Having a second material possibility is excellent. Having used printers with support material, I consider this a must have feature.
Can you expand some on your theory of operation, why you chose printing in arcs, for example? And tell us more about what the Propeller is doing and how?
Regarding the "voxelated" surface. Is there any plans to support a "hybrid" printing mode? I.e. run one or two vector passes around the surface of the part, and use the parallel voxel printing for in-fill?
That is similar to one of the options that we are looking into and we even have some code to do this, but it needs some more testing and refining before we let it out into the wild.
Having a second material possibility is excellent. Having used printers with support material, I consider this a must have feature.
Can you expand some on your theory of operation, why you chose printing in arcs, for example? And tell us more about what the Propeller is doing and how?
We're also big fans of support material, which is why we're slating it for the production Lionhead release.
Let me provide a brief overview of the most common 3D printing technique for printers using filament as a material. Models are defined as a surface of vertices and triangles, and the models are sliced into layers. Each layer has two parts: an outline of the object's surface and the infill that makes up the interior of the object. This is a great technique that has a few issues from our perspective, but the most important is speed: Outlines are easiest to make with one printhead, and solid infills likewise take more time than a percentage infill.
By using arcs in parallel, we're able to print with multiple printheads simultaneously at 100% infill with no additional time required. This further allows us to slice our objects quickly since we have fewer calculations to perform, also leading to faster print times.
After we generate the arcs on the host PC, we send the data to the Propeller which controls all the actual hardware. Specifically, we have cogs for
1. Serial transmission,
2. Interpreting text commands from the user for testing/experimentation or binary PC data for actual printing,
3. Reading in button data for things like calibration, board identification, and safety (e.g., so motors don't try to move hardware too far during testing),
4. Reading in data from other boards—the Lionhead controllers each control eight motors and can be chained together. Our Bunnies will have one board (to control 4 base motors + 4 extruder motors), and the Lionheads two (to control the 4 additional extruders),
5. Stepping motors & updating motor state such as direction,
6. Reading in extruder temperatures and controlling heaters, and
7. For executing local or remote query commands such as asking how much space is available in the motor control queue. Each snippet is written in assembly, and using a dedicated cog lets us share the code easily between the main board and any chained boards.
The eighth cog is currently used for debugging. Of these, the interpreter and button data cogs are written in Spin and the remainder in PASM.
Large overhangs will still require supports or support material, which we plan to support in the production Lionhead as the second material.
I wasn't commenting on large overhangs but small. These are important when printing things like this Yoda: http://www.thingiverse.com/thing:10752 and the chin.
You should print that - it's a good test of overhang and resolution.
I'll be honest, not a huge fan of 100% fill. Still, you can always fill with concentric circles :-)
I wasn't commenting on large overhangs but small. These are important when printing things like this Yoda: http://www.thingiverse.com/thing:10752 and the chin.
You should print that - it's a good test of overhang and resolution.
Could you elaborate on why you're not a fan of 100% fill? Supporting alternate percentages is something we're considering. Our approach thus far has been to make to printing fast enough that lower fill percentages are less important unless used in something like a draft print mode.
Yoda is on our radar of things to print. We're currently focusing on other objects while working on general quality, but we'd love to see him printed, too!
I was off by a digit. Thought the voxel was .002 when it's actually .02" (approx)
So that is kind of a largish voxel. Not unworkable though. Do you have plans to reduce this? Frankly, if you could get it to .005-.010", I feel that's about the cross over from simple consumer model to a more serious model print capability.
Gave the software a try. Honestly, I was kind of put off for a bit. Now, I've run a ton of different CAD software over the years. The video game interface is unfamiliar. Rather than slog around making some basic shapes, I went ahead and exported a few solid models as STL to play with. The big difficulty is picturing scale and alignment. One can look at the numbers for scale, and knowing something about the source model. Alignment is a bit harder with perspective view, but not impossible.
Once I did that and created a few pattern tool instances in the workspace, I began to have a much better time. Flying around on keyboard, using shift key to go fast, holding left or right, tapping up and moving the mouse "orbits" around models just fine! Pixel clean up reminded me of cleaning old bitmap images in simple editors. Decided that square holes made great sense, and it was trivial to do some cleanup operations on the model.
What I like about this the most is a really little kid will get it. Voxels are known things, easy to think about, move, etc...
Ended up kind of impressed. The software is simple and somebody can do a lot, particularly given a few well thought out shapes added to the primitives! I think you've got something special here with this.
Now... if only I could back a printer. I really want one. Nicely done.
100% infill means that you will be using far more material.
Most of the print time on my printers is spent on the infill. If you could use just one nozzle to lay down the perimeter shells then do an infill with multiple nozzles using arcs it should greatly reduce print times while making a strong part with a smooth finish.
