HBot using fishing line

I've been messing with an HBot design, after seeing Sublimes work using fishing line in place of the timing belts I thought I'd give it a try.
The test doesn't run the line through the pulley, though I may eventually go that way, it's just wrapped around 4 times, there is a design problem as it stands right now, there isn't enough space on the drive pulley for the line to walk across the entire extents of the motion, and I didn't give the line enough clearance from the Y axis pieces, but I think the initial result is promising.
The test is running at 150mm/s and 9000mm/s^2 acceleration, it'll run faster, but the line starts slipping, I'll adjust the design and print some more pulleys tomorrow.

HBot Video

The video looks very promising. I know that tech was used for decades for radio tuners and the like, before they were completely migrated away from mechanical adjustments (aside from volume and tone pots) but I still was surprised at how well it moved. Have you tried it with some weight on the center section, to mimic the inertial loading of an extruder assembly? I'm wondering what a "substantial" inertial load will do to the accuracy under a rapid change in acceleration. BTW I liked your carriage system. clean...

My intent is to use a Bowden style extruder, the original design goal was to build something with similar moving mass to an ultimaker, but without the need to use bushings because of the rotational and linear motion on the same rods.
Having said that I'll probably try mounting a conventional extruder on there at some point, but I'd expect to have to reduce the acceleration setting.
I have a couple of ideas which I might get to this weekend which should let me improve the performance further.

I think you pretty much answered the basic concern I had by saying Bowden, since that's one of the obvious benefits to a Bowden style...low inertial loading. In any case, looks really nice in motion and it's incredibly "clean" and trim. With an extrusion head mounted, no matter the type, it will still carry a higher moment than it does now, but with a Bowden by far less than with a 340g (3/4 lbs) nema17 sitting atop, and with far less radial torque from the offset motor mount.

I'm still wonder how the system will behave over time, whether there might be imparted "drift" from the discontinuities, but I suppose some of that will be discovered when I get my system up and running and can start building my own secondary testbeds.

nice work.

Edited 1 time(s). Last edit at 09/01/2012 11:41PM by xiando.

Looks really nice. One thing to think of is the smaller the pulleys the cable wraps around the tighter is will cinch around it. This will reduce the slipping and increase the torque and resolution.

---

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

Rather than using spectra, has anyone considered using kevlar line? From my years of stunt kite flying, I know that kevlar will give you much more friction, which is bad for kiting, but I assume it would be good in this hbot application.

Yeah the spectra is pretty slick (as in slippy), but it is braided, so it has some bite.
I'm considering using a rubber material, on the outside of the pulley, but I don't know how it would wear. ABS might also be a better choice of a material, they are currently PLA.
The current driven pulleys are 12mm in diameter so there is some scope for possibly making them smaller, the issue becomes how tall they then need to be. There has to be enough space for the line to walk as it wraps on or off, unless I use a moving screw as a guide and that's getting complicated for an initial implementation.

Of course, I forgot about how much cable you have to move to get the head to move that far. Maybe just the fixed point is needed so it can not slip no matter what.

I am just glad to see it works so well and is being adopted by others. Good work.

---

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

I had this same idea, very cool to see someone else thinking along similar lines and actually getting it working

Have you had any issues with the line "climbing" over itself on the pulleys? Could you post some close up pictures of the pulleys you're using, and how the line wraps around them?

I've tried 3 different types of pulleys, they look like you would expect. The first one was way to small and didn't accomodate the walking, as a result the line would wrap over itself, though with only 4 wraps, the steppers were quite capable of pulling the line out of the wraps. in fact this hid a bigger issue with the HBot design and line - see below.

As long as there is space for the line to climb and sufficient tension on the line, generally it doesn't wrap over itself.
I did also try a pulley with a screw thread that I cut on my lathe, it was very promising, at stopping the line overrunning itself, and the AL had more friction than the PLA pulleys I printed, but doesn't address the issue below.

