Meet "Spike"

[Tray]

I've been plodding toward building a more conventional printer, but an alternate design keeps popping into my head...

What I really I want to build is a printer with these goals:

• extremely simple mechanical build
• lightweight effector for fast speeds
• scales easily and cheaply to large dimension parts.

I didn't see what I wanted in existing designs, so meet "Spike", sketched below. Feedback welcomed, since I hope to leverage your collective experience to minimize mistakes or toss it if others have already tried this and found it an unworkable approach.

https://docs.google.com/file/d/0Byh7txlH45lddXExRl9xVUxXcG8/edit?usp=sharing

The picture in the above link is better than any written description that I can give, but roughly speaking, from the perimeter where the motors are, three uphaul lines locate the bottom end of the spike in X, Y & Z. Three downhaul lines also locate the top of the spike in X, Y, & Z. A lightweight spike separates the two sets of lines, and a hot end is mounted on the bottom of the spike. Coordinating the movements of the top & bottom of the spike keeps it oriented correctly and tensions the system.

(To keep the sketch simple, I didn't draw most of the frame, Bowden tube, air supply tube, or hot end shroud.)

Pros:

* Wicked simple, low precision mechanical build

* Very low inertia

* Scales cheaply

Cons/Risks:

* Requires 6 drivers (plus extruder). AFAIK, that isn't supported off the shelf by any HW controller out there.

* Low usable build volume compared to printer size.

* The equations of motion are, um, interesting due to the spools.

* Threaded drive spools might not handle the +-25 degree approach angle without skipping.

* To keep the lines tensioned correctly, either one of the motor mounts will require a strain sensor or add a spring to the spike.

* Can't use standard end stops for homing.

* Requires parking the effector to shut it off.

* Requires a radially balanced hot end to prevent Z-axis rotations.

Well, that's plenty of challenges, but none are insurmountable. (What's missing in that list?)

To get enough outputs and computes, I'd probably end up with an mbed running drivers on separate breakout boards & hack up smoothieware to run it. 

The threaded spools might not work by themselves, so I might have to add a guide for the filament. (Steve Graber's spools look promising though.) A guide might simplify the EoM a bit as well.

 

For homing, I dunno - maybe use light colored spectra and attack part of it with a black sharpie so I can use an optical pickup by the spool? I'd do it so the transition was near the parked position. Anyone done something like that? Better ideas?

I haven't yet plotted the resolution over the print space, and might look into raising the uphaul motors quite a ways above the downhaul motors to improve print volume. The nice thing about this design is such experiments would be easy

Any other potential pitfalls/ helpful suggestions?

[Tray]

Reattempting image...

[Brad Hopper]

Very cool

[Leslie]

Interesting system, but I think your idea for mechanical simplicity is a false impression.

I think trying to use line for control arms is going to be the biggest problem. Line (and belts) stretch, and in this case, you aren't simply driving something with it, you are actively trying to control something only with string. Even one slightly loose belt on a Delta can cause alignment issues and slack, on this, it means your extruder may not even sit straight. Ultimately, I think the filament is going to be a serious nightmare and something you will never get stabilized enough for any consistency, and this is before you even extrude any plastic.

[Brandon Heller]

The fundamental challenge with purely string-based systems is that each motor does not fully constrain motion in one axis, unlike with a cartesian printer.  That's not to say you can't use just 3 motors - think a deltabot with string, using gravity to hold to the effector down - but that you need something extra to constrain the motion, because the axes don't provide it automatically.  Either you get your force from gravity or a spring, or you add one or more extra motors.   

Your setup could work, but I think there's a simpler option that's halfway between a delta and your design, which would retain the lightweight effector, require no precision linear motion, use string, BUT require one or more springs.   You hinted at it.

