Timing belts - what's the story?

Is there really a difference between timing belt performance?

Has somebody got empirical evidence that one profile produces better prints, is more reliable or easier to work with?

Thanks!

There is a substantial difference between timing belt tooth pitches. Most obvious difference is that larger pitches will have more "cogging" on the small pulleys we use.

Another difference is that belts designed for conveying have a large gap between the teeth in the belt and the teeth in the pulley. This leaves some play which creates backlash. Newer tooth profiles reduce or eliminate that space so they are much more accurate. You can see a good overview of tooth pitch selection here: [www.gatesmectrol.com]

Another cause of poor accuracy is belt tension. Belts used for linear positioning are supposed to be under fairly extreme tension to eliminate stretch. I'd guess that a typical reprap would need 20-30lb of tension to eliminate that error. That kind of force would probably destroy the frame.


All these errors are real, and fixable, but I suspect that the belts contribute only a small amount of the total error in a typical 3d printer so time/money would be better spent making the frame more rigid and accurate.

I recently was at the 3d print event in Eindhoven. This was my first time to have real life face to face talks with other people making/using open source DIY printers.
I had a few talks with the Ultimaker people. They use mxl timing belts and pulleys and asked about quality differences in the prints. They say the quality is not that different not enough tension on the belts is the quaity killer.
When you have higher quality available it does not hurt.

A main concern of mine regarding belt drive systems on repraps is the accuracy of the printed toothed pulleys. The ones I've seen (including those on my Mendel) suffer from several issues: hole too big for shaft causing pulley to be off-center, hole itself off-center causing the same issue, inconsistent pitch diameter caused by differences in tooth size/shape, and vibrations caused by poor material finishes on the teeth. I'm leaning toward ditching all the printed pulleys in favor of the much more accurate pre-made kind (most likely made of metal).

Also, I have concerns about periodic surface distortions caused by the existence of the teeth altogether. I should think that a toothless belt with good friction and well-tightened would perform better in this regard. Will look into this as well as time allows.

Of course, ball-type lead screws are even better yet, but I'm not ready for that extensive a rebuild...

the problem with toothless belts pretty much comes down to how to provide adequate tension.
It's much the same issue as cable drives, they can be very good, if you can maintain adequate tension.
I have some ongoing experiments I'm doing with fishing line in place of belts on a H-Bot, but I think I'm gradually coming to the conclusion it isn't worth the investment in time. I have a workable solution currently, but the lack of slippage comes down to holding very high tension on the line (relative to reprap belts), and I'm not sure how well the system will maintain that tension over extended periods.

There are no inherent limitations of timing belt systems, it is all in the implementation:

[bell-everman.com]
Has belt systems that can go to 4m/s with 4um/m accuracy.

There are 8x4 foot routers that use 2" wide steel reinforced timing belts and can rip through metal with accuracy in the thou range.

H-belt systems are incredibly fast and precise as well: [www.youtube.com]


You just need to use belts properly.

Screws of any type (including ballscrews) are not a good choice because the inertia of the screw makes high accelerations impractical.

Quote
Polygonhell
I have some ongoing experiments I'm doing with fishing line in place of belts on a H-Bot, but I think I'm gradually coming to the conclusion it isn't worth the investment in time. I have a workable solution currently, but the lack of slippage comes down to holding very high tension on the line (relative to reprap belts), and I'm not sure how well the system will maintain that tension over extended periods.

I built an h bot gantry this weekend that uses braided fishing line. With a few wraps around the drive pulley I was able to get satisfactory movement out of it with no more tension than I hold on my toothed belts (GT2). Experiments and improvements are ongoing, but I see a lot of promise.

crispy1 Wrote:
-------------------------------------------------------
> > I have some ongoing experiments I'm doing with
> fishing line in place of belts on a H-Bot, but I
> think I'm gradually coming to the conclusion it
> isn't worth the investment in time. I have a
> workable solution currently, but the lack of
> slippage comes down to holding very high tension
> on the line (relative to reprap belts), and I'm
> not sure how well the system will maintain that
> tension over extended periods.
>
>
> I built an h bot gantry this weekend that uses
> braided fishing line. With a few wraps around the
> drive pulley I was able to get satisfactory
> movement out of it with no more tension than I
> hold on my toothed belts (GT2). Experiments and
> improvements are ongoing, but I see a lot of
> promise.

