Thursday, September 17, 2009


Mendel Apollo

While preparing the Mendel files, we’re all building prototype machines in the Bath Uni RepRap lab to get a mini factory up and running. Whilst the other guys are building three standard issue Mendels, I’m going to scratch an itch I’ve had for a long time: it’s that "how big can this thing go", itch ;-) My single motivation is to have a large enough build volume to manufacture all of the printed parts in one single shot. This, in my mind, makes for a much more independent self-manufacturing machine.

I laid up what a single shot build might look like and found the area for printing all the parts, with enough spare room for evolutions, to be 440mm square (marked with a blue line in pic below). That’s 5 times bigger than the standard Mendel build area.

I’m going for a build area of 350 mm x 550 mm. Whilst this achieves the same area as 440mm square, it yields a more manageable footprint out of the Mendel architecture and caters for the fact that most print jobs are rectangular, not square. I modelled the variant design (I’ve dubbed it Mendel Apollo after the 440 coincidence) and here’s a pic comparing the two volumes. Note: the Apollo model is fairly naked: electronics/brace plates need to be added.

So the Apollo’s a big machine, and with its size will come all sorts of wonderful problems. Structural for sure… how will the M8 components fair? No doubt reinforcements will ensue. Will NEMA 17’s pull the bed around OK? There will also be a huge test of reliability for the one-shot build, specifically for the software and extruder mechanics.

I think this will be interesting. I believe it will be a bit of a milestone if we ever get that single shot replication and in doing so it may establish some mechanical limits. But before I can scratch the itch, I need to keep going on getting the standard Mendel files finished… grrr

what if you were to figure a way to stack vertically instead of horizontally?

Or, perhaps you could modify a standard Mendel with a moving platform ( a "b" axis of sorts )
that would allow the build platform to index sideways.
Dave, intersting, but not sure how that would work with the bed movements- please ellaborate. One thing I'd love to try one day is a conveyor devlopment for the bed... then it could literaly spit parts off one by one.
Hey, do you guys always do your modeling with the Y axis vertical? I've been having to rotate almost every part I print for the Darwin; perhaps it would be faster to just re-wire my axis so Y, instead of Z, is vertical.

I like the idea of using a belt sander as the Y axis, and having parts spit off into a bucket.

Alternatively, you could try using bridge spanning to create mid-air rafts above existing parts, and continue to print more parts vertically.

Unfortunately, I have trouble keeping my extruder alive for more than about 9 parts at once. If the parts are built in parallel, that means I get 9 half built, usually wrecked parts. If you can build them serially, then you get n-1 good parts for a failure at the nth part. Head collisions are much more likely when you're building multiple parts serially on the same bed.

This comment has been removed by the author.
Just looking at the Mendel design I find myself wondering how stiff the structure will be. Two of the axes appear to be quite rigid (where triangles are formed). The remaining axis may get a bit sloppy at larger scale.
Do you think it may be necessary to add bracing?
The next question is "how do you brace it without reducing the available build volume?"
Awesome. I'm into building big machines too. I'm probably going to make my first post this weekend about a design I'm working on.

My design uses DC gear motors and some optical mouse sensors to record the angle. I'm going to print a strip on a CD label (118mm diameter) and then use the sensor to read this strip to determine the angle. The sensors I'm using can theoretically get 5000 cpi (counts per inch). $2 sensors can get 1600 cpi. This is about 14 - 17 bits or angle resolution depending on various things.

You could also use it for linear motion or 2-d linear motion (x and y travel). If you want to know more, I'll fill you in, or just look for a post this weekend.

If you need a larger build area make the bottom tray of the Mendel longer, and use a toothed gear strip along the bottom to feed it through the machine.

Could have a very long axis provided that the tray was supported on the exterior of the machine.
I have the same itch of setting the RAPrep off and coming back with latter with everything printed.

However I think the way forward has already been solved in the 2d printer world. A large print is cut up into A4 pages and collected at the end. Not printed on a single huge bit of paper.

After all smaller repraps are cheaper to build and more reliable.

So To summaries I believe we need

1) Compiling Software that batches up all the pieces into "A4 chunks" or whatever space is available. I am thinking of the way they batch up model airplane kits. All the parts are held together with just a tiny bit of plastic into a standard plastic square frame.

