timing Belt driven sliders and I wonder if anyone ever has questioned this system.

A spindle driven slider can be much more precise as a timing belt driven one. Especially when printing small parts in the center of the print bed of a Delta printer, precise movements do matter.
Maybe you have noticed that when printing small parts in the print bed center with holes of about 4 mm (0.157"), the holes are not round anymore. (Timing belt driven delta) There is a reason for that behavior and it has nothing to do with your build but more with the quantity of steps per unit the system needs.
In the center all three parallel arms will move much more aggressively per unit as out of center. (At least two of them) (See schematics attached)
For a given belt driven system, for 1 mm movement it takes 5 MOTOR steps (1/16th driver makes it 80 micro steps); 0.1 mm takes 0.5 MOTOR step, what won't happen, it will be skipped till the next full step, due to many variables working together with Murphy.
A spindle system with a spindle pitch of 2 mm will take 100 MOTOR steps for the movement of 1 mm (1/16th driver makes it 1600 micro steps); 0.1 mm takes 10 motor steps and the system will do most probably his job.

See attached schematics and new design picture below

Yes a spindle-driven slider is more precise in theory; but a belt drive is just about precise enough. I recommend using 0.9deg/step motors in new designs for greater precision and greater torque per unit of position error. The improvement over 1.8deg/step motors is small - see [miscsolutions.wordpress.com].

IMO, avoiding backlash and friction in the joints is a bigger issue if you are looking for high precision.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

I was just thinking about this and mailed with openbuilds because they offer a nice vslot + leadscrew bundle but didn't get a straight answer from them.

Seeing that industrial pick and place machine always use lead screw i think its time to test this. My only concern is how the motors handle the higher speeds that are required to drive the screws, but since the movement of the effector is always divided partially over 3 carriages the speed of the individual carriages is a bit slower that a straight cartesian movement, so that gives a little advantage already.

The trinus printer that was recently on kickstarter is a screw driven cartesian machine and they recommend speed of only 70mms which sounds a bit slow for my taste.

I think an increase in current to motor (with active cooling of the drivers or DVR8825) could get us started but might not be enough. A 2nd idea could be to use something like the duet that has the ability to bridge the existing drivers with external ones that might be used on a desktop cnc, together with some high torque nema17s

Och, by "spindle" you mean a nut on a leadscrew. Ok, enjoy your precise but slow delta
But if you want a slow printer then a cartesian is a better option.
I though this is a spindle

I got he was talking about a leadscrew from the start, he may be talking about a ballscrew or a buttress threaded screw, but the problems are the same.

But besides the obvious problems, lets take a little trip down this rabbit hole and see what we find.

At first actually building the machine is trivial. leadscrews are not really an inventive idea, and in may respects they are easier since you don't need belt tensioning, probably need a flexible coupling and since it will need to run at maximum speeds bearings to prevent the screw suffering from whip, its really shocking how nobody every thought of this before.

Preventing the screw from imparting a distortion into the carriage motion can be a problem, but higher build quality will help resolve this,

so by now you have a reasonably accurate build and a super position capability.

Once you put to control system on you have to accommodate the problem of the pulse frequency being 160x higher, many controllers can operate in this range, but watch out for increased heat.

Heat problems will get worst because your going to have to run your processor 160x faster order to process the higher step resolution, but most of that will be wasted because your GCode file still has a very high level of faceting and all that will happen is you get a more precise representation of the GCode faceting.

To resolve this you go to your slicer and instruct it to slice at a higher XY resolution, this will signifigantly increase your file size and also mean that it takes significantly longer to upload the file to the printer, if your not using stand alone this may not be an issue, but beware when drip feeding, because its possible that your speed can still be limited by the Baud rate and its not possible to upload the file fast enough..... but then again your machine is probably slow so it doesnt need fast drip feed time.

However, you still have faceting, why?, well because your STL file is faceted, all those files you download from Thingiverse are now pointless, so to make best use of your slow but super accurate machine, you resolve to only use files you generate yourself, and can process at a super high accuracy, this of course takes a lot longer to create the STL file, and longer to upload it into your slicer and longer to actually slice.....but your slow printer creates loads of time anyway.