I had some initial success with a couple of wraps around pulleys, there is a video of it in another thread.
The issue I came across as I tried to refine the design was that either the pulleys had enough space that the fishing line walked enough that there was a significant change in tension at various points in the X/Y positioning leading to intermittent slipping, or the pulleys constrained the motion of the line so it couldn't walk and the line would sometimes "trap itself" leading to momentary positional inaccuracy.
I redesigned the driving ends so that neither can happen, and it's functional, but in order to get enough friction on the drive ends I needed decent tension on the braided line (I'd guess in the 10lb range) and it's difficult to get that tension on the line without a tensioner, my current tensioner design is difficult to print and needs another pass, depending on how my weekend goes I might post an update to the other thread, if it doesn't happen this weekend it'll be a couple of weeks, too much travelling going on at the moment to get much done.

Don't get me wrong, this all cool but it isn't answering my question:

I see the assertions regarding the value of different tooth profiles/pitches, e.g. buying 2mm pitch belt and commercial pulleys (http://www.reprap.org/wiki/Prusa and [blog.reprap.org] and others) instead of printed pulleys and T5, but I've not seen anything to back up the assertions.

I'm not willing to fork over the money for expensive pulleys and new belts without evidence that it makes a real difference.

You can get good quality prints with T5 belts and pulleys. It just takes a lot more work to get there. Plenty of people did it before T2.5 and GT2 became commonplace.

As long as the printed pulleys are perfectly round, dimensionally accurate, and the teeth are well formed, they will work fine. But hitting all three of these points is nontrivial even with a well calibrated printer.

The recommendation to use commercial pulleys and GT2 or T2.5 belts is because it takes less work to get working nicely, and saves you a lot of time and energy. If your money is more important to you than your time and emotional wellbeing then by all means use cheaper belts and pulleys and feel free to experiment.

Nice jab.

I find the challenge of greater value than the money. Easy is rarely worth it.

Thank you Crispy1, you've helped me make up my mind. I'll stick with the original intent of RepRap and enjoy the challenge of printing my own pulleys and gears. Besides, my prints look great. I never liked store bought cookies anyway.

10 tooth T5 Aluminium pulleys are pretty cheap from China if you look on Aliexpress and they have a 5mm bore, I know I have about 20 sitting on my shelf

Edited 1 time(s). Last edit at 11/01/2012 02:38AM by NelsonRap.

---

__________________________________________________________________________
Experimenting in 3D in New Zealand

There are inherent limitations in belt drive systems for the simple reason that there are inherent limitations in any kind of system that could ever be conceived or produced. The only reasonable approach to engineering is to understand that there is no such thing as a perfect solution and that the choice between options is a matter of weighing the inevitable tradeoffs that each option presents.

A quick survey of high precision automated machine tools reveals that screw drive systems are vastly preferred over belt drive systems in cases where high precision is critical. Many decades of work in this area have clearly established that belt drives cannot match up to screw drives for the task of guaranteeing precise motion.

Consider that precise linear motion is not just a matter of moving from point A to point B with fine control over the final distance moved. It also concerns ensuring that the path traveled does not deviate from a theoretical straight line between the start and end points by more than an acceptable tolerance. Think of it as the difference between a straight line between A and B and a line that starts and ends at the exact same points but wobbles around as it makes its way. The former would create a very smooth finish and a surface that maintains its intended form and the latter would not.

I've seen, used, repaired and designed any number of high precision machines over the years and I can't recall a single one that didn't use lead screw drive in the high precision axes. In cases where belts may provide special capabilities, such as applications requiring very long travel, the designer may have been forced to accept less precision and/or to adopt more stringent methods to minimize the inherent inaccuracy in belt systems, but surely the decision was not due to a concern about the mass imposed by a screw drive system. The screw itself is only rotated and has a very low rotational inertia (due to the small diameter) and the bulk of the mass is in the moving stage and any load it carries, which would not be any different by using a belt drive.

