Monday, July 16, 2007


Mendel Preview

We are all working like crazy on RepRap 1.0 "Darwin" of course. But I have an MSc student, Arman Ghandizadehdezfuli, who is looking at Mendel for the future. Here's a video of the prototype extruder changer that he's designed being tested in Darwin. (The slight jerkiness is all from the cheapo webcam software; the actual movements were clean and smooth.)

The idea is to have a row of extruders along one (or more) sides of the machine, and have it pick each up and use it in turn so that many different materials can be incorporated in a single built object.

I had originally thought that each extruder would have its own RepRap controller attached to it. Docking and undocking an extruder would momentarily break, and then re-route, the token ring that allows all the controllers to communicate. But now I think that that would be unreliable. Better to have one controller on the moving carriage, and to connect the drives for the motors and heaters from that to each extruder when it is picked up. The extruder machinery doesn't need clean data during the moment of connection, and so this would be far more robust. Any extruder that needs a controller when it is parked (to keep it at temperature, for example) can have its own controller attached to its docking station that makes contacts in the same way.

Another change will be that the X, Y, and Z flags for the opto endstops will be on each extruder (rather than having just one set on the carriage), so the machine can always zero with precision and get the registration between materials exactly right. The picker-upper grip on the extruder in the carriage is firm, and it doesn't move. But this zeroing scheme would also mean that the grip wouldn't have to grab an extruder in exactly the same place each time.

Labels: ,

Very nice!
Looks like a good start! Now if you widen the footprint of Darwin to accommodate that rather deep connector so that you don't take a big hunk out of your xy working area you will be good to go. :-)
Yup - we can probably make the thing more compact. But it's easy to extend the axes - we just cut the rods longer...
If it works don't fiddle with it, at least for now. As you say making the rods a little bit longer is a hell of a lot easier. :-D
A good improvement.

There are two dominant issues to contemplate:

Need for multiple materials in layup process, and simplicity of design.

If the goal is replication under any circumstance, the design must be loose and free from limited-supply objects, such as your eol switches and counters.

We have an entire set of "food service" objects operating with two specific designs, which are directly interrelated.

The objects are made of most economic material, flour paste and white cement, white glue and hot glue, etc.

Each has less than 3 total control mechanisms, one of which is at most a binary toggle.

We all know you can control infinitely many systems with a transversal actuator, but the selection of optimal design must be the most simple that works.

All items have an irda receiver set to 8bit wide parallel out. Batteries or power based on what it controls.

One double h bridge and a binary relay, possibly more if desired.

The control unit, costing 3-5usd, handles any number of mechanical drives with its two motor/actuator system and a set of actuators for control mechanism selection.

Irda speeds exceed 20ksps full frame so control loops are handled by the external controller, not any on-board circuits. You go put an hbridge on a parallel port for 35 cents ;)

replication is a lowest power requirement, notice that the friction from a gear rod is higher than a gravity mounted cog?

cogs can be any size and are quite easy to mass produce.

threadrod has vulnerability issues.

head transversal, especially for small scale targets, is finite

mount your heads to a transversal system that precludes limitaions, and mount everything for control (the 3 motors, heater) on board.

Now the simplest part: find a ball in a picture.

notice said ball and 1usd webcam have 3d resolutions far higher than your optos. Now smack your device with a stick and see if it is still aligned.

Typing on a fone again, so will cut it short.

A pic controller that can find balls in an img frame, generate vector control data, and spit out a data feed (hint irda blinkie) with more than enough memory and processing capacity to handle a 2000 control actuator system is around 3usd.

multiple blinkies are no different.

feel free to msg me or aim/msn/yp/ etc "wilfredguerin"

Im busy burning flour into a 3m wide control arm today.

ok.. i have to say it...:

i've used lawn tractors to position pivoting I-Beams in a huge warehouse with first-level accuracy of 3d location for control head within a quarter inch, usually closer to mm resolution.

You've used the antique caliper device that helps you to scale a drawing.

We speculated that a football field sized system could etch silicon (without magnifying lenses) at a resolution comperable to high end fabs using nothing more than scrap metal and a handfull of laser pointers.

oh, and at 10 cent each retail, said laser pointers also cut metal in less cost than your opto counters ;)

Dont forget to use your balls!

I'm concerned that there my one day exist a design that allows the head to become glued to the working surface, just before tracking in the direction that disconnects it.

I think I'd be more comfortable if the carriage either actually latched the extruder electromechanically, or the extruder cradle unlatched the carriage when the head was being seated/unseated.

The latter would be easy enough. One spring loaded catch on the cradle, that was interrupted when it was within half an inch of seated in the cradle, catching a spur on the print head.

I like the idea of attaching a webcam for alignment. That'd have an additional advantage in you could use it to continue work on a previously interrupted project. (If it had a camera, and remembered where it left off, it could find the object where you arbitrarily laid it on the working surface, and realign the electronic model to correspond with the partially completed physical one.)

Having a single controller would certainly be cheaper. As for having to do so to avoid tangling cables, however, if you used an overhead array of telephone handset cables, I don't think they'd become tangled on their own. Each one would only be moving around under the control of the single carriage, so no two would be twisted relative to each other before being returned to the storage cradles.

If you want to get really simple, you might be able to use a single controller to maintain all the non-active heads. Have one of the stepper controls outputting a binary address, and translate that into which common conductor wire to, well, connect, leaving the remaining heads off by having the common conductor unconnected. This would allow you to cycle between the parked heads for maintenance.

Since the steppers have four coils, I'm assuming the universal controller could be instructed to output any of 16 addresses.

Would any theoretical head need more than a moments attention, every few seconds, to check status, and apply corrections? Could a heater keep the material soft enough to extrude, if it were only on one second out of sixteen? If not, could a capacitor across the coil, (electronically, actually mounted on the cradle,) allow the heater to provide consistant heat?
Thanks for the exciting preview!
"I think I'd be more comfortable if the carriage either actually latched the extruder electromechanically, or the extruder cradle unlatched the carriage when the head was being seated/unseated."

Hey Sean! If you really feel that way get yourself a Darwin or some other kind of RepRap going and design one like that! :-)

That's what the RepRap revolution is all about, really. It's about creating what amounts to a Cambrian explosion of design creativity! :-D
gah. academics.
Good stuff...

I don't know the complete history behind what all you've accomplished so please do delete this comment if it doesn't make any sense or if it is out of place...

I'm just curious.... Have you ever thought about using a low cost cnc machine to work with. Sherline has some real cheap ones that actually come with computers and some linux based software. Here is the URL I broke it up so that this comment would not be considered as spam...


*I was think that a CNC machine will allow you traverse all required planes....and that the spindle speed could be used to control fluid flow during the fabrication process...

Also...I run a site called we have some concepts that you might find interesting. We created some very unique interlocking technolgies that are well suited to robotics. Basically we have some scalable cubic blocks that can be made from anything...they simply interlock on placement...real easy for robots to work with....
Post a Comment

<< Home

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

Subscribe to
Posts [Atom]