by makeme

Archive for November, 2011|Monthly archive page

Professionals Do It On Powder Beds

In Uncategorized on 23, Nov, 2011 at 12:00

There are a lot of approaches to additive manufacturing. Fundamentally, however, they all break down into a few categories:

  • points
  • lines
  • planes

Planes means using sheets of material. This approach is usually more of a combination of additive and subtractive, since the excess sheet has to be trimmed away. Lines means using long thin noodles (floppy or rigid) to build up the part. Points means using powder, either laid down in a bed or shot out of a nozzle.

I’ve been trying to come up with a way to take Fused Filament Fabrication (FFF, AKA: the copyright free version of FDM) to the next level by removing the need for a support structure. If you can free up the extruder nozzle to move in more dimensions relative to the build surface you can print “overhangs” right onto the part instead of printing them onto thin air.

3d print FFF without overhangs

This approach would definitely allow an extruder-style printer to print nearly-arbitrary shapes faster and with less waste. However, it would not allow printing of truly arbitrary shapes. The easiest example of this limitation is the spiral. It doesn’t matter where you start from, an extruder-style printer is going to have to use support material to print a spiral.

cant print a spiral

Ultimately, if you want to print arbitrary shapes, you’re going to have to hold everything in place until you can get the entire print finished so that every little bit is attached to every other little bit. Either you use support material, or there are some things you just don’t print. That being the case, might as well embrace support material, and nothing is supportier than a powder bed.

CandyFab is a good open-source example. Here’s a great overview of the process by How It’s Made (my favorite show evaaar!)

There is a sort of natural difference here between hobby 3D printers and professional 3D printers. Powder beds allow for truly arbitrary shapes, but they require a lot more of the printer and the environment the printer is in. I think this means that hobby 3D printers will be limited to nearly-arbitrary shapes. Maybe in a decade there will be pro-sumer printers next to the drill presses at home improvement stores that will use powder beds. It’s at least a possibility.

The result of my investigation, and the point of this post, is that I don’t think the open-source hardware movement is going to drive the development of powder bed printers. I think those are going to be left for the professionals.

Time and Resolution

In Uncategorized on 22, Nov, 2011 at 19:36

I was discussing ways of making better prints with a guy in my office. What follows is the result of that discussion.

A big problem for 3D printing is that you have to trade resolution and time. If you want a cleaner and more accurate part you have to wait longer for it. This is because the printers can’t cover a large area at once. Whatever method is used for the smallest areas has to be used for the largest ones. Maybe you can change tools, but that’s got issues.

Wouldn’t it be nice to be able to form an entire layer at once? Yes. Yes it would. Here is a theoretical method for doing that.

Start with a build surface. Fill that build surface with a grid of tiny holes and in between those holes put tiny electromagnets. Next, flip that surface upside down so that the electromagnets are on the bottom. Now when you inject ferrofluid through the holes you can move it around on the underside of the build surface with the electromagnets. By turning the electromagnets on in a controlled order you can arrange the ferrofluid so that it outlines the exact shape of your first layer.

magnetic fluid 3d printing

Now put that into a tank so that the bottom of the ferrofluid, which is hanging off of the build surface, is just touching the bottom of the tank.

magnetic fluid 3d printing 2

Inject the printing substance through the build surface into the cavities formed by the ferrofluid. Even if the substance is liquid, it will be constrained by the floor of the tank and the ferrofluid.

magnetic fluid 3d printing 3

Harden the printing substance in some way. Heat, UV light, catalyst…kind words. Whatever works. Then raise the build surface by one layer height. Fill the tank with enough of a support substance to just reach the depth of one layer height. This support substance needs to be denser than the printing substance.

magnetic fluid 3d printing 4

Inject more printing substance through the build surface surface to fill the cavity formed by the ferrofluid and the previously hardened printing substance. Any overhangs will rest on the support substance.

magnetic fluid 3d printing 5

Repeat this procedure for each layer, rearranging the ferrofluid when necessary.

magnetic fluid 3d printing 6

This process, or something similar, could open up a paradigm in which you don’t have to trade time and resolution. Each area of detail can be resolved at the same time by just controlling all the relevant electromagnets, then the open space can simply be filled with whatever it is you’re using to print. It doesn’t seem like the control electronics would be all that complicated, either. Basically you’re just drawing on an LCD readout. The complicated part of this idea is the various substances. It’s more of a chemistry problem than an electrical problem.

How Low Can You Go

In Uncategorized on 20, Nov, 2011 at 14:34

When it comes to 3D printing, the most expensive part of the system is the electronics.

Makerbot wants $370 for their Gen4 electronics. With their Gen6 stepper extruder (and the driver for it) costing $165, and a set of X-Y-Z motors costing $105, that puts the complete cost of electronics at around $640. I figure this is a good upper bound on what 3D printer electronics should cost since Makerbot’s electronics are probably the most professional and full-featured. I’m not going to include a heated bed in this comparison because it’s not strictly necessary to get started, it’s just a performance upgrade.

Obviously, when there are complete 3D printer kits starting at $500, $640 for just the electronics is unacceptable for my purposes.

The confounding thing is that when you move away from Makerbot (and complete kits in general) you start to have to source from multiple vendors. There is not, as yet, a clearing house for open-source 3D printer components. Sellers tend to focus on one or two options. Additionally, they tend to be located in Europe, so that whole “You want HOW MUCH for shipping?!?” thing gets reversed.

RAMPS and Gen6 are mid-range in terms of performance flexibility and cost. So, you know, whatever. All the electronics kits I know of use pretty much (if not exactly) the same stepper drivers, and they all use USB, and these days they all have an SD-card option, so unless one of them tends to spit out errors more often they all have the same 3D printing performance. 

Sanguinololu seems to be the strongest attempt to whittle the electronics down to just the bare necessities. eMakerShop sells the whole thing (including drivers & firmware tested) for about $170 shipped, they want $95 for the motors (shipped) and $75 for the extruder (shipped). That all comes to around $340. Solidoodle sells the whole thing (without drivers) for $105 (shipped), but doesn’t sell anything else. LulzBot can make up the difference with four Pololus for $65, four motors for $75, and a 12V 25A 300W power supply for $35. They have a hot end that they want $75 for. That all comes to around $350.

There are rumors of people pushing the electronics cost even lower. For example, something called GenL offloads most of the computation to the host computer by using more USB bandwidth. Another example is Repic by Mark Feldman, but I can’t find much information about it.

As you can see, the electronics is the most expensive part of the printer and the stepper motor/driver combination is the most expensive part of the electronics. The need for four bi-polar steppers and four microstepping drivers demands $120 minimum. You might be able to get that down under $100 if you get really lucky on sales or start salvaging parts. The biggest barrier is that these particular parts can’t be made much cheaper. The Sanguinololu board can be brought down to around $60 if you buy the bare PCB, then the components, then burn the bootloader with something you already had (or maybe you can find a chip that’s already burned). But the motors and driver prices aren’t going anywhere.

Some blue-sky ideas for lowering the cost even further involve basically starting from scratch and creating a new family of electronics. The primary reason steppers are so popular is that they don’t require any feedback for accurate positioning. It’s possible that coupling a feedback mechanism (linear resistor, optical encoder, whatever) with a standard DC motor to create a servo would be cheaper. It would also be possible to simulate the entire electronics board on an FPGA for a one-chip solution; just a PCB with the FPGA, its interface, and a bunch of transistors for amplification. Maybe the motors, and their complicated drivers, could be replaced with solenoids and some clockwork. Running all the high-power functions off of AC (out of the wall) might eliminate the need for a power supply (get logic power from the USB).