Jump to content
Ultimaker Community of 3D Printing Experts

Recommended Posts

I know that the very early UMOs came with 8mm carriage shafts and then were switched over to 6mm, going forward.

 

Has anyone "gone backwards" on this?

 

It is all a balance between lightweight (for inertial benefit) vs stiffness (for positioning repeatability). The 6mm carriage shafts are very easy to deflect, especially near gantry center. This includes the Z direction from varying force imparted by the bowden/extruder drive; and also in the XY directions from inertial forces when changing the velocity/direction of the print head.

 

As you tune the machine for better and better print quality, your eyes get drawn to smaller anomalies in the printed surface. Even with slow outer perimeter speeds and acceleration, I see ripples at sharp direction changes. I also see slight Z banding that I believe is from the bowden tube changing the print head height slightly.

 

Supporting the print head with 8mm rods instead, will certainly increase stiffness. But also increase the inertial forces. My guess is that there is an optimal diameter... I just am not sure exactly where that lies. Has anyone put some serious thought into this?

 

My calculations show that a 6mm rod weighs ~60g, while 8mm is ~105g. That is a total weight gain of ~90g. This weight gain is distributed evenly along the length of the shaft though, so not all of it contributes to inertial forces that deflect the shafts. Some is put fairly directly into the XY blocks themselves. I think it would be fair to say only ~75% of the increased inertial force works to bend the shafts. Misumi does have hollow 8mm linear motion shaft. The ID is 3mm, which doesn't reduce the weight all that much. Each shaft would drop from 105g to 90g. Not a bad reduction though.

 

LM6UU = 8g   and  LM8UU = 13g. So another 10g of weight gain there, entirely on the head.

 

So, we have added ~55g of weight to the head. (using hollow shaft)

 

The next step would be to calculate and compare the increase in stiffness (reduced deflection), against the increase in acceleration forces, for a given realistic printer move...

Share this post


Link to post
Share on other sites

I've taken the other approach to solving the ripple at corners problem, by removing all the drive-belts and replacing with a helical-shaft direct-drive system. I'm now printing at 60mm/sec, and I've reduced the accelerations in Cura to 750 mm/sec**2 . Pics in the thread here, on page 9:

See what you think; I still have the 6mm shafts & bowden cable. I don't think they're the worst offender on UMO, I think the belts are.

 

I don't seem to have any Z layering problems - even when I still had belt drives. I put that down to changing the table support, I replaced 2 of the wooden arms with aluminium ones, so the HBP is now firmly located to the Z bearings via the aluminium bits.

Edited by andywalter
typo
  • Like 1

Share this post


Link to post
Share on other sites
22 hours ago, andywalter said:

I've taken the other approach to solving the ripple at corners problem, by removing all the drive-belts and replacing with a helical-shaft direct-drive system. I'm now printing at 60mm/sec, and I've reduced the accelerations in Cura to 750 mm/sec**2 . Pics in the thread here, on page 9:

See what you think; I still have the 6mm shafts & bowden cable. I don't think they're the worst offender on UMO, I think the belts are.

 

I don't seem to have any Z layering problems - even when I still had belt drives. I put htat down to changing the table support, I replaced 2 of the wooden arms with aluminium ones, so the HBP is now firmly located to the Z bearings via the aluminium bits.

 

 

Wow, that is quite a modification you made there! 

 

I'm surprised with the lengths you went to on the motion control, that you are still using the stock print head. Looking through your pictures, it seems like you have inconsistent extrusion and could benefit from some improvements there. Are you running the original extruder drive?

 

I'm still not so convinced that most ringing is caused by the belts. For sure, running the short belts contributes a good bit. But with a direct drive setup and GT2 belts (a big improvement over MXL), I don't think belt stretch is a significant factor. For example, with the printer stopped, I can push the head in an X or Y direction, and see noticeable deflection of the head form the 6mm rods bending, but essentially no movement of the XY blocks:

 

 

 

 

Also, the bronze 8mm bushings aren't a bad thing in my eyes either (so long as they fit well and are kept clean). Yes, they suffer from more static friction to overcome during starts/reversals. Once moving though, they move quite nicely (as long as they are not put into bind by an improperly squared machine). Since they are driven directly from the belts, their friction will not increase ringing. It increases the load on the steppers, and at the very worst, could affect micro stepping precision. But I think the amount of force they contribute is negligible on a well tuned machine. They are super easy to deform though. So, if your XY bocks clamp too aggressively, they will bind like crazy.

 

The print head bearings on the other hand can really mess with head precision and ringing, since any friction in them acts with inertial forces, making them worse. Since linear ball bearings will always have some radial lash, I think that minimizing the distance from the bearings, down to the nozzle tip is an important aspect. Radial bearing freeplay causes angular misalignment of the print head. So, the longer the head is, the greater the imprecision of the nozzle's XY position. My print head is nearly 1" shorter than the stock head! This is about the practical minimum height, as I am nearly out of Z travel:

 

image.thumb.png.f312ef6a49bf56a94ab5c88e74fc8a51.png

 

Edited by lars86
  • Like 1

Share this post


Link to post
Share on other sites

All I can say is, I think I've pretty much eliminated the ringing I used to see. Maybe reducing my acceleration has helped that a lot, but I also have far better precision than before, with each layer sitting very neatly exactly above the earlier ones.

