Monday, August 13, 2012

Printer tuning

I’ve just spent several days working on my primary RepRap printer, and the net result is that my prints are now way, way better than they were before.  I am going to quickly cover the changes that I made here. Some of these changes made a huge difference, some very little, but they’re all improvements IMO.
1. New version of Slic3r
The newest (0.9.1 as of this posting) version is really, really good.  Alessandro not only did a really awesome rework of the GUI and added a ton of great new features, but the output is now even more accurate.  I think that I had been suffering from over-extrusion with previous versions.  The previous version already produced better output than I could get with Skeinforge, but with the new version the output is, IMO, perfect.  I really can’t see ANY problems with the print, at all.  Really good, accurate slicing is a hard problem, and this version does it right.
2. New firmware
I don’t know if this improved things or not, but I moved up to Marlin 1.0 RC2.  I had been on RC1 for months.  I don’t believe it made much difference really, but it can’t hurt to be on the latest version.  If you’re still running Sprinter, then yeah, switch to Marlin.  Really, I mean it.
3. Run the M303 command
In hindsight, I think a lot of the trouble that I’ve been having with layer consistency has been due to poor temperature control. With my old MakerGear hotend and without proper tuning, I would have +/- 8 to 12 degree swings in hotend temperatures.  The M303 command got that down to about +/- 4 degrees with the MakerGear hotend, and +/- 0.5 degrees with the J-Head.
4. Switch to J-Head hotend
I bought a J-Head hotend several months ago and gave it a go, but I was not happy and shelved it quickly.  The J-Head has a protruding nose at the nozzle and this tended to “plough up” plastic as it moved around the print.  However, I now believe that was 100% caused by over-extrusion due to software problems – it simply isn’t a problem now that Slic3r is producing the awesome output that it now puts out.
Advantages (over MG hotend, the only other hotend I’ve used)
Better temp control - there’s no barrier between the nozzle and where the heat gets injected by the power resistor, and the thermistor is in a hole instead of being taped to the side of the nozzle, so that works better too.  With the J-Head, Marlin/PID (after tuning) can keep the head within +/- 0.5 degrees.  This is 10x better accuracy than I was able to obtain with the MG hotend.
Less jamming - I think that a lot of the jamming that I got in the past was due to the plastic softening/melting far up the barrel.  The barrel in the MG hotend is brass most of the way up, and the heating element is ABOVE the nozzle quite a ways.  This just necessarily is going to have a lot of heat high up.  The J-Head has all the metal and all the heat in the bottom 15mm or so.
Less dribbling and better control – The J-Head has quite a long narrow segment at the nozzle, which helps keep the plastic from dribbling out between extrudes.  Also due to the very short melt zone, control seems to be very good.  I have essentially zero strings hanging between separate objects on a print, and at the end of a print and hotend cool-down, I’ll have perhaps 10mm of dribbled PLA hanging down, and I think I only get that much because I hadn’t previously thought to put a filament retraction command into my post-print GCode (I’ll have to do that tonight).  On the MakerGear hotend, I would typically have a whole pile of PLA dangling out of the hotend and sitting on the print bed after a print – I think all the plastic in the bottom 20mm or so of the nozzle would dribble out (I had to extrude 20 to 25mm of filament when starting the next print before anything would start coming out).  This really speaks to the fact that there is clearly a large blob of molten plastic in the bottom, and that dribbling control could be better.
The only downside to the J-Head is that it really needs a cooling fan blowing on the hotend to keep PLA from jamming, but that’s a trivial price to pay.  It’s actually cheaper than the MG, and it takes about 2 minutes to “assemble” rather than an hour or two for the MG hotend.  I also just get the feeling that the J-Head is likely to keep printing without maintenance or repair for longer than a MG hotend.
I don’t want to leave the impression that the MG hotend is bad.  I used them for a year and they’ve done OK by me.  I’ve just decided that the J-Head is enough better that I’ve ordered J-Heads to replace both of my MG hotends, which will go into the junk box as spares.  The J-Head is now my recommended hotend.
5. Switch to machined LM8UU linear bearings
I think that PLA printed bushings are OK, but I have found that over time they do wear out faster than I’d thought.  LM8UU off eBay are really pretty cheap and I now think that they’re definitely worth the $20 or so you pay for a set.  Delivery from China can take a few weeks but they’re good.  I would recommend getting a few extra – I have always had one or two that make an annoying popping sound as they move on the smooth rod, I assume that’s stuck balls in there and not only is it irritating, it’s likely to damage the smooth rod over time.  Have some extras to replace the bad bearings.
6. Go away from the Aleph objects X ends
These seemed like a fantastic idea, they vastly simplify assembly over the standard Prusa press fit ends (which I hate intensely), but the hard clamped single drive nut in the trap telegraphed every irregularity in the Z axis drive screws to the print, and the little open-sided PLA bushings did not have the authority to hold the ends steady.  I simply could not get smooth prints with these ends.  They weren’t terrible, but the regularly spaced bumps really bothered me.  Luckily this X-end has now been posted to Thingiverse – it goes back to the floating/sprung pair of nuts but retains the clamp type smooth rod interface, AND it uses LM8UU bearings, and has enough space to fit the GT2 36 tooth pulleys.  Huge win.
7. Added a fan
This is a big deal I think. I did it in the midst of a bunch of other changes so I can’t isolate how much this changed things on its own, but when I added a fan to my 2nd printer it made a big difference.  I’m going to always look for a fan mount on my X carriages from now on.
8. Switched to optical endstop on Z axis
The X and Y endstops aren’t in any way critical, but the Z is.  I was having significant problems with repeatability on my Z axis, and always had to dork around with the Z screws at the beginning of a print.  I had some Ultimachine optical endstops that I just threw into an order a few months back, and decided to try them.  Huge improvement.  I can actually just hit “print” and the print will start perfectly now.  Wow.
9. Slow down
I slowed way the heck down, to the default speeds in Slic3r.  I had been printing pretty fast.  I don’t know if this made a significant difference or not, once I get done with some of the stuff I’m doing now that needs very good output I’ll try some faster prints.  I had a significant amount of X axis wobble in the prints that was clearly a result of frame wobble, but it’s possible that the frame stiffening that I’ve done will allow me to pick the speed back up and not have that.

In the end, I have prints that are better than I could have dreamed of making a week ago.  Really, I just don’t see anything I can improve at this point.  The only problem is that I have a ton of work to do updating my build instructions on the site now.

No comments: