All Things BOXZY *Unofficial*
Register Calendar Latest Topics
 
 
 


Reply
  Author   Comment  
gregket1

Member
Registered:
Posts: 30
Reply with quote  #1 
Any ideas or suggestions on this one?

If we have an object that has to print on the diagonal due to length, all the straight-lines become waves instead. I thought this might be the stepper drivers so I just completed the upgrade to the 8825s and we get the same error.

I should mention that all updates have been applied. And this isn't a case of inconsistent extrusion. If you look at the magnified photo you'll see that all the lines are exactly the same width, They are just laid down "other than straight".

Attached Images
jpeg image.jpeg (249.05 KB, 37 views)
jpeg image.jpeg (134.96 KB, 35 views)

0
dewhisna

Avatar / Picture

Senior Member
Registered:
Posts: 158
Reply with quote  #2 
I have the same problem too, and I know several others I've talk to on these forums have mentioned having it as well...

Tightening the belts helped a little, as did lubing the linear bearings, but I think (at least for my case), I need to completely replace the linear bearings and rods (there even feels to be a flat spot on the underside of the Y-axis guide rod, the one on the very front of the BoXZY).  I have new rods and bearings ready to go on, but am waiting for an opportune time to tear things down and replace them.

It's like a sinusoidal oscillation throughout the drive system.  In fact, if you watch the linear bearings carefully while it's printing (stare at the front linear bearing and blur your eyes slightly to bring the main quick change carrier into your view), you will see it oscillating like that.

My theory is that the linear bearings aren't rolling smoothly and is binding a bit until it builds enough spring tension in the rod to pull it forward.  I think one side of the sine wave printed is it binding up and the other side is it springing free.  It could even be a rod alignment issue too, I suppose.

Some print directions and angles seem worse than others and even at different spots on the bed makes a difference.  The flat spot I feel on the rods, for example, is near the center, interestingly about where the majority of the prints end up -- since the slicer defaults to centering the object.  Sometimes moving it to a different part of the bed helps, though not as much as I thought it would have.

Another general design flaw is that the X and Y axis are interlinked.  Any error on one will couple into the other.  The better design would be to keep them moving independent, like move the bed itself in/out for one axis and move the head back and forth for the other (like the RepRap Prusa design, for example).  But that's difficult to get in the footprint of the BoXZY and be hefty enough for milling, so I understand their design choices.

0
nox771

Member
Registered:
Posts: 52
Reply with quote  #3 
I get this also.  I've measured the pitch of the "waves" at 5mm, which is the same as the ballscrew pitch.  I'm wondering if the screw rods are rotating off-center, causing the carriage to "sway" as it moves along the axis.  Mine is slightly better on one axis than the other, but I see this in both directions.

I thought at first this might be something like off-center gear on the extruder, but it is very consistent along the length of the X/Y axis.  That is, if you build up a thin wall it lays directly on top of the layer below.  I would think if it was the extruder it would be out of sync with the axis travel.

When I get a chance I'm going to check the carriage, rods, and print head with a dial gauge to see if it oscillates along the axis.
0
dewhisna

Avatar / Picture

Senior Member
Registered:
Posts: 158
Reply with quote  #4 
Bingo!  I think you are on to something.  Probably wobble in the ballscrews relative to their parallel linear guide rods -- very similar to the Z-axis wobble problem on the Prusa i3.  Theirs is a normal threaded rod instead of acme threaded and is on the Z-axis instead of the X and Y axis, but watch the threaded rod in this clip as it wobbles relative to the guide rod:



The fix is to remove the T-nut's hard mount point to the axis assembly and replace it with a thrust plate that mounts the nut unattached from the axis assembly, letting it "float", with little arms around the linear guide rod to keep it from rotating so it can still provide thrust.  That way, rotation on the threaded rod will push the thrust plate up and raise the axis up and gravity handles the down direction as the thrust plate is lowered, and the threaded rod can wobble freely relative to the guide rod without moving the X-axis from side-to-side.  Here's what the fix piece for that printer looks like:

https://www.thingiverse.com/thing:1431036

Now, how to apply something similar to the BoXZY?  Hmmm....  Seems we'd have to make some sort of "soft mount" for the blocks on the ballscrews so that the linear guide rods can freely slide in and out as the ballscrews wobble relative to the corresponding parallel guide rod.

Too bad I don't have any UHMW lying around.  If I had a block of UHMW handy here, I'd see if I could design and machine such a block to replace the aluminum ones on the ballscrews...  I think that would be a good candidate material.  Nylon may work, but is probably too soft.


0
nox771

Member
Registered:
Posts: 52
Reply with quote  #5 
I've confirmed that the axis has wobble on it.  It's difficult to measure given what I have, but I managed to put a dial on the Y-bearing mount and measure some spots along the axis.

I read roughly 1.5mil wobble (TIR) near the ends by the bearings, and up to 4mil near the center.  Here is an example of a 3mil reading near center.

screenshot.581.jpg    screenshot.582.jpg 

I wasn't exactly sure how these axis are constructed, so I pulled it apart on the Y-axis.  This was more difficult than expected, due to them having put what looks like some adhesive between the pulley and shaft (wtf).  I had to use some old paint can openers to lever the pulleys off (pry both sides at once).  It was difficult enough that I even managed to bend one of the openers (I was about ready to go buy a gear puller).

screenshot.584.jpg 

Eventually I got the pulleys off, and once apart I found that the ballscrew threads have been ground down to an 8mm shaft.  It looks like this:

screenshot.583.jpg 

What you see here is the shaft (8mm, with some "residue" from the pulley on it), two bearings, a white plastic "thrust bearing", and the ballscrew threads.

The fit on bearings is tight, so the 1.5mil error I'm reading must be due to the machining quality.  The 4mil near center must be due to a bow along the length of the rod (kind of unbelievable given how short these rods are, but I've replicated this measurement several times).

The 4mil error near center is bad enough to cause consistent bad quality on prints.  Notice the perimeter and edges on this (the short axis is my Y-axis, the one I measured):

screenshot.585.jpg 

A 4 mil error is ~0.1mm which on a 0.4mm line is 1/4 the width (and it may be getting amplified by the smooth axis acting as a "pivot" for the carriage, not sure).  It is easily noticeable on all prints. 

Frankly I'm pretty annoyed by the boxzy team for this.  There is no way they haven't seen it.  Looking at the gallery shots it's obvious in many of the prints.  And given that they have yet to fulfill orders (laser panels, etc) to everyone I really doubt a fix from them is in the future.  The more I think about it, using the threaded rod as both a linear bearing for the carriage, as well as providing the force for movement is a serious design flaw.  It requires exact precision, which is obviously not what we got here.

I don't think there is an adjustment that can be made to these screws to compensate this.  I thought maybe if the error was due to machining on one of the bearings axis only, I could maybe loosen the bearing holder slightly (allow it to "sway" as the rod turned), and that might help.  But given that the rod is bowed I don't think this will help much (it might create a backlash effect).

I'm now very seriously considering pulling the ballscrew rods out entirely, and putting in an 1/2" acme rod with delrin anti-backlash nuts.  Instead of machining the rod down I would get a 1/2" bore pulley and bearings, and use the rod stock from the factory (cut to length of course).  This would hopefully eliminate error introduced by machining the rod down.  It would require new bearing mounts (I think existing mounts are too small for 1/2" bore bearing, but I might hunt around VXB and see if I can find a roller bearing with 1/2" bore that can fit).  However the printer as-is could probably print a bearing mount that could fit (wobbly bearing mounts of course [rolleyes]).

I've been searching around sdp-si.com, and I could probably get a belt/pulley system from there to connect the 1/2" shaft to the 1/4" stepper shaft. 

It's just highly irritating that we have to go to this amount of effort on a machine of this cost.  I've already gone through the DRV8825 replacement (12V supply is on the radar too).  It's like we are very slowly building a new machine (I suppose it's a hazard in this hobby).


0
dewhisna

Avatar / Picture

Senior Member
Registered:
Posts: 158
Reply with quote  #6 
Nice investigation work!  You've confirmed the wobble I visually see in mine.  And thanks for the photos of those details -- I was wondering how that was put together as well, but didn't want to tear into it yet.

