Problems with thin walls

I've recently been trying to print some components from Repetier Host 1.6.2, using the included Cura slicer. The components have been carefully designed in a CAD package as they need to fit up and interlock with some existing components. I remember seeing somewhere that the shell thickness set in Cura should be a multiple of the nozzle diameter. I use a 0.4 mm diameter nozzle, so I usually leave the shell thickness in Cura set at 1.2 mm for most of my prints. The internal walls within the CAD model were also designed as being 1.2 mm thick on the basis that this was again an exact multiple of the nozzle diameter and the same as the shell thickness. The reasoning was that this should enable the most accurate reproduction of the design.

Several test prints of the interlocking end part of the model produced parts that would not fit up to the originals without really heavy and excessive cutting and trimming. All the test pieces measured up at the correct overall dimensions but all the wall thicknesses (external and internal) measured up at almost 1.6 mm. This was the reason that the parts would not fit together since the outside of some of the external walls needed to fit against some of the internal ones.

I suppose that I could simply have redesigned the model shifting things so that the resulting parts would fit. However, being a retired engineer this approach niggled so I decided to do some tests to identify the cause of the problem. My Anet A8 printer is pretty accurate and the overall dimensions of the parts that I had printed were correct to 0.01 mm, so I discounted the printer as being the cause and decided to have a look at the slicer.

The test prints that I designed for this were bottomless, topless and hollow 30 mm x 30 mm x 5 mm high boxes. These would enable me to measure the wall thicknesses at several points along each side and determine the average wall thickness for a side and for the whole perimeter of the box. The wall thicknesses should not be affected by the bed (which in my case is aluminium coated with super-hold hairspray) and I used some 1.75 mm PLA printed at its optimum bed-stick settings of Extruder 200C Bed 60C with cooling except for the first layer. All were printed at 0.2 mm layer thickness. I printed a set of three test boxes each with a different wall thickness at the 1.2 mm shell thickness in Cura with 20% infill (this setting shouldn't matter as the walls are too thin to require infill). I then followed with a further set of three with Cura set to 0.8 mm shell thickness and 100% infill ( the 0.8 mm is smaller than the setting recommended in my Cura manual and again the % infill shouldn't matter).

The results were not what I had expected and the detail is given in the attached PDF. Using a 1.2 mm shell thickness to print 1.2 mm thick walls gave the WORST result with the average thickness measuring at 1.56 mm. For the box with designed 0.8 mm walls the measured value was 0.93 mm, and the one with 1.0 mm walls measured up at 1.17 mm. It seems that as the wall thickness increases the error between the designed thickness and the measured result increases. I might carry out another test for walls of 1.6 mm, 2.0 mm and 2.4 mm at these settings to see what happens. Above 2.4 mm wall thickness there will be infill between shells.

The second set of three boxes (with the shell thickness set at 0.8 mm in Cura) produced much better results. For the 0.8 mm wall the measured average was 0.92 mm, the 1.0 mm wall measured at 1.13 mm and the 1.2 mm wall at 1.32 mm. For this set of tests the error was almost a constant at +0.12 mm.

The results in the second test give me a sound basis on which to modify my design. The main point however, is that the results indicate that something strange is happening in the operation of the slicer, and which appears to be contrary to intuition and common sense.

Any thoughts or comments would be much appreciated.

Edited 1 time(s). Last edit at 11/05/2016 12:13PM by Supermec.