What I like about this the most is a really little kid will get it. Voxels are known things, easy to think about, move, etc... Ended up kind of impressed. The software is simple and somebody can do a lot, particularly given a few well thought out shapes added to the primitives! I think you've got something special here with this.
.
I think this is most important for the intended audience. I volunteer regularly in our high school Engineering Technology course. They're programming ActivityBots this Fall. In this class there's barely time to do anything, and once a student ties up the single 3D printer it's occupied for hours or days, and the process repeats itself if there's a botched print. We'd gladly trade out precision for completion in this class, and spread the experience across more students. Educational users are very different from people like us on these forums who now have experience with CNC, molding, and designs of all kinds.
I was off by a digit. Thought the voxel was .002 when it's actually .02" (approx)
So that is kind of a largish voxel. Not unworkable though. Do you have plans to reduce this? Frankly, if you could get it to .005-.010", I feel that's about the cross over from simple consumer model to a more serious model print capability.
Gave the software a try. Honestly, I was kind of put off for a bit. Now, I've run a ton of different CAD software over the years. The video game interface is unfamiliar. Rather than slog around making some basic shapes, I went ahead and exported a few solid models as STL to play with. The big difficulty is picturing scale and alignment. One can look at the numbers for scale, and knowing something about the source model. Alignment is a bit harder with perspective view, but not impossible.
Once I did that and created a few pattern tool instances in the workspace, I began to have a much better time. Flying around on keyboard, using shift key to go fast, holding left or right, tapping up and moving the mouse "orbits" around models just fine! Pixel clean up reminded me of cleaning old bitmap images in simple editors. Decided that square holes made great sense, and it was trivial to do some cleanup operations on the model.
What I like about this the most is a really little kid will get it. Voxels are known things, easy to think about, move, etc...
Ended up kind of impressed. The software is simple and somebody can do a lot, particularly given a few well thought out shapes added to the primitives! I think you've got something special here with this.
Now... if only I could back a printer. I really want one. Nicely done.
Thanks for the kind words!
We're currently experimenting with smaller voxel sizes (0.25mm), and so far the results have been promising.
For Li, starting out and scale are two issues we're also working on. We have the help screen (accessed by pressing F1; this isn't as visible as it should be) and some videos (with more planned) to help ease people into using the system.
For scale, we currently have two tools that help but don't quite solve the problem. One is the ruler tool that you should be able to access by pressing the corresponding number in the toolbar (it should be on zero by default; if you've moved tools into your inventory accessed by E, it might be there instead. Just click it and replace an tool on the toolbar to access it while modeling). Using it should tell you the distance between objects that are relatively close together—though I may have just found a bug making it more problematic for larger distances.
There's also zooming out & rotating around the object: Press & hold tab to do so.
100% infill means that you will be using far more material.
Most of the print time on my printers is spent on the infill. If you could use just one nozzle to lay down the perimeter shells then do an infill with multiple nozzles using arcs it should greatly reduce print times while making a strong part with a smooth finish.
We've seen long delays on infills with many printers, too, and there's a tension between getting a part quickly and getting one that has the desired mechanical properties. What infill percentage do you use most frequently? And is that from wanting to save material or time?
As Nathan mentioned, we're looking to the hybrid solution you described, but it's not quite ready yet.
I can chime in on the fills I have some experience with:
Usually, it's saving material. Solid fills just are not required for a lot of parts. On parts with extreme detail, it can be about saving time and or getting accurate prints. There can be a lot of head changing and stopping and starting of the support material required for partial fill parts that have lots of details. A full fill here may well make great sense, even if it does have a higher material cost.
Once I printed some fractal shape I found, and a partial fill would have been a 13 hour print. Full fill would have cut that by a third. I guess this depends on geometry. This example had many smaller shapes that touched to form a larger shape. A pathological example for sure, but I've seen some parts with detail exhibit smaller examples of this dynamic.
Sometimes it's all about mechanical properties. Solid fill, thin wall parts behave much differently than full fill ones do. They can be lighter and stiffer than solid fill ones. They may also better resist twisting forces too.
Now, this depends A LOT on part size, printer extruder output size, orientation and so forth, but the general idea is the interior structures can significantly change the mechanics of a part and this can often be favorable.
On the printer I have access to, we have a few coarse settings. Light, heavy, full fill. I often wish I could tweak this and get some better control.
Honestly, I think voxels have some advantage here. A full part can be detailed in the system. Then patterns could be subtracted from it on a voxel by voxel basis by simply starting at the bottom and stepping through the part! Doing this on solid models converted to STL can be complicated and it's error prone.
Seems to me, once the shape is in voxel land, applying fill could almost be something a person can do to the model rather than have it be a print option. Even if it is a print option, processing through the model layer by layer seems fairly easy like most graphics fill / pattern type operations are easy. Possibilities are here that are worth thinking on over time, IMHO.
