Hi, author of the Vibration / Ripple / Shadow / Ghosting test thing here. Funny thing, i accidentally came across this vid on youtube and immediately spotted one thing - the changed orientation. The initial orientation it loads in is actually correct, just not intuitive :) The idea is that the letter with the ghosting tells you which *axis is exhibiting vibration* on direction change, not which axis the face of the "cube" is parallel to. So X is actually on the Y side, but if it's exhibiting ghosting, that means the X axis is responsible (ripples are oscillations along the X axis in that orientation).
So what he is telling me in this video is that the ghosting is not from sharp turns like a car fishtailing. It is from acceleration like it's backfiring as it goes in a straight line less and less as it gets further away from an object. But it still doesn't make sense to me, why would that only happen after a sharp corner and not all over the object. I still think I agree with you that it is in the wrong position and is from x axis motion while the y axis should be the only side moving.
@@chromeaviation You don't seem to know what acceleration is. When there's a direction change (any acute corner) then the printer needs to decelerate and accelerate. The more open the corner is, the less the printer needs to slow down, thus the effects of acceleration are less visible.
@@alejandroperez5368 ya i kindof have it figured out now. Still a little confused why you don't see ringing during decelerations. But I get the acceleration ringing.
@@chromeaviation you get ringing when there's a direction change, and that involves 2 axis, think about the nozzle wanting to keep it's previous trajectory and it oscillates until it stops, that's when the ringing ripples cease. Ringing on only 1 axis movement by accelerating doesn't just happen, or at least I'm not aware of it. But you do get other problems related to nozzle pressure delay (check Linear Advance).
You Can also turn the object a little in cura so it does not move only along one axis but always along both axis at the same time. That also improves your printquality in the corners.
you referring to s^2 as per second per second threw me of at first, but because of that i thought about it and understand now why s^2 is being used for acceleration amounts. And yes great vid as always
Had a scroll down the comments to see if this was mentioned, but it appears that nobody had considered the bed. As a cantilever component it's far more likely to be moving due to vibration than the gantry. While your testing is valid, the steel Rods result in more ghosting, it's not the Rods that cause it. It's because the bed is an supported cantilever and the heavy moving mass is causing that to wobble. That'd be my theory anyway. Thanks for the video
No More BS Please removing the weight that was causing the inertia will obvisly stop the ripples that doesn't mean it's the problem, that's why slowing the acceleration down also worked. Bolt the whole machine to a concrete floor and use a fixed bed instead of a floating one will also stop it. if all the motion control components are solid and have no play then the wobble has to come from somewhere else. Simple to test wobble the table when its printing and see what happens. Exactly the same issue with cnc routers and lasers have if the bed or the machine tabke isn't solid.
@@NoMoreBsPlease Correlation vs. causation (Freakonomics) Gantry mass *induces* bed vibration, in turn, resulting in ghosting, etc. Correlation: Cantilever bed vibration. Resolution: Modify bed design. Don't stop at the first apparent cause. Sometimes you'll find that the tail does, indeed, wag the dog. Going further, as the clip of the water glass shows, the mass of the steel rods induces vibration. These vibrations are transmitted to the build plate and manifested as ghosting in the test object. Cause: Gantry mass. Resolution: Reduce gantry mass. Result: Fast print speed and rapid acceleration with high precision.
Oh, and you've absolutely nailed it w/ respect to acceleration / jerk being the key configuration variable to affect this. I once asked Richard Horne to test a printer using this, and he didn't, berating me (in a vid, nonetheless) for suggesting printing such a small object as a test, as it's "clearly too small for the printer to reach the speeds where that would matter" (or something to that accord). Which is clearly incorrect, as it's the speed vector *change* that's responsible, not the speed (i.e. acceleration and jerk).
Neat video! I like your content. Also, damping not dampening. No one is getting wet here. And underdamped oscillation for that matter. Harmonic oscillations relative to rigidity of the gantry design and induced harmonics from the printer movement. Try a steel tube as opposed to a steel rod as well...
You could also get rid of the ripples by lowering your jerk setting to something like 0.5 instead of 10, but you also wouldn't be able to achieve the same top speeds during a small print like that. Print that same cube at 200x200mm and lower jerk to 1 on both x and y so the print has time to reach max speed on each line and those ripples probably won't show on either axis.
