One problem I have had, on and off, with Mendel is a tendency for the infill not to meet the outline. This was particularly bad with PLA. I have combated this by having some infill overlap and also extruding the plastic slightly faster than it should be, so that the solid layers are well stuffed. I don't have to do either of these things with HydraRaptor. I couldn't figure out what the difference was until I changed a reel of plastic recently. The first print on the new reel came out like this: -
The gap is always at one side like this, it is as if the infill is not centred within the outline. The reason, I have come to realise, is that the belt has some play in it because it is not infinitely taught. When the extruder pulls filament off a reel it exerts a force on the carriage, which displaces it slightly from where it would come to rest without any external force. Because the carriage moves, the filament only gets pulled from the reel at the local extremes of movement, the rest of the time it is slack. This causes small offsets in the filament paths. In particular, when it is doing zigzag infill it is using filament relatively quickly, so at the end of the zigzag furthest from the centre of the bed it is likely to give a little tug of the reel each time, causing the zigzag to stop short of the outline.
The conclusion is that the filament feed for a belt-driven moving-head machine needs to be very low drag. The hanging basket technique that I used on HydraRaptor is no good because it has to pull plastic out from under its own weight. Making the feed point high above the machine reduces the lateral drag on the carriage, but you can easily get enough vertical drag to deflect the x-bars upwards, or even lift the z-axis slightly because the backlash in the thread is only taken up by the weight of the x-axis.
The reason it was bad with PLA was because I was pulling it from a hanging basket and being very stiff, even a small coil needs a lot of tug.
The system I now use is a vertically mounted spool big enough to take 5kg coils, which last me a couple of weeks.
The bearing is just a stainless steel axle running in PLA bushes, lubricated with some lithium grease. It is low friction, but not as frictionless as a ball bearing. It needs a little friction to stop any in-balance in the coil causing the spool to spin to its low point. Also the faces of the spool need to be quite big to stop a loose loop of filament coming over the side. It pulls tight and jams if that happens.
I can take the spool apart to insert a new coil of plastic.
In general though I have to wind it all off and on again to get it tight and balanced enough to wind off smoothly. Since 5kg is about 800m it takes a long time to wind it onto a garden hose reel and then back on again. Someday I will get round to making a machine to do it for me. In the meantime I will make a second identical spool so that I can just mount the coil on one spool and wind it onto a second one to use.