So much depends upon what your proposed market is. Ken has laid out a defense of voxelated surfaces based upon the educational environment he's familiar with. If that's your primary market-segment target, then whatever disadvantages more commercial interests may see with voxelation are subsumed by the simplicity of voxel-based construction.
So go with it.
OTOH, if you want to attract a more commercial user base, smooth surfaces are definitely a prerequisite. Still though, the exclusive use of PLA versus ABS may already constitute a tilt in favor of the educational market.
If they can get to a .005" or so voxel, it's a very good split. It all affects scale. Bigger things work now, though the print volume isn't too big. A .007 or even .010 voxel would bring many more shapes into the volume with enough detail to be very useful, IMHO.
Honestly, the better commercial technology coming out isn't worth competing with. Well, unless you want to get bought out and that's fair and valid. I'm not knocking say, Makerbot. In any case, the major players are well beyond this. And they should be. There are Boeing parts being 3D printed now, others are mixing alloys in the same model, like sort of a gradient, or to simulate a part that has been hardened on one end...
A firm education target is an open field from what I can see. Toys are too, though I know of at least two products produced in Asia aimed at the toy segment. Kind of pricey, and very limited extruders right now.
The way I see it, the voxel models can be finished post printing easily enough. Trade fast, easy, cheap for quality, expensive, potentially slow and suddenly somebody who has time, but may not have a lot of money may well find printing models on voxels well worth it, simply because they can finish up the job where it makes sense.
Kids, and the educational market in general are surprisingly good at making it all work too. Personally, I think the .020" voxels are just a bit coarse. .010" ones would be much better, and .005 would be excellent!
In commercial land, it's always "will it do what I want?" and always pushing, "does it look good?"
Education really only needs, "will it do what I want?" with the rest just a matter of creativity, with a lot being optional. I spent some time today drawing up and importing some shapes I wanted to dork around with in STL.
I think squarish LEGO type blocks will work with voxels. Somebody will probably print a few of those. With everything quantized like it is, design ends up defaulting to some basics too. Younger people will pick up on that right away. Having the basic primitives in there is good. I think some other common shapes might be better. That's what I was toying with today. Found myself wanting a "scale snap" or default where I could build up a nice library of things that always fit together. If it were me, this would be in the next software iteration.
Hooks, joints, shells, pins, etc.. all sized just right so objects can be assembled quickly. Use the primitives for some more organic things...
Heh, I did model up a voxel for use in my CAD. Just stack those up. An STL conversion ends up near perfect! That was kind of fun. I also have a program I wrote that voxelized bitmaps in the CAD, creating unit blocks out of the pixels. Used it for some odd job that required converting bitmaps into something useful in solid modeling land. That thing could take shaded pictures and create "emboss" pieces or elevations for use on the printer. Put your name on the model, print out data for tactile visualization, etc... All kinds of educational type fun possible with these things.
Two other things:
One, it might be smart to incorporate the 3Dconnexion motion controllers. This would improve accesability. The other would be to potentially incorporate the haptic tech, which is also voxel based, that Sensible has had on the market for some time. Sensible produces a "digital clay" software that is insane fun, but was pretty expensive back in the day. If they are still around, perhaps that has all dropped in cost, or they are looking for new markets. That one allowed you to "carve" and "paint" on models in an extremely realistic way. The haptic gave the user feedback very similar to what one would feel using a pen to color a solid object, or when using a simple tool to carve on same.
What infill percentage do you use most frequently? And is that from wanting to save material or time?
I have used infill from 0% to 100%. Generally though it is between 10% and 50%. It depends upon the part. A lot of things I print are fairly large, a low infill saves time and material.
I can chime in on the fills I have some experience with:
Usually, it's saving material. Solid fills just are not required for a lot of parts. On parts with extreme detail, it can be about saving time and or getting accurate prints. There can be a lot of head changing and stopping and starting of the support material required for partial fill parts that have lots of details. A full fill here may well make great sense, even if it does have a higher material cost.
Once I printed some fractal shape I found, and a partial fill would have been a 13 hour print. Full fill would have cut that by a third. I guess this depends on geometry. This example had many smaller shapes that touched to form a larger shape. A pathological example for sure, but I've seen some parts with detail exhibit smaller examples of this dynamic.
Sometimes it's all about mechanical properties. Solid fill, thin wall parts behave much differently than full fill ones do. They can be lighter and stiffer than solid fill ones. They may also better resist twisting forces too.
Now, this depends A LOT on part size, printer extruder output size, orientation and so forth, but the general idea is the interior structures can significantly change the mechanics of a part and this can often be favorable.
On the printer I have access to, we have a few coarse settings. Light, heavy, full fill. I often wish I could tweak this and get some better control.
Honestly, I think voxels have some advantage here. A full part can be detailed in the system. Then patterns could be subtracted from it on a voxel by voxel basis by simply starting at the bottom and stepping through the part! Doing this on solid models converted to STL can be complicated and it's error prone.
Seems to me, once the shape is in voxel land, applying fill could almost be something a person can do to the model rather than have it be a print option. Even if it is a print option, processing through the model layer by layer seems fairly easy like most graphics fill / pattern type operations are easy. Possibilities are here that are worth thinking on over time, IMHO.
Thanks for the additional information on your prints. Your point about fills being something that a person can do to a model is exactly how it currently works in Li: If you want a 50% infill pattern, you can carve it out of the model. STL models are imported as solid objects by default, though, and though I can think of some ways to make infill tools, I'd imagine an automated system would save people's time if they don't want solid parts.
So much depends upon what your proposed market is. Ken has laid out a defense of voxelated surfaces based upon the educational environment he's familiar with. If that's your primary market-segment target, then whatever disadvantages more commercial interests may see with voxelation are subsumed by the simplicity of voxel-based construction.
So go with it.
OTOH, if you want to attract a more commercial user base, smooth surfaces are definitely a prerequisite. Still though, the exclusive use of PLA versus ABS may already constitute a tilt in favor of the educational market.
-Phil
Hi Phil—
Our target market is currently a mix of enthusiasts, education, and small businesses. There's a number of people who are interested in 3D printing who have the hardware skills to build or maintain a 3D printer but, once assembled, lack the CAD skills to design or modify parts. Instead, they're reliant on downloading parts from sites like Thingiverse. Others in small business have bought into the 3D printing hype and are disappointed that they can't get their printer to work correctly or model parts. And in education, we agree with everything Ken posted: Fast, reliable, and simple 3D printing has a special place in education in part due to class, program, and club structures.
If people are already comfortable with creating 3D models and are using their own 3D printer well already, we don't expect to convert them with our initial offering. We'll continue to refine our smooth surface code, but at the moment we're not competing with the highly competitive (and potentially saturated market) of 3D printers for highly skilled enthusiasts and professionals.
Automated system.. . Li Should get a Python interface. That is possible now, in that Python can write an STL easy enough. I would just write out lots of voxels expressed as STL cubes. I tried STL files containing multiple solid volumes and LI filled it perfectly. So normals and such are super easy.
People can author programs that output to LI easy that way. I think LI can and should do the reverse too. Output STL in the same way. Many systems will just deal with the many enclosed shapes. Some won't, but I don't think it's a big deal. Maybe simplify output some by culling interior voxel polygons. This would allow some interaction with external programs and code that people might write. Again, for education, I think this would prove useful.
If they can get to a .005" or so voxel, it's a very good split. It all affects scale. Bigger things work now, though the print volume isn't too big. A .007 or even .010 voxel would bring many more shapes into the volume with enough detail to be very useful, IMHO.
Honestly, the better commercial technology coming out isn't worth competing with. Well, unless you want to get bought out and that's fair and valid. I'm not knocking say, Makerbot. In any case, the major players are well beyond this. And they should be. There are Boeing parts being 3D printed now, others are mixing alloys in the same model, like sort of a gradient, or to simulate a part that has been hardened on one end...
A firm education target is an open field from what I can see. Toys are too, though I know of at least two products produced in Asia aimed at the toy segment. Kind of pricey, and very limited extruders right now.
The way I see it, the voxel models can be finished post printing easily enough. Trade fast, easy, cheap for quality, expensive, potentially slow and suddenly somebody who has time, but may not have a lot of money may well find printing models on voxels well worth it, simply because they can finish up the job where it makes sense.
Kids, and the educational market in general are surprisingly good at making it all work too. Personally, I think the .020" voxels are just a bit coarse. .010" ones would be much better, and .005 would be excellent!
In commercial land, it's always "will it do what I want?" and always pushing, "does it look good?"
Education really only needs, "will it do what I want?" with the rest just a matter of creativity, with a lot being optional. I spent some time today drawing up and importing some shapes I wanted to dork around with in STL.
I think squarish LEGO type blocks will work with voxels. Somebody will probably print a few of those. With everything quantized like it is, design ends up defaulting to some basics too. Younger people will pick up on that right away. Having the basic primitives in there is good. I think some other common shapes might be better. That's what I was toying with today. Found myself wanting a "scale snap" or default where I could build up a nice library of things that always fit together. If it were me, this would be in the next software iteration.
Hooks, joints, shells, pins, etc.. all sized just right so objects can be assembled quickly. Use the primitives for some more organic things...
Heh, I did model up a voxel for use in my CAD. Just stack those up. An STL conversion ends up near perfect! That was kind of fun. I also have a program I wrote that voxelized bitmaps in the CAD, creating unit blocks out of the pixels. Used it for some odd job that required converting bitmaps into something useful in solid modeling land. That thing could take shaded pictures and create "emboss" pieces or elevations for use on the printer. Put your name on the model, print out data for tactile visualization, etc... All kinds of educational type fun possible with these things.