 

Coincidentally, I've just repaired a Makerbot 2X, which I believe has the GT belts you mention - as far as I can read what's on the belts, they're Gates GT 2MR 283  which I can't see in their catalogue. The tooth shape looks much nicer than the MXL; MXL looks very trapezoidal, while GTs look more like the circular-ish ones designed to fit a curved-tooth pulley rather than the trapezoidal MXLs. When I worked with the trapezoidal belts years ago, they had a small amount of backlash as the belt teeth were a slightly loose fit in the pulleys; but the later HTD belts with a curved rounded profile were designed to sit tightly into the pulley at the tip of the rubber teeth, and as the belt transmitted greater power, the teeth progressively sheared a tiny bit more, bring more of the tooth into contact with the pulley. At max power you then had the entire tooth face with equal pressure at all points. The trapezoidals didn't do that, the teeth matched the pulley shape exactly, so with increasign power, the pressure increased most at the base of the tooth close to the fibreglass belt, and the tip of the tooth was under almost no stress. So these belts tended to rip the teeth from the belt, starting at the base of the tooth. The HTD belts had zero backlash in theory, and thanks to the better stress-distribution they had better power capability for the same dimensioned belt. The GT belts look to have this better profile, as far as I can see.

 

Having repaired the MakerBot, I ran the same resonance-test box-thing. The print speed was slower than I use, 40mm/sec, and there's definitely some ringing-like features showing. But the precision was good - the box has a wall 1.0mm thick, which gets printed as 2 walls of 0.4mm each with 0.2mm gap between; on my UMO the gap exists everywhere (except at corners); in the Makerbot the gap was there mostly, but a few places the 2 walls had bridged. But definitely better than I ever got using the original MXL belts.

 

I like the look of your printhead,  is it aluminium? Must admit, I hadn't given any thought to the nozzle wobbling around. I've got plenty of 8mm bearing shaft around, so I'm rather tempted now to try your 8mm idea with a custom printhead designed to minimise vertical distance. I'm not yet 100% happy with the circularity of my printed cylinders, there's some radial error which looks a bit like backlash, as I'm getting that slight flattening effect on the 0, 90, 180, 270 degree directions which is where the X & Y axes reverse direction. Maybe some of that is springing of the thin 6mm shafts as you suggest.

 

Is thare a cooling fan on your head, and do you direct any cooling air around the nozzle itself? I've got a rather badly made & bodged-up circular diffuser thing which could be improved.

 

My extruder is a replacement for the nasty original, uses 2 geared pinch-rollers. I think it's pretty accurate & repeatable. Yes there's a dodgy corner in my test-box print, not sure what caused that. I'm not sure I'm getting the layers to bond properly; last night's MakerBot was extruding very molten plastic at 230C, and getting the layers very well fused together. I tend to suffer delamination, so I acetone-vapour my prints later to melt the surface into a strong skin, but my print walls tend to be matt-black when just off the HBP, whereas the MakerBot print came off looking very glossy, which I think means I'm not yet melting my plastic as I should. But I print at 240C, and I'm using Voltivo ExcelFil  ABS, while the MakerBot is probably on MakerBot's own stuff. Could that be the difference? I might try setting my nozzle to run at 250C for most of my print, cooling it as I get to the last few layers as I don't want to bake any plastic left inside the nozzle after the print has ended. Even the MakerBot nozzles had some charcoal-like burnt-plastic when  cleaned them out, so >240C is to be avoided if poss, I think.

  • Like 1

Share this post


Link to post
Share on other sites
On 1/7/2018 at 8:43 AM, andywalter said:

All I can say is, I think I've pretty much eliminated the ringing I used to see. Maybe reducing my acceleration has helped that a lot, but I also have far better precision than before, with each layer sitting very neatly exactly above the earlier ones.

 

 

Reducing max acceleration has a huge impact on ringing. Remember, Force = Mass * Acceleration. Deflection (ringing) is proportional to force. You won't see a true comparison of your modification unless you match the print settings from your older tests.

 

 

On 1/7/2018 at 8:43 AM, andywalter said:

 

Having repaired the MakerBot, I ran the same resonance-test box-thing. The print speed was slower than I use, 40mm/sec, and there's definitely some ringing-like features showing. But the precision was good - the box has a wall 1.0mm thick, which gets printed as 2 walls of 0.4mm each with 0.2mm gap between; on my UMO the gap exists everywhere (except at corners); in the Makerbot the gap was there mostly, but a few places the 2 walls had bridged. But definitely better than I ever got using the original MXL belts.

 

 

 

Makerbots do use GT2 belts but they are direct drive printers. With all that extra weight on the print head, it's no wonder you see ringing.

