New mouthpiece for our Artist Series Irish Whistle

In part IV we will look at preparing the 3D printing of our Artist Series mouthpiece. The advantages of 3D printing are clear: we can choose between different materials, different colors and finishing methods make it easy to create a prototype or (more or less) finished product. In addition to this, it is becoming cheaper and cheaper to print in 3D.

3D printing is done with a filament; a round ‘plastic’ wire of 2.85 or 1.75 mm thin is fed through a hot-end. This hot-end melts the filament at a temperature of 190-270 C (depending on the type of filament). Meanwhile, the printhead moves and deposits a pattern of melted filament, layer by layer according to x, y and z values, in a three-dimensional coordinate system, as we all do know from our mathematics class (or do we?). The printer quickly reads these values ​​from a text file (the g-code) and controls the print head and other parts and settings.

Disadvantages of 3D printing are also evident: it takes a long time before you have a print. Layer resolution can be as fine as 0.1 mm, although there are already printers that can print much finer. Filaments are available in many types, ranging from ABS and Nylon to PLA. PLA (polylactic acid) is a non-toxic biodegradable 3D printing material which is still the most widely used. Downside is that prints can be somewhat brittle, and the fact that their deformation temperature of 50C is quite low.

Part I – Introduction | Part II – Het Design | Part III – Simulations | Part IV – 3D Printing

Written by Ruud Roelofsen

Converting a solid to 3D

Our model that we have made with our drawing program is a so-called solid as you have learnt in previous parts of this series. However, I will briefly explain solid modelling again.


Known packages for doing 3D modelling are AutoCAD, Autodesk Fusion360 and SolidWorks. There are countless packages that are free, such as GrabCAD.

Solid modelling is the most advanced method to create a geometric model in three dimensions. It is a representation of solid parts (solids) that you create using a computer. A typical geometric model consists of a series of wire models that are made up of vectors. A vector has a direction, a length and a path. A mathematical representation is converted into a line. An object therefore consists of one or more different lines, which together form a wire model. These wire models can be two-dimensional, but also two-and-a-half-dimensional or three-dimensional. If we add surface properties to these wire models in the multi-dimensional space, this results in a so-called solid. Forms, shapes and surfaces can be extruded or cut unto other solids as we have seen in part I and II. This is called solid-modelling.


Solid modelling is a well-known application of CAD (Computer Aided Design) and has been on the rise in recent years. The solid modelling CAD software helps the designer to see what he has designed as if it were a produced object. It can be viewed from different angles and in various states (transparency settings, cross-sections, colours, material properties, etc.). It helps the designer to check whether it is going to be as he really meant. Some software also help create joints, holes, threading, molds, welding, folding, CNC milling etc.

So we have seen that a solid consists of vectors, composed of lines in a plane in space. Its representation is a set of mathematical values. You can scale a vector infinitely, which is a big advantage over an object with a fixed dimension.

You can compare the difference between a vector and a fixed ‘object’ with the difference between a vector image and a JPEG image. A vector model is scalable without losing the resolution; after all, the file consists of vectors, and whether they are displayed very small or very large, makes no difference. A JEPG, on the other hand, uses compression. The advantage is that the file size remains small, but JPEG uses static information; if you increase a JEPG, you will eventually see ‘blocks’. The JPEG compression always tries to keep the impression of the original as good as possible.

Left is the original (of course this is not a vector). On the right is the compressed JPEG.

Export from a solid to a 3D file

Some software packages allow to directly print to a 3D printer, but conversion still has to be done. The 3D print file is called an .STL file. STL stands for Standard Tessellation Language or Standard Triangle Language, and originates from lithography, a principle from the graphic design and print world.

The main purpose of an STL file format is to encode the surface geometry of a 3D object. It encodes this information using a simple concept called “tessellation”. Tessellation is the process of ’tiling’ a surface with one or more geometric shapes, such that there are no overlaps or openings. If you have ever seen a tiled floor or wall, that is a good example. It looks like a mosaic. We now have our conversion principle, but that is only based on the surfaces, not on the inside (the infill). Moreover, it is not very accurate, because our ‘solid’ will therefore be converted into ’tiles’ (sometimes called polygons, as they are not really triangles – as you have seen already, of course).

If we convert the Mona Lisa in tiles or polygons, we get this:

Left: Mona Lisa as tesselation-model. Right: the orginal.

Tessellation (also called triangulation nowadays) is also used in geographical position determination.

Now that we know all this, we have actually come across the biggest drawback of 3D printing: it is always a compromise between the original (the solid) and the triangulated model. Moreover, only the surface is converted, so we need to compensate for this as well. The triangulated model is also called a mesh – a network of polygons consisting of all the triangles it calculated. Incidentally, these triangulation calculations are vectors again, so they can be scaled well, but the resolution would decrease because the number of polygons does not change, they only become larger. Now we can of course convert the model into MANY triangles, to get an accurate model for our printer. The picture below shows this clearly: the more, triangles, the more accurate the model becomes.

Different triangulation resolutions

A disadvantage of the multitude of triangles is that the file becomes very large, making the software of the 3D printer causing to have trouble with the calculation of the layers (this conversion to layers is also known as ‘slicing’ of a model, building a 3D model in print layers). The slice method is a important factor.

Solid Artist to STL Artist

We are now going to convert our Artist’s mouthpiece into a .STL file. As a reminder, here is the original:

If we convert our model into a low resolution STL, our 3D print preview of the mesh model looks like this:

Triangulation resolution

The above image is not very encouraging; there are far too few triangles (7224 to be exact), making the model look too coarse. We will now have the 3D printing software build up this model into a preview. We use Simplify3D software for this, it has a very good slicer module, a lot of options and supports many printer models.

This is what the rough model looks like in the Simplify3D software previewer. The surfaces from which the model is constructed can be clearly seen. The print does not look very accurate (it is not a smooth circular shaped object), and will be difficult to finish by hand, if it is printed from material that should be sanded and polished.

We therefore want our model to look as accurate as possible. We are now exporting the model to an .STL file of our solid with the highest possible resolution.

The above model now consists of more than 1 million triangles. The density is so high that the model now looks black due to the multiplicity and density of polygons! There is now a tolerance of less than 0.5 degrees in each ’tile’ that represents the surface structure, which is why we need so many tiles. In the software of Simplify3D the preview looks much better:

Although we have now improved the quality of our 3D model, there are still a number of challenges. The first challenge is that our mesh must be printed with an infill of 100% to make it massive. We can specify this in the software, but where the printer starts to print more slowly, especially around the wind way, the 3D model becomes quite warm as the printhead is more than 200 degrees. We must therefore use cooling. This is done by built-in fans in the printer and can be controlled by the software. However we also have to make sure layers will stick on top of each other. Another problem is that if we print the model in this way, the ‘roof’ of the room may collapse; the filament (the thread with which it is printed) horizontally bridges the window, so it will have to be supported. We do this with help of the software using support structures. These are very thin ‘turrets’ that are co-printed to support the horizontal overhang. These can later be easily removed with a thin set of pliers. You’ll see this in Part V as we discuss the print results. The program automatically calculates when and where this support should be placed, based on our settings, however we can override the placement and amount of supports.

Supports are coloured dark orange in the above cross section.

In the software we can add or remove supports. The amount of parameters in the software that controls the printer often makes the process quite long to get to a usable model, as often times it requires trial and error. Issues such as temperature of the print head, print speeds, resolution, layer thickness, placement of supports, cooling, material, retraction of the filament to prevent droplets or blobs and the amount of filament feed can cause frustration and confusion. We now have a few years of experience with 3D printing at deQuelery, and this is really necessary to be able to estimate what it takes to produce a usable print that matches the original design (the solid).

In the end, the software converts the model into layers, while taking into account all the other settings that we have applied. In the image below you can see print speeds, the duration of the print (7 hours and 53 minutes), the number of layers that make up the print (1005 layers of 0.1mm) and the amount of filament needed to print the entire model: 11236.6mm , or more than 11 meters. The final weight will be about 33 grams. Other filaments, especially metal infills are much heavier. More of this in Part V.

The ring around the model is called a ‘brim’, and is needed to allow the right amount of filament to flow into the printhead before the model is actually printed. The thin platform on which the model is positioned is called a ‘raft’. First a few layers are printed to create a ‘foundation’, resulting in a nicer print.

The code (the so-called g-code) that the software produces controls the printer and consists of 652992 lines of code. This code controls temperatures of the print bed and extruder, where the print head should go and how much filament should flow through the nozzle. The fans are also controlled with this code. Below is an example of a few lines of g code:

G1 X0 Y0 F2400 ; move to the X=0 Y=0 position on the bed at a speed of 2400 mm/min
G1 Z10 F1200 ; move the Z-axis to Z=10mm at a slower speed of 1200 mm/min
G1 X30 E10 F1800 ; push 10mm of filament into the nozzle while moving to the X=30 position at the same time

This was part IV of a series about designing and creating our new Artist mouthpiece. In part V we will go deeper into materials and finishes of the printed models. What filaments are there, how is finishing done (and is this really always necessary) and what does the end result look like. Does it make sense to use heavier filaments with metal or wood? Do they affect the sound?

These and more issues will be discussed next time!

Part I – Introduction | Part II – The Design | Part III – Simulations | Deel IV – 3D Printen

This post is also available in: Dutch