Imagine a deltabot frame like that on a Kossel: a triangular prism.  At the top 3 corners, mount the spools to a spherical joint (like a single Traxxas rod end), such that the spool always pulls on the same spot. At the center of the top triangle, place another spherical joint.  At the center of the effector (just a triangle), place a spherical joint.  Connect them with an extendable rod, ideally using a constant-force spring, but probably a regular Hooke's law spring would be fine too.  

-Minimal changes to firmware, with positioning based on the desired line length, rather than the linear axis location.  

-NO board changes or many-motor hacks.

-NO precision linear axes.

You would need to make the effector joint coincident with the hotend and filament input, but this is easy enough to fix, with a wide u-joint with interior cutout, or simply offseting the hotend down.  You'd also need to make a simple L-bracket to connect the motor's mounting holes to the spherical corner joint; with traxxas ends, this is literally an L shape with 4 NEMA17 holes and one for the set screw in the traxxas joint. 

Inertia would be lower than a normal delta if the spring wasn't very heavy.  You could also use CF springs connected at the bottom to add the necessary anti-backlash force.  Would make a nice repstrap - only the spool needs to be precision.

I've worked on string-based motion control before, and have a story that shows this kind of thing is possible, dating to 2001.  This is the MMID:

http://www.i-a-i.com/view.asp?tid=144

It was intended to be an astronaut exercise machine on the space station, but ended up as a spacesuit flexibility measurement device.  Anyway, take a cube, then at each vertex, put an orb with a spool and inline force sensor connected to a spherical joint.  Connect four vertices on each side to a single joint.  Connect the two joints with an exercise bar.  Voila, 6DOF exercise motion, with configurable forces, in a something that compresses to a single drawer.  If you used a triangular prism instead of a cube, you'd only need 6 motors, and it'd effectively be like your spike design but with the 6 motors farther apart.  It seemed to work pretty well; I was just an intern at the time, adding some software diagnostics, not the builder.  It used either PCI or ISA-based motion control boards, one per axis.

Other,

For homing, you could use a ferrite bead + reed switch with software offset to tune.  Or a microswitch design with the same bead.

Not sure how you'd handle the motion of the line on the spool, though.

-b

[Tray]

I appreciate your concerns, however the drive for simplicity isn't intended to make it trivial, but rather minimize the number of failure points. Taking your example, in a rostock style delta, there's number of parts - belts, slides, ball joints and so on that all must operate precisely for the printer to work well. This reduces the list considerably, although it relies on spectra (or perhaps wire) which I don't have experience with. Yes stretch exists, but if a regular delta can manage reasonably x, y & z with filament, this should be able to handle a few more.

[Tray]

Hi Brandon. That's an interesting setup from a decade ago. Printer's a laid out a little differently to keep the lines from hitting the print. With the three string setup, I don't understand how hot end would stay perpendicular to the base. I must be missing something.

[Brandon H]

On Tuesday, June 11, 2013 5:09:13 PM UTC-7, Tray wrote:

Hi Brandon. That's an interesting setup from a decade ago. Printer's a laid out a little differently to keep the lines from hitting the print. With the three string setup, I don't understand how hot end would stay perpendicular to the base. I must be missing something.

You're right that there'd be no hard constraint to ensure the effector stays parallel to the build platform.  If a spring was always pushing the effector, it'd find a stable point, but would be rotated relative to the build platform when the spring force was not perfectly perpendicular to the effector - which is at all positions but directly below. 

But, I still assert that you can use 3 motors and 3 spools, with something at the top of the delta to constrain XY motion, with a piston pushing down on the effector to tension the lines and keep the effector parallel.  You could use a standard X-Y stage, a two-bar linkage at the top of the delta to get arbitrary but constrained X-Y motion like in Nicolas Seward's two-bar, two-string XY idea, or even use ball casters against a plate.  Of course, this does not meet your goals for a simple build or lightweight effector, so it doesn't seem like too useful a thought.

