From T5's to GT2's to Control line.

Hi All,

Im just curious on what peoples thoughts are on control line driven printers. I originally build my printer using aluminium pulleys with T5 Belts. Later I switched to GT2's (slight improvement) but i've spent the last week or so experimenting with control wire (on the Y and X axis) and so far im finding this far superior in almost every respect. Im using idlers at both ends of the axis with the motor spindle in the middle. But it seems that the control wire option is much cheaper, all the parts are easily printable (unlike pulley's), with the spindle in the middle you can up the tension quite considerably higher and yet the current draw on the steppers has actually dropped (possibly cause the belts are no longer pulling the motor shaft sideways?), theres no flex at all anywhere in the mechanism and no backlash.

So im curious, what are the downsides to such a setup?

What sort line are you using?

It does seem to be the way to go. I will be trying it soon.

There is always some backlash unless you have motors with infinite torque or zero friction bearings.

---

[www.hydraraptor.blogspot.com]

On my X axis, im using control line salvaged from an old printer. However, on the Y axis i didnt have enough of that so i had to improvise and my other half's mum has some cotton thread designed for really heavy work (demin stitch or something). Seems to be just as good (very very strong, and no flex in it at all).

I dont quite get how things ended up going down the timing belt path (in general) when control wire seems so much easier to implement, performs better, its cheaper and the parts print simply... there must be a downside i cant quite fathom yet.

Interesting you replied Nophead, im using a m90 derivative, so i made a mod to the motor side of the X axis to add an idler just past the motor., i'll be uploading everything to thingiverse once its complete.

By control line, I take it to mean what might be called a "cable drive" i.e. a cable or line wrapped around the driving "drum". One end of cable would be hard fixed to the drum. This setup in contrast to what might be called a "drive belt" drive i.e. not wrapped around the driving pulley and not fixed to the pulley...like a fan belt in a old car.

So given I have the right idea and the question is why not a "cable drive", I think some of the general pros and cons of cable drives:

* lack of periodic positional error. Cogged belt drives "telegraph" some amount of their cogged nature through to the movement resulting in periodic positional error. Accordingly, cable drives have been popular for telescope drives where periodic error is a problem for photography (long exposures requiring tracking).

I believe this accounts for why cable drives are so common in photocopies and printers (??).

* They can provide transmission of motion in all kinds of crazy angles and planes since the cable has no "sided-ness".

* Inherently zero backlash (in a static sense) but subject to dynamic cable stretch under acceleration.

* IF you need to provide simple rotation to linear translation (like an axis on a cartesian mover), you don't need the flexibility described above. A cogged belt can be very wide and provide high stiffness. Equivalently, a cable would have to be increased in diameter and would get much more difficult to bend.
Of course, it would be possible to use multiple cables in parallel to get both stiffness and flexibility but would be more complex.

* Depending on the number of turns of the cable on the "drum", a cable drive will have some error due to the fact that the angle of the cable to the drum changes as more or less cable is wound on the drum. This may be addressable with more complex drum designs (????).

Bottom line: IMHO, they may be as good or in fact, better drive system for reprap machines do to the virtually zero backlash and their lack of periodic error. The only concern would be the practically achievable stiffness which might be a issue for high inertia, high acceleration moves (heavy extruder head?) where the edge might go to a cogged belt. Tradeoffs, tradeoffs....

Interesting points, and yes, cable drive is the word i was looking for... I was thinking of getting a dual shaft stepper, or designing a new drum with two spools for the reasons above, but dropped the idea cause tyring to keep a constant tension across two different cables would be very tricky... The printer cable i have is quite heavy (its from a pretty old printer), its a little less flexible then the denim cotton line (or whatever that stuff is). Both of them though seem to have almost 0 stretch to them, though i'll take one to my old work and see if they'll let me use the cable load test machine to see what each does under load (though its really designed to test much larger cables than these).

As for the deflection, i actually did some calcs of this earlier and it comes up being almost insignificant. Assuming you have a 30cm range a motion with a motor in the center of your axis and a drum with a 5mm radius, your pulling 31mm per turn, or approx 5 revolutions for the 15cm total distance you'll move (from the center). The printer cable is .28mm wide, the denim cotton is .17mm wide, so total deflection its up to being 1.4mm... if you go up to a worst case scenario, i.e. motor right next to idler and .3mm cable, your up to 3mm of deflection and add to that a minumim distance between idler and drum of 50mm, the line length difference is less then 0.1mm - which would be poor design choices ultimately, at 100mm distance, that drops to 0.05mm. Keep in mind, its also very easy for a deflection type error to occur exactly the same way with timing belts if the idler/motor exit point on the pully differs in height from the attach point on the bed.

As for weight, my big weight item is my y-axis (420x450 print bed, mdf + cork + 3mm glass is not light) and thats using the heavy thread (not the printer cable). With the timing belts (t5, gt2), this was actually pushing my stepper towards a 40c increase in temp and even then i had to allow a little slack in the timing belts - with the thread, i've upped the tension considerably and my stepper now runs at only about a 10c increase in temp.

But, all things being equal, what i dont quite understand is the why of timing belts on reprap's. Even if your not using a dual-ended idler system, cable drive is cheaper, the parts are much easier to print (+ less prone to error), and it offers alot of flexibility. To me, this just seems more in-line with the general reprap philosophy (if you know what i mean) which leads me to believe someone, somewhere did this at one point and it was discounted for a really good reason and that i would love to know. So far i believe cable drive is MUCH better then timing belts for the printers we're using but i feel like i must be missing something obvious here (or something bad is going to happen that will make me go "oh now i get it"). Im going to say with the exception of perhaps something like the rostock where theres going to be alot of tension pulling constantly against all your axis at once and stretch might just cause you grief on a cable drive system on those things.

If you look at Ed Sell's thesis I think cable drive was tried and rejected for reasons I can't remember.

Edited 1 time(s). Last edit at 08/03/2012 08:04AM by nophead.

---

[www.hydraraptor.blogspot.com]

Conjecture on why the prevelance of cogged belt drives on repraps: I spoke with an mechanical engineer friend with lots of experience designing cartesian "movers" in this size and accuracy range. His "go to" design would be cogged belts...straightforward, readily available parts, well understand (not perfect, but understood). I asked about cable drive....offered tradeoffs I described above but not a "vanilla" solution...definitely something that could be made to work but more "fiddly". Note: these were his general thoughts, based on years of experience designed devices under schedule pressure and commercial constraints and *not* in the context of the reprap world of "self replicating" machines. I would think printing a drum could well be easier than printing a cogged pulley and cotton thread is a pretty clever, low tech cable source...and closer to achieving the goal of self replication.

As far as why a cable drive would not work on a reprap machine, it is not apparent to me. Possibly, the answer to this question resides in a better understanding of what it means to "not work". I have seen cable drives, moving masses over distances at speeds and required accuracies within the performance envelope of a typical reprap machine work just fine in commercial settings. I have not, however, any experience with machines using reprap'd parts and low tech cable materials...but who knows unless you try!

In the most general case, I think the choice is cogged belt for stiffness (you need to sling significant mass at a good clip) and if you can stand some level of periodic error and you need to build it and great ...and cables if you have the time to fuss some with the design and implementation and periodic error is tantamount. YMMV.

Steve

Well, I read through the thesis and its interesting that i initially had the same problems he had (though he was focused more on a z-drive system), though i found them quite simple to solve also. He also tried a number of materials and concluded "The best transmission cable was found to be steel fishing wire (nylon cord, although better for grip, was found to be too elastic).", so i think i'll have to get my hands on some steel fishing wire and having a quick look around it seems quite easy to acquire.

@mars bonfire, im originally an engineer myself (though its been 10 years since i worked in the field and work in IT these days).

I assumed you had already heard of Tantillus and how I used cables to drive it but I now think you may not be aware that there is a RepRap that uses cables to drive the X and Y axis. I am using Spectra fibre fishing line which is 15 times stronger than steel, has almost zero stretch while still having the grip of the nylon cord. For more info check my blog post [geometricobjectdepositiontool.blogspot.ca]

---

FFF Settings Calculator	Gcode post processors	Geometric Object Deposition Tool Blog
Tantillus.org	Mini Printable Lathe	How NOT to install a Pololu driver

Presumably grip is irrelevant because you anchor the cord to the pulley (8 mm rod)?

