Monday, November 24, 2008

 

Idea for reduced cartesian structure


I've been playing with ideas on part count reduction for Darwin. Having an elevated XY table requires a fair bit of support structure. After brainstorming with Adrian we came up with this as a concept - I will try to do a feasibility on it at some point. Interestingly the rolling table concept could lend itself well to a large-batch production machine: if the table was pimped with a conveyor we'd have an infintely long work area.

Labels: ,


Comments:
Ohh I like this design. I may go this route with my repstrap. The conveyor would be neat if it self de-seats parts without breaking them. Maybe sandpaper from a belt sander as the surface (high grit).
 
Nice. I'm guessing the Y motor will need to be a bit bulkier to move the axis at comparable speeds.
 
Yeah, I'm concerned at the ability to move the Y axis at speed. The current Z is certainly bulky, but also only needs to move at low speed. I'm not keen on the reliance on roller-baerings either.

Also, we're looking at 2 more stepper motors in this design - I'd like to reduce the number if anything.

Finally, it's going to take up twice the bench area of a Darwin, and in my workshop bench space is at a premium.

If we want to simplify the design, we could cantilever out the Darwin Z axis so that it only needs 2 supports, much like they do in the Fab@home machine.

Not sure 2 bars are going to give the X axis all the support it needs either. Ian's use of 4 bars on the X axis is a nice design and suited well to head interchange.

Vik :v)
 
If I understand correctly, the y-axis loses the mass of the x-motor, x-axis and extruder. Instead it has the mass of the table and the object being constructed. That could well be less if a strong light material is used for the table.
 
I just happen to have an X axis off the rails at the moment (probably not the only thing off the rails, but that's another story). It weighs spot on 2000g including the extruder and docking mechanism.

A sheet of 9mm MDF for the bed weighs 860g. Add on a trolley mechanism and it might well be a bit under.

But the area taken up by the machine is still twice the size of a Darwin. That concerns me for some reason.

Vik :v)
 
Yes that is a concern for me as well. One could always make the y build area half that of x, so that the overall size is about the same.

It is not often that you would want to make something that is both long and wide. It takes too long!
 
You know, even a build area the size of a belt sander would let you build all the Darwin parts.

If you could keep the warping under control (heated base?), a way to automatically remove the parts would greatly speed up the time it takes to replicate, as well as reduce the size of the machine.

My Darwin built all its parts in roughly a month, but most of that time it was sitting idle with several parts on the bed, waiting for me to come and remove the parts, clean the bed, and program the next build. I didn't do much optimizing of my build area though.

Removing the parts was a bit time consuming too, especially when I was overzealous on getting good adhesion between the part and the bed. Lots of work with chisels and knives.

A conveyor belt would be great for part removal in that it would peel the parts off the bed - you'd only have to break a small section of raft at a time, as opposed to right now where the entire raft has to break at once during part removal.

Imagine a belt sander sized Reprap spitting parts one at a time into a bucket. You could very quickly build a complete set of parts that way, you'd just be limited to small parts.
 
Building on a flexible belt would be nice for removal, but the bed has to be stiff to resist the warping. I can't see how it can be both.
 
Table issues (non-conveyor):

> I'm guessing the Y motor will need to be a bit bulkier to move the axis at comparable speeds. I'm concerned at the ability to move the Y axis at speed. A sheet of 9mm MDF for the bed weighs 860g.

How about perspex sheet with a bolsar wood top?

> It's going to take up twice the bench area of a Darwin.

Disagree, we know there's a large amount of dead space in Dar's current configuration. Knowing what we know now, I'm 99% sure I can optimise this design to fit the existing footprint and still achieve the current work area. If there is any compromise, nop's point is valid: "It is not often that you would want to make something that is both long and wide"


Conveyor bed:

> Maybe sandpaper from a belt sander as the surface (high grit).

Repstrap wet dream ;-)

> Building on a flexible belt would be nice for removal, but the bed has to be stiff to resist the warping. I can't see how it can be both.

Elevated bed mandril slightly, so belt is pulled down over it perhaps - always a taught surface? Obviously geometry needs to be played with to optimise friction.


Extra:

> we could cantilever out the Darwin Z axis so that it only needs 2 supports, much like they do in the Fab@home machine.

The first thing that breaks on the fab@home is the catelevered bed. At this level of fab we need to avoid cantilevered designs if poss, imo.

> Ian's use of 4 bars on the X axis is a nice design and suited well to head interchange.

Agree.
 