[Brad Hopper]

Yeah, I think 6 strings and motors is perfectly constrained. If instead of a stick, you have a triangular effector on each end the strings will also constrain rotation. The challenge will be getting the initial tension and home position repeatably. Can't wait to see this working.

--
You received this message because you are subscribed to a topic in the Google Groups "Delta robot 3D printers" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/deltabot/-r8AKjQz0oQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to deltabot+u...@googlegroups.com.
To post to this group, send email to delt...@googlegroups.com.
Visit this group at http://groups.google.com/group/deltabot.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Tray]

Hi Brandon.  One thing I don't know in all this is how perpendicular you need to keep the print head. Fairly, I'd guess, but I don't know the tolerance. And that tolerance is probably different along/perpendicular to the path, and near overhangs. 

I agree you could mechanically constrain orientation (for example, something like a folding door), at some increase in build complexity and effector weight. Perhaps I worry too much about effector weight. At some point you're constrained by extrusion speed, and I hear for some materials you need to keep that low for proper adhesion. The 6 motor idea still has a certain appeal though.

[Tray]

Hi Brad.

I think a single plate at the end of the stick would work as well as two. (Actually, you can see a small plate at the top end in my picture, for that very purpose. However, I was pretty meek about it. I haven't sorted through whether it's a good idea or not. While it would help control Z-axis rotation, it would also induce Z-axis rotation on horizontal moves. I don't know if that's a net gain. (Speaking of gain, that control of z-axis would have a variable gain depending on the head position - I'd hate to build an oscillator! :)

I can't wait to see this working either. My available build time is pretty limited, but I'll get there eventually.  I'm also such a noob it'll probably take me a while.

[Anatoly Makarevich]

Hi Tray,

It is interesting but while making a more or less conventional delta type printer I had a very similar idea.

Have you made any progress since your first post?

[Tray]

Hi Anatoly.

I've made a little bit of progress between work, travel and family visits:

• I have a fairly stock version of smoothieware running on a controller (NXP LPC1768 MiniDK2)  This is an ARM cortex3 board.
  
• Derived the arm solution required for the motors. Coded and unit tested it against test vectors on the host. (More on this later.)
• Made 6 motor mods to smoothieware (coded & compiled, haven't started debug yet.)
• Cerberus drive spools _finally_ arrived (vendor had some delays - no gripes there, life happens and he kept me well informed.) Now that I have them in hand, the thread spacing (lead) is way too tight (.5 mm) to expect them to work off axis running directly to the spike. I suspect an M16-2 or M16-1.5 spool might work, but I lack the setup to make these accurately. So my backup plan is to use those and run the line over some small V pulleys that I just got in, which look like they will handle the +-25 degrees I'm looking for.
• I banged together a sturdy (AKA probably overbuilt) chassis
• I think a hall rotary position sensor might function homing, if I first get the spike near the home position. (Plan on parking it in a "parking spot")

So lots and lots left to do:

• Make some motor/pulley mounting brackets (don't like the first ones I tried - machining at midnight is a bad idea.) A printer would be handy for that!
• Wiring up the stepper drive boards to the controller and motors (missing some headers at the moment)
• Debug of the 6 motor mods
• Integration
• Haven't even started the hot end design or mount - thought I'd see how the other bits go first.
• Haven't worked out the tension measurement

A word on the arm solution. Having coded it up - in my opinion, there's too many coefficents to measure/enter (or mismeasure/mistype). So my new plan is now that I know what the solution terms are, manually position the head at a few points and run a regression to produce the coefficients.

[Brandon H]

On Wednesday, August 7, 2013 11:39:44 AM UTC-7, Tray wrote:

Hi Anatoly.

I've made a little bit of progress between work, travel and family visits:

• I have a fairly stock version of smoothieware running on a controller (NXP LPC1768 MiniDK2)  This is an ARM cortex3 board.
  
