Mike's Oud Forums

Experiments with a 3D printer - Sound Hole Rosettes

jdowning - 11-8-2019 at 05:17 PM

Having purchased a low cost 3D filament printer (a.k.a. FDM or fused deposition modelling printer) about 18 months ago I am curious to find out if it might be used to print sound hole rosettes of reasonable complexity and quality. A preliminary discussion on this forum can be found here:


As a novice in 3D printing I have much to learn so at this point cannot predict the outcome of the proposed experiments. I am interested in testing two approaches to the task the first being to start with a scanned 2D drawing of a rosette as the initial input and later to create a rosette using a basic free CAD software program. The former approach should be by far the simplest if it works. Both methods will create an .stl
( stereolithography) file of the rosette model that must then be processed further by 'slicing' software that can be read by the 3D printer that builds up the model in layers or slices.

The printer extrudes layers of thermoplastic from a spool of filament fed through a heated nozzle on to the heated bed of the printer - movement of the nozzle being controlled in 3 dimensions, the X, Y, Z axes of the printer. Mechanical set up of the printer is critical to production of satisfactory prints there being a number of variables such as print speed, filament melt temperature, bed temperature etc as well as bed levelling, setting of nozzle to bed distance to consider. All variables depend upon the model geometry as well so there may be a lot of trial and error involved to obtain optimum results.

So let's make a start.

jdowning - 11-9-2019 at 03:48 PM

Lute rosettes are - unlike oud rosettes - traditionally pierced and carved directly into the sound board using a paper pattern glued underneath the sound board as a guide. Sound board thickness at the sound hole location is generally around 1 to 1.5 mm thick to facilitate cutting the rosette after which small reinforcing bars are glued under the rosette for strength.

Oud rosettes are thicker and cut from a variety of materials different and separate from that of the sound board and are attached with glue under the sound hole periphery. Although it is possible to inset a lute rosette into and flush with the sound board (sometimes done in restoration work) a 3D printed rosette would be more historically appropriate for an oud than it would be for a lute.

I do not have a 2D line drawing of an oud rosette but do have a full size paper pattern of reasonable complexity for a lute - drawn from an original by renowned luthier of the early 16th C Laux Maler. So this will be used as a starting point for this project. Note that a clear photographic image of a rosette might also be used as a starting point - something to be tested later.

The scanned image of the rosette pattern is first edited in Photoshop (any other photo editing software will do) to correct any obvious minor imperfections in the hand drawn image and then edited with a black back ground to produce a sharp maximum contrast black and white image necessary for the next conversion stage. The edited image is then saved as a .jpeg image file.

jdowning - 11-9-2019 at 04:09 PM

The edited .jpeg image must next be converted to a 3D .stl (stereolithography) file suitable for being further edited and 'sliced' with software ready for input to the 3D printer. The conversion is done quickly and freely on line here:


A depth value is assigned to the image as well as the outside dimensions. All white areas of the image will be printed full depth whereas all black areas will be printed zero height - hence the black and white image. As it is possible to adjust dimensions further in the slicing software a slightly oversize image has been converted and saved to my PC as the required .stl file.

jdowning - 11-9-2019 at 04:26 PM

The slicing software that came with the printer is CURA - freely available as are a number of other slicing software. A range of printing parameters is available. As a start I am using the parameters of my last print on the printer made 18 months ago as a convenient starting point but which may not be the optimum setting for this print.
Opening the .stl file into CURA displays the rosette 3D image on a working platform of size limit related to the Creality Ender 2 printer. The dimensions of the rosette model are then adjusted slightly and a generated .gcode file saved to a micro SD card for insertion to the printer. The time to make a print of these dimensions is also displayed as 6 hours 44 minutes.

Now to make the print.

jdowning - 11-10-2019 at 02:58 PM

The printer platform has been modified with the addition of a piece of thin mirror glass held in place with three clips. This provides a very smooth and flat hard surface that may be removed to facilitate removal of the print and cleaning of the surface. Adhesion of the print is important to prevent lifting of the print during the printing process. A thin coating of water soluble glue stick on the bed prior to printing provides good adhesion. Platform adhesion may further be improved by printing a temporary thin flat rim (a.k.a.- a Brim) around the periphery of the print that is removed on completion of the print.

No mechanical adjustments have been made to the printer since the last print was run 18 months ago. A spool of PLA (Polylactic Acid) thermoplastic filament already loaded on the printer will be used for these trials.

