In his article: x-idler-bracket-continued Vik Olliver alluded to the fact that you can extrude filament with a smaller diameter than the hole in the nozzle. I did some experiments to see how fine I could go. In fact the final filament diameter is simply determined by the feed rate of the extruder and the travel rate of the nozzle, or in my case the bed. The filament stretches to the length that matches the rate of travel while it is still liquid. You can then calculate the mean diameter from the volume of material extruded. The nozzle height has to be a bit less than that mean diameter and then the width becomes a bit wider.
Here are three 20 x 20 x 20 open cubes with different wall thicknesses :-
The first was 1mm diameter filament extruded at 4mm per second with a height of 0.8mm giving a wall thickness of about 1.2mm.
The second was the same feed rate but with the extruder traveling over the bed at 16mm per second to give 0.5mm filament, the same as the nozzle hole diameter. The height was set to 0.4mm giving a wall thickness of about 0.6mm. As you can see it warps more but I expect it would behave if it was building a solid object. The bottom layer which was stuck to the table has better corner definition.
The third attempt was 0.35mm filament extruded at 16mm per second with a hight of 0.28mm and a width of about 0.5mm. As you can see holes started appearing but I think that was just because the sides buckled so badly. Interestingly the holes can be bridged by filament above that needs no support. Again, I think this would be OK making solid objects, or at least objects with thicker walls.
This is really good news as it means I can get down to the sort of resolution commercial machines get (0.25mm) without having to have a very small nozzle aperture, which would limit the flow rate. It remains to be seen what effect stretching has on the polymer but as it is still liquid at that point I think it wont increase the contraction much, if at all. It does mean I need very fast head movement to keep up the deposition rate, about 64mm per second. I think my machine will do that if I reconfigure the steppers for speed rather than torque, a simple one wire change.