Two other things:
One, it might be smart to incorporate the 3Dconnexion motion controllers. This would improve accesability. The other would be to potentially incorporate the haptic tech, which is also voxel based, that Sensible has had on the market for some time. Sensible produces a "digital clay" software that is insane fun, but was pretty expensive back in the day. If they are still around, perhaps that has all dropped in cost, or they are looking for new markets. That one allowed you to "carve" and "paint" on models in an extremely realistic way. The haptic gave the user feedback very similar to what one would feel using a pen to color a solid object, or when using a simple tool to carve on same.
These are great points about markets. Expect more about voxel sizes later today or tomorrow.
We haven't investigated 3Dconnexion or Sensible's digital clay, thanks for passing them along! We'll take a look into them.
I have used infill from 0% to 100%. Generally though it is between 10% and 50%. It depends upon the part. A lot of things I print are fairly large, a low infill saves time and material.
What I don't understand is that you have a round table that spins and you should would be able to do near perfect curves/splines etc
But you still go with pixels (eg voxels), not that I have any 3D experience so maybe I don't get it.
Automated system.. . Li Should get a Python interface. That is possible now, in that Python can write an STL easy enough. I would just write out lots of voxels expressed as STL cubes. I tried STL files containing multiple solid volumes and LI filled it perfectly. So normals and such are super easy.
People can author programs that output to LI easy that way. I think LI can and should do the reverse too. Output STL in the same way. Many systems will just deal with the many enclosed shapes. Some won't, but I don't think it's a big deal. Maybe simplify output some by culling interior voxel polygons. This would allow some interaction with external programs and code that people might write. Again, for education, I think this would prove useful.
Some scripting would be a great, and STL output is another feature we're considering. We'll also be publishing details on the .radiant format closer to the time we ship the printers.
What I don't understand is that you have a round table that spins and you should would be able to do near perfect curves/splines etc
But you still go with pixels (eg voxels), not that I have any 3D experience so maybe I don't get it.
Yep, we can get very nice curves and splines on the printer. However, we decided to go with voxels for the editor (and the resulting blocky look for prints, at least initially) because voxel-based modeling is usually easier to pick up than mesh-based modeling. One reason for this is that voxels are either present or absent, and users can draw analogies from using Legos or building blocks to create models. The first-person interface also lets people reason with their bodies in 3D space much like Logo's turtle did for 2D graphics. In contrast, meshes have complications such as degenerate triangles and intersecting faces that make using them harder to reason about without significant practice.
- I really appreciate KevinH's open attitude and responsiveness on the forum. This is a very good sign.
- The resolution and current lack of smoothing will doom the project if it isn't addressed as part of development. that crude looking pawn in the video turned me off big time
- Variable infill is important but is secondary to resolution and surface smoothing
- You may need to support mesh + voxel in the future or have a voxel surface smoothing mode or something
- don't forget windows 8.1 3D printer driver architecture (later on).
Nobody here is talking about the scanning. Built in scanning sounds really super cool. I'd love to hear more about it.
PS If you can fix the above issues and have a printer that doesn't require a league of voodoo priests to get it printing right, then this might be a big hit!
- I really appreciate KevinH's open attitude and responsiveness on the forum. This is a very good sign.
- The resolution and current lack of smoothing will doom the project if it isn't addressed as part of development. that crude looking pawn in the video turned me off big time
- Variable infill is important but is secondary to resolution and surface smoothing
- You may need to support mesh + voxel in the future or have a voxel surface smoothing mode or something
- don't forget windows 8.1 3D printer driver architecture (later on).
Nobody here is talking about the scanning. Built in scanning sounds really super cool. I'd love to hear more about it.
PS If you can fix the above issues and have a printer that doesn't require a league of voodoo priests to get it printing right, then this might be a big hit!
Thanks for your comments! We're focusing on resolution at the moment and gathering the infill information for future development.
For scanning, we're currently using silhouette scanning with two cameras mounted between the spool holders. Adding laser line scanning would let us scan a wider range of items more reliably without much of a increase in scanning or processing time. Voxels are again useful here as we they make reconstructing the 3D object simpler than using a point cloud and mesh.
As someone who is considering a 3D printer for making casting patterns for my hobby(metalcasting), the lack of smoothness is a real show stopper for me, since I'd have to finish the pawn or any object with a Dremel tool and hand sanding which sort of defeats the purpose of having a 3D printer.
If you can correct this issue, it could be a big hit with artists. Especially if you don't need to be a ring-zero software guru to operate the machine.
As someone who is considering a 3D printer for making casting patterns for my hobby(metalcasting), the lack of smoothness is a real show stopper for me, since I'd have to finish the pawn or any object with a Dremel tool and hand sanding which sort of defeats the purpose of having a 3D printer.
If you can correct this issue, it could be a big hit with artists. Especially if you don't need to be a ring-zero software guru to operate the machine.