2) A network spool to store our "A4 pages" to be printed.

3) A mechanism to move the printed pages away from the system so it can print more. Possibly the square frame in 1) can be grabbed and z axis moved down. separating the bits from the z axis.

4) A storage mechanism to hold the "pages". again much like a printer.

I think the way to do this is to print the parts not in an outside square frame like the airplane kits mentions in 1) but an outside wire "cube". That way they would be easily stackable. with all the fiddly bits safely held in the cube for storage.

All you would need is a box to store them in for later retrieval.

I suppose the software used to compile the parts would group the parts into similar heights for optimisation sake

In this fashion the reprap would become a real factory working 24 hrs a day 7 days a week to keep me in the manner I wish to become accustom to.


Stephen George
reprap (@) stephengeorge . org

I was also wondering about the lack of y stiffness. Just after reading your post, I noticed that the electronics are on a board connected to the triangles. That board is adding to the y stiffness.

I don't know if it's enough. If it isn't, an extra board or sheet could be added on the other side to add more y stiffness. If that still isn't enough, a big upside down U shaped board with the cutout wide enough to let the table and model pass through could replace the existing rectangular board.
I was talking with someone on the forum about a similar proposal, but with the possibilty of making a bicycle frame.
On the Y, its not a problem, what will be a problem is the X axis, especially when the carriage is sitting in the middle, here its so critical, that you know, to maintain a constant level, that even a bow of just 1mm could mean destroying your extruder.

My suggestion, easy and cheap, you might want to think about getting box aluminium extrusion (or two), and have that take the load, running on bearings as well.

Keep us updated please Ed, be neat to watch a video of it in progress.
One way to really increase your build speed would be to hack the firmware so that you have two extruders printing in parallel( ie they move exactly the same, with two filaments being made about 15 cm from each other). Just make them print the duplicate parts and it should go faster.
@Wade: that must be a pain. I'll ask if there's a way we can orientate the STLs better in AOI. Which extruder do you use? Pinchwheel? Like bridgespanning idea, but may suffer from rigitity issues witha mobile bed.

@Marauder & @ Enrique: braceplates have been not been modelled yet but are essential for the design. As with standard version, they wil mount on the top and support the electronics too. This will provide all the rigidity it needs.

@Brian: Looking forward to it! GOod luck... go big or go home innit ;-)

@mfsamuel: Nice idea in theroy, but it's all about practical footprints in this game, our lab's not that big! Could be a good idea for large productions though.

@stephen: Large printers that print on large peices of paper do exist ;-) Disagree with 1 & 2: manual approach is extremely simple, and also necessary wrt build orientation, and only has to be done once. 3 defintely - disposable bed would be the most successful I reckon, unless you can reliaby part the parts with a doctor blade or something. 4: bin! Either way I share your desire to make replication as automatic as poss, so look forward to seeing your developments.

@Groygan: yes agree with extrusion profiles on this as second MOI will be brililant. Luckily for us we have a Vik who's on the case...!

@letsburn: yeah that's a pretty good trick in a factory siutation
Since everyone is making suggestions, I'll throw mine out there as well. Would it be possible to build vertically, above the first layer, using some kind of easily broken sprue columns as supports?

Basically, you would try to arrange the parts so that the top surface of the bottom part and the bottom surface of the top part are both not dimensionally vital. Then, since my understanding is that the Reprap can print overhangs up to ~45 degrees, you would create a 2d grid of hourglass shaped columns coming out of the top of the bottom pieces and serving as support material for the top pieces.

The idea would be that the center of the hourglass shape would be thin enough to be easily broken after the print job is done. Sure, it would add some volume to the job, and it might not be pretty, but it could let you print all the parts in one go within the print volume of the standard Mendal.

Also, as to speeding up print speed, I've long thought that a two print head system was the way to go. However, my thought was that one head should be much thinner than the present head (something like a high guage hypodermic needle with a steel hub) and the second head should be much larger.

The idea would be to print the first few vertical layers with the high detail head and then switch to printing only the first few concentric layers of the horizontal detail. Once those layers are as high as one layer of the large diameter head, the large diameter head would be used to fill the inner part of the model.