With the printer switched on and micro SD card installed, the print head is moved to the start position on the platform and the .gcode file selected to start the print. The printer bed and nozzle temperatures are then automatically raised to the preset chosen temperatures after which the print automatically starts and continues to completion without further intervention in this case 6hours 44 minutes later.

The glass plate is then removed and the print released with a thin spatula.

jdowning - 11-10-2019 at 04:17 PM

The final print is less than perfect with thin flashing partially obstructing open spaces on the underside, a rough upper surface and filament 'whiskers' across the open spaces. The latter is due to insufficient filament retraction at the nozzle when crossing the open spaces. This can be corrected for future test prints to achieve optimum results.

The print had been improved by sanding both sides on a sanding block to remove the flashing and whiskers. Any remaining whiskers were removed with a craft knife. A total of about 1 mm in thickness of material has been removed taking about 2 hours work. The print was sanded using a block with rubber pad to firmly and uniformly grip the thin print. The plastic is hard and firm so may be easily carved for further improvement. However the objective next is to produce a print requiring little or no finishing work.

jdowning - 11-14-2019 at 02:50 PM

To test printing from a photographic .jpeg image an image of a Nahat eyun rosette, found on this forum, has been edited to a high contrast black and white image before conversion to a .stl file ready for editing in CURA prior to printing.

jdowning - 11-14-2019 at 03:01 PM

The print of the eyun took just over an hour - outside diameter 55 mm. Adhesion to the printer platform was good without need for a Brim - just a thin coat of glue stick was used.

The print has three obvious basic faults
1) There are gaps in the print in areas where the printer was unable to resolve the fine image detail.
2) The open areas under the print are closed up due to 'elephant footing' of the first layer.
3) There is heavy 'whiskering' across the open areas.

Comparing the original .jpeg image with the print (after sanding and clean up of whiskers) it can be seen that some of the very fine lines in the original have not been reproduced in the print - leaving gaps. This might be corrected by further editing of the .jpeg image to thicken up the problematic areas.

The 'elephant footing' may be caused by the first layer of the print spreading out in contact with the heated printer platform. This may be corrected in part by reducing the platform temperature but may also require adjustment to platform levelling and nozzle height.

The 'whiskering' may again be caused by a non level platform and nozzle height as well as filament retraction and temperature. The CURA software is supposed to move the nozzle in such a path as to avoid crossing over open spaces.

So the next step will be to edit the original image to eliminate any very fine lines. Then re-level the platform and adjust nozzle height and reduce platform temperature and filament temperature before the next test print. Adjustments will be made step by step to determine the optimum settings for the print.

pouryatorabi - 11-15-2019 at 07:04 AM

It was very interesting to see you use a 3d printer, I'm actually a co-founder of a 3d printing startup company, but our main field is DLP printers which are a little different from FDMs but I have also built and worked with FDMs before. This kind of 2d parts can be produced better and easier with laser cutters. You can easily produce one from wood or Plexi Glass.
I also wanted to print the whole oud with a FDM 3d printer once but I got too busy for that:( but I might try it one day.:)
For this part, It's good you are using PLA. PLA is really easier to print with and tougher than ABS, but I think 200 degree for 1.75mm PLA is a little high, lowering it to 190 degree may help you get better prints, And also as you mentioned, and as I see from your images, you truly have retraction problem, try higher values for retraction.
By the way why are you using 0.1 mm for layer height, I think 0.2 or 0.3 is good for this part. If use 0.3, I think you might have less retraction problem, This will also decrease the printing time.

jdowning - 11-15-2019 at 04:32 PM

Thank you for your helpful observations and recommendations. As a newcomer to 3D printing I have much experience to gain before I can select the optimum print settings for a particular print design.
I purchased a relatively low cost entry level FDM printer as an educational tool to learn more about CAD/CAM and CNC technology and find out what the possibilities might be for making useful prints. I likely will not advance to DLP printers or laser cutters once these trials are completed and may give the printer to my grandson to play with while I continue to manually cut rosettes in the traditional manner (should I ever get around to building another lute!)

