@Kevon_Daines You can run them with a few bearing balls missing, and it’s not a huge deal because 3D printer loads and speeds are pretty light compared to the bearing ratings. Not ideal but it works. Obviously the more you lose the more likely you are to have problems though.
@mcdanlj Understood, but based on the tolerance mentioned a 0.01 mm digital precision caliper will be sufficient.
Indeed , if the tolerance were to require a higher precision (ex. 0.001μ) then a micrometer would be needed.
I have never seen anyone with this concept before. Which probably raises the question why this endeavour. I should undoubtedly look further to more of this project.
Look into hydrostatic linear motion. 
@Ryan_Carlyle it will likely work nonetheless, but regretfully ant the end on the missing bearings there will be a print error probably. Not ideal but works.
@Kevon_Daines these linear blocks have something like 20-30 bearings in contact on each side of the rail at any given time. It’s not like a rotary ball bearing where only 2-4 are loaded at a time. The other balls will carry the load without deflecting any measurable amount. In fact, in cheap rails with slightly inconsistent ball sizes, you often get uneven ball loading and only the larger balls carry load at all. I probably wouldn’t worry about it until 5-10% of the balls are missing.
@Kevon_Daines the tolerances I mentioned were for the parts I was machining and do not apply to the linear rail system. I don’t have spec sheets for the knock-off rail I bought, so I don’t know their specified tolerances or bearing sizes. As I say, if/when it is a problem, if the system is otherwise working well, an upgrade to real hiwin should address that problem. 
Someone in the RailCore community replaced their original Chiwin balls with nicer ones and got much nicer results. I’ll see if I can ask around.
That person used these balls: https://www.ebay.com/itm/3-32-Inch-Si3N4-Silicon-Nitride-Ceramic-Ball-Bearings-G5/291641117031
https://www.ebay.com/itm/3-32-Inch-Si3N4-Silicon-Nitride-Ceramic-Ball-Bearings-G5/291641117031?ssPageName=STRK:MEBIDX:IT&var=591370209822&_trksid=p2060353.m2749.l2649
Apparently some rails use 2.5mm balls. You would need to measure a couple samples.
I slowed down playing with this project as I learned machining (see the parts above I machined for the project) and then just as I was about to pick it up again, Google announced the Google+ shutdown and I spent a lot of my spare time working instead of moving content to MakerForums and improving this site. Every time I walked past the half-empty skeleton of the X5S, it mocked me. Then the junk “bed-flinger” hictop that got me into this whole 3d printing experiment really started to wear out, so I finally started work again.
I made a few mistakes, but so far I have a SKR Mini E3 V1.2 and TFT35-E3 connected to it, a 750W 120V SSR-switched and TCO-protected keenovo heater heating up a 1/4" / 6.25mm cast aluminum bed, a 24V DC supply, and a single Z motor running single-start lead screw with a timing belt keeping both sides in sync, as well as the linear rail I started the post with. It bears almost no resemblance to the X5S I started with, just the frame and some parts of the Z stage remain.
It’s spitting plastic again for the first time today. I have a lot more to do. I don’t even have a cooling fan mounted, and I’m sick of the bowden feed and will replace with direct feed. It turns out that the cooling fan on the dying bed-flinging printer wasn’t running anyway, so that’s not first priority here; switching out the bowden for direct drive is the first thing I need to do.
I’m currently printing on glass on top of the aluminum bed while I decide whether to install the thick PEI sheet I bought earlier, or get a PEI-coated steel sheet instead:
https://www.aliexpress.com/item/4000138916012.html
I bought the hollow-core polycarbonate sheets that Mark Rehorst recommends for an enclosure, but I’ll wait to build that because how high I make it depends on what I do to ditch the bowden for direct drive.
1 Like
I converted from bowden to direct drive. I had purchased a cheap clone titan extruder that Mark Rehorst had mentioned, and modified it to add a filament guide as he suggested. It’s temporary while I decide how I want to mount the bondtech I got.
I did have problems with the extruder not working. It would sometimes extrude a few mm of filament and then make noise and the motor would jitter noisily. I carefully went through every possibility I could think of, swapping out stepper motors, wiring harnesses, stepper controllers. I finally discovered that there were two physical problems: the set screw on the drive gear was very slightly interfering with the driven gear (solved by moving the drive gear about 1 mm further down the stepper shaft), and the screw providing the axle for the driven gear cause binding when it was too tight. Simple in retrospect, but carefully trying one thing at a time took hours of debugging.
I picked a random recent model to do a test print, and the quality is so much better than the “bed-flinger” ever was! I have a lot more I can do to improve here, but this is so much of an improvement I’m about ready to tear apart the bed-flinger and make something better out of it; I just didn’t want to be without an at least somewhat working printer if I could easily avoid it.
But I should design and print a fan shroud for the newly-rebuilt printer before too long.
By far the loudest part of the printer is the 30mm fan that came with the hotend. Now I discover that Noctua don’t make a 30mm fan, so I need to print a 40mm to 30mm duct/adapter as well as a cooling fan shroud.
Also, I found that I had an idler pulley binding sometimes, which resulted in some spaghetti printing. The beautiful print I started last night turned into spaghetti after about 28mm of Z. I removed a shim washer that seems to have been responsible for the binding I saw.
Sadly, metal from the binding idlers got into the MGN12H carriage and… now the carriage binds. So I guess I order new carriages and see whether they run smooth on the existing rails after a deep cleaning.
1 Like
I first tried buying new 3/32" Inch Chrome Steel Bearing Balls G25 that I saw recommended for a different clone MGN12H carriage, just to try the replacement. It looks like my existing cheap set was designed for a different tolerance class, so the G25 3/32" bearing balls were a little bit too large and the rail didn’t run smooth. However, I was pretty sure I had damaged the existing balls for which I had no specs, so no harm in trying a workaround; at worst I’ve spent some time and learned something.
I spent some time stoning the slot in the rail larger to give the new balls more room. I did my best to apply constant pressure from one end of the slot to the other, and started each round with a coarse diamond grit DMT plate and worked down to an arkansas stone, and every 10-20 strokes each on top and bottom of each slot, worked smooth, I cleaned it thoroughly and tried running the carriage. I got bored before I got it as loose as the other rails are now, but it doesn’t hang now. The stiffness is fairly uniform along the length, so I was being careful enough, but I lost patience before it was ready to use in a printer. I’ll keep it around in case I get bored, I guess.
I ordered a set of three of some clone rails with MGN12H carriages that are still not hiwin, but cost more than the original set, and have at least some good reviews. They also come with plugs in the ends to avoid the carriages falling off by accident, and include two extra balls just in case. (The are"free return" so if they don’t run well I can return them without more than loss of some time. I have another use for the two more-or-less working rails anyway.) All three of the new rails slide freely. I’m sure there are uses for which they would be too loose, but they are probably better than the belt I’m using, so they aren’t the weakest link in the chain.
Also, to better tolerate expansion in the gantry, I replaced the drop-in T-nuts holding the gantry on one side with slide-in T-nuts. Of course, I only had M5 slide-in T-nuts handy, so I had to make M3 slide-in T-nuts. I cut a couple 1cm long pieces of 1/8" x 1" mild steel stock, milled them into a T shape, tapped them M3 near one end, and filed them smooth. I put threadlock on the screws, screwed them down tight into the extrusion, then backed them off to just loose enough to move. Now I’m reasonably confident that the gantry can expand as it heats without causing binding.
The first print with these rails and the loose T-nuts to tolerate expansion was the highest quality print I’ve ever done. A few more prints and I’ll have confidence to move forward with starting over on the old printer to rebuild it into something better as well. 
I haven’t done speed tests yet to see what the limits are. Right now I’m printing at 80mm/sec and it looks like I’m being conservative. 
but so far this is great.
Last night I started a roughly 13-hour print, and woke up in the morning to air-printing — a few cm above what had printed fine, until it suddenly quit extruding. I’m glad it was still trying to print, because I didn’t have to start by assuming that it had clogged, so diagnosis wasn’t so hard. It was made a little harder by the quiet trinamic drivers, though; I didn’t hear the extruder motor turning until I started going through diagnosis and had everything else turned off.
Turned out that the drive gear in the extruder was loose. The set screw in the drive gear had to be so far back that it didn’t engage the flat of the D shaft, so it didn’t take much for it to come loose.
Tonight I wrapped the motor to keep metal dust out, and used a diamond file to extend the D profile closer to the motor. I put threadlock on the set screw, and tightened it as much as I reasonably could, and it’s printing again.
3 Likes
Air-printing again this morning, with the parts nearly complete. This time because the screw holding the idler pressing the filament against the hobbed gear came out. 
Two failures from loose screws in a row. Makes me wonder if the real loose screw is in my head! I haven’t used threadlock because I’m still playing with the design, but at some point it’s time to commit.