Yes every CNC Mill uses a screw to drive it, but that's because it's the best solution for that particular set of constraints
Go look take apart an inkjet printer, it has accuracy requirements that are just as extreme as most CNC mills, and no way to tollerate backlash and they use ... belts

It's important to understand the differences between something like a Mill and a 3D Printer, they both have 3D positioning systems, but they don't share the same constraints.
Above and beyond the obvious positional accuracy issues which are common.

Mills usually make long continuous cuts a 3D printers tend to make many small motions
Mills have considerable side load on the head, 3D printers have none

Above a certain point (and it isn't very high) the acceleration of the axis has a large effect on the printing speed, which makes the rotational inertia of larger steppers and ball screws work against what you are trying to achieve. Yes my 1000lb mill can rapid at 200+mm/s, but if I set it up to oscillate over about 10mm it won't approach 1/10th of that speed because it needs 10x that distance to hit it's maximum rapid speed.
Now the table on my mill is >150lbs so there is more to it than just the mass of the screws etc,

The point though is printers aren't mills and there is nothing that makes belts inherently inaccurate, they are cheap, mechanically efficient and don't require complicated mechanisms to eliminate backlash.
Tension on the belt is important and I personally wouldn't use a printed pulley, the waves that some people see on flat surfaces are usually a function of a course toothed belt running over a smooth idler pulley, either twist the belt or buy a toothed idler.

To say that an inkjet printer has the same accuracy requirements as a precision milling machine is simply incorrect. I guarantee you that no decent machine manufacturer would tolerate the amount of multi-axis vibrational instability or the lack of structural stiffness that is present in an inkjet printer.

Suppose an inkjet printer prints a line and the head deviates from a perfect path by +/- 0.0016mm. Would that be a problem? Of course not. But it would in a precision mill or lathe. That amount of deviation translates into a surface that would vary from the intended plane by a total of 64 microinches (ui), which is considered far too rough for many applications. 16ui and even 8ui surface finish callouts are common in machined parts. Do you suppose there's a reason why machines requiring that level of precision don't use belts to drive their linear ways?

Another consideration in precision motion is the consistency of the speed at which the carriage moves. Even if the motor is spinning at a very consistent rate, every time a tooth on the belt disengages the corresponding tooth on the pulley, there is friction rubbing (a source of vibration) and then the sudden release of drive force as the two teeth separate completely, causing a transient pulse along the length of the belt. The result is a periodic variation in the drive rate of the tool, which shows up in the surface finish of the part, and the same disruptions are occurring on the other side of the pulley where a tooth comes into engagement with it, and multiplied by as many pulleys as are in the system. It can add up quickly.

It may well be that none of this is of particular interest to someone building a home-made 3D printer, and it's quite possible that belt drive (properly implemented) provides plenty of precision for this application, but to say that there's no difference between toothed belts and ball-screws from the standpoint of precision is simply ludicrous. I'll stake over 25 years of mechanical engineering and machining experience on that.

Screw systems are favoured on machines with large moving masses that require precision under extreme cutting forces. Neither is the case for a 3d printer.

Many systems in the 0.001-0.005" range use belts, and I even provided a link to a belt system with 4 micron accuracy. Its just that belts tend to be equally or more expensive than screws and it is harder to get the required reduction for high force applications without introducing backlash.

Certainly, both systems have their advantages but belt drives are the standard for pick and place, waterjet, laser, and dozens of other similar applications for a reason.


You are correct that a screw has relatively low rotational inertia due to its small diameter, but consider that the screw is coupled directly to the motor shaft whereas the final load is being moved through the reduction of the screw pitch.

For a 5mm pitch/12mm diameter screw, the load moment is M*(0.005/(2pi))^2 or M*0.00000063
Ballscrew moment is M*(0.012^2)/2 or M*0.000072

Each gram that the screw weighs has ~100 times the effect of a gram on the load.