---

[www.hydraraptor.blogspot.com]

@sublime, no i didnt see that before.. alot of what turned up when I went searching for stuff was mostly fairly "belts are better" type stuff. Spectra fibre sounds great, i'll have to give that a shot. Tbh, i wish i'd read your blog before designing my parts cause i ended up solving the problems i had with the cable drive the same way you have (hole thru the middle, tied to the rod to avoid slip and wrapping in opposite directions). Havent had any slip so far myself that way. Though in my case "rod" is the shaft of the stepper.

Edited 1 time(s). Last edit at 08/04/2012 05:20AM by Takigama.

nophead Wrote:
-------------------------------------------------------
> Presumably grip is irrelevant because you anchor
> the cord to the pulley (8 mm rod)?

Its true, but the idler side does not have a hole and when first tensioning the cable you can tell the cable grips the rod really well and would most likely not need the hole. The hole was just in my original design and I never tried it without.

---

FFF Settings Calculator	Gcode post processors	Geometric Object Deposition Tool Blog
Tantillus.org	Mini Printable Lathe	How NOT to install a Pololu driver

If I understand what is being proposed, slip should not be an issue since the cable is anchored on one of the drums. The subtley there might be one of multiple turns on a given drum "slipping" or better put as jumping over another turn. This can be tamed, I believe with using a cable with reasonable diameter (not overly "fine") so that it lays up on the drum in a smooth, constant diameter fashion or fancy drum profiles that incorporate a spiral track in the drum surface.

Many cable drives I have seen have a spring incorporated at one end termination to provide constant, uniform tension. Of course, the down side is the spring is can be a source of dynamic stretch under forces induced by acceleration. Just need a stiff enough spring. Concerning Spectra or other polymer cables, they can be very stiff (Spectra certainly is). The only thing to watch for is polymers do typically have a much greater thermal coefficient of expansion and they can creep over time. Again, should not be a big problem for a reprap, particularly with an appropriately designed spring return for constant tension. Stainless fishing "line" sounds pretty good to me...pretty close to what is used commercial applications.
I looked at an example of a cable drive I have in an old mechanism...pretty much metal cable like fishing cable about 1/32inch in diameter covered with what looks like a vinyl overcoat (very thin).

If you want to gild the lily, one could also parallel the tensioning spring with some implementation of a "shock absorber" to provide dynamic stiffness.

I also read the referenced thesis and I don't think that while the author abandoned his cable drive, it in any way establishes that cable drives shouldn't be considered for repraps (IMHO). The more I think about it, the better it sounds (I am most of the way through a cogged belt build based on repurposing an XYZ "robot" and I am too far in to change horses).

One final thought is that I think a practical implementation of a cable drive may require more tension on the drive than a cogged belt (to promote proper layup and accomodate stretch). Just need to be careful to stay within the load limits of the various bearings and drums involved in the drive components (stepper bearings, idlers, etc.) and shafting (avoiding shaft flex fatigue). Fun stuff to think about!

Well, this was the evolution of my own design. The first was:



Very simple design, also quite large (long spool), i think its about 14mm diameter (smaller then the pulley's i was using). This slipped occasionally (most notably when homing), went through a few ideas before i found this one was quite suitable:



I made the spool a little too short on this one, but it has a larger diameter to accommodate the hole where i was tying off the cable. This one works (no more slip), however, i had to up the diameter considerably (about 19mm i think) to accommodate the hole (this one just passes thru the spindle and doesn't connect with the axis shaft at all.

I played a bit with sizing for a little while, then i eventually got to this design (this is not quite what im currently using):



Hole goes thru to the shaft, then the shaft has two conduits on either side for the cable (cable hole is also smaller with this one), this one is also much narrower (these go straight on my stepper, which is 5mm diameter, and the spindle diameter is 10mm) but taller. Basically the way it goes together, you feed the line all the way thru the spindle hole, push it to one end of the shaft hole, tie it off on the shaft, then feed the shaft in (hope that makes sense).