Thanks for the feedback! Smoother prints are something we're investigating.
As I had mentioned previously, we had hoped to show some positive results for higher quality prints; unfortunately, we haven't yet been able to reach the quality level we've heard people wanting and that we were happy with. As such, we've elected to cancel the kickstarter campaign until we've developed something that our potential backers would be delighted to see. Thanks for the support & questions!
Wow, this is a little shocking. I just discovered it as I was about to contribute to the kickstarter. I'm sure you guys can figure out how to deal with the voxelated look, but it may require a more traditional software approach in addition to the way that Li works.
I still think your controller approach with 4-8 extruders is a big deal and that the scanner is awesome too. After working on it a bit, please consider returning to the kickstarter or other approach. It sure would be great to see the lionhead printer
Comments
Having a second material possibility is excellent. Having used printers with support material, I consider this a must have feature.
Can you expand some on your theory of operation, why you chose printing in arcs, for example? And tell us more about what the Propeller is doing and how?
Marty
That is similar to one of the options that we are looking into and we even have some code to do this, but it needs some more testing and refining before we let it out into the wild.
-Nathan
We're also big fans of support material, which is why we're slating it for the production Lionhead release.
Let me provide a brief overview of the most common 3D printing technique for printers using filament as a material. Models are defined as a surface of vertices and triangles, and the models are sliced into layers. Each layer has two parts: an outline of the object's surface and the infill that makes up the interior of the object. This is a great technique that has a few issues from our perspective, but the most important is speed: Outlines are easiest to make with one printhead, and solid infills likewise take more time than a percentage infill.
By using arcs in parallel, we're able to print with multiple printheads simultaneously at 100% infill with no additional time required. This further allows us to slice our objects quickly since we have fewer calculations to perform, also leading to faster print times.
After we generate the arcs on the host PC, we send the data to the Propeller which controls all the actual hardware. Specifically, we have cogs for
1. Serial transmission,
2. Interpreting text commands from the user for testing/experimentation or binary PC data for actual printing,
3. Reading in button data for things like calibration, board identification, and safety (e.g., so motors don't try to move hardware too far during testing),
4. Reading in data from other boards—the Lionhead controllers each control eight motors and can be chained together. Our Bunnies will have one board (to control 4 base motors + 4 extruder motors), and the Lionheads two (to control the 4 additional extruders),
5. Stepping motors & updating motor state such as direction,
6. Reading in extruder temperatures and controlling heaters, and
7. For executing local or remote query commands such as asking how much space is available in the motor control queue. Each snippet is written in assembly, and using a dedicated cog lets us share the code easily between the main board and any chained boards.
The eighth cog is currently used for debugging. Of these, the interpreter and button data cogs are written in Spin and the remainder in PASM.
I wasn't commenting on large overhangs but small. These are important when printing things like this Yoda: http://www.thingiverse.com/thing:10752 and the chin.
You should print that - it's a good test of overhang and resolution.
Could you elaborate on why you're not a fan of 100% fill? Supporting alternate percentages is something we're considering. Our approach thus far has been to make to printing fast enough that lower fill percentages are less important unless used in something like a draft print mode.
Yoda is on our radar of things to print. We're currently focusing on other objects while working on general quality, but we'd love to see him printed, too!
So that is kind of a largish voxel. Not unworkable though. Do you have plans to reduce this? Frankly, if you could get it to .005-.010", I feel that's about the cross over from simple consumer model to a more serious model print capability.
Gave the software a try. Honestly, I was kind of put off for a bit. Now, I've run a ton of different CAD software over the years. The video game interface is unfamiliar. Rather than slog around making some basic shapes, I went ahead and exported a few solid models as STL to play with. The big difficulty is picturing scale and alignment. One can look at the numbers for scale, and knowing something about the source model. Alignment is a bit harder with perspective view, but not impossible.
Once I did that and created a few pattern tool instances in the workspace, I began to have a much better time. Flying around on keyboard, using shift key to go fast, holding left or right, tapping up and moving the mouse "orbits" around models just fine! Pixel clean up reminded me of cleaning old bitmap images in simple editors. Decided that square holes made great sense, and it was trivial to do some cleanup operations on the model.
What I like about this the most is a really little kid will get it. Voxels are known things, easy to think about, move, etc...
Ended up kind of impressed. The software is simple and somebody can do a lot, particularly given a few well thought out shapes added to the primitives! I think you've got something special here with this.
Now... if only I could back a printer. I really want one. Nicely done.
Most of the print time on my printers is spent on the infill. If you could use just one nozzle to lay down the perimeter shells then do an infill with multiple nozzles using arcs it should greatly reduce print times while making a strong part with a smooth finish.