• Derived the arm solution required for the motors. Coded and unit tested it against test vectors on the host. (More on this later.)
• Made 6 motor mods to smoothieware (coded & compiled, haven't started debug yet.)
• Cerberus drive spools _finally_ arrived (vendor had some delays - no gripes there, life happens and he kept me well informed.) Now that I have them in hand, the thread spacing (lead) is way too tight (.5 mm) to expect them to work off axis running directly to the spike. I suspect an M16-2 or M16-1.5 spool might work, but I lack the setup to make these accurately. So my backup plan is to use those and run the line over some small V pulleys that I just got in, which look like they will handle the +-25 degrees I'm looking for.
• I banged together a sturdy (AKA probably overbuilt) chassis
• I think a hall rotary position sensor might function homing, if I first get the spike near the home position. (Plan on parking it in a "parking spot")

So lots and lots left to do:

• Make some motor/pulley mounting brackets (don't like the first ones I tried - machining at midnight is a bad idea.) A printer would be handy for that!
• Wiring up the stepper drive boards to the controller and motors (missing some headers at the moment)
• Debug of the 6 motor mods
• Integration
• Haven't even started the hot end design or mount - thought I'd see how the other bits go first.
• Haven't worked out the tension measurement

A word on the arm solution. Having coded it up - in my opinion, there's too many coefficents to measure/enter (or mismeasure/mistype). So my new plan is now that I know what the solution terms are, manually position the head at a few points and run a regression to produce the coefficients.

Very cool - that's a ton of progress.  Definitely keep this group in the loop - I'd love to see pics as it develops.  I'm interested in the motor mount design, and I have spools & spare motors to try out ideas.

[Tray]

A bit of playing at the hardware store with some spectra (65 lb. 0.45 mm) makes me think standard bolt thread profiles aren't going to work for direct drive reels. I don't have the capability to cut (or print) custom thread profiles, so for now it looks like I'll add a V pulley between the reel and spike. I have some 624VV pulleys that look like they'll work, if they don't have too much bearing slop.

It turns out it won't affect the calculations too much - there's still line unreeling from a spool and tangential contact with a cylinder from different angles, just rearranged.

[Tray]

On Wednesday, August 7, 2013 3:00:50 PM UTC-4, Brandon H wrote:

Very cool - that's a ton of progress.  Definitely keep this group in the loop - I'd love to see pics as it develops.  I'm interested in the motor mount design, and I have spools & spare motors to try out ideas.

Hi Brandon. I'll have to get in the habit of photos. First round of motor mounts was a late-nighter before a meetup. My hole saw hadn't arrived, so I lopped off a bunch of pieces of angle iron with a saws-all and drilled a pair of holes in each. Then screwed pairs of those to the motors. Seems like it has the potential to slip. I wonder what pressure is required to get liquid locknut to seize? (I haven't looked at reprap.org yet to see how the big kids do it.) If I weren't so cheap, I'd just plonk down the $10 apiece for premade mounts.

>I have spools & spare motors to try out ideas.

Try out ideas? Please do! It'll go faster that way.

[Anatoly Makarevich]

Wow that is really a lot of a progress.

Why have you chosen to go with ARM based solution instead of a more common 3d printer's boards/firmwares?

Another point that bothers me is the need for 6 steppers which is a bit of overkill from my perpective. That leads to double the power consumption. So you need a better power source which is again is an addition to the price.

I am positive that for a 3DOF machine 3 motors is enough. Have you thought of a simplier solution? I have some ideas. If you are interested I will create a scetch I am thinking of for quite some time.

[Tray]

Hi Anatoly.

The positioning solution requires more computing - a few vector products and square roots per motor, so I wanted a faster processor. Also I needed more stepper drivers than common boards support.

>6 steppers which is a bit of overkill

It's only overkill if you can remove some motors and still meets the objectives. :) I'm trying for a mechanically minimalist build, not numerically smallest build. The extra 3 motors seems like a simple, low mass way to orient and tension the system. I'd be thrilled to hear of a better way.