LOL! That's basically Tommelise 1.0. :-P
 
Elevated bed mandril slightly, so belt is pulled down over it perhaps - always a taught surface?

Imagine the force require to hold down a piece of ABS say 10mm thick to stop the corner curling up by say 0.5mm. Now imaging the tension on a belt required to prevent the middle moving 0.5mm when a force of that magnitude is applied upwards. The required tension would tend towards infinity.
 
2/3rds Tommelise, 2/3rds McWire, 2/3rds Hydraraptor, 1/3rd Dar, 1/3rd DaWitch ;-)
 
> Elevated bed mandril slightly, so belt is pulled down over it perhaps? >>> The required tension would tend towards infinity.

Good point. Wonder if metareprap's heating idea could be incorporated somehow... need conductive belt.
 
"Wonder if metareprap's heating idea could be incorporated somehow... need conductive belt."

Why? I'm just using a cartridge heater in a light, insulated box with a sheet of aluminum or copper on the top. Leads to run the cartridge heater are no big deal. :-)
 
Is there an issue with violently jerking the still molten plastic in the workpiece as we change the deposition direction?

I could imagine that causing some deformation that we should be working hard to avoid.
 
There is only one way to be sure...

(No, Aliens fans, it is not "take off and nuke the site from orbit.")

Vik :v)
 
HydraRaptor voilently jerks its table in both x and y directions and I haven't noticed any deformation due to it. The plastic seems too light and viscous to be affected.

It might become a problem at higher speeds though, or when depositing less viscous materials.
 
Rather than two Z motors I think we should link the screw drives across the top. I originally thought that a toothed belt would be the thing as we have at the moment, but that gives all the joining problems and expense etc.

We know we can do chunky gears, so lets just have a cross shaft with two right-angled pinions on the ends. A few degrees backlash in that would be no problem in the Z screw drives.
 
If you lightly spring load the gear train you won't have backlash. :-)
 
I'm wondering how the extruder interchange is going to work. With the head now only moving in one axis, we've lost the ability to manoeuvre it onto one of several positions to wipe the nozzle or pick up a new print head.

Vik :v)
 
> to wipe the nozzle or pick up a new print head.

We could add a nozzle wipe to the bed.

As for head exchange, we have 2 options I think:

1. Dock heads on the bed. For the industrial/conveyor option this would need a second rolling tool table. For the domestic rolling bed option, we could dock the heads on the side of the existing bed - perhaps up to 2 of them before we get inertial problems.

2. Dock the heads at the top of the z. I can imagine a cradle at the top and slightly in front of the axis. Vertical picking might require more thought, but it has the advantage of working for both bed types.

Admittedly though, this doesn't feel as natural as the existing cartesian offerings...

The question as to whether this is worth persuing (assuming we can tick all the boxes) is will the part count reduction be significant, if at all? I'll have a stab at some point.

Keep the concerns coming, it's geting stronger ;-)
 
Swapping out tools would be easy if you'd make the top support rotateable. You'd easily be able to put 6-10 tools which could be put into place by a rotating ring.

The ring would also give the top weight, which will make the rig more accurate.
 
> make the top support rotateable.

I kind of get where you're going with this but can you illustrate it/detail it further please?
 
I guess we could stack the heads vertically?

Vik :v)
 
All you have to do is make the x-axis wider than the xy printing surface by as much as it needs to accomodate however many extruders you want to use. They mount the lot of them side by side on the x-axis. Viola! No problem. :-)
 
On belt vs rigidity:
what about a tank tread like link belt? Easy to fab the pieces and the it could be made rigid in the needed directions (up and sideways) and flexible in the down direction. also a possible aid to detaching as only one link at a time releases , and that falls off rather sharply.

On the printhead changer:
mounting the changer up top seems like a good idea as it keeps the mechanism out of the way of the business of printing so the changer can take whatever form it needs to. Simple format: a bar with a series of docking connectors running along the x axis, when it's time to change the printhead carriage moves in front of an empty dock, the bar moves along the Y axis and grabs the printhead, which is released by the carriage, the bar retracts, the carriage moves in front of the next printhead and the process repeats in reverse. You can have 2 printheads or 50,000 because the entire active machinery of the changer sits up at the apex out of every bodies way.

On motor count:
I would imagine the Z axis could work much as it does now with a single motor driving 2 (instead of 4) linked shafts, given the mech advantage of screws I doubt we would need to up motor as Z movement is minor and so not speed critical. What would be the weight/balance tradeoffs of moving the X motor onto the printhead carriage?

