Here I've changed the advance value on the fly a couple of times, to show the effect that advance has. These cubes were printed at 100 mm/s, with an acceleration value of around 300mm/s/s, which shows the problems with the current version of smoothie (ie, no advance possible). You can see the blobbing on the corners as the head slows down, as well as the filament starvation as the head accelerates away from a corner. Changing the advance value to about 1.2 or so on my machine seems to really clean up the print.
The algorithm doesn't handle different acceleration values all that well, so next up is to try the 'proper' algorithm as shown here:
My algorithm also only works when segments are used, so is better suited to deltas than other machines at the moment.
If anyone is interested in testing these changes, the github repo is here:
I can post some test binaries if anyone wants to try them.
Edited 1 time(s). Last edit at 04/23/2017 06:52AM by nebbian.