A funny thought popped in my head while reading your post and your comment of: "The fit on bearings is tight, so the 1.5mil error I'm reading must be due to the machining quality.".  I had the thought that it must have been machined on another BoXZY and the error transcribed itself.  The line from the movie "Multiplicity" came to mind: "... when you make a copy of a copy, it's not as sharp as ... the original."  LOL

What I was thinking about trying on mine, after standing there staring at it for about ten minutes while it was printing and watching it wobble, is to leave the ballscrews as-is, and get a block of UHMW (ultra-high-molecular-weight polyethylene for anyone reading who may not be familiar with it) and machine replacement pieces for the aluminum mounts in the perpendicular rods where they connect to the mounts for the linear bearings on the far parallel guide rod.  The pieces like this in this picture here that I borrowed from the Dozuki site (i.e. machine new replacement UHMW pieces for the piece he is holding in his hand):

BoXZY-Perpendicular-Mount.jpeg 

Instead of a "hard" attachment for the perpendicular rod (the one that goes through the quick-change carriage) with the set-screw, instead, simply have a hole there where the rod is free to slide in-and-out.  That way, as the ballscrew on the far opposite side wobbles from side-to-side, this perpendicular rod is free to move in and out on the new mount, so it won't be trying to bow and move the carriage itself as it is doing today, and it will still support that end and allow forward/backward movement on that linear guide rod as it should.

I believe the UHMW would provide enough self-lubrication for the perpendicular rod to slide freely in-and-out and yet be stiff and strong enough to properly support things.  This may require a slightly longer perpendicular rod, so that it can extend completely through the UHMW piece  -- I haven't taken it apart yet to see just how deep it currently goes.  I'm tempted now to just remove the set-screw holding that rod, but fear that the aluminum would want to bind and that the current rod setup isn't deep/long enough.

And of course, make a new replacement piece for both the X and Y axis.  There will also need to be a nut put into the UHMW with a set-screw to hold the LM12UU bearing in place (which would then be mounted into the new UHMW piece), and there's probably some more details on the UHMW machining work that I haven't discovered yet, but I think this may be a viable solution -- and possibly an easy and somewhat inexpensive fix.

I ordered a 12"x12"x3/4" block of UHMW to experiment with -- at least I hope the 3/4" will be thick enough.  There was quite a price jump between the 3/4" and 1" blocks.  Though given how my experiments usually go, I'll get the 3/4" piece only to find out it wasn't thick enough and have to order the 1" anyway and end up spending double (oh well, we'll see)...

I have lots of concurrent projects currently in progress, so it may take me a while to machine the pieces and get them on (so if anyone else wants to give it a go too, go for it; you'll probably finish before I do).  I should actually be able to use the BoXZY's CNC for making them -- something I've been somewhat avoiding because of shavings destroying the bearings, etc.  UHMW, though, I don't think would hurt them too bad -- it's only a step away from teflon shavings...

Does anyone see anything wrong with my "fix" that I might have missed?  Or have any ideas or input on how to do it or do it better?  If so, please yell at me before I spin my wheels trying it...


0
nox771

Member
Registered:
Posts: 52
Reply with quote  #7 
Well, it is an interesting idea to try, and it may provide some improvement.

One thing to note is that the error is not only in the X/Y direction.  If you look closely at the part I printed, you'll notice there is error in the Z-direction also (it's visible on the starting perimeter, you can see a Z-error).  I think this is because the smooth axis is acting as a pivot (imagine one side being fixed, and the other side attached to a rotating wheel, with the carriage in the middle).

I think the two axes are both doing this which leads to a really complex motion.  Z-axis maybe not as much because it is a heavier screw (although it might also have a machining error).

I'm going to try and swap an axis for an 0.5in acme screw.  Mostly because I already have some of the parts (not all though).  I've located some bearings which should work with the current mounts (they are actually smaller OD, but I'm going to print a spacer).  I've also worked out the pulley requirements. 

I'm still considering what to do about the thrust bearing.  They were using the machined edge to act against the plastic "thrust bearing", but with a non-machined 0.5in ID that won't work.  I have an idea how to deal with this but I need to study it more.