>I have some ideas. If you are interested...

Most definitely.

[Anatoly Makarevich]

Tray,

here is what I'm thinking about. The idea is to move both ends of the 'spike' simultaneously by the same motor/spool. the Blue strings should be simultaneously winded and unwinded on the same spool attached to the motor. The upper thing is a simple idler.

This eliminates the need for a tension control between 2 motors and greatly simplifies the calculations complexity. I am afraid that for 2 motors scheme you will need a tension sensor. In other works a kind of a feedback system.

Of cource there should be some force applied to the spike downwards. So there are 3 strings or springs in Red that make the spike stable.

So the only problem as I see is to find a balance between red string tensions and the motor strength and the speed of effector movement. But I am positive this is doable.

After (and if) I create a delta printer I will try to implement the above scheme. It should be very cheap - this is my goal. I understand that you have slightly differen goals but from my point of view the above is the simplest solution.

[Ray Tice]

Hi Anatoly.

I quite like your parallelogram layout. Those red tensioning springs worry me though - they need quite lot of travel and with the some random mitoiin could develop some weird dynamics.  As drawn they might intrude into the build volume during extreme movements.

Another way of tensioning it would be to... add another motor. (But just one.) Instead of the red strings, run a string from the top of the spike over an idler near one of the other three motors, and then connect to a fourth motor. This tensions everything. You can add some stretchiness to this last string and maybe skip the feedback. (I'm a little foggy this morning, but I think that''s workable.) There are common boards out there that will drive 4 motors.

A few words on the parallelogram - this increases the need for build precision a bit. Also, to maintain the parallelogram's integrity, you'd need to consider the effect of the unwinding along the spool - it's different for the two strings here. I haven't calculated whether it's significant.enough to worry about. Adding yet another pulley and repositioning the motor would help that. Some thought on how the string winds onto the spool would be in order - maybe a two start thread on the spool?

Keep up the great ideas!

[Anatoly Makarevich]

You are right, the need for springs makes the solution non-perfect :) It does bother me so may be there is an even more elegant solution.

I think that you should put a little hole before the spool and put the string through the hole. That will eliminate the movement of the thread on the spool as well as the change in the angle while the effector moves. It is even easier to control the position of 'holes' instead of motors and pulleys. IMHO, of course.

Please let us know when you make any progress!

[Tray]

I considered eyelets earlier in this thread - I'm still concerned about wear, but I haven't done any experiments on that. Also it occurs to me the the you proposed is overconstrained - the third parallelogram (and thus the third upper line) isn't needed if you tension the rig appropriately.

Have you given any thought to the idler mounts in this system? Their placement must be precise and probably adjustable. Any deviations would cause the hot end to tip. I don't know the precision requirements in hot end orientation,

[Brandon Heller]

This is genius in its simplicity - thanks for posting!  Of course there are potential practical concerns like tensioning and getting the line lengths exactly right, plus possibly needing bearing mounts on motors to avoid needing eyelets.  Feels like synchromesh would be a perfect fit here - with a small pulley, the side to side line twist as the extruder moves might be negligible.

Also, take a look at the Simpson design, which has a similar need for concentric bearings on its effector.  My hunch is that by modeling the string mounting offsets on the effector in the position-control math, these can be avoided..

[James Pottebaum]

Actually one chip ARM processors are cheap - -  about the same as the 8 bit that is commonly used

if I could use 6 steppers and get rid of the slides - - the cost will be lower

Jim P

[James Pottebaum]

being and engineer  - - - I find most of the chassis are a bit weak - - - - fine for small size and very limited for fast moving- - too much flex

[James Pottebaum]

as for homing -- - in many industrial applications there is a dual sensor
one is on the encoder - - - for the fine position and a once a revolution of the encoder - - - - This works with the limit switch on the bed. or travel - -

could add a optical or other sensor for the spool, and another on the line or other place

On Wednesday, August 7, 2013 11:39:44 AM UTC-7, Tray wrote:

[James Pottebaum]

I have been playing around with designing a high end 3D printer set of control electronics
High currents - - 5 amps, and multiple drivers - - - starting at 6 and expandable to 10 - - - - just for applications like this
or multiple extruders

Jim P

[James Pottebaum]

actually, I was at NIST 12 years ago doings some certifications on a load cell and electronics.