I have now edited the 2D .jpeg image to thicken up the very thin white lines that the printer is unable to resolve in order to eliminate the gaps evident in the last print. The printer platform (bed) has been re- levelled and some changes made to the settings in Cura such as nozzle temperature 190C, bed temperature 45C, layer height 0.2 with increased retraction to 15mm. The print failed due to detachment from the platform. However 'elephant footing' appears to have been eliminated (hence reduced area under the print allowing the print to detach?) as well as the gaps in the print - so some progress perhaps due partly to re-levelling the platform? I will run another print tomorrow increasing the bed temperature to 60C and adding a brim for extra adhesion (as well as using glue stick adhesive). Print time was just under an hour.

I wonder if part of the 'whiskering' problem may be due to the relatively high proportion of open areas in this model of rosette compared to the Maler lute rosette initially tested in this thread?

MattOud - 11-16-2019 at 08:40 AM

I really like this thread. Thanx for posting this valuable experiment.

"The CURA software is supposed to move the nozzle in such a path as to avoid crossing over open spaces."

I have another thought on this....
Have you tried a PNG(which retains transparency info) rather than JPG(which is a compressed image which does not retain info).? Maybe CURA is printing the 'black' spaces even though there is nothing to print in those areas...the software does not know this when seeing JPG, but if you make the black transparent, and save as a PNG, it may yield better results?
Hope it works out. Great stuff!

jdowning - 11-17-2019 at 03:02 PM

Thank you MattOud.

I may be wrong about the CURA software being able to move the nozzle on a path to avoid crossing open spaces. There is slicing software that claims to do this but it is not free - costing US$150 - so I do not plan to test it!

My latest print shows an elimination of 'elephant foot' but resulting in a print adhesion problem on the printer platform. So, before going further I have checked out that there is no mechanical looseness of the printer moving parts and re set the platform levelling - all critical to producing good prints. On the Ender 2 printer the platform is 'levelled' with respect to the three X, Y, Z axes by a three point adjustment - a bit fiddly to adjust using a paper strip as a 'feeler gauge' to set the critical nozzle height above the platform. Set too high may create problems with adhesion of the first layer. The next test run at the new levelled settings will be with nozzle temperature of 190C, platform temperature 60C, layer height 0.2 mm and retraction of 20mm.

I plan to further improve the B/W image by editing in Photoshop to eliminate all gaps in the print so will save that as a PNG file for conversion to a .stl file to see if that makes any difference.

jdowning - 11-20-2019 at 05:14 PM

I am still finding my way in CURA by testing several settings and running a few prints for comparison.

There is a setting in the 'Expert' tab that is supposed to prevent the extruder head from crossing open areas in the print - known as 'combing'. So far this retraction setting alone has not eliminated 'whiskering' so I suspect that for a print with a high proportion of open areas (as is the case with the current eyun print) it may not be possible to avoid whiskering across the open spaces? This idea is to be tested later with another model - the central section of the Maler rosette printed earlier in this thread.

The eyun B/W image has been further enhanced in Photoshop to sharpen the image and thicken up all thin line areas to eliminate gaps in the print - saved as both .jpeg and .png files for conversion to .stl files for input to CURA.
The printer platform (bed) has also been re levelled over the three suspension points to ensure optimum print quality. Use of glue stick adhesive on the platform may not be necessary as a few prints on the glass platform surface alone (well cleaned with alcohol) have adhered without problem and have been easy to release.

pouryatorabi - 11-26-2019 at 10:23 AM

I really like to see some pictures of your latest prints if you can share, besides, as I know there's no need for platform heating for PLA but it will help and 45 degree would be enough. For ABS, we used to use kapton tapes and for PLA we used paper tape (or painters' tapes) and add glue adhesive on top of that. If you want to print large files, your prints may separate from bed because of warping, using glue and heating bed may prevent that.
If you still have 'whiskering ', try to increase the print speed.
Take a look at these:

jdowning - 11-26-2019 at 12:58 PM

Thank you pouryatorabi.

I have had problems with the last couple of test prints that I think may be partly due to one of the three V guide wheels on the platform support base failing. The bearings moves slightly within the outer hard plastic V wheel causing the platform move out of level with a clicking sound when pressed on one corner. I have ordered some spare replacement wheels that should be delivered within a week. No point in running more tests until that problem is fixed and the platform re levelled.

jdowning - 12-2-2019 at 05:48 PM

Here is one of the three guide wheels from the printer platform - running in a V slot on the Y axis rail. The wheel assembly that was faulty was moved to a warmer room in the house so that I might measure it to order the correct replacement. When picked up a few hours later it just fell apart (!) revealing the individual components of the assembly - the outer V wheel, two ball bearings and bearing spacer. The bearings are supposed to be a tight press fit in the V wheel so there is no relative movement so I am not sure why the assembly failed. The outer V wheel is made from POM (polyoxymethylene) a.k.a acetal a hard stable plastic with low friction characteristics, high wear resistance and low thermal expansion. It is easily machined. The precise POM composition of the original Ender 2 wheels is not known so I ordered replacement wheel assemblies made from Delrin POM costing CAN$1.99 each(!) a dollar cheaper than the Ender 2 OEM wheels.