I think this is most important for the intended audience. I volunteer regularly in our high school Engineering Technology course. They're programming ActivityBots this Fall. In this class there's barely time to do anything, and once a student ties up the single 3D printer it's occupied for hours or days, and the process repeats itself if there's a botched print. We'd gladly trade out precision for completion in this class, and spread the experience across more students. Educational users are very different from people like us on these forums who now have experience with CNC, molding, and designs of all kinds.
Thanks for the kind words!
We're currently experimenting with smaller voxel sizes (0.25mm), and so far the results have been promising.
For Li, starting out and scale are two issues we're also working on. We have the help screen (accessed by pressing F1; this isn't as visible as it should be) and some videos (with more planned) to help ease people into using the system.
For scale, we currently have two tools that help but don't quite solve the problem. One is the ruler tool that you should be able to access by pressing the corresponding number in the toolbar (it should be on zero by default; if you've moved tools into your inventory accessed by E, it might be there instead. Just click it and replace an tool on the toolbar to access it while modeling). Using it should tell you the distance between objects that are relatively close together—though I may have just found a bug making it more problematic for larger distances.
There's also zooming out & rotating around the object: Press & hold tab to do so.
We've seen long delays on infills with many printers, too, and there's a tension between getting a part quickly and getting one that has the desired mechanical properties. What infill percentage do you use most frequently? And is that from wanting to save material or time?
As Nathan mentioned, we're looking to the hybrid solution you described, but it's not quite ready yet.
Usually, it's saving material. Solid fills just are not required for a lot of parts. On parts with extreme detail, it can be about saving time and or getting accurate prints. There can be a lot of head changing and stopping and starting of the support material required for partial fill parts that have lots of details. A full fill here may well make great sense, even if it does have a higher material cost.
Once I printed some fractal shape I found, and a partial fill would have been a 13 hour print. Full fill would have cut that by a third. I guess this depends on geometry. This example had many smaller shapes that touched to form a larger shape. A pathological example for sure, but I've seen some parts with detail exhibit smaller examples of this dynamic.
Sometimes it's all about mechanical properties. Solid fill, thin wall parts behave much differently than full fill ones do. They can be lighter and stiffer than solid fill ones. They may also better resist twisting forces too.
Now, this depends A LOT on part size, printer extruder output size, orientation and so forth, but the general idea is the interior structures can significantly change the mechanics of a part and this can often be favorable.
On the printer I have access to, we have a few coarse settings. Light, heavy, full fill. I often wish I could tweak this and get some better control.
Honestly, I think voxels have some advantage here. A full part can be detailed in the system. Then patterns could be subtracted from it on a voxel by voxel basis by simply starting at the bottom and stepping through the part! Doing this on solid models converted to STL can be complicated and it's error prone.
Seems to me, once the shape is in voxel land, applying fill could almost be something a person can do to the model rather than have it be a print option. Even if it is a print option, processing through the model layer by layer seems fairly easy like most graphics fill / pattern type operations are easy. Possibilities are here that are worth thinking on over time, IMHO.
Kevin,
So much depends upon what your proposed market is. Ken has laid out a defense of voxelated surfaces based upon the educational environment he's familiar with. If that's your primary market-segment target, then whatever disadvantages more commercial interests may see with voxelation are subsumed by the simplicity of voxel-based construction.
So go with it.
OTOH, if you want to attract a more commercial user base, smooth surfaces are definitely a prerequisite. Still though, the exclusive use of PLA versus ABS may already constitute a tilt in favor of the educational market.
-Phil
Honestly, the better commercial technology coming out isn't worth competing with. Well, unless you want to get bought out and that's fair and valid. I'm not knocking say, Makerbot. In any case, the major players are well beyond this. And they should be. There are Boeing parts being 3D printed now, others are mixing alloys in the same model, like sort of a gradient, or to simulate a part that has been hardened on one end...
A firm education target is an open field from what I can see. Toys are too, though I know of at least two products produced in Asia aimed at the toy segment. Kind of pricey, and very limited extruders right now.
The way I see it, the voxel models can be finished post printing easily enough. Trade fast, easy, cheap for quality, expensive, potentially slow and suddenly somebody who has time, but may not have a lot of money may well find printing models on voxels well worth it, simply because they can finish up the job where it makes sense.
Kids, and the educational market in general are surprisingly good at making it all work too. Personally, I think the .020" voxels are just a bit coarse. .010" ones would be much better, and .005 would be excellent!
In commercial land, it's always "will it do what I want?" and always pushing, "does it look good?"
Education really only needs, "will it do what I want?" with the rest just a matter of creativity, with a lot being optional. I spent some time today drawing up and importing some shapes I wanted to dork around with in STL.
I think squarish LEGO type blocks will work with voxels.
Hooks, joints, shells, pins, etc.. all sized just right so objects can be assembled quickly. Use the primitives for some more organic things...