Fantastic. I've been trying to eliminate my corner bulge for ages and found that decreasing the acceleration made it worse. I thought i was going crazy because everyone was telling me accel was too high. This fits perfectly with my experience as well.
The load on these axis is very little, if it is properly setupand high quality carbon rods it shouldnt wear that fast at all. Epoxy is actually quite hard and, like i said, the load that these bearings have to transmit is very low.
Slowing down accelleration requires retuning "advance" or pressure compensation as well or you get those bulges on the corners... There is no reason to get those with correct setup where extrusion will compensate for pressure in melt chamber.
Nice video. It was cool to see the differences and carbon fiber tubes would really pimp up my printer. Added to my long todo list. FWIW - I have a stock Anet A8 which has steel rods and a direct drive extruder so there's a bit of mass moving around. I didn't look at weight factors but I did observe a substantial reduction in ghosting, similar to your qualitative assessment, by tightening the X and Y drive belts. I printed tensioners for X and Y and now they are taught. When I had ghosting I could literally see a wave propagate along the belts with each direction change. Now the belts just move back and forth with no judder at all.
i have never owned a 3d printer but i am pretty sure its about the resonance and harmonics of the thing. putting something floppy on the nozzle would proly help alot
On my P3Steel Toolson Mk2 with steel rods I do not get the letter's ghosting at 60mm/2. I use sinter bronze bearing. But those corner bulges. They also occur with your carbon fire tubes. They can be reduced by printing the exterior perimeter with around 25mm/s while maintaining a speed of 60mm/s for the inner perimeters.
Why not use steel tubes? Thin wall precision tubes are available, and a tube is always going to be stiffer than a rod of the same diameter and material, as long as the walls aren't too thin.
Also, one interesting tidbit from my own battle with ripple back in the day... I had a Y axis with a movable bed, and an X gantry. You'd expect the Y axis to exhibit the effect more, right? The bed being heavier and all (at least mine was, it was a solid aluminium plate with another plate on that, then glass, plus a bunch of V wheels for the Makerslide to keep the thing solid). Well, it turned out the X axis was exhibiting way more ripple than Y. It's not just the mass, it's actually a combination of mass and rigidness. If you're familiar with the spring oscillations in physics / mechanics, there are three components to it - the dampening factor (friction), the mass, and the spring constant. The spring constant, in this case, is the rigidness of the whole system, and the belt tension. You can reduce the frequency of the oscillations (making them invisible, or less noticeable) by reducing belt tension, as long as that doesn't induce slop. That was the case with my Y/bed axis. You can also increase the dampening in some cases, by, for example, tightening the V wheels to the rail (or whatever the equivalent for your motion system of choice). This makes the steppers work harder, but there's always some margin left if they're not overheating to begin with.
This analysis of the ringing of your 3D printer is really cool. Reducing the moving mass is always a good idea. However, I am puzzled if it was possible to counteract the ringing in software: It should be possible to measure the ringing amplitudes at different positions of your print bed. In the next step, one can generate a model for the ringing amplitude. Finally, the G-code generator can be modified to add a ringing with inverted amplitude to the G code. As a result, the ringing should be canceled out. This should allow significantly higher print speeds without changing the hardware. So it would allow for higher printing speeds for cheap printers but could also further speed up fast printers. I am new to 3D printing, and I am not sure if this is very difficult to implement, but as far as I see the G-code is very modular.
Its is oriented the right way X is supposed to be on the Y axis, and Y on the x axis. to make it easier to see witch axis to work whit. This is because movements on the X axis is making the rippel on y axis and movement on y axis is causing the rippel on the x side. It's been my experience anyways but i might be wrong :-D
I've watched all your videos and am very impressed with the science you apply to all your testing. Indeed, weight is the enemy of all 3D printers and lugging around steel is a very bad thing to do. On Prusa i3 type printers, I've never understood why people put glass on the bed because this can weigh around 300g in itself. I've printed your new X carriage design for an E3D hotend and been very impressed with the results. What kind of time improvement do you get with the carbon fibre tubes whilst still maintaining an equivalent print quality comparable to the steel rods?