I’m not ready yet; I need to design a fan mount, and I should probably swap out the titan clone I hacked for the bondtech I already bought, too. I’ve just been hesitant to put the good stuff on while I’m running experiments.
In the past week, the problems I have had were only from
- I still don’t have a cooling fan
- I had bad filament that clogged the nozzle, which with XCR3D hotends is an easy fix
- I knocked the bed off the kinematic mount and didn’t notice.
The new rails are working great. I still haven’t done a speed test, but printing at 100mm/s doesn’t seem to reduce quality, with DEFAULT_ACCELERATION 2000 and JUNCTION_DEVIATION 0.25. The motors are running cool at 1A RMS.
2 Likes
The 30mm cooling fan on the XCR3D hotend is junk. Mark Rehorst noted that his died quickly, and mine did too. Mine quickly developed a bad bearing. I kept it going for a while by dragging a fingernail lightly against the fan as it spun up until the bad bearing seated and it quit whining, but as one would expect that was getting harder and harder to do.
I bought a set of 24V 40mm axial and radial fans, because 24V 30mm fans are hard to come by. After I bought them, Mark Rehorst pointed out that just using a buck converter to supply 12V would let me use a quality 12V 30mm fan, but I decided to first try making an adapter, since I needed to figure out how to also add a print cooling fan anyway.
I designed an adapter that provides an attachment point for a cooling fan and it’s working well so far. The 40mm fan moves lots of air and keeps it cool, and adding part cooling has resolved the normal issue of PLA curling up when printing overhangs and as a result knocking parts off the print bed.
I’ve been printing lately at 80mm/sec maximum, but probably at some point I’ll experiment to see how fast I can push this. I think the limit might be the low thermal mass of the small XCR3D hot end. I’m almost tempted to make a larger block and pid tune again, but probably not tempted enough to actually do it. Getting higher quality at 100mm/s on this printer than I did at 30mm/s on the “bed flinger” makes me happy already.
2 Likes
I have just finished tuning up 3 of my Di3 printers. Rebuilt all the hotends, rersquared the gantries, oiled bearings, replaced build plate surfaces, and recalibrated extrusion rates, PID tuning on the beds and extruders. Now I need to figure out why my PrintrBot simple Pro performs correctly on my office PC, but using the identical settings on my print farm PC it wants to drive the hot end into the bed. Temporary fix is to have the printer on a rolling cart and wheeling it into the office. Not a viable long term solution by any stretch of the imagination. I have pulled apart my large format printer and am redesigning it, as I was not happy with my original design. Always something to do with these printers. Oh, and did I mention I am also designing a portable small foot print printer that will fold down and fit into a pelican style hard case for transport
?
2 Likes
Something about the print farm PC must be confusing the induction sensor, then? I fixed a simple metal for a friend, whose induction sensor wasn’t working on 5V, but by jumping it to the 3.3V bus it worked fine.
Now that this rebuilt printer is better in every respect than the bed flinger, I’ve been challenging myself to design a new bed-flinger that I can build as much as possible with parts I already have lying around. So far the only part I havehad to order was a short closed-loop GT2 belt, and I chose also to buy some optical limit switches even though I have microswitches available. It’s a design that I haven’t seen anywhere else, so it’s probably a terrible design. But if I take that on, that will be a different topic! This one’s long enough. 
1 Like
I might just pull the Simple Pro apart and install a spare Cohesion3D mini board (smoothie) and call it a day. I am not a fan of the TinyG board that is in the Simple Pro, and the LCD is useless without the web app for it. I have an MKS LCD touch screen that could replace that LCD easily and give me back some headless functionality.
2 Likes
Back when I started this project, I kept an eye out for 330mm square PEI because I have a 330x330mm (nominal 13") bed, whereas a lot of similar printers have a 300x300mm (nominal 12") bed instead. I finally found a sheet of 13"x13"x 0.8mm PEI listed on amazon and bought it, and set it aside for when I was ready to mount it.
It didn’t occur to me to measure it.
Today I wanted to start printing ABS, and overall the printer was working well, so I went to install the permanent bed surface.
It was labeled 13"x13".
It measured 12"x12".
Return window closed on May 22, 2018
And now I understand why that particular item hasn’t been available for a long time.
So I’m back to printing on glass for the moment, and considering my options.
I guess I could cut the undersized piece down to size for the smaller bed for the gantry cantilevered design I’m considering if I do decide make that and not just think about it.
1 Like