These arrived today so the faulty wheel assembly will be replaced tomorrow and the printer bed re levelled ready for printing more trial prints.

The ready availability of replacement wheels suggests that this mode of failure is not uncommon so 6 spares were ordered.

More precise and costly printers use rails and linear sliding bearings for their guide systems.

jdowning - 12-10-2019 at 07:08 AM

The replacement wheel for the printer has now been installed and adjustments made to the printer mechanical systems such as re levelling the bed, re setting nozzle height over the platform etc. At the same time upgrade components have been installed to improve performance and make adjustments easier such as alignment washers for the platform springs and increased diameter platform height adjustment thumb wheels. All of these useful components were printed on this printer from files freely available on line. An advantage of a 3D printer!

jdowning - 12-11-2019 at 05:08 PM

Having adjusted the printer several small test prints have been made to try to determine the optimum settings in the CURA sllcing software. For the full size Mahler rosette there is some improvement over the first attempted print but still requiring work to refinish to remove the whiskers (not quite as bad as before) and to produce a smooth upper surface finish by sanding - the printed finish being uneven and irregular in places.

jdowning - 12-11-2019 at 05:29 PM

Further test prints have been made of the Nahat eyun with some improvement but again with the same problem of clean up work necessary to remove whiskers and uneven upper surface requiring smoothing by sanding.

Undeterred a test print of the main Nahat rosette has been attempted with a diameter of 118 mm plus the brim diameter (for maximum print adhesion to the platform) - just about the maximum size for this printer.
A JPEG black and white high contrast image was prepared from an original photograph of the rosette and then converted to an .stl file for slicing in the usual manner.
There is a lot of fine detail in the rosette design that the printer was not able to perfectly resolve so again some whiskering and upper surface roughness at the current slicer settings. I increased nozzle temperature to 210C (as recommended in CURA for PLA plastic) hoping that it would bond the filament better and result in a smoother upper surface of the print but it has not made much difference.

jdowning - 1-13-2020 at 04:28 PM

Having run a number of other test prints of both the Maler and Nahat designs on different settings of the Cura software (all using a 0.4 mm nozzle and PLA filament) I have been unable to obtain a print with a smooth upper surface and free of stringing or whiskers across the open spaces. Although the prints may be cleaned up by fine sanding the upper surface and cutting out the whiskers with a scalpel this entails significant extra work which is undesirable. Better results might be possible with a more costly 3D printer and different slicing software (?) but I have concluded that this printer is not designed to handle rosettes of this design and complexity. I am therefore concluding this series of tests.

Currently I am now interested in experimenting with a low cost desktop CNC router for milling out rosettes from wood using very small diameter milling cutters. The machines are mechanically basically the same as a 3D printer but with a motor and cutter in place of the heated filament. I have ordered a self assembly kit - a 3018 Pro version of the machine subject of this video:


The machine ordered has an upgraded main board and with addition of an off line controller so can be operated without being connected to a computer. Cost is under $300 CAN. All part of my ongoing education in this technology.

As with the low cost 3D printer subject of this thread I cannot predict the outcome of future trials in milling rosettes but will report my future experiments as a new and separate topic on this forum. Delivery is supposed to be by the end of this week so I should have something to report sometime next week.

Greg - 1-14-2020 at 04:16 AM

A very big thank you John. Your series are always so detailed and informative, as was this one. I look forward to your milling experiments.
You say the cost to you to purchase the equipment and materials in order to perform these experiments is part of your ongoing education.
I can say with certainty that you play an important part in the ongoing education of many (perhaps most) of us. Many thanks.


jdowning - 1-14-2020 at 02:37 PM

Thank you Greg! It has always been a pleasure over the years to contribute to this excellent forum and - in turn - to learn from other forum members.
One is never too old to learn something new each day!

This new educational tool is now scheduled for delivery next Monday so I am looking forward to that and then to discover what it can do at relatively low cost.

Happy new year to all forum members and visitors.