Thanks for the video.. Very interesting. I'm building a hypercube based machine, but have started making changes to the design and so it'll barely resemble yours. Because of this I've decided to build two! One much like yours: lightweight, open and good for PLA/PETG and my own machine that will be slower, hotter and enclosed with a direct drive head. Hopefully better for difficult materials. I had no luck with the carbon tubes I bought. They were poor in both linearity and diameter all the way along. I've since found a better source so will get some more after this video. I also bought a Titan Aero because I prefer direct drive; but this video has convinced me that I should also make the lighter machine a bowden type and give it a go. More videos like this one please! Thanks again.
Okay, Tech2C. You've sold me. I've been - theoretically - building a 3d printer with a 500mm X 500mm bed and a 450mm lift. But it's taken so long for parts to arrive from China to Canada - about 14 weeks, on average - that I'd kind of lost interest. I had been considering 8mm hardened steel rods for the 500mm span, but I saw your earlier video on the CF tubes you installed, so I held off. The trouble with a 500mm bed is that you need a 600mm tube; and you can get 600mm hardened steel tubes, but you can't get 600mm carbon fiber tubes. The best deal I could find was 5 - 1000mm tubes... delivered, to Canada, for $75.00. Ouch. So, I ordered 5 4mm X 8mm X 1000mm pultruded tubes. Pultruded has less crush and torsional resistance than woven, but it is stronger at 90 degrees - if you try to bend it, for example. And the walls are 2mm thick, so it ought to work. Geez. You have been absolutely the worst thing for my bank account. But now I have enough carbon fiber tubing to build a bunch of printers. Sigh.
haha I thought your printer would have been finished ages ago! Oh well the fun is yet to begin when you actually start printing... For such a large span of the CF tubes I'm glad you chose a thicker version. I haven't had shipping delays from China to Oz in a awhile, but 14 weeks sounds ridiculous!
Yep. Used to be about 6 weeks, but I ordered some stuff around December 5... and that's when the delays started. Canada Post used to provide tracking support for China registered air mail, but apparently they've gotten into a spat. CanPost now prominently states on their website that they no longer provide tracking support for China registered mail... and it appears that they're also 'storing' China shipments for a couple months in Customs. Or so I suspect. I had also finished the design of the printer, right down to every single part, bearing and bushing... and through sheer, unforgivable, utter stupidity I formatted the drive and lost it all. Tried recovery software, but got nothing. And that was just about the time that Autodesk announced that they were pulling the plug on 123Design, and that I'd have to shift over to Fusion360. Sigh. Just got a little bummed out for a while. But you got me back on track, so I've ordered the carbon rods - and the 8mm bronze bushings to go with them. So, lets see. It's the middle of May...I'm sure they'll be here by late August, early September. Sigh. At least it gives me a bunch of time to learn Fusion 360 and redesign everything!
I am in that same situation myself and worse. My print area will be 300x300x360ish which is already huge enough to require 450 watt mains powered heated bed. I don't like having to do the mains powered heated bed at all.
If you can find a discarded desktop computer somewhere, you can cannibalize the power supply. Failing that, you can buy one for maybe $50 or so. They usually can supply 400 - 500 watts on the 12 volt rail, so they make a great cheap power supply. No need to go with a mains supply, a cheap fiberglass mat heater (usually red in color) is typically around $10 - $20 and has resistances designed for 12 volts. You can use them with a plate of glass, or bond them to the underside of a sheet of aluminum. I don't like having mains voltage in a system, either. A decent 500 watt computer power supply should get you rockin pretty quickly. Use a solid state relay triggered from the 12V heater output on your motherboard to switch the power supply in and out. Good luck!
charles worton you don't need it to be mains powered for that size, i am using a 300x300 24v silicone heater that is rated at 360w but i have been running it at 400w+ using a 24v 20A supply and it got hot enough to crack piece of ordinary glass that i was using originally because i had my thermistor in the wrong place so it was seeing about 70 degrees when it was really at 140°c, i have since moved the thermistor to the center of the bed and I'm now using a tempered piece of glass and it will go over 100°c easily.