 

 

 

On 1/7/2018 at 8:43 AM, andywalter said:

 

I like the look of your printhead,  is it aluminium? Must admit, I hadn't given any thought to the nozzle wobbling around. I've got plenty of 8mm bearing shaft around, so I'm rather tempted now to try your 8mm idea with a custom printhead designed to minimise vertical distance. I'm not yet 100% happy with the circularity of my printed cylinders, there's some radial error which looks a bit like backlash, as I'm getting that slight flattening effect on the 0, 90, 180, 270 degree directions which is where the X & Y axes reverse direction. Maybe some of that is springing of the thin 6mm shafts as you suggest.

 

 

My print head is a printed assembly. I'm working on a new revision of it to hold an E3D V6. I have been running a sort of Franken-hot-end with a stock heater block & nozzle, but a custom stainless heat break and heatsink. I actually designed an entire hot end from scratch and made all the parts on my CNC machines. But was shocked that my press fit connections between the heat break and heater block leaked molten filament! I never came back around to remake it.

 

If I were you, I would print some big circles at very slow speed (20mm/s). This should remove any inertial flex. If you still get the flat spots, then you likely just have too much backlash in your spiral followers.

 

 

 

On 1/7/2018 at 8:43 AM, andywalter said:

 

Is thare a cooling fan on your head, and do you direct any cooling air around the nozzle itself? I've got a rather badly made & bodged-up circular diffuser thing which could be improved.

 

My extruder is a replacement for the nasty original, uses 2 geared pinch-rollers. I think it's pretty accurate & repeatable. Yes there's a dodgy corner in my test-box print, not sure what caused that. I'm not sure I'm getting the layers to bond properly; last night's MakerBot was extruding very molten plastic at 230C, and getting the layers very well fused together. I tend to suffer delamination, so I acetone-vapour my prints later to melt the surface into a strong skin, but my print walls tend to be matt-black when just off the HBP, whereas the MakerBot print came off looking very glossy, which I think means I'm not yet melting my plastic as I should. But I print at 240C, and I'm using Voltivo ExcelFil  ABS, while the MakerBot is probably on MakerBot's own stuff. Could that be the difference? I might try setting my nozzle to run at 250C for most of my print, cooling it as I get to the last few layers as I don't want to bake any plastic left inside the nozzle after the print has ended. Even the MakerBot nozzles had some charcoal-like burnt-plastic when  cleaned them out, so >240C is to be avoided if poss, I think.

 

 

 

Yes. That photo doesn't show the fan configuration. I currently run two 30mm radial fans with light ducting to direct flow. I print almost exclusively high-end PLA and feel like cooling is very key to high quality prints. 

 

In addition to not wanting to breathe styrene gas, generally I find printing ABS a nightmare with very little upside. Modified PLAs tend to have better stiffness, less shrinkage, similar strength, great bed and inter-layer adhesion, etc. I just don't see the allure of ABS. Having prints that look good, but peel apart along layers does not interest me. Also, not being able to run adequate cooling means overhangs will be garbage.

 

If I were you, I would grab some Polymax or Protopasta PLA and do some test prints to get your printer dialed in. You can run much cooler nozzle temps 200 or less, full speed fans, prints stay stuck to even a cold bed (with blue tape). This will make sure you aren't trying to calibrate a printer off warped ABS.  $0.02

  • Like 1

Share this post


Link to post
Share on other sites

Hi Lars86,

 

Yes, I've been going the other way-, trying to decrease the weight and eventually the "moment of inertia" for the whole gantry (sure not easy, but fun, -because it is all about those small little details).. :)

Thought about hollow, or gun drilled shafts and even thin walled straight stainless tubes. The latter one is interesting, as a 8 mm shaft with wall thickness of 0.25 mm (same length) will weight only 11.3 Gram!

Seems unreal?  Well, had two 8 mm shafts scrapped due to too much play.

 

There's some smart things to do, -then we can use such shaft's in your UM's... So, what can we do???

 

Short 8 mm shaft weight; 131.4 Gram and length 337 mm.  (I have a precision weight that's calibrated!)

Long 8 mm shaft weight; 135.9 Gram and length 347 mm.

 

I'll fully agree with you that it is the self resonant frequency of the extruder head is what cause "most" of the ringing in the print and NOT the short belt. Well, -as I do not have an UMO (the grand father of UM2) but an "UM2 ext", I'll believe this issue is valid for all UM's -more or less if the speed is set to high.. Hint; there is a wider belt to try!

 

As I've never seen this test object Andy refer too, I just downloaded the square one (for resonance test) to see what this print would look's like.

To be fair, I downloaded the new Cura ver. 3.1.0. Used all default values, and let it go in order to see what it could be. Also, I just recently got a mail from S3D about there was a new ver. 4 out to download for me. So, downloaded this one as well. Did a monitor view of the model and saw that the writhed text was very bad "melted and over extruded". So, deselected the retraction and print looked good.

 

So I printed the object sliced by the two different slicers, Cura ver. 3.1.0 and Simplyfi3D ver. 4.0.1

 

The left one is S3D and right is Cura.

5a593400c3148_20180112_2254041.thumb.jpg.c8dc0cb34e484b7d6146c2cd3234f35e.jpg

 

Funny, but the one made with S3D is darker than the right done with Cura.