There is a bigger issue with line used like this specific to the Hbot mechanics.
In something like Tantillus, the line always walks in the same direction on the opposing "pulleys" (or rods on Tantillus). On an HBot however this is only true of motion in the Y direction, in X they move in opposite directions this results in the length of line between the pulleys effectively changing and as a result the tension on the line. This is what causes the slipping at higher speeds, as it moves in a direction where the line tension is reduced, the line tries to equalize the tension on the loop, but it's not instantaneous.
With fast moves the change in tension is significant enough that the line loses traction on the pulley and it slips.

Even if it didn't slip, it would cause unwanted motion in the HBot mechanics, that leaves only 2 options for something like an HBot with line, somehow get enough traction with a very small number of loops around the pulley and have the pulley constrain the position of the line, I haven't been able to get enough traction to make this feasible, or using a moving screw to drive the line, the only ways I can see to build this mechanism are "complex". So for now I'm shelving the idea on the HBot.
I have a somewhat different design that uses a different belt layout that may work better, but as of right now it's just a sketch.

On something with conventional loop like belts on the axis, I think it's completely feasible, I'd go with the screw thread pulleys, it's probably not necessary, but I'd probably do what Sublime did and pass the line through the pulley, this requires more space for the line to walk, but i guarantees no slipping.

Thanks for the really in depth response.

Could you expand on your third paragraph? I can't picture in my head what you're talking about. The only difference that I can think of between tantillus and an H bot is that the idler pulleys on the ends of the X beam are "flying", i.e. they change position relative to the Y end pulleys. But I can't see how this would impact the cord tension...

On a design like Tantillus, there is only a single drive and a driven end, they always move in a known relationship to each other, i.e. when one end goes clockwise the other end goes clockwise this means you can arrange line so that it walks in the same direction at both ends, say towards the outside, retaining fairly constant tension. In addition any error you introduce is in the direction of motion and likely consistent for motion over the same range.

On an HBot you effectively have a long loop with two driving motors, they move either in the same direction or in opposite directions depending on which direction of movement is required, that means there is no way to guarantee that the line walks in the same direction. As they walk in opposite directions the tension on the line changes, in practice the tension change can be enough that even if the line feels very tight in some locations it can be much less tight in others, and slip.
It's exacerbated at the ends of travel because the vertical (in my case) offset in the line causing tension changes as the YCarriage moves to the ends.
You can reduce the latter effect by moving the driving steppers further away from the end of motion of the carriage, and I had considered mounting the steppers on the outside of the box.

Perhaps one solution is to have some sort of dynamic tensioning of the line, but the issue would be that the change in tension would then be converted into motion of the head and it's unclear to me how this error would present itself.

I see what you're saying now, thanks for the clarification.

I am surprised the change in vertical position of the incoming/exiting line is large enough to impact the tension in the line.

Could the issue be that there is too much resistance in the pulleys/cable handling on the side away from the drive motors?

crispy1 Wrote:
> Could the issue be that there is too much
> resistance in the pulleys/cable handling on the
> side away from the drive motors?

I don't think so I have 9mm bearings inside the printed pulleys, and they spin very freely.

What about just pre-tensioning them with lot of load using something like a guitar tuning peg. You should be able to get enough tension on it that it can never go slack. Or as you say use active tensioning with a spring or something. I know plotters that use cables have springs for tensioning.

---

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

I'd think just putting one of the idler pulleys on a threaded rod to give it some adjustability would be simpler than a tuning peg.

There was certainly some scope to increase the tension, though I'd have to attach the steppers more securely than the current arrangement.
probably the easiest approach for testing is some sort of screw tensioner at the carriage.
Another possible option is to use rosin on the line, it's used to increase the friction on violin strings, but I don't know how it would hold up.
I think though first I want to remount the steppers horizontally to remove the current space restriction.

Please keep us posted, very cool to see how much progress you've made.

If I find enough time in the next few months I have plans to develop a makerslide-based H-bot that is cable/fishing line driven. Currently I think a completely fixed bed and an X-Y gantry that moves in Z is the way to go.