its not unusual that a caliper is plus/minus 0,1mm, for a more exact measurement you'll need a micrometer. Even if the retailer said that the rod is "exakt" 10mm, axels are always plus/minus a few 1/100mm.
now what is oscillating? the belt? the motors? the frame? i think the MK2 has defaults at 1500 acceleration, does not seem to ring as bad as this.... perhaps the frame just isn't solid. some aluminum panels covers might help a bunch. maybe just wood or acrylic would make it stronger even.
brilliant.. i do with you would check more how jerk affects the quality (it is the the speed to which it decelerates when changing direction) - low acceleration, high jerk?
I am a bit confused. The rippling should occur due to the sharp cornering while newton told the weight to go straight on, resulting in an oscillation on the face after the corner. This means that weight moving in the Y direction should affect the faces perpendicular to it. The heavy Y-gantry should affect the face that is marked X in your print because you rotated it. Then Y oh Y is there still an obvious difference in ghosting where it shouldn't be?
Because theory and practice sometimes don't match, well at least in this example. The ripples you see originate from stationary to acceleration moves only. Hence why you don't see ripples on both sides of the Y letter.
Thanks for this test, very useful to see the differences in quality control settings. Would you ever consider testing/reviewing the various changes that people have made to the Hypercube design? Perhaps even include some of them into your designs if they turn out to improve the build or print quality?
I note that you did rotate the object (cube) to make it 'flat' to the y axis. What would happen if you only rotated it 45 degrees? so it now sits diagonal on the build plate?
Great explanation, thank you. Would you mind watching the MakersMuse video and making the tolerance test that he created? I have printed a few with my HyperCube and found a 0.20 tolerance to be the best I can get at this moment. I can imagine that printing it with Carbon Fiber rods will give better results, but I only have anodized Aluminum here.
Ray Jones thay are "only " 70 € I got one that will do 0.001 mm but I use it for school so if you don't have one 0.05 mm is good enough for your purpose
Another factor here you could experiment with for the bulging corners is called linear advance. Thats its name in marlin firmware but it's also been implemented in some others like smoothie. Basically it gradually lowers the pressure in the nozzle when you approach a point of zero extrusion so that none will ooze out. Might be worth looking into or even doing a video on!
Hey mate, I tried lin_advance, but I didn't notice an improvement at any speed or acceleration. I'd set it up in configuration_adv.h and also using M900 commands, and could hear the extruder motor skipping if the K factor was over 200, but it just didn't improve the bulging corners. I did read that this feature is best suited to direct drive, and that Bowden may be less useful? Bit of a let down. Have you used it?
Hmm.. did you follow the calibration instructions on the documentation website? I mean doing the test cube with the specific values it specifies and making sure you have wipe and coasting and all of those other features turned off. I haven't personally used it yet because prusa marlin hasn't implemented it. However, I have seen really really nice results from prints done on a tevo tarantula using lin_advance which is bowden
Linear advance should be more useful with bowden setups. You might get better results in direct drive though only because you're already 90% of the way there as far as having better pressure control due to having a short filament path. The whole purpose of linear advance is to model the actual pressure in the hotend to provide consistent extrusion due to the springy nature of the filament as it's pushed and compressed into the hotend by the extruder. The longer the path is from the extruder to the hotend, the more compression force will have to build up in order to actually push the filament through the hotend, so then when you stop (or slow down) extruding, you get extra extrusion at the end as that compression force is then released out the nozzle. Linear advance attempts to regulate this by slowing down the extruder or even reversing it if needed, essentially decoupling it from the X and Y axes.
Very interesting demo -- quick question though. Have you ever calculated the total weight of the gantry? i.e. sum up the carriage, extruder, fan, bushings, Y-rail LM8 bearings and connectors, etc.
Thanks for this video, it provided a lot of really helpful information for me to address my bulging corner issue. It's surprising to me that slicers still don't take the corner slowdown into account and change the extruder feed rate to compensate. I imagine that someone could create a slicer that would adjust the speed, acceleration and extruder feed rate dynamically during the print to optimize speed without causing any artifacts. The slicer has all of the information it would need about the geometry of each line or curve it will have to draw and should be able to use physics to determine where to slow down. When I'm driving my car I would never use cruise control on curvy country roads because I'd fly off the road on the corners. Instead I adjust my speed and acceleration on the fly. 3D printers should be able to apply the same logic.