One way or another I'm going to make this machine move in a straight line... or it's going to die trying [biggrin]


0
Gary

Avatar / Picture

Senior Member
Registered:
Posts: 116
Reply with quote  #8 
I also wonder if the screws could have been bent in shipping, what with all the stuff packed in around the gantry.
Could also be a reason for the early bearing failures - impact damage to the races in the bearings?
Or they are just crappy and bowed...
0
nox771

Member
Registered:
Posts: 52
Reply with quote  #9 
I have confirmed that the wobble in the X-Y movement is due to the ballscrews.  I've successfully swapped the Y-axis out for a 1/2" Acme screw with delrin anti-backlash nut and the wavy behavior is now gone in one direction.

I'm still refining my solution.  Once I get a final solution I'll post a parts list and what's involved (some parts can be printed, as you'll see below).  But for now here are my current findings.

What I did was obtain some new bearings, 1/2" precision Acme rod, delrin nut, new pulleys and belt, and printed some parts to make it all fit together.  The rod is specifically 1/2"-10 5-start, which is a 2-turns per inch rod (5-start means it has 5 interwoven threads).  I'm using this because I already had the delrin nuts for it (they are very common). 

I'm not machining the 1/2" Acme rod down, so all parts that fit the rod will be 1/2" ID.  For the bearings I'm using 1/2" ID, 3/4" OD bearings.  These are smaller OD than the 22mm original bearings, so I'm printing a bearing holder to fit inside the existing bearing mount (this enlarges the 3/4" OD to 22mm OD).  This ends up being somewhat thick and large because it serves a 2nd function in holding the top acrylic piece on the machine (the top screws into it).

This is what it looks like with two bearings in it (left) and mounted to machine (right) - note I'm bootstrapping this, so any parts I print now will be "wavy" due to the horrible machine inaccuracy:
screenshot.611.jpg  screenshot.626.jpg 

Those bearings hold the rod in place vertically, but they don't prevent axial movement.  For that I'm using thrust bearings mounted on the outside of the machine.  These are 1/2" ID needle bearings, like this (they use specialized thrust washers, the assembly consists of a sandwich of: thrust washer - bearing - thrust washer):

screenshot.606.jpg 

On the non-pulley side this is held in place with a 1/2" ID shaft collar (I had to use an extra washer on this side so the collar would clear the large hex heads):

screenshot.620.jpg 

On the pulley side, the pulley acts as the collar (so there is a thrust bearing behind the pulley).  For the pulley I'm using two new pulleys and belt.  This is a 5mm GT2 belt, using a 28 groove pulley on the 1/2" side, and a 14 groove pulley on the 1/4" stepper side.  This gives a 2:1 step ratio.  Mechanically given the 2-turn per inch rod, this gives 6.35mm movement per full stepper turn (1/4in per stepper turn).  This is a little faster than the ballscrew which was 5mm per turn.

screenshot.619.jpg 

As you can see, the belt is too large.  I was able to re-rig the tensioner mechanism to work (this required some additional parts), but even so the belt is a little loose.  I'm going to obtain a better sized belt.  FYI - the X-axis has NO tensioner mechanism, so one will have to be printed to deal with that (highly unlikely any new belt will be an exact fit).

I printed an adapter plate to attach the delrin nut to the existing ballscrew mount.  It looks like this:
screenshot.612.jpg   screenshot.613.jpg

It uses 4 captive hex nuts to hold the delrin nut in place.  I mount this to the ballscrew mount using the existing screws.  The whole assembly looks like this:

screenshot.616.jpg   screenshot.617.jpg 

So that's the basic assembly, now the real question is how well does it work.  In fact the motion is surprisingly smooth.  The stepper had no problem moving the Y-axis the full length of the machine all the way to the outer stops.  I was concerned the loose belt would give some backlash, but I don't notice it at all (maybe because the new belt is somewhat stiff yet).

So on the original ballscrews my print quality looked like this, wavy lines everywhere and lumpy circles:

screenshot.623.jpg 

Now with the Y-axis swapped I get this (notice the long dimension wavy is gone, see more below):

screenshot.622.jpg 

My X-axis still has a ballscrew, so it still has wavy behavior, but to me this confirms the problem and fix is to get rid of the ballscrews.

Comparison photos - Original ballscrew on left, Acme screw on right:

screenshot.624.jpg    screenshot.625.jpg   

It may not be obvious in the photos, but the wavy parts are really dimensionally garbage.  The wavy motion really throws off all accuracy.  I had to tweak multiple dimensions to get things like the bearings and screws to fit properly.  Even with the Y-axis fix, the X-axis is enough to prevent it from moving in a straight line diagonally.

I'm going to do a similar swap on X-axis, and I'll post when I get a result.

0
dewhisna

Avatar / Picture

Senior Member
Registered:
Posts: 158
Reply with quote  #10 
Nice!  So did you tweak your "steps per mm" settings on your firmware's EEPROM setup to match the new number of turns?  Or did you just rescale things in your model to match as a stop-gap?

You should be able to use the M205 and M206 g-code commands on the "manual control" tab to set them if you haven't already:
https://www.repetier.com/documentation/repetier-firmware/rf-installation/

It wouldn't hurt to eventually update your actual firmware too, so if your EEPROM ever gets nuked and reset back to the firmware defaults, you won't have to remember to change them again.

I was also curious if you think you'll be able to crank the top-end speed up a bit and/or acceleration values with these new Acme screws?

I ask because recently I got a used FlashForge Creator Pro printer "as-is for parts only" (factory return) on eBay for a couple hundred bucks.  Its only problem was a crashed Y-axis (it was jammed on the gear and the belt was knocked loose).  It only had 41h28m on its meter (so nearly new).

After fixing it and getting it to print, my first task was to print an exact model as I had printed on my BoXZY for comparison.  Other than changing the coordinates (it's 0,0 is the center of the build plate instead of the corner), I ran everything else exactly as-is for slicer settings (as I wanted to see how the BoXZY's settings ran on it) -- same infill, same speed settings, same slicer, same temperatures, same everything...

Well, it printed it in 1h08m (including the time to heat the bed and extruder, and it's a slow DC heater that takes a good 5-10 minutes to get to temp), and my BoXZY with the exact same print takes 3h38m not counting heating time (as I use a manual AC bed heater on the BoXZY and had it preheated already).

So 3.5 to 4 times the difference in speed -- and no wavy lines or anything either (it was a much cleaner print).

It would be nice to speed the BoXZY up a little in the process of fixing the wavy lines -- maybe not 3.5-4 times, but maybe a 20-50% speed improvement?  It may be something else for you to experiment with once you finish with your rework...

But great progress on solving this!

0
nox771

Member
Registered:
Posts: 52
Reply with quote  #11 
Yes I adjusted the steps-per-mm setting in EEPROM.  Since the Acme screw is inch based the number is fractional, but everything seems to print correctly (circles are still circular).

Specifically I'm using 1/8 microstep settings on X and Y, and these are the settings I have in EEPROM:

X-axis steps-per-mm: 320.0000    (ballscrew)
Y-axis steps-per-mm: 251.9685    (acme)

Right now I'm printing a new tensioner to fix the loose belt.  I think it is giving some overshoot (backlash) in the Y-axis due to being a little loose.  For the new tensioner I'm using one of the old "skate" bearings (8mm ID / 22mm OD) from the old ballscrew.

Once I get this movement cleaned up then I'll be able to adjust for extruder settings and such (right now everything is obscured by the bad movement).

As far as top-speed, it could be made larger, however from what I've been reading the limit is actually due to the max extruder speed.  The extruder can only physically push a max amount of filament out per given amount of time, based on the nozzle size (especially on bowden tube extruders).  I've read that if you exceed the capability of the extruder you will get insufficient fill (prints will look stringy).

One of my original purposes for the machine was to make mechanical parts, so I've been focused on trying to get accuracy up over speed.  At the moment everything gets lower priority than fixing these ballscrews.

A modification I'm thinking about farther out is to swap out the hotend for an all-metal with changeable nozzle.  All-metal requires an always running fan, and the challenge is coming up with something that fits in the quick-change metal tube. 

IMO they sort of built the machine backwards.  Instead of basing the quick change head on the Makita, they should have made an adapter for the Makita to a more generic quick change system.

So maybe replacing the entire carriage assembly is in the future also.  Despite Boxzy's carriage looking like it has some rigidity, it in fact doesn't.  In a milling scenario all the forces on the head have to be transferred down the 12mm rod to the leadscrew (2 out of the 3 rods in either the X or Y direction apply no force, they are only guide rails, so eventually force must transfer to the leadscrew, which for both X and Y is way on one side). 

It doesn't take a lot of force to flex that 12mm rod (eg. you can move the head to the diagonally opposite corner from home position, and push the head around by hand).  Fixing that problem would take a lot of rework (eg. dual leadscrews driving from both sides).  At that point it might be better to use Boxzy to make parts for a better machine.

Anyway one thing at a time.

0
nox771

Member
Registered:
Posts: 52
Reply with quote  #12 
New tensioner printed.  I guess those old bearings are good for something after all.

screenshot.627.jpg    screenshot.628.jpg 


0
dewhisna

Avatar / Picture

Senior Member
Registered:
Posts: 158
Reply with quote  #13 
Well, we are still a long way from exceeding the extrusion limit.  That other printer not only has the same 0.4mm size nozzle, but the motor is actually slightly smaller than that of the BoXZY's extruder (just slightly), and it was not only fast, but quite precise in its print.  I still have calibration work to do on it, but was quite pleased with that first print.

Though it's broke at the moment -- in my attempt at customizing my gcode generation for the dual extruder head, I accidentally missed an "if" statement in my code and created a file that, when I sent it to the printer, heated up one extruder ready for printing and then proceeded to switch tool selections and try to ram filament through the other extruder unheated...  I cleared the jam, but it isn't feeding -- I'm hoping just a set-screw on the gear or something, but am tearing it back apart again...

I actually get more stringing on my BoXZY from the fact that the BoXZY firmware doesn't use a PID control loop for the heat control -- it uses dead-band and a poorly tuned one at that.  Mine does good to hold the temperature within a 6 degree window, and running filaments too hot is an easy way to get stringing.

If I set my BoXZY to 230 degrees, it is actually running it between 232 and 237 degrees most of the time and is all over the map.  You don't notice it so bad on the computer display, due to the time lapse and sample rate for its update, but you certainly notice it if you've added the LCD to the BoXZY.

I too am looking at making an all-metal hot-end for the BoXZY.  I actually found an interesting all-metal dual extruder that will fit right inside the BoXZY quick-change holder.  (even found a 3-in-1 mixer that will fit it too)...  I will be working on building some custom print-heads for my BoXZY from them.

For the extruder fan, I'm planning to use the existing fan, but flip it around so that it pulls air up rather than blowing it down on your print and change its function setting in the firmware.  That way, it can run (as a correct extruder fan) and cool the extruder and not affect the print itself.  Then, I'll add a second fan as an actual print cooling fan (as it should be).  The dual extruder assembly I found shares a common heatsink block and can use a single fan for the heat-break, so I think it will work well for the BoXZY.

I'm also planning to spend some time doing PID loop calibrations and figure that out for both the new print-heads and the existing print-head (both in it's original stock form and with its fan flipped around to make a proper extruder cooling fan).  And then maybe it will be able to hold the heat constant.

But I agree with your plan of getting things working well first over speed...  Though I bet once you get everything printing perfectly, you'll want to speed it up a little.  I was here daydreaming about how nice it would have been for some of those really long prints that I've already done on my BoXZY (one that was 17 hours, one 26 hours, and another that was over 30 hours) had they been run on that other printer -- they would have been more like 4 hours, 6 hours, 7-1/2 hours, respectively, printing in under a day what took my BoXZY over 3 days to print...  ....  well, actually, it took my BoXZY longer than 3 days on those prints -- 3/4ths the way through two of them, I had an axis shift and had to scrap the entire print and start over...

I like the new tensioner design, BTW...

0
Drunken Boxzer

Avatar / Picture

Moderator
Registered:
Posts: 327
Reply with quote  #14 
Geeze, there's some serious hardware fix work going on in this thread! I, like pretty much everyone, also have the oscillating problem and have done nothing to fix it. I really just want it to work at this point and not have to continuously fiddle with it, but that's life (plus I get to keep learning all of this cool stuff you guys are doing to fix things).

I'm completely in on the changeable head idea, I have wanted to try different nozzles for a while now. Originally I had thought it would be cool to just make entire swap-able heads with quick attach power clips, but this sounds like a good idea too. I figured the quick attach would remove some of the risk of filament jam through the removable nozzle insert.
0
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.