...and after you have done all this you find yourself still looking at parts that are not really showing the benefits of your machine, there are irregularities in the extrusion, backlash and wear issues.

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

It's not shocking, and it has been "though of before". I want to say there is video of a screw based delta around somewhere, I want to say there is one by Seward. Pretty sure I've also seen video of a screw based 6 axis delta somewhere too.
There are ballscrew (and leadscrews of various types) with higher leads as well, so the resolution of the system can be tuned for the desired step rates. Have a look at Roton.com for some of their higher lead solutions. Screw solutions are completely attainable, but if you do it right they cost a lot more. Screw based drives need axial and radial bearing support on one end, and generally radial on the other end. Screw whip will likely be at issue given the lengths involved in a linear delta, although creative placement of the screw can at least minimize the length.

Quote
Koko76
It's not shocking, and it has been "though of before".

Sometime I forget that sarcasm is wasted on those who speak English as a foreign language

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

Quote
bgkdavis

Quote
Koko76
It's not shocking, and it has been "though of before".

Sometime I forget that sarcasm is wasted on those who speak English as a foreign language

It's Ok dude, sometimes I forget that posting useful information like links to websites with good practical information is wasted on those who would rather be smart mouthed jerks. You have a lovely one ok bud.

If you guys want to read professional guidelines for applying lead screws (ball screws) please visit the THK site and walk through the tutorials.
A little calculation tells me that a 1004 ball screw and a Nema17 1.7A 4200gr/cm will do the job perfectly, also speed wise, and don't worry about the coupler and bearing design, I'm doing this for 35 years, i know what I'm doing.

Quote
evandene
I'm doing this for 35 years, i know what I'm doing.

Don't make the mistake of thinking your alone.....

As I said earlier, the mechanical construction is trivial, but don't forget a 200 step per mm printer working at 100mm/s with a 2mm pitch ballscrew will have to do 50 rev per second at 200 pulse per rev and 1/16 microstepping the controller will need to operate at a stepping frequency of 160khz most controllers cant do more than 40kz so your maximum speeds going to be 25mm.sec the highest stepping controller I've identified is the smoothie and that's going to be limited to 75mm/s.....and that's flat out!

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

A lead screw 1004 means 4mm pitch and 10mm diameter.
DRV8825 on a BeBoPr++ and a BBB computer runs 1/32 micro steps. So, should work perfectly, right?

Edited 2 time(s). Last edit at 05/21/2016 01:31AM by evandene.

1/32 micro stepping with a 4mm pitch is still 160kHz at 100mm/s, and you still don't want to run a machine constantly flat out, plus this speed is only your stepper speed, you still have to consider the block processing speed of the controller.

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

Yeah, my mistake, I should step down to 1/8. I remember that test are performed on the BeBoPr++ at 70kHz successfully , need to lookup that part.

Well, now your down to 400 steps/mm and 40kHz, at 100mm/s I can still easily upgrade my KosselXL to 0.9deg/step motors and get 200 step/mm.

Remember the whole context that you started this thread with, was how much better 1600step/mm is than 80 step per mm and why you saw the ballscrew as a solution, with a 16 tooth drive gear 0.9 degree step motor an a 5:1 reduction box you could achieve 1000 steps per mm without completely redesigning the entire printer. use a 14:1 reduction box and your at 2800 steps/mm

It sounds like what you really want to do is build a printer with ballscrews and everything else is backward justification, which is fine, but if your only reason to do that is to improve your steps per mm then there are easier ways of doing it, and your still going to run into all the speed, model and processing issues

And don't forget 100mm/s is only the printing speed, you should be aiming at a 200mm/s so you have a decent rapid traverse speed.

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

I followed the thread with great interest, but came to the conclusion, it doesn't make much sense to use leadscrews for a delta.
But I have a few DVD-stepper motors with a 80mm long spindle shaft and got a vision of a tiny toy delta made with those.
Maybe use it as a laser engraver instead off FDM printer....
Thanks for involuntarily planting this idea

Edited 1 time(s). Last edit at 05/21/2016 03:07AM by o_lampe.