I was shown around the place.  I was this positioning system - - - it was 3 post with motors to control the line (steel cable) to what is best remembered as a weight
The post were about 15 to 20 feet a part and 10 to 12 feet high

I kind of joked about what precision could it have - - - note this is the National Institute of Standards Technology - - - 1/4 inch or so. I got the response of as this was 12 years ago. I do not remember if it was 0.001 inch or 0.0001 inch - - - I gave it a second look in amazement - - -

So I do not have a problem with the position ability of his system.

if he is using braided fishing line - - - it is good. -- - - I am looking at building a delta on the large size and picked up off of Ebay a spool of braided 100 lb test fishing line for it.

[James Pottebaum]

the point that is missing is the 3 lines on the top keep the extruder vertical - - - or at any desired angle - - -
the 3 lines at the top provide the tension needed.

the only constraint that I see is that the top must be higher than 3 top line control motors - - - to keep a steady downward push/positioning control

[James Pottebaum]

variable gain - - - this is not an analog system

The math is a variation on Delta printers - - - instead of the rods being fixed length. - - and the slide moves up and down. - -
the rod length (Line) is a variable into the same type of formulas.

The top lines have the same math involved - - just inverted

the only design constraint is to match the speeds of the motors so that over tightening or overly loose conditions occur. - - - overly tight - - stepper motor skipping.
 this could be handled by some small springs to give a bit of flex on the top strings.

[James Pottebaum]

I hate to be the one to burst your bubble.
this looks good and works only if the stick is moved up and down - - if you move it near one of the verticals - - the length of the blue string had to change to pull it in and the other have to get longer - - - which is not possible  by this concept - - -

Jim P

[James Pottebaum]

hum - - - being a fisherman. - just getting into it.
the casting reels - - - have a means of laying the line across the spool to make nice layers
a small gear at the end of the spool to a larger gear on a shaft that is attached to a shaft that controls
the laying of the line - - - thus the feed into the spool - - - - - a bit more complicated - but will solve the problem of laying the line that is coming to the spool from different angles.

and other would be to feed the line over a V-slot pulley and then with the motor and spool spaced a bit away - - - the feed to the spool will be uniform independent of stick position.

Jim P

[James Pottebaum]

I have been playing around with development of and advanced set of electronics complete with a custom designed Hardware based motor control chip - - the result will be higher current drive, and more motor controls. - - -

as was pointed out by an other post talking about how microstepping does not work - - - - he was totally right about the magnetic forces act like a spring - - and with physical load - - torque on the motor from a load - weight - spring etc.  the rotor will never go into perfect alignment. - - - - Some very high end motor controllers are doing this.  by High end, I mean  on $100,000 Plus industrial equipment - - - - The concept that I was looking at for the electronics and motor drive, this ability is possible. Knowing some information about the stepper motor, measuring some parameters - - the displacement - which is related to the motor magnetic - - torque - - drive current, and external force applied.

why my comments - - - here is the spring tension system - - - add a bit more math - - - allow the top strings motor control not to be a fixed position - - go 1000 steps to here.
but go to about 1000 steps until the spring tension is - - - and microsteps  before or after desired position to keep top string tension as desired.

the cost of manufacture is low and the expected price will be on the high side of current electronics - - -

I would like to get feedback about interest - - if lots of interest - - it will push me to bring it to market

Jim P

On Monday, June 10, 2013 1:23:46 PM UTC-7, Leslie wrote:

Interesting system, but I think your idea for mechanical simplicity is a false impression.