Heh, I did model up a voxel for use in my CAD. Just stack those up. An STL conversion ends up near perfect! That was kind of fun. I also have a program I wrote that voxelized bitmaps in the CAD, creating unit blocks out of the pixels. Used it for some odd job that required converting bitmaps into something useful in solid modeling land. That thing could take shaded pictures and create "emboss" pieces or elevations for use on the printer. Put your name on the model, print out data for tactile visualization, etc... All kinds of educational type fun possible with these things.
Two other things:
One, it might be smart to incorporate the 3Dconnexion motion controllers. This would improve accesability. The other would be to potentially incorporate the haptic tech, which is also voxel based, that Sensible has had on the market for some time. Sensible produces a "digital clay" software that is insane fun, but was pretty expensive back in the day. If they are still around, perhaps that has all dropped in cost, or they are looking for new markets. That one allowed you to "carve" and "paint" on models in an extremely realistic way. The haptic gave the user feedback very similar to what one would feel using a pen to color a solid object, or when using a simple tool to carve on same.
I have used infill from 0% to 100%. Generally though it is between 10% and 50%. It depends upon the part. A lot of things I print are fairly large, a low infill saves time and material.
Thanks for the additional information on your prints. Your point about fills being something that a person can do to a model is exactly how it currently works in Li: If you want a 50% infill pattern, you can carve it out of the model. STL models are imported as solid objects by default, though, and though I can think of some ways to make infill tools, I'd imagine an automated system would save people's time if they don't want solid parts.
Hi Phil—
Our target market is currently a mix of enthusiasts, education, and small businesses. There's a number of people who are interested in 3D printing who have the hardware skills to build or maintain a 3D printer but, once assembled, lack the CAD skills to design or modify parts. Instead, they're reliant on downloading parts from sites like Thingiverse. Others in small business have bought into the 3D printing hype and are disappointed that they can't get their printer to work correctly or model parts. And in education, we agree with everything Ken posted: Fast, reliable, and simple 3D printing has a special place in education in part due to class, program, and club structures.
If people are already comfortable with creating 3D models and are using their own 3D printer well already, we don't expect to convert them with our initial offering. We'll continue to refine our smooth surface code, but at the moment we're not competing with the highly competitive (and potentially saturated market) of 3D printers for highly skilled enthusiasts and professionals.
—Kevin
People can author programs that output to LI easy that way. I think LI can and should do the reverse too. Output STL in the same way. Many systems will just deal with the many enclosed shapes. Some won't, but I don't think it's a big deal. Maybe simplify output some by culling interior voxel polygons. This would allow some interaction with external programs and code that people might write. Again, for education, I think this would prove useful.
These are great points about markets. Expect more about voxel sizes later today or tomorrow.
We haven't investigated 3Dconnexion or Sensible's digital clay, thanks for passing them along! We'll take a look into them.
Thanks for the information!
But you still go with pixels (eg voxels), not that I have any 3D experience so maybe I don't get it.
Some scripting would be a great, and STL output is another feature we're considering. We'll also be publishing details on the .radiant format closer to the time we ship the printers.
Yep, we can get very nice curves and splines on the printer. However, we decided to go with voxels for the editor (and the resulting blocky look for prints, at least initially) because voxel-based modeling is usually easier to pick up than mesh-based modeling. One reason for this is that voxels are either present or absent, and users can draw analogies from using Legos or building blocks to create models. The first-person interface also lets people reason with their bodies in 3D space much like Logo's turtle did for 2D graphics. In contrast, meshes have complications such as degenerate triangles and intersecting faces that make using them harder to reason about without significant practice.
- I really appreciate KevinH's open attitude and responsiveness on the forum. This is a very good sign.
- The resolution and current lack of smoothing will doom the project if it isn't addressed as part of development. that crude looking pawn in the video turned me off big time
- Variable infill is important but is secondary to resolution and surface smoothing
- You may need to support mesh + voxel in the future or have a voxel surface smoothing mode or something
- don't forget windows 8.1 3D printer driver architecture (later on).
Nobody here is talking about the scanning. Built in scanning sounds really super cool. I'd love to hear more about it.
PS If you can fix the above issues and have a printer that doesn't require a league of voodoo priests to get it printing right, then this might be a big hit!
Thanks for your comments! We're focusing on resolution at the moment and gathering the infill information for future development.
For scanning, we're currently using silhouette scanning with two cameras mounted between the spool holders. Adding laser line scanning would let us scan a wider range of items more reliably without much of a increase in scanning or processing time. Voxels are again useful here as we they make reconstructing the 3D object simpler than using a point cloud and mesh.
If you can correct this issue, it could be a big hit with artists. Especially if you don't need to be a ring-zero software guru to operate the machine.
Thanks for the feedback! Smoother prints are something we're investigating.
I still think your controller approach with 4-8 extruders is a big deal and that the scanner is awesome too. After working on it a bit, please consider returning to the kickstarter or other approach. It sure would be great to see the lionhead printer