I like the idea of a completely static bed, but I think there will be less mass to move in the bed than to move an XY stage, and if either one is moving you have the same fundamental leveling problem.
I do think if you can get the steppers off the head by whatever means be it Ultimaker style, HBot or some other belt layout, and you're willing to use manufacturing methods that allow you to build something square relatively easily, then having an XY gantry is a better solution, than having Y on the bed.

My thinking is it's equally mechanically complicated to lift the bed or the gantry. I don't think the weight has much of an impact since the Z axis moves so slowly. So I'd rather completely fix the bed and just remove that variable from the alignment, setup, calibration, etc entirely.

I'd also like to move away from leadscrews since they are so difficult to get straight and remove wobble. Right now in my head I'm picturing a belt or cable drive arrangement that lifts at multiple corners driven by a single or two motors.

Here's a short video of my prototype running: [vimeo.com]

I solved the cable wrap issue by only taking 2 wraps around the drive pulley. I found with 2 wraps it would not overlap itself when it ran out of vertical travel on the pulley. The disadvantage of this is it requires higher line tension.

I don't think the line is slipping on the drive pulleys with the current speeds and accelerations, but it is hard to tell. I don't think it has a lot of traction margin at the current tension. I plan to make a design change to the idlers to allow more adjustment range.

It looks good, I really like the use of tslot.
My issue was two fold with the original design.

The first was that although the line appeared to have good tension, as the line walked on the two motor pulley that tension would change, leading to slipping at aggressive speeds/accelleration. This is probably hard for you to see, it's easiest to see the change in tension by plucking the line one side (known length) and listening to the pitch variation. It should be noted the impact of this effect diminishes as the H is scaled up. It didn't help the original has no tension adjustment.

The second issue was the line walking on itself, even with two wraps I would see this, and coupled with too low a tension I would see the line momentarily on one of the motor pulleys come off the pulley at a slightly different point, this cause a momentary positional inaccuracy. This could probably be resolved with more consistent higher tension.

My redesign, which if I get a chance this weekend I'll show pictures of, basically changes the driver end to have both the motor and an idler pulley, the line wraps once exactly 360 degrees around the motor pulley then continues to the idler pulley, maximizing the contact area for a single wrap, since it's just one wrap I can remove all the vertical motion from the drive pulley and it can't walk on itself. However it still needs enough tension on the line to grip the drive pulley adequately.

My tensioner design is far from ideal, it would be better tou tension by moving the idler pulleys, but my original box is a bit on the small side to do that.

I did finally get some time to work on this, latest changes can be seen here HBot2 Video.
The redesigned drive pulleys prevent the two issues I mentioned in the previous post, I've added a trivial tensioner, to the carriage, basically it's a rod with a hole in the center, the bottom of which has a square end that fits in a square hole when pushed downwards, lift and turn the peg to increase the tension, push down to lock.

With moderate tension there is no slippage in limited testing, I'll decrease the size of the drive pulleys in the next iteration, I don't think it will impact the friction, and it'll mean I can increase the steps/mm somewhat.

There is a mechanical issue, the moving cross piece is flexing slightly as the carriage runs along it, I think the issue is that the plastic idlers on the moving part are slightly out of round, which effectively slightly changes the amount of cable on either side causing some twist, it might be better to have the line run on the bearings, Ill have to think about it.

Quote

There is a mechanical issue, the moving cross piece is flexing slightly as the carriage runs along it.

I have a similar (maybe identical) issue and it's driving me nuts: X moves cause the X crossbar to shift/tilt in -Y for +X moves, and vice versa, only at the start of the move. I think it is because the cable tension on the sides becomes unequal, so the reaction force on the idler pulley at the end of the X crossbar changes, which causes the bar to shift.

My original theory was that the change in cable tension during X movement should be very low compared to the static tension in the line, but this appears to be a bad assumption. Adding more cable tension seems to make the problem worse.