I think trying to use line for control arms is going to be the biggest problem. Line (and belts) stretch, and in this case, you aren't simply driving something with it, you are actively trying to control something only with string. Even one slightly loose belt on a Delta can cause alignment issues and slack, on this, it means your extruder may not even sit straight. Ultimately, I think the filament is going to be a serious nightmare and something you will never get stabilized enough for any consistency, and this is before you even extrude any plastic.

[Tray]

On Friday, August 9, 2013 3:17:28 PM UTC-4, James Pottebaum wrote:

variable gain - - - this is not an analog system

Sure it is. The hot end has mass, the strings will stretch, rigid parts will flex. And as the print head moves around the relative lengths change of the strings change, which changes the response. Hopefully I can these issues to a minimum.   

[Tray]

Hi James. Yeah, the 60 degree thread angle on most bolts doesn't work for this. 90 degree might and a more rounded thread profile would probably work better too. I don't have the tooling to play with either so I'll run the line over a v-pulley and onto a threaded spool. 

[Tray]

I was worried that the lateral movement of the winding might throw your spike out of alignment, since it will contribute differently to the upper and lower strings as the spool unwinds. However a back-of-an-envelope analysis says the tip that induces would be very small and hopefully not worth worrying about.

[James Pottebaum]

this is digital control - - - -

I understand flex.  Here you set a desired position, and the system will settle at that position, no feedback.

to me gain means feedback - - as in analog systems - - - opamps etc.

I think what you are talking about is the way that the term gain is used in analog motor control systems.
this is normally used in feedback control systems.  

In collage, in my engineering science course - - we had to analyze a plotter control loop - - before digital processing (no PC's, no Apple II, and no microprocessors)
this consisted of a DC voltage controlled motor, tachometer for speed feedback,  We were to analyze for response stability - - to find the critically damped performance
 - too much gain - - overshoot - - too little gain - - overly damped response.

finding the critical response gain was to control the respond for fasted settling or under damped response by adjusting the gain

but here it is all digital - - - with controlled acceleration and deceleration with no possibility of overshoot or under damped
so the control system will position precisely.  

But the flex in the system is a different aspect - - - -
I believe what you are calling gain, I am calling it the acceleration and deceleration factors as with many items the old term is carried forward
to new technology.  

higher acceleration means faster movement to position - - and will result in higher dynamic loads and this more flex and stretch.

Sorry about the confusion on my part

Jim P

[James Pottebaum]

more that I think about your design the more that I like it.

I was looking to build a standard delta. - on the larger size for some upcoming projects.

with proper top tension, I see only one problem - - - the string to the spool angle variation as you commented on.
along with a bit more software cal's,

the positioning dynamics are very flexible - - - including tilting the stick in any direction for some interesting print options

I was looking at using 2 motors per vertical for a total of 6 motors. Why?  This will allow for independent control of each connection rod.
I have seen CNC mills using delta version  - - - In that system the control the RODs were lead screws and moved in and out. - -
with the result of being able to tilt the cutting head in a wide range of angles.   This would allow the same ability and would be interesting to try.

In the current delta, any rod end vertical position results in a valid head position.
but with each rod being independent controlled. I will have the same aspects as you do.
only with solid rods  it is very possible that combinations of  end vertical positions that could result in binding or simply not possible - - so need to tie the motor actions closely together that is not needed on the standard delta.  In your case the need to keep from over tensioning the strings

think about this - - - might try this concept - - - - as many aspects are the same between what I am thinking about - - - if one concept does not work, I can try the other one.
this project is going to push forward the multiple stepper motor driver system - - 6 only for the motors and then extruders and want to try new versions that are lighter weight along with
multiple colors - - - so - - - and like you said - - need more than the standard processor - - need 32 bits processing power - - and higher speeds.

[Tray]

The settling you refer to is the result of an analog feedback system at work. (even a weight hanging on a string is a feedback system.) In this case it's nonlinear feedback, and the system coefficients depend on where the print head is in the print head.