After nice tips, misunderstandings and good idea's, I have decided for; a ball screw 1004 or 1204 with a stroke of 350 mm and a max print speed of 100mm/sec and max travel speed of 150mm/sec.
The stiffness, durability and accuracy is preferred; more speed and acceleration is easy to achieve with other motor drive systems.
Thanks guys

Don't forget also about accelerations too. The point is that a delta which can move quickly will be still printing slow if the maximum accelerations are low. You want to achieve at least 5 m/s², preferably 10 m/s² or more. Stepper rotor and rod inertia will not be negligible when you are going to spin your motors about 10 times more quickly than the rest of us. A simple Nema17 we typically use would waste about quarter of its torque only to spin up/down its rotor and the rod. If you want to use micro-stepping for precision too then it lowers your usable torque very badly. And you may need to go with rather high current motors (low impedance) to still have the needed torque at high speeds. Not even sure whether 40 V power source will be enough.

You need high accelerations for quick printing because slicers will generate a lot of short segments with big direction changes ... especially for infill of models which have a lot of holes (i.e. short infill lines) ... or for hexagonal infill ... if anybody uses it at all.

I think that you are going to create a pretty printer which will have a really bad price/performance ratio. Or you are going to build a slow printer in which case you should be rather building a cartesian. Anyway if you enjoy building it then just go for it. And post pictures!

Edited 1 time(s). Last edit at 05/21/2016 06:41PM by hercek.

Quote
hercek

I think that you are going to create a pretty printer which will have a really bad price/performance ratio. Or you are going to build a slow printer in which case you should be rather building a cartesian. Anyway if you enjoy building it then just go for it. And post pictures!

X2


oh, and take care of it, if you get this running at speed those ball screws are going to pack a helluva punch, belts are soft and squishy and absorb the impact of a crash quite nicely, your ball screws will shatter any glass/plastic/fingers that offers an obstruction

Edited 1 time(s). Last edit at 05/21/2016 08:30PM by bgkdavis.

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

Steppers are not build for speed neither timing belts. Did somebody ever saw a professional cnc milling machine with a timing belt driven?
I'm very familiar with ball screws and they are without backlash and for that there is no play the balls are traveling trough and for that reason the whole system will take care of the kinetic energy and proper acceleration and declaration curves from driver setups will keep everything under control.

Having spent most my professional life building and controlling some of the biggest and most complex CNC machines in the world, I can honestly say, yes, I have seen timing belts used several times also seen hydraulics and even pneumatic drives, but generally we normally use servo drives and not stepper motors, plus if we actually want accuracy we use dual scale and encoder feedback systems and then laser and thermally compensate the machine to get even more accuracy, plus have an entire bagfull of tricks for managing backlash.


.... you know, I've just had an epiphany, I realise that when YOU see a ball screw in a machine you think its about accuracy, and your looking to emulate that, the problem is, this is an incorrect perception.

When talking about CNC machines most people picture a metal cutting machine, these machines generally have to drive a cutting tool through a resistant material, and require a lot of thrust and accuracy under thrust.

That's what the ballscrew is really about, thrust, prior to the ballscrew we had plenty of other very accurate screws that could produce thrust, but under high thrust they also suffered from friction, and that tended to cause judder, and when your trying to do accurate machining this is pretty much exactly what your dont want, this is why you have never seen a belt drive on a CNC axis and I have, because Ive seen CNC axes used in low thrust positioning applications, where using a ballscrew is just nuts.

This is where the 3D printer is different, its a low thrust machine, in fact in order to minimise crash damage you want as little thrust as possible

Edited 2 time(s). Last edit at 05/22/2016 04:34AM by bgkdavis.

---

RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD

Quote
evandene
After nice tips, misunderstandings and good idea's, I have decided for; a ball screw 1004 or 1204 with a stroke of 350 mm and a max print speed of 100mm/sec and max travel speed of 150mm/sec.
The stiffness, durability and accuracy is preferred; more speed and acceleration is easy to achieve with other motor drive systems.
Thanks guys

So you will have 4mm carriage travel per motor revolution. Assuming 1.8deg/step motors and 16x microstepping, that means you will have 800 steps/mm. To get 150mm/sec XY travel speed, you will need to move the carriages at speeds up to about 300mm/sec near the centre of the bed (more near the edges). That's a step pulse rate of 240kHz. I don't know any existing electronics/firmware combination that can maintain that rate on more than one motor, although it's possible that the Smoothieboard 2 Pro might when it comes out because it uses an FPGA for step generation. The existing Smoothieboard is said to be good up to about 100kHz. The current Duet is about the same - I have my Kossel configured for maximum travel speed of 300mm/sec @ 160 steps/mm, which works out at about 96kHz near the centre of the bed.