Both is smooth, but S3D has better smoothness..  However, Cura's measuring is way better; 50.0 mm X 50.0 mm and 16.0 mm height (with 0.1 res perfect!) and finally wall thickness 1.0 mm same accuracy (amazing)..

 

It's a little unfair over S3D as I did not do any correction for the line width etc. with S3d, but been using Cura as a standard for my print of course..

 

Some more details from the Cura printed object.

5a593c20bc46f_20180111_1038261.thumb.jpg.87561f9364aa9b48fea310b1f5f87b43.jpg

The texture inside the box.

 

5a593c8e866cd_20180111_1039201.thumb.jpg.ebaa84c0d5a66274fcf6cb3aa58fda6c.jpg

See the small bubbles in the middle, oscillation.. Cura 4.0.1

 

5a593cdfb6f75_20180111_1043461.thumb.jpg.9bb1acda8297964b94d8f35c7690ebfa.jpg

Corner with bubbles. Cura 4.0.1

 

5a593e8614cf7_20180111_1043461.thumb.jpg.8c4e761d8f815a5606bac6b748cc311e.jpg

Same here, but otherwise very good.. Cura 4.0.1    :)

 

Ps. All pictures is high res. so just zoom!

 

 

Keep up the good works (with an Ultimaker).

 

Thanks.

 

Torgeir.

 

 

 

Edited by Torgeir

Share this post


Link to post
Share on other sites

Hi Togeir, thanks for the reply!

 

It looks like the S3D print might be a much thicker layer height, which could explain the color difference.

 

Maybe we should start a thread to compare resonant test prints at comparable speeds and max accelerations. It would be interesting to benchmark different modifications.

Share this post


Link to post
Share on other sites

Hi, Torgeir (& Lars86),

Nice to see your 50mm square box. The "bubbly corner" is interesting, as you're only seeing that effect for half the way up the corner. I get a similar defect on my UMO, Cura 3.1.0.

 

Background digression, hopefully useful info for someone:

I've just discovered that my printhead isn't actually getting anything like hot enough. Thinks it's 250C but actual is 175C. That's why my prints are weak, and not glossy. So the above pics and tests will need re-doing when I've repaired it. Spent a couple of days analysing it & studying the heating controls, and my temp sensor chip AD597A is defective, replacement chip + spare on order (RadioSpares 230646). I'll make a better thermocouple & relocate the tiny preamplifier board holding the AD597A & put it near the motherboard, it has a built-in standard-temperature generator so "knows" what the room-temperature is; surely it can't be good to have it sitting above the heated nozzle head gently roasting if you leave m/c on preheat for any length of time, plus that board, wires, connectors, led, is all adding mass to printhead. I will now make a super-short, super-light thermocouple & take wires & preamp elsewhere. I prefer thermocouples, rather than thermistors. 

 

AD597 calibration accuracy is +- 4C at 60C, and it's meant to be laser-trimmed to be accurate to within 1C over range -20C to +350C (seems to be a bit of conflict in the docs as I read them) so, whatever, should be within +- 5C at the very worst. At any room temperature the machine is in. New thermocouple will be super-short bit of stainless steel tube maybe 4mm long, drilled, RadioSpares 6212158 K-type welded bead thermocouple wrapped in a tiny bit of Kapton tape to insulate, crimp the end up & poke it in the heater block 3mm dia hole. The Existing thermocuple is perfect, but I htink I can reduce mass with a new one. Arduino Mega 2560 is doing a perfect job, converts the preamp's output of, say, 2.40 Volts to temp=240C.

Couldn't be a simpler conversion, and the Arduino screen temp-display exactly matches my digital voltmeter measuring that voltage signal from preamp.

 

Back to the test-shape. I wonder if the bubbly corner only extending 1/2 way up the wall is connected with whether the print at the zig-zag bit fills in the hole (lower half), or leaves a groove (top half). It seems too much of a coincidence that both our prints have rather different top & bottom halves! The top half of all our prints looks better than the bottom half; there's no infill to be done at the zigzag place. So top half is just printing continuous walls sections, immediately before, and after, a Z-step. If, like me, you have Cura's Z-step Place set to Nearest, then the printer hardly needs to stop moving as it Z-steps. Even if you have a fixed place set, the nozzle is going to be in the right place after printing 2 walls, inner & outer.

 

On the bottom half, there's an infill to do. My print-sequence is infill-before-walls, inner-wall, outer wall, Z-step, infill, inner-wall, outer-wall, Z-step, ...

so I'm wondering if on the bottom half of the print, a side-effect of the extra extrude, & maybe a retract, to do the infill, means your PLA feed gets slowed-down a bit compared to the top half, and the hotter PLA is outgassing/boiling/oozing a bit/whatever at that corner? Maybe my UMO does my inner wall, outer wall, gets to a uter corner, then traverses across to the other side of box to to the infill, and it's that traverse which is tearing/pulling my almost-solid too-cool filament with it? I will know more once m/c is fixed & I'll watch the print sequence like a hawk next time !!!