One thing I like about this idea is it's inherent rigidity. Triangle GOOD!
 
Why even have the second support ? It's not like the extruder is that heavy. Just have one, and hold the triangle in place with more sturdy supports. 2 L joints should be quite capable of doing this.

If truly necessary, a second support could be achieved by having 2 smooth bars that slide into the Z axis machine* that are simply not powered, merely meant to stabilize it.

Basically what I'm suggesting is : cut the proposal in 2 along the X axis. Throw away one of the parts and don't replace it.
 
On belt vs rigidity:

I think a linked belt would suffer from wear and tolerance issues, however, it would be great to fire hot air up through (as a substitute for the hotbed).

On the printhead changer:

I think it would be best if the system was passive. ie. relying on existing mechanics to achieve motion. I'm sure we can do this using a vertical loading system (also we take advantage of the screw drive which will be really handy). I agree docking space on the top is ideal!

On motor count:

I can feel the motor debate coming on ;-)

> What would be the weight/balance tradeoffs of moving the X motor onto the printhead carriage?

Imo the clincher here is dead space i.e. keeping the x-motor out of the useful space (i.e. docked on the z).

> Why even have the second support ?

Needed. Cantilevers at this level of fab = big trouble.
 
Ed, you know how to raise and lower that x-axis. Remember the pulleys and cables that you used on just about the first cartesian robot you designed. Think about how a drafting table's parallel bar works, again. It worked fine for powering Godzilla from one side a few years ago. :-D
 
I was looking at using the top part of RepRap as a PCB CNC driller/router and noticed this possible design change and how it might complicate using RepRap for other projects. The current design has the X-axis, Y-axis, and print head all in the top few inches of the machine and that's all that would be needed for a mod to a PCB driller/router. Replace the print head with a high speed drill/cutter RepRap gets another use while using most of the same parts and possibly the software too. With a solenoid and hydraulic dampened spring, the z-axis is reduced to a binary operation.

So basically, a subset of the whole RepRap machine could be used to make the circuit boards for other RepRap machines. I would think that if this can be demonstrated or proved possible, redesigns should be considered for how the new design might limit the PCB making capabilities.

I also saw a forum post where someone was looking for the cheapest way to get a RepRap machine an again, I thought of building just the top portion and for the Z-axis, some kind of manual system for raising a base placed in the center. This might then allow the owner to make the remaining pieces to build the full RepRap structure. So the original material count for custom pieces goes way down and the owner of the RepRap Basic Top portion makes his/her own parts for the full version.

I could look into this further if it's something of interest to the RepRap folks. I'm doing an Arduino PID system right now and making my own PCB board is what got me here.

Doug
 
eD:
I don't know that wear would be a problem with a linked belt as I don't see it taking any huge loads or speeds. Even if there were some wear replacement tracks would be easy to fab and sub in over time. I like the idea of a ventilated belt with a hot air blower. First, we are already making and wasting lots of hot air; simply moving our power supply and motor drive boards and getting creative with some duct work we can turn that into a resource! Bucky Fuller would be proud! Second, that same creative duct work and a few valves makes it simple(er) to selectively heat sub-areas of the bed, thereby reducing the overall energy demand, at least for smaller prints. Third, waste heat is common throughout the third world so even if the heated dep surface needs more energy than our electronics supply it would not be mandatory to waste watts to generate heat, that could instead be scavenged from the environment. Obviously, the treads would have to be made of a nonthermoplastic but either an epoxy depositing head or a milling head could take care of that.
 
nophead wrote:
"Building on a flexible belt would be nice for removal, but the bed has to be stiff to resist the warping. I can't see how it can be both."

How about someting like a caterpillar tread? Stiff plates jointed together. It would likely mean a significant increase in weight, though.
 
Ah, I see the suggestion has already been made.
 
On the subject of link-belts, how about bicycle chain (and sprockets.) Although not reprappable, they're widely available, not expensive (unless you buy Campy....) and pretty easy to shorten or repair. And they can move much faster (and transmit much more power) than we need.

-- Larry
 
An infinitely long work surface

hmm. if the tool head could "feel" deposited plastic, you could make a mechanical model of a turing machine...
 
on the subject of link belts, how about aluminum security shutters?

http://www.alutechsecurity.com/security-products/rc37-roll-shutter.htm
http://www.alibaba.com/sellinglead-gs/202687681/Sell_SLLP_120_aluminum_profile/showimage.html

they also come in PVC and motorized versions, and presumably other plastics as well. what is the ideal bed material?
 
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]