Currently my idler pulleys are not mounted on bearings, it's just a PLA pulley with a hole drilled in the center riding on a steel shaft. The friction between the pulley and the shaft may be making this problem worse. But it looks like your pulleys are on bearings and you still have this issue.

Quote

I'll decrease the size of the drive pulleys in the next iteration, I don't think it will impact the friction, and it'll mean I can increase the steps/mm somewhat.

Smaller diameter pulleys should have higher gripping force, all other things being equal.

The smaller = more gripping force is true for a given tension, but there is more contact area with the larger pulley which also factors in.
Anyway I can run the current version at > 700mm/s with acceleration @9000 without it slipping, so I don't think slippage is much of an issue with reasonable tension.

Part of the skew issue is the box I'm mounting it in is skewing as it moves, I need to add some corner braces to remove that, I didn't see much of the same effect with the first prototype, but the rods were a lot tighter in the holes, which probably means the box was effectively a lot stiffer. Having said that I can't see you setup flexing much.

My idlers are printed V Grooves over 9mm bearings, to they have very smooth motion, but they are not perfectly round.

The effect also seems to be a lot less pronounced at higher speeds.

I discovered my printer was not square and everything circular I had printed was actually an oval by .05" or more. So I spent Sunday rebuilding the frame, which fortunately is pretty easy with an ord bot. Now my circular prints are only out by .005" or so. I will need to re-print all my idlers and drive pulleys, I'm sure all the eccentricity was not helping the drive issues.

I also redesigned my X idler pulleys for the third time, so they will mount on bearings. I will see if this improves the x bar shifting.

Mounting the X idlers on bearings seems to have solved the cocking at the start of the X move. However, when I reprinted my drive pulleys I bored the holes out off-center so now I can see the x carriage wobble a bit during Y moves. Fortunately they are quick to print - only about an hour to print and clean them up.

I'm not 100% confident I solved the cocking because I rebuilt the gantry, so it's possible I got the bearing alignment *perfect* for once and there's no slop... But I think this is kind of unlikely.

Hi guys,
I am really impressed with your progress (congratulations!) and very interested in the h-bot design.

Polygonhell, can you please share your design files (including of the first version)?
I would like to take a look at them.

Thanks in advance.

I'm travelling all of this week, I will share the files when the more obvious issues have been addressed.

I'm not sure I still have all of the V1 design files still, but I'll check when I get back, I probably at least have them in STL format, but there are issues with a number of the V1 designs, like incorrectly sized nut traps among other things.

At the moment development has been at the level of design piece print piece, refine piece print piece, since everything is a one off it's often easier to fix something with a knife than it is to fix the design and reprint.

I did some more work on my prototype. The main problem was all of my idlers had been printed with lots of out-of-round-ness, and this effectively changed the cable path length and resulted in the head wobbling during Y moves. I decided that instead of trying to print perfectly circular idlers, I would just run the cable directly on the bearings. In order to keep the cable on the bearing, I designed little "keepers" that I superglued to the bearing's faces.

I also re-designed and re-printed my drive pulleys, so that they had double opposing nut traps. Unfortunately I am still using an oversized SAE drill instead of a proper 5.0 or 5.1mm drill because it is all I have. To combat this I wrapped the motor shafts in 1 layer of kapton to make them a press fit to install. They are circular within 0.003", which I consider good enough for my purposes.

With all of these changes I see nearly zero head wobble when I manually push the x bar assembly to and fro in Y. This is a big improvement. I have not been able to test with actually running the motors yet, because the computer I use is tied up on my working printer doing a 5 hour print.

My conclusion is this style of bot is feasible. The biggest hurdle is X-Y axis coupling, which my experience indicates is caused by poor cable path (too much friction, idlers/pulleys aren't round, etc). This coupling exposes weak points that might otherwise get overlooked in a mendel-style bot where the axes are driven separately. My wishful thinking is that this coupling forces the designer to make the design inherently better, because problems in the cable path are not masked.