 

Info about a MakerBot 2X Replicator making the same box resonance-test-shape using default 0.2mm settings:

I'm fixing one for a school nearby, just used default ABS at 250C. That's sliced with MakerBot Desktop, and has a lower-half wall section immediately after the filled-in zig-zag bit, where the inner & outer walls look as if someone has touched the wall with sticky tape & dragged the surface a bit, leaving it a rough & a bit like sandpaper. A sort of "torn" roughness. I think this m/c is set too hot, and during the slower zig-zag infill maybe the plastic is oozing out a bit, or overheating a bit more as the tip stays in that place for maybe a second more, and immediately after that, when it does the inner wall, the plastic temp is a bit wrong, maybe too runny, and it's tearing rather than melting & blending in nicely. But the print on this m/c comes out really strong, and nice & glossy, as it should. My UMO obviously is nothing like that, and that's what made me realise my nozzle temps have been slowly falling over time without me realising what the problem was! So I now have a few reels of other ABS which weren't working at all, while the Voltivo ExcelFil I have been managing with (very high quality stuff, no nasty phthalates etc in it by the way!) presumably has a slightly lower melting point;  Voltivo Black ABS is recommended to print between 211C & 219C; I've been forcing it out at 176C or close to that !!!

Share this post


Link to post
Share on other sites
15 hours ago, Torgeir said:

... I'll fully agree with you that it is the self resonant frequency of the extruder head is what cause "most" of the ringing in the print and NOT the short belt... Hint; there is a wider belt to try! ... Torgeir.

 

Torgeir, a few thoughts:

Wider belts would be nice to try, but the increases the effective mass of the printhead! And the wider pulleys add inertia.

 

I have my original UMO belts, just weighed these on my calibrated lab scales, all belts are 6.5mm wide:

Long belt B300MXL weighs 5.900 gr

Short belt "B100 MXL" weighs 1.744 gr lets call it 1.75 gr.

 

Each axis has 2 x Long +  1 Short so that's =13.55gr mass. Call it 13.6 gr.

 

I also have the aluminium pulleys, so could do an accurate inertia calculation on those. But it's probably good enough to simply weigh them; as most of the inertia comes from the outer dia where the speed is close to the belt speed, and that's also where most of the mass is, this will give a good estimate of the equivalent mass, after converting rotational moment of inertia to linear. Each axis has 6 pulleys, these weigh 35.004 gr,  so approximate this to 35 gr linear-equivalent mass.

13.6 + 35 = 48.6 gr  so call it =49 gr for belts+pulleys.

 

So a double-width set of pulleys & belts would add about another 45 gr to this. (screw-hub & grubscrew not duplicated)

 

I've weighed the plywood blocks plus bolts & nuts + oilite bearings which clamp drive-belt to the 6mm rods, 2 of these per axis. 6 blocks + bits weigh 54 gr. This mass only affects the belt-resonance deflection, not the 6mm shaft deflection.

 

So we're now up to 49+54 = 103 gr for the belt load. Next we need mass of the central printhead thing, plus equivalent mass (inertia) of the long 8mm rods, mass of 6mm rods, and inertia of the stepper rotor. Anything I've missed? I need to calc the elasticity of the 6mm shaft under bending; knowing this, and the printhead mass, we can calculate the out-of-straight deflection distance of the middle of the 6mm shaft when the stepper is accelerating at whatever mm/sec**2 we set in Cura's speed controls. That should be the max deflection we get from resonance due to a 90 degree sharp corner, no? Should we measure the belt-stretch & tooth-deflection elasticities as well, because those have to accelerate the mass of the 6mm shaft + blocks, as well as the printhead, so I think the drive belts might be as springy as the 6mm shaft maybe?

 

If we measure the peak-to-peak distance of the resonance ripples after a sharp corner, and we know the travel speed, we can calculate he resonant frequency. That should line up with some combination of masses and elasticities form the above bits.

 

There could be several things happening:

1) Suppose the step-motor has a large inertia, and a smallish stationary holding torque, and everything after that (belts, oilites, 6 & 8mm shafts, printhead) are of negligible mass and infinitely stiff in comparison; then we should see a resonant frequency & amplitude matching a step-motor oscillation.

2) Suppose the step-motor is perfect (doesn't oscillate & stops perfectly at the sharp corner with zero overshoot and has infinite stationary holding torque). Also suppose the drivebelts & pulleys are perfectly rigid & zero mass; then we are left with the 6mm shaft bending acc to the mass of some of the 6mm shaft itself ?half? plus mass of printhead. Hopefully would have different freq & amplitute to case 1) above.

3) Suppose step motor is perfect, 6mm shaft is perfectly rigid, and it's the belts that are flexy. We should see a different frequency, I hope, based on mass-equivalent of 2x8mm shaft + 5 pulleys + 2.5 belts, then the added mass of 6mm shaft + blocks + printhead, all oscillating, depending of stiffness of these belts. (Motor pulley is perfectly stationary, and we should probably halve the short-belt mass, as one end is locked motionless!).

 

4) Real life will be some superimposed combination of the above,  or maybe the belts are flexy-but-well-damped, so it's all a bit too complicated to calculate.