I should clarify that I probably should not have said belts have no inherent disadvantages. What I meant was that teeth offered no appreciable loss of accuracy compared to cable/toothless belts (even that is not entirely true, I know some telescope systems prefer cable drive). I agree with everything you have said about ballscrews being superior in many ways. Its just that acceleration is of prime importance in 3d printers.

Edited 3 time(s). Last edit at 11/01/2012 06:30PM by 691175002.

Well, not to beat the issue to death, but using lead screws need not add any appreciable mass to the system. The mass of interest is that which is being driven, and the vast majority of that (in a 3D printer as well as in machine tools) is in the carriage itself, not the lead screw. The acceleration load of the screw is limited to rotating it, which is not a big deal since they are small in diameter, therefore the inertia (which is really what you're concerned about) is quite low. I could easily design a lead screw driven axis that had no more mass load (inertia) than a comparable belt system. The much greater mass in a mill or lathe is due to the massive steel structures they use, which would be wasted on a 3D printer.

The big issue for repraps is that belts are cheap, but there will always be a limit to the quality of motion with belts. I also agree that using printed gears/pulleys introduces motion error, and I'll be looking to address that soon in my own machine.

SOMEDAY... (goes into daydream mode) I'll whip up a printer with linear slides and lead screws. I expect it would run just fine on the same motors and electronics used on my humble Mendel, but would be much more capable of high quality printing. I'll post back when I've got something to show

I think this question comes up every two months "why not use a screw drive? They are better and all the grown up machines use them".

Plenty of people with similar claims to experience with screw drives have tried to build printers with screw drives and they all come out more expensive and slower. The extra resolution provided does not provide a useful cost/benefit ratio.

There are other practical advantges of screw drives in mills, one of which is that they are much more tolerant of swarf. But there seems to be a pervasive "real engineers use screw drives" meme. Real engineers use the appropriate solution for the problem (which you did point out), and the appropriate solution in many cases is a belt drive.

Quote
Polygonhell
Go look take apart an inkjet printer, it has accuracy requirements that are just as extreme as most CNC mills, and no way to tollerate backlash and they use ... belts

But most inkjets use belts only as means of drive, not to achieve positional accuracy. Positional accuracy is achieved by electro-optical sensing of position. They also doesn't change direction of travel during print and move only one axis at a time (except non-printing CR/LF travel)

I know the stratasys fortus 900 uses ballscrews.
I suspect that they have custom, or at least $$$ screws with exceptionally high pitch and some hefty servos to drive them though.


The calculations I did took into account that the screw is only rotating. Spinning up a 10 gram screw is just as hard as moving a 1Kg mass once you take into account the mechanical reduction that a typical screw provides.

I have been a fan of this printed rack and pinion design for a while now: Rack and Pinion RepRap

I would love to see this kind of thing further developed

And what do you think about this?

[www.youtube.com]

Thirty years ago my Juki daisy wheel printer had a linear stepper motor driving the head. In 1982 I went to the Robot show in Detroit. For me the most exciting thing at that show was Motoman's RobotWorld. An inverted etched steel platen with 4 - 2d stepper motor heads running independently all over the platen. Two heads worked together to pick a circuit board off a conveyor and orient it on a fixture. Then all 4 heads changed grippers and picked up parts and placed them accurately on the circuit board.

A company in NJ called Megamation had a similar system. Northern Magnetics also made the 2 axis steppers. They were bought by Baldor who apparently still makes them.