Work would proceed by alternating between the thin head printing the outer shell of the object up to the height of one layer of the thick head's output and the thick head doing the fill job. Of course, the two head sizes would have to be paired such that the output from the thick head was a multiple of the thin head's output and all other settings (acceleration, heating, etc.) would have to be tuned to ensure that it remained that way.

This should allow much higher surface detail on printed parts while, potentially, speeding up the process. It, also, might be a little more forgiving of slight misalignments between the two heads where two heads of the same size would have to be able to their threads on, exactly, the same spot every time.
@ Letsburn
I have been working more than 9 months with a club, we have got a 3-extruder head (with five-head capabilities for the future) nearing completion. It is not standard at all, and is very much a RepStrap. But we intend to make a proper automatic factory, complete with some kind of part-to-bin offloading.
As the parts look a little smaler than the parts for Darwin. Are the new parts for Mendel less likely to warp using ABS?
You commented on practical footprints, but wouldn't the footprint ultimately be determined by the build area?

Also, couldn't a compromise between Apollo and Mendel with a longer y axis save you from some of the major structural issues you may have to deal with?
This comment has been removed by the author.
Wow! You have no way of keeping the yz plane rigid save through the stiffness of the connectors. That's just not going to work.
@everyone who thinks there is a problem with sagging on Y axis.

My first assumption to get around that problem was imply to add more blocks between the plate underneath and the printing bed itself. I realize now that wont work and would make things worse.

Though the solution is still just as easy, just add more, maybe just two more Y rods with bearings. Those would have blocks underneath them to the bed without any screw holes needed, this I think would all that be needed to fix sagging on X and Y axes.
For X axis remedy see my previous post.

Have a great weekend
I might actually go one further and say, that with the additional blocks underneath the bed, that are sitting on rollers, I would design them such that a spring would be added to compensate for sagging of the Y rods.

If that doesn't workout then possibly having to use box extrusion on this axis as well would be needed to stop any sagging
OT re name Apollo:

I had to google to find out the association between Apollo with 440, as I didn't know.

This reminded me of another band, Apollo 100, apparently unrelated to 440. They were something of a 1-hit wonder that did a variation on a Bach Composition ("Joy"); this was popular on the US radio around 1972. I found it on youtube and gave it a listen for old-times sake.

-- Larry

First post here ever and probably nothing that will change the world. My thought was that some parts ar identical and prehaps can stack? If you print one part, then print a raft on top of that and then another part on top of that. I don't know about stability problems but prehaps this can be used for two parts before things get unstable? After the print is finished you just separate the parts at the raft. I imagine that not all parts will have to be all clean whitout working properly?

Really don't have a clue, just my spontainious thought...
This `Mendel Apollo' idea seems to me as a good idea.

It could be be a significant price reduction compared to the original design and a quicker spreading of the technology.
Hmm guys, I like the idea of having the bed as a conveyor type belt.

You cold print the objects in rows with enough inter row spacing to make sure that there are no head collisions.

Where the bed belt curves around an end roller it should rip the belt off of the bottom of the otherwise flat component.

As printing could be done by rows a failed print would only mangle a rows worth of objects rather than a whole beds worth.

I guess it would all go pear shaped though due to the belt not being stiff enough to resist warping.......

Thought for what they are worth.

You guys rock!
Keep it 440 guys!
Hello, First time.... In some industries they would use a 2X longer Y axis to facilitate part removal from one end while the other end is burning. The cycle time is sufficiently long enough to allow for that. I'm sure that your build time will require that checks are made on material usage and extruder function. The design, while very innovative, still requires attendance to keep things moving. Also, a simple switch could be set to halt production if the previous parts are not removed before a transition is required. Just a thought.
Hi Guys
I have just started building an oversized Mendel - I have just made a test assembly of the frame with 1m rods (12mm thick). It actually seems quite stable with the nuts just handtightened. I will cut it a bit down from it's current size, but will still be more than the souble of the standard Mendel, and with a build area that is bigger than the Apollo. Please let me know if anyone wants me to post info about the progress of my project.
Gert, I'd love to see progress on your Mendel Apollo. Do you have a blog or anything? Keep me posted!
I am impressed with the concept of being able to print and entire set of parts on one bed. A suggestion for increasing the rigidity of the Y access is to add webbing to the connectors on the front and back squares to keep the studding more rigidly at a 90 degree angle.

For the bed, I would suggest making it more rigid by making two layers and gluing a honeycomb in between. That makes for something very rigid and very light. Although that kind of undoes the print one full set in one pass, because you will want to print the honeycomb for the bed, too.

I have also been thinking about a single extruder with multiple nozzles that you can select what size filament to extrude. As mentioned above, the finest setting would make the outline for the next few layers. An intermediate nozzle would reinforce 2 or 3 fine layers, then a really fat, fast nozzle would fill the bulk of each piece. But I am not sure what changes need to be made in skienforge to make that work.

Just an update - My MegaMendel is coming to a state where I can start testing things. By now I'm only waiting for the pulleys and timing belts. It has been quite a process, but fun too. Some ideas have come out of it - Like adding more extruders to print more copies simultanously. That would of course reduce the working area of each extruder, but it would add to the production rate. I had some trouble with manufacturing/getting all the mechanic bits and pieces, so I got this idea of trying to create a mechanical parts kit to save others the trouble. I've also rethought the whole construction a bit. I think that by now my ideas should find a place of there own on the wiki as this is getting a bit out of the Apollo concept. I will post an update once I get things properly documented.
Wow Gert, that's wonderful. Do you have a blog or anything? I'd like to follow your progress/see some pictures!
Surely - if you want a rectangular area - the best axis to extend is the Y axis - there is scope for an almost infinite 'belt' in that direction - sure, you might have to lay it on top - and loose some zspace - but there is loads of spare Z space with these components.
The MegaMendel has now found it's home on the wiki:
It is incomplete at this moment, but I will strive to have it complete in a few days time.
Gert, I have reviewed your progress and to say the least, I am impressed with the MegaMendel. You have a great concept and it will be very useful for larger and larger builds. Are you finding that the NEMA 23's are all that necessary? I can see it used for the Y axis as you are moving a lot of mass, but I question the X & Z axis. The Z axis may be as simple as a gearing issue and unless you use multiple extruders, the X axis is linear and the extruder should be similar to a standard Mendel. Other than that, great job!!!! Ron
Ron, I know I've been a bit slow in getting the wiki page completed. You see - I had to take things a bit apart to make the photos for the wiki.
That occasion came with the rework of the Y-chassis, that work is almost complete. So I may have the page complete in a day or two.
The NEMA 23 is certainly a good precaution at the Y-axis in order not to lose steps with large builds.
And yes - I do have a plan about running more than one extruder in parallel, the NEMA 17 may have been able to handle that too.
As for the Z-axis - It already has quite a gearing ratio already due to the lead screws (712 steps/mm as opposed to the X an Y's 8 steps/mm).
But I do see a big advantage in using the same stepper set up in all drives - I only need to keep one type of stepper in stock in case I want to have a spare stepper ready :-)
Have a look at the wiki page in some two day's time, it should be complete by then - Actually "up to date" is more the term as I have further developments planned.
Thanks for the feedback - It's great to get an idea of what needs some more in-depth description.
Thanks for the feedback and it all makes perfect sense. I look forward to your posts with great interest... Carry On!! Ron
Hi all.
this is my first ever forum post so I apologise if I'm not using the proper etiquette/location/whatever else.

This question isn't specific to mendel apollo but rather the extruder design in general.
I'm only in my first year of a degree in mechanical engineering and am relatively new to 3d printing so correct me if I'm wrong.
Most (all?) polymers start as liquid monomers and are combined with a catalyst that solidifies them correct?
Has anyone considered using this approach - combining the two in a nozzle - instead of melting and extruding plastic rods?
I haven't yet researched the matter but I would have thought this would allow for faster print speeds and, if the curing time was long enough (I don't know much about that,) larger printed pieces at lower temperatures as the previous layer would stay soft for longer?

Again I don't know if this has been tried and failed or just wouldn't work for whatever reason but thought it worth mentioning.

P.S I love the whole Dawin/Mendel idea.

Post a Comment

<< Home

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

Subscribe to
Posts [Atom]