 

What I can do is print the testcase on my UMO, and use identical feeds & speeds as you use on your m/c; maybe if we crank the speeds up high we can get the ripples really high amplitude; then if I have ripples with stiff drive & 6mm shafts, and you have none with belts & stiff 8mm shafts, that might really prove what's the wobbly bit. I think we need to agree a set of speeds & accelerations & get matching Cura profiles as far as possible.

 

It might be possible to design a test-shape that deliberately provokes the worst accuracy & maximum rippling, once we know some frequencies & stiffnesses! Sounds like a lot of fun!!

 

Edited by andywalter
more info & fix typos.

Share this post


Link to post
Share on other sites
19 hours ago, andywalter said:

Hi, Torgeir (& Lars86),

Nice to see your 50mm square box. The "bubbly corner" is interesting, as you're only seeing that effect for half the way up the corner. I get a similar defect on my UMO, Cura 3.1.0.

 

Hi Andy,

 

Thanks. Very busy for the time being, so reply will always take some time.  :)

 

Just come to think about what caused this small bubbles, -realized that this is the corner were the extruder head make a 180 turn -at speed. So this seems to be the reason for this remains at the corners.. 

 

 

19 hours ago, andywalter said:

Background digression, hopefully useful info for someone:

I've just discovered that my printhead isn't actually getting anything like hot enough. Thinks it's 250C but actual is 175C. That's why my prints are weak, and not glossy. So the above pics and tests will need re-doing when I've repaired it. Spent a couple of days analysing it & studying the heating controls, and my temp sensor chip AD597A is defective, replacement chip + spare on order (RadioSpares 230646). I'll make a better thermocouple & relocate the tiny preamplifier board holding the AD597A & put it near the motherboard, it has a built-in standard-temperature generator so "knows" what the room-temperature is; surely it can't be good to have it sitting above the heated nozzle head gently roasting if you leave m/c on preheat for any length of time, plus that board, wires, connectors, led, is all adding mass to printhead. I will now make a super-short, super-light thermocouple & take wires & preamp elsewhere. I prefer thermocouples, rather than thermistors. 

 

AD597 calibration accuracy is +- 4C at 60C, and it's meant to be laser-trimmed to be accurate to within 1C over range -20C to +350C (seems to be a bit of conflict in the docs as I read them) so, whatever, should be within +- 5C at the very worst. At any room temperature the machine is in. New thermocouple will be super-short bit of stainless steel tube maybe 4mm long, drilled, RadioSpares 6212158 K-type welded bead thermocouple wrapped in a tiny bit of Kapton tape to insulate, crimp the end up & poke it in the heater block 3mm dia hole. The Existing thermocuple is perfect, but I htink I can reduce mass with a new one. Arduino Mega 2560 is doing a perfect job, converts the preamp's output of, say, 2.40 Volts to temp=240C.

Couldn't be a simpler conversion, and the Arduino screen temp-display exactly matches my digital voltmeter measuring that voltage signal from preamp.

 

Great way of doing it. K element (Chromel/Alumel) have been used for a very long time in the aviation, esp. for ITT/EGT temperatur monitoring and is still used here. Believe it or not, but today PT100 is used as an absolute temp sensor in modern jet/turbine engines. 

By the way, the PT100 is now available as a (-200 to 600) deg. Celsius version in Class A. 

 

 

19 hours ago, andywalter said:

Back to the test-shape. I wonder if the bubbly corner only extending 1/2 way up the wall is connected with whether the print at the zig-zag bit fills in the hole (lower half), or leaves a groove (top half). It seems too much of a coincidence that both our prints have rather different top & bottom halves! The top half of all our prints looks better than the bottom half; there's no infill to be done at the zigzag place. So top half is just printing continuous walls sections, immediately before, and after, a Z-step. If, like me, you have Cura's Z-step Place set to Nearest, then the printer hardly needs to stop moving as it Z-steps. Even if you have a fixed place set, the nozzle is going to be in the right place after printing 2 walls, inner & outer.

 

Same thing happen (bubbly lines) were the little squared box made inside the lower part of the "U" profile of the wall. (Can only be seen on the Cura print).

Only seen the half way up the wall. (Have a look into the sliced model using the monitor.)

 

19 hours ago, andywalter said:

On the bottom half, there's an infill to do. My print-sequence is infill-before-walls, inner-wall, outer wall, Z-step, infill, inner-wall, outer-wall, Z-step, ...

so I'm wondering if on the bottom half of the print, a side-effect of the extra extrude, & maybe a retract, to do the infill, means your PLA feed gets slowed-down a bit compared to the top half, and the hotter PLA is outgassing/boiling/oozing a bit/whatever at that corner? Maybe my UMO does my inner wall, outer wall, gets to a uter corner, then traverses across to the other side of box to to the infill, and it's that traverse which is tearing/pulling my almost-solid too-cool filament with it? I will know more once m/c is fixed & I'll watch the print sequence like a hawk next time !!!