My current design is the same, except the spool is a little longer again (though its not really needed) and the diameter smaller again. The reason for this was ultimately quite simple, as the diameter gets smaller you gain both pulling power and resolution. The downside is you promote some re-wrap, you lose some speed (though i havent really pushed my printer past ~250mm/s for prints, which is roughly double the default speed) and you wind more around the spindle (affects deflection).

But, so far the re-wrap problem hasnt caused any problems - part of this has to do with the way it winds and unwinds at the same time so they never end up over lapping (the opposing directions that is). The deflection (assuming it wrapped all the way in one direction) is almost 0 on my printer. Consider my axis moves 450mm (on the y), i have two idlers at each end of the Y and my stepper in the middle. Ultimately the deflection along the spool is around 5mm (I think), it adds up to a total error (from Y 0 to Y max) of 0.05mm roughly. To me, 450 +/- 0.05 is pretty decent (or at least, the error it will introduce compared to other parts of the printer is fairly negligible.

The last thing is that the current draw/temperature of my stepper (especially on my y axis where it was getting a little too high) has reduced considerably, and i believe so has the wear and tear on the motor. This isnt because of the cable design per se but because the stepper is no longer one end of my axis anymore, very simple to do with a cable design, much harder with belts. But because the tension pulls equally against either side of the motor, i believe this likely lowers the strain on the motor (where a single-ended design translates all the tension directly to the motor's shaft). The tension on the wire is considerably higher then it was with the belts, but i believe the tension on the motor is actually lower with that design.

I do have a forth design in mind for the pully should the cable wrapping over itself ever prove to be an issue aand it would simple be to have two sindles on the motor with some wire guides (something like a rotated W), i havent really gotten too enthusiastic about this one though as it'll be complex to print, i can see it causing as many issues as it solves, the thesis did mention his use of a spreader bar as being problematic and the spindle i have currently is proving quite adequate.

My Y tensioner does have the capacity for adding springs, though i havent done so yet, but i might experiment a bit with that, see what some form of shock absorber design does, though i think it would translate into error (much like how Ed Sells thesis mentions that nylon fishing wire is too elastic).

I really gotta stop writing posts at 3am, i get all blabby and my mind is mush.

Edited 1 time(s). Last edit at 08/04/2012 01:49PM by Takigama.

I don't understand how the motor can draw less current. Do you mean you can reduce the current setting? The current should be pretty much constant, regardless of load, with a constant current drive.

---

[www.hydraraptor.blogspot.com]

nophead: yeah, just reduced the setting basically

sublime: btw, how come 64lb spectra line? what lead you to that particular figure?

Takigama Wrote:
-------------------------------------------------------
> sublime: btw, how come 64lb spectra line? what
> lead you to that particular figure?

I intended to use a little lighter but when I saw how thin it was I went with one that was .5mm in diameter and that happened to be 65lb test.

---

FFF Settings Calculator	Gcode post processors	Geometric Object Deposition Tool Blog
Tantillus.org	Mini Printable Lathe	How NOT to install a Pololu driver

Was reading this thread with great interest. After some googling about cable drives i came across this patent i'd like to share. I don't own a 3d printer, though it's one big goal for the future.

BR

I presume it is necessary to have the cable under tension, is this done with a coil spring or some other method?

Considerably more tension then you apply with belts and im just using a screw-tightening tensioner head, slightly modified version of this:



I tried putting springs in, but it impacts the print too much really (or at least, i couldnt get this to work well, i might give rubber grommets a go next).

Edited 1 time(s). Last edit at 08/21/2012 03:16AM by Takigama.

EDIT #2:
Okay, I figured it out, I think. The spindle for the cable would be threaded like a screw, to keep the cable at the same r position, and the motor would be mounted in a threaded drum that would translate the spindle to keep the cable at the same z position. This way, the cable only changes in theta. The spindle would have the cable mount on the furthest point from the motor, a set screw, and then it would flare out to slightly larger than the motor diameter, with internal threads that match the motor housing's external threads. I don't know how much error this might introduce, especially if it's RP, but it's a design starting point, anyway.

Edited 2 time(s). Last edit at 08/21/2012 01:59PM by Acumen.