So you will need to use lower microstepping - but that will make the machine noisier at low speeds. With a Duet you could get 120mm/sec travel speed using x8 microstepping. I guess the Smoothieboard would do about the same. Dynamically varying microstepping would be a solution, but I don't know of any electronics and firmware combination that supports that yet.

I admire your attempt to get greater precision, but you don't seem to have considered what precision you actually need in a 3D printer. As I said in my earlier post, timing belts are precise enough for 3D printing, especially if you use 0.9deg/step motors.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

All,
Traveling from A B with a stepper driven system Works well if no speed or accurate positioning is required.
Please do following test with your delta printer;
1) go from A to B in Z height like 25 mm (1") and measure the real distance. Most probably all it's within a few hundreds of a mm.
2) do this again but now Z= 0.1mm (0.00393"). Asuming you use micro stepping of 1/16 or 1/32 you will notice that sometime the requested positioning is slipped and sometimes double or more executed. This is normal behavior for stepper systems.
See Wikipedia for steppers, 5th paragraph.
I can't have that in my system.

... you can get higher stepping rates without problems!

I'm driving my pastedispensers at max. 400mm/s with 170kHz (425 steps per mm on the X-axis) wth the RADDS/RAPS128 combo ... and have more than 100kHz 'reserve', what's not usable because of the short traveling ranges (only max. 200mm) and 'safe' acceleration settings. So it can clock with around 300kHz max. for XYZ+E synchronized.

And have a BeagleBoneBlack based XY-'laserplotter' with 500kHz stepping frequenzies.

My fastest 1/256 microstep-drivers can handle up to 10MHz clocking ranges - I've tested this with a frequnzy generator, but didn't find a contoller fast enough.

On one of my jobs they were assembling high-accuracy and high-speed milling machines with controllers and drivers with up to 2MHz clocking speeds ... but this machine was >150k€, so actually not in my budget range

*** EDIT ***
... here is the link to the 'high-speed' cutting mill - [www.imes-icore.de]

I've then sourced and programmed the 6-axis "toolhandling"-robot for this machine

Edited 1 time(s). Last edit at 05/22/2016 01:22PM by VDX.

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

Quote
evandene
All,
Traveling from A B with a stepper driven system Works well if no speed or accurate positioning is required.
Please do following test with your delta printer;
1) go from A to B in Z height like 25 mm (1") and measure the real distance. Most probably all it's within a few hundreds of a mm.
2) do this again but now Z= 0.1mm (0.00393"). Asuming you use micro stepping of 1/16 or 1/32 you will notice that sometime the requested positioning is slipped and sometimes double or more executed. This is normal behavior for stepper systems.
See Wikipedia for steppers, 5th paragraph.
I can't have that in my system.

My DTI does not agree.

A screw/nut may be interesting if you are running servomotors. There is work in progress on servo drivers here [forums.reprap.org] here [forums.reprap.org] and here [forums.reprap.org]

Due to high rotation speed, you will have inertia problems with difficulties to get accelerations, however servo controlled motors (or hybrid steppers) allow temporary current pulse much higher than what you can do on an open loop stepper which use continuously the same current, as short current pulses will build less heat than continuous operation. That allows much higher transient torque than open loop systems and will help to mitigate the high inertia.

However, in some cases, if you use honeycomb infill (which I do most of the time), the acceleration and decceleration will be permanent during the infill and the motor and driver may build heat, much more than what you can have on any other servo driven CNC machine.

Motor_control_loop [reprap.org]

About step output rate, it is possible that redeem can have a large one, as it is using real time units for step generation (in the beagle bone black), but I did not found data.

---

Pierre

- Safety [reprap.org]
- Embedded help system for Duet and RepRap Firmware [forums.reprap.org]
- Enclosed delta printers Lily [rouzeau.net] and Lily Big [rouzeau.net]
- OpenScad delta printer simulator [github.com]
- 3D printing on my site [www.rouzeau.net]