 

I'll do the same thing, but hawk eyes..  hmm not any more for me. :)

  

 

19 hours ago, andywalter said:

Info about a MakerBot 2X Replicator making the same box resonance-test-shape using default 0.2mm settings:

I'm fixing one for a school nearby, just used default ABS at 250C. That's sliced with MakerBot Desktop, and has a lower-half wall section immediately after the filled-in zig-zag bit, where the inner & outer walls look as if someone has touched the wall with sticky tape & dragged the surface a bit, leaving it a rough & a bit like sandpaper. A sort of "torn" roughness. I think this m/c is set too hot, and during the slower zig-zag infill maybe the plastic is oozing out a bit, or overheating a bit more as the tip stays in that place for maybe a second more, and immediately after that, when it does the inner wall, the plastic temp is a bit wrong, maybe too runny, and it's tearing rather than melting & blending in nicely. But the print on this m/c comes out really strong, and nice & glossy, as it should. My UMO obviously is nothing like that, and that's what made me realise my nozzle temps have been slowly falling over time without me realising what the problem was! So I now have a few reels of other ABS which weren't working at all, while the Voltivo ExcelFil I have been managing with (very high quality stuff, no nasty phthalates etc in it by the way!) presumably has a slightly lower melting point;  Voltivo Black ABS is recommended to print between 211C & 219C; I've been forcing it out at 176C or close to that !!!

 

I'm sure your printer will be as good as UMO should be, I've seen UMO prints looks as good as the UM2 can do, but need a little more "attention" over time.

 

I'm still only using PLA, as it is predictable and very good for prototyping models etc.

Later I'll need something that is stronger and not as brittle as PLA.

 

Thanks

 

Torgeir

 

 

 

Share this post


Link to post
Share on other sites

Hi Andy,


 

I've been investigated some over those ringing marks we often see on our prints, if the speed is set to high. -I've learned that this was due to flexing short belts.

However, after some experimenting I found that the flexing mostly come from the extruder itself, and is caused by the weight and flexible attachment of the extruder assembly.

Another contributor to this is the CT (center of gravity) of the extruder as it is located under the center of the gantry.

When our printing object have details as writing letters or repetitive pattern the printer head (extruder) may be very close to resonance, even if the print speed is «relatively» slow.

This ringing often propagating to the metal housing for the two fans and the tiny metal walls in here start making additional noise.

Due to the spring effect in the gantry setup, the print head continues for a short while (muted oscillations) and appears as an echo in the print object as it disappears.

In order to prevent this to happen we can reduce the print head weight, move the center of gravity up toward the center of the two 6 mm shafts and this will reduce the ringing (or say increasing the resonant frequency). However, if we increase the printing speed enough, the ringing problem will reappear.

As I see you already have 6.5 mm belts installed on your printer, -my printer have only 6.0 mm GT2 belt, but can take 6.5 mm belt..

The printer head in the UMO seems to be much lighter than the one used in UM2, are you using hollow shafts?

UM2 head weight complete is about 127 Gram, and the two 6 mm shafts (127 / 117.4) Gram.

The four 8 mm shafts weight is; (2 X 131.4 / 2 X 135.9) Gram. Belts and pulleys approx. same as yours. 

 

Thanks

Torgeir.

Share this post


Link to post
Share on other sites

Torgeir, my 6mm shafts are the originals, as is my UMO head. I think I've pretty-much eliminated any resonance in my m/c from the belts, as I don't use them! The spiral shaft+plastic nuts is a very rigid system. I suspect hollow 6mm shafts would be too flexible. Hollow 8mm shafts does sound interesting though.

 

Very interesting about the head C og G being so low; I'm starting to think about redesigning this lot using 8 mm shafts placed really low, as close to the table as possible. Then the nozzle+block could sit in the corner of the 2 shafts, rather than dangling vertically below the intersection area. My HBP can easily go 40mm higher. Maybe it's time to think about making one of those extruded-aluminium frames, and also about keeping the heat in the print-area like MakerBot does with the lid & enclosed sides. Sometimes I've wrapped a towel carefully round my UMO on large prints to try to keep everything warm, and it's not the ideal solution!

 

In Cura I've reduced all the accelerations to 750 mm/sec**2. This may be the biggest factor in reducing my ripples. But I do need to improve my positional accuracy, and the flexy head you've discovered won't be helping my accuracy.

 

 

Share this post


Link to post
Share on other sites
On 1/13/2018 at 7:57 AM, andywalter said:

 

Torgeir, a few thoughts:

Wider belts would be nice to try, but the increases the effective mass of the printhead! And the wider pulleys add inertia.

 

I have my original UMO belts, just weighed these on my calibrated lab scales, all belts are 6.5mm wide:

Long belt B300MXL weighs 5.900 gr

Short belt "B100 MXL" weighs 1.744 gr lets call it 1.75 gr.

 

Each axis has 2 x Long +  1 Short so that's =13.55gr mass. Call it 13.6 gr.

 

I also have the aluminium pulleys, so could do an accurate inertia calculation on those. But it's probably good enough to simply weigh them; as most of the inertia comes from the outer dia where the speed is close to the belt speed, and that's also where most of the mass is, this will give a good estimate of the equivalent mass, after converting rotational moment of inertia to linear. Each axis has 6 pulleys, these weigh 35.004 gr,  so approximate this to 35 gr linear-equivalent mass.