The college experience you described is closed loop gain, which is one form of gain. Stability is achieved when the total gain in the loop is <1 for all frequencies. Oscillation occurs if the system is underdamped and results in one or more overshoots of the system in response to an input.

In the case of the z-axis of this printer, I'm concerned it will overshoot/oscillate due to torque applied from acceleration if the center of the mass in not aligned with the center of effect (the spike). Obviously I should try to get it lined up to minimize the torque, but reality is imperfect and filled with tradeoffs.

[Tray]

I'm in the same boat James - I had been planning on a more traditional printer, but then this popped up, and somehow it skipped to front of my glacially slow build queue.

For tensioning, I bought a cheap pressure sensor for my 6 motor design, but I haven't decided how to best integrate it to control tension. Instead I might start by using monofilament fishing line for 1 of the upper strings (or at least part of the string). The stretchiness will put some compliance in the system.

[Tray]

Well, life intervened there for a while. But it's calmed back down now. I sketched at hybridizing Spike with Anatoly's parallelograms...

Not sure about that. To get the same print height, the posts are 50% longer. Then, to keep tension at all points in the print envelope, the spike needs to be lengthened as well, with the tension line attached to the top of the spike. This introduces a bending torque on the spike, since this line is above the other two upper lines. I could move the other lines up as well to reduce the torque, but that increases the distance between the lower lines and hot end and might lead to a wobbly print. Seems like some experimenting might be in order. 

Or maybe I'm missing something clever here.

To unsubscribe from this group and all its topics, send an email to deltabot+unsubscribe@googlegroups.com.

[Anatoly Makarevich]

Tray,

You know I think you are right that the stability of the spike is best to control by an additional motor.

But I'm afraid you're current sketch won't work. The forces applied to the spike will bend it and introduce errors that can't be controlled.

Look at what I've been thinking for the past week or so. This setup even may eliminate the twist of the spike around the Z axis which is a problem. to do this we need to tweak the line setups a little bit off the center of the spike.

3 out of 4 motors control the position of the spike while the 4th (red) one is just a 'passive' motors which reacts to the tension change. We may use a Hall sensor, a metal spring and a pair of magnets to do this.

Let me know what you think. I hope Brandon will also be interested.

By the way just curious what software do you use for 3d modeling? 

вторник, 24 сентября 2013 г., 2:10:42 UTC+4 пользователь Tray написал:

[Tray]

Yes, the bending force concerned me too - making the shaft rigid enough might make it too heavy. (It also just "looks" wrong.)

The four parallelogram is interesting - it would increase the usable print area. However, the build height would be very short, for two reasons: 

First, the lines would interfere with the print object. You could extend the nozzle further below the bottom hub, to help this, but then you need a very rigid shaft again to control wobbling as the print head moves.

Second, the bottom hub has to stay vertically between the upper and lower pairs of motors. You could increase the offset between the pairs of motors, however eventually this brings in sharp angles on the control lines, which reduces control.

(Coincidentally, my primary goal objects of the large format version of this are very flat (~10:1 aspect ratio) But the remaining items are not.)

For software, I've been learning/using OpenSCAD. I yearn for a better mix of procedural and interactive though.

[Anatoly Makarevich]

Here is an update. This way the spike won't rotate around Z axis so the position of an effector is 100% under control.

Hope this makes sense.

[Anatoly Makarevich]

It's interesting. I also use Openscad but your pictures are much more prettier :). Could you please share the code for the above pictures?

[Tray]

On Tuesday, September 24, 2013 6:19:37 AM UTC-4, Anatoly Makarevich wrote:

... Could you please share the code for the above pictures?

 

It was my very first scad program, so its style might not be so good :). but you're welcome to it...

https://docs.google.com/file/d/0Byh7txlH45ldNHM3eVBBNjd1RkE/edit?usp=sharing