@csorrows yup - actually have an update to that end
so I've been working through settings for the past few days, methodically going through everything that could improve my accuracy based on various forums, blogs, and my experience with CNC's, and here is what I've found to dramatically improve my build accuracy -I'll leave this here for future reference and anyone who is curious
for reference, I'm using the FT5 with Titan extruder head and a V6 .40 mm nozzle, configured for direct feed 3mm filament. all experiments so far have been on low-grade inland PLA with moderate dye load. my extruder is configured for the steps-per-mm recommended in the Titan documentation - 418.5 or thereabouts, i don't remember exactly.
1) @PaulS is correct on the upscaling in whatever slicer software you use - I'm running at 101 - 102% upscale on X and Y and it's getting within +/- .10 mm now, even on very large dimensions, much better than before.
2) I determine the scaling factor by printing the first few layers of an unscaled part, taking measurements, scaling the part in the slicing software, and restarting the print. this adds about 20 min. per print.
- prints around 20 mm outside dimensions are doing well with 102 % scale X and Y, nothing in Z
- prints around 50 mm are doing well with 101% scale X and Y, nothing in Z
- larger than that typically goes down to around 100.5% or lower scale X and Y, nothing in Z
3) I've set the extrusion width to match the nozzle width (.40 mm from the default .48 mm recommended in simplify3D), and set the extruder to very slightly over extrude - 1.05 or 105% - thought this may seem a bit self defeating, it's not once you consider how the slicer parses the STL to generate Gcode
- This change gave me dramatic improvements on my smaller internal dimensions.
2) Lowering speed didn't help me much at first, so I ran through inertia calculations on the X and Y gantry/head combo, and found that, for the added mass of the titan extruder at least, the default XY acceleration and Jerk settings in the Marlin firmware were way too high - i downed the XY-from-start and max XY to 2000 from 2500, and my corners have improved immensely - stepper oscillation is down and ringing artifacts have nearly disappeared from my prints. I'm going to go even lower and see if it gets any better.
- I've downed XY jerk as well, down to 12 from the default 15
- people elsewhere have expressed concerns that downing the acceleration and jerk would lead to blobbing on the corners, and they appeared to be right on my first test print, so I upped the accel and jerk on the extruder and blobbing has nearly disappeared.
- I've since increased my print speed in testing, and i am currently printing out a part at 70 mm/sec that's coming out clean as a whistle, cleanest print I've done so far.
3) moved over to glass bed on top of the FT5 heated base, coated with hairspray, and holy crap, the difference in unbelievable. Best surface I've tested so far, so good i'm actually considering not testing anything else. When people say a good first layer is massively important, they're not kidding - having a super clean first layer has dramatically improved the quality of the whole print.
4) sort of a general design tip that I learned the hard way - when doing super thin features, around the scale of the extruder width, you have to remember that the slicer software can only raster the edge as some multiple of the set extrusion width - account for this in design by making your feature thickness as a multiple of your set extruder width, or bulk out the part enough so you can accomodate the outside shells plus some internal fill - this can improve your dimensional accuracy by as much as 1/2 of the extruder width - generally +/- .20 mm. with the improvements listed above, following this design idea gave me overall improvements in the neighborhood of +/- .50 mm. This may have been intuitive to some people but I clearly wasn't thinking on my first few runs.
I've got a lot more testing to go - a casual google search brings up people pulling +/- .025 mm accuracy out of their machines, and that's my final goal.