13.6 + 35 = 48.6 gr  so call it =49 gr for belts+pulleys.

 

So a double-width set of pulleys & belts would add about another 45 gr to this. (screw-hub & grubscrew not duplicated)

 

I've weighed the plywood blocks plus bolts & nuts + oilite bearings which clamp drive-belt to the 6mm rods, 2 of these per axis. 6 blocks + bits weigh 54 gr. This mass only affects the belt-resonance deflection, not the 6mm shaft deflection.

 

So we're now up to 49+54 = 103 gr for the belt load. Next we need mass of the central printhead thing, plus equivalent mass (inertia) of the long 8mm rods, mass of 6mm rods, and inertia of the stepper rotor. Anything I've missed? I need to calc the elasticity of the 6mm shaft under bending; knowing this, and the printhead mass, we can calculate the out-of-straight deflection distance of the middle of the 6mm shaft when the stepper is accelerating at whatever mm/sec**2 we set in Cura's speed controls. That should be the max deflection we get from resonance due to a 90 degree sharp corner, no? Should we measure the belt-stretch & tooth-deflection elasticities as well, because those have to accelerate the mass of the 6mm shaft + blocks, as well as the printhead, so I think the drive belts might be as springy as the 6mm shaft maybe?

 

If we measure the peak-to-peak distance of the resonance ripples after a sharp corner, and we know the travel speed, we can calculate he resonant frequency. That should line up with some combination of masses and elasticities form the above bits.

 

There could be several things happening:

1) Suppose the step-motor has a large inertia, and a smallish stationary holding torque, and everything after that (belts, oilites, 6 & 8mm shafts, printhead) are of negligible mass and infinitely stiff in comparison; then we should see a resonant frequency & amplitude matching a step-motor oscillation.

2) Suppose the step-motor is perfect (doesn't oscillate & stops perfectly at the sharp corner with zero overshoot and has infinite stationary holding torque). Also suppose the drivebelts & pulleys are perfectly rigid & zero mass; then we are left with the 6mm shaft bending acc to the mass of some of the 6mm shaft itself ?half? plus mass of printhead. Hopefully would have different freq & amplitute to case 1) above.

3) Suppose step motor is perfect, 6mm shaft is perfectly rigid, and it's the belts that are flexy. We should see a different frequency, I hope, based on mass-equivalent of 2x8mm shaft + 5 pulleys + 2.5 belts, then the added mass of 6mm shaft + blocks + printhead, all oscillating, depending of stiffness of these belts. (Motor pulley is perfectly stationary, and we should probably halve the short-belt mass, as one end is locked motionless!).

 

4) Real life will be some superimposed combination of the above,  or maybe the belts are flexy-but-well-damped, so it's all a bit too complicated to calculate.

 

What I can do is print the testcase on my UMO, and use identical feeds & speeds as you use on your m/c; maybe if we crank the speeds up high we can get the ripples really high amplitude; then if I have ripples with stiff drive & 6mm shafts, and you have none with belts & stiff 8mm shafts, that might really prove what's the wobbly bit. I think we need to agree a set of speeds & accelerations & get matching Cura profiles as far as possible.

 

It might be possible to design a test-shape that deliberately provokes the worst accuracy & maximum rippling, once we know some frequencies & stiffnesses! Sounds like a lot of fun!!

 

 

 

 

 

 

Very thorough!

 

A couple of thoughts:

 

- "Jerk" in Marlin is not the second derivative of velocity. It is a maximum allowable instantaneous change in velocity. It certainly isn't ideal, but with the limited computational power of 8-bit boards, and no provision for allowable path deviation to smooth path motion, it is a necessary evil. You can imagine the jerk setting can have a lot to do with balancing motion stutter (as the planner tries to move around a faceted model and connect all the "dots"), and accelerations above the max acceleration setting. I guess you would have to compare the forces exerted by two cases: max acceleration & maximum stepper torque (a jerk move); to see which is higher.

 

- It seems like you could save a lot of time by deciding which parts deflect a relevant amount. Pretty easy to estimate max force exerted on the print head. Then compare elastic elongation of the belts, with point loaded deflection of a shaft.

Share this post


Link to post
Share on other sites

Pointer to append I made, planning to replace 6mm rods by a construction of thin-but-long steel rulers! Almost twice the stiffness & half the mass of the 6mm rod, that's my target.

Details here: https://community.ultimaker.com/topic/1114-throw-away-your-short-belts-direct-drive/?do=findComment&comment=197722

Edited by andywalter
improve link

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Our picks

    • Taking Advantage of DfAM
      This is a statement that’s often made about AM/3DP. I'll focus on the way DfAM can take advantage of some of the unique capabilities that AM and 3DP have to offer. I personally think that the use of AM/3DP for light-weighting is one of it’s most exciting possibilities and one that could play a key part in the sustainability of design and manufacturing in the future.
        • Like
      • 3 replies
×

Important Information

Welcome to the Ultimaker Community of 3D printing experts. Visit the following links to read more about our Terms of Use or our Privacy Policy. Thank you!