... for me it was a Brother typewheel-printer, maybe 40 years old: [forums.reprap.org]

And a 'multi-stations' SMT-assembly-system on a fair, where on a big steelplate with raster on it 4 XY-stages were interchanging and moving synchrone

Here is a link to a German company building and selling 'planar motor applications' - [www.direktantriebe.de]〈=en

Edited 1 time(s). Last edit at 05/16/2013 09:38AM 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]

screw drives work fine if you have the cpu cycles to support the acceleration ramp and steps per mm/second. also there is the issue of acceleration in between moves. there are micro stutters in arduino firmware that make an actual acceleration ramp not possible. the speed changes are compensated for by the absorbing into the belts and use of inertia to keep mass moving as next acceleration is being approximated. in a near zero slop lead screw system, with tolerances tight to keep backlash at a minimum, the arduino reprap firmware would not be optimal. using mach3 or emc2 with a decent modern cpu wold be advised for these kind of precision systems. and likely a high end stepper controller such as a gecko.


it has been argued before that people have had successful machines that are lead screw based. but here me out. they are not based on current reprap firmware, not the ones that produce good prints.

I've used T2.5 T5 GT2 and various others with printed pulleys and aluminum ones, the only real difference i found was in the price,

on a prusa there are bigger issues to contend with such as frame rigidity and extruder design, well before anything cause by the belts can be seen,

---

-=( blog )=- -=( thingiverse )=- -=( 3Dindustries )=- -=( Aluhotend - mostly metal hotend)=--=( Facebook )=-

the tooth profile of the t5 is not meant to travel in both directions, only one and the small the pulley the worse the backlash. also the idler on the other end also makes a difference. it should be a toothed pulley, because otherwise the tension actually changes between teeth, so you will get little sort of ridges in the print that correspond with the belt. now the reason gt2 is better is because the tooth profile is designed for reciprocating motion, it wont have backlash if its properly tightened. additionally since the teeth are rounded when it goes over the smooth idler it does change the belt tension nearly as much as the bigger t5 square tooth.

agreed T5 was never meant for this kind of use, i just found that there were other factors which have greater influence on the print quality before starting to worry about the belts, the cogging that some people have noticed with T5 can be migitated by simply not tightening up the belts to the full guitar string tension, ideally if you have it just tight enough so that you remove the noticable looseness what you'll find is that the cogging effect will dissapear,

the printed pulleys i used are a little unorthadox they are printed with a rounded tooth profile (12tooth) and they have to be pressed on the shafts (i literally don't need setscrews)

somthing i'm itching to try out is a tigertail wire drawn system with ptfe idlers, on the dual carriage x axis mod i'm builting at the moment,

aduy Wrote:
-------------------------------------------------------
> the tooth profile of the t5 is not meant to travel
> in both directions, only one and the small the
> pulley the worse the backlash. also the idler on
> the other end also makes a difference. it should
> be a toothed pulley, because otherwise the tension
> actually changes between teeth, so you will get
> little sort of ridges in the print that correspond
> with the belt. now the reason gt2 is better is
> because the tooth profile is designed for
> reciprocating motion, it wont have backlash if its
> properly tightened. additionally since the teeth
> are rounded when it goes over the smooth idler it
> does change the belt tension nearly as much as the
> bigger t5 square tooth.

---

-=( blog )=- -=( thingiverse )=- -=( 3Dindustries )=- -=( Aluhotend - mostly metal hotend)=--=( Facebook )=-

From my build, I learned: the higher the speed, the more "grip" is needed. Also, the higher the load, the more "grip" is needed.

I tried 150 mm/s on identical axis but with different result See video on the blog. I ended thinking, that GT2 is good until a certain speed, beyond that speed, one must reduce the speed or the load or simply change from GT2 to T2.5.

I also think that the idler on the other end makes a difference. Still have to test it.

[www.eventorbot.com]

For the 4m/s ~ ±4µm/m error quoted above - The spec says it's the linear optical encoder having ±4µm/m accuracy, not the motor.
But at any case, I agree timing belts is fairly accurate, as long as it's under tension.

I have T5 belt + "Chinese" metal timing pulley. (Disclaimer: I do live in China and 5 USD can buy you a pulley freshly fabricated).
So far so good.
The pulley I made has 14-tooth - the maker says 12 is the smallest I can make. I saw some design is using a 8-tooth printed pulley for T5?
IIRC that's out-of-spec for the T5 belt I have bought - turning radius being too small.
Maybe a larger pulley will have positive contribution to the accuracy/repeatability, YMMV.