I noticed that your cork insulation appears to have an air gap from the heated plate. If you use 100% silicone sealant between the cork and heated bed surface it works excellent as an adhesive. I put about 10-12 small globs and press the cork against the globs to flatten them out. Let it sit overnight and you are golden!
Pressure advance FTW.... low acceleration, sharp corners, even the 1500mm/2 accel can benefit (usually does). Though it's another dimensions to tweak per filament. unfortunately it ain't so good on a bowden.
Replacing steel rod is good for reducing vibration. But I think it is bad for accuracy. Carbon or Aluminum pipe is no precision. It is necessary to think about a structure to suppress vibration of the bed.
I have 8mm steel rods for the gantry and I am printing with 220c, 150mms speed, and acceleration 3000mms and I am having a little rippling effect! and btw, my printer is literally dancing when printing with those speeds, not just shaking and my neighbor told me that he is hearing my printer's vibrations in his room :lol
Heyya, great video, thanks :D I'd just like to clarify, that it's just your printer interprets the X axis as Y. In cura, and most other software, X is on the Red axis. you can verify this by clicking the move tool.
The letter X should be facing the Y axis, since each face is measuring the motion perpendicular to that face. You should have seen the difference on the Y face.
oof. I'll be using (if I actually make one) steel chrome plated ones cause they're the only ones that work with a ball bearing from what i've heard 0_0 brass bushings and everything are just too much effort to get working properly
This video is now outdated cura is using M204 Sxxx.xx for acceleration in later versions but marlin is using M204 Pxxx.xx Rxxx.xx Txxx.xx (print, retraction, travel) so you can set the values in cura but they wont affect the acceleration in actuall printing. To get this to work with cura modify the start Gcode to include: M204 P1500 T1500; Set acceleration Keep the settings in cura set as this will keep your print time estimates more accurate and only increases gcode file size which is not a big deal. As for corners becoming out of shape you need to enable linear advance in marlin then you can use a K value to fix the issue. Looking forward to a update video.
Iv thought about it before with the directions of the x y it’s looks wrong in the slicer but I think you should put the x on the y axis because because the face on the y axis will only show ghosting in the x axis as in side to side in relation to the y axis. Damm Iv not said that well
Good vid. Have you tested this the other way where you use the lightest of the rods, then turn up the speed to see the limitations? My opinion because I'm impatient like most people that have a backload of stuff to print and just want to go faster.
Perhaps the print head should have an accelerometer on it and the system go through a setup phase so that it can detect those issues and subsequently negate them through better print head control, i..e through dynamic software.
12:01 There is most definitely ghosting on the print on the right...You can see it to the right of the Y. Watch it on a desktop, not a phone. YES the print looks better, but ghosting (The repeating pattern of the Y) is clearly visible.
there's a lot to be said about the way cameras see the world vs the human eye. I'd wager that the difference is imperceptible even with the tongue test and that what you see is most likely an effect of the contrast and slight difference in angles to the camera.
Video after Video, your content is teaching me new information! Before I use my new printer I have been learning as much as I can to improve the quality of my prints and printer. You have much experience and I appreciate your channel. Thank you, Subscribed
Interesting video. I bet the steel rods are close to advertised size. Digital calipers cannot replace high quality micrometer for really fussy measurement. Even then it takes some practise to get the feel of using a micrometer. Lots of beginners try to use them like a C clamp : )
Excellent video. Now this means I will tinker with my settings, a bit more. Before, was not sure what settings affected what outcome. Now a little less confusing. I also want to make parts that fit together, and this does require much more knowledge. Please, please continue on this theme, I am sure you will make us all pro, very soon, regards
I appreciate all the effort in the testing, but the whole time I was looking at the edges of all the 60mm/s prints like, is he going to address what a mess these edges are? Only the 30mm/s had decent edges. It may be a solvable problem, but until it's solved there's no point cutting weight to go high speed when all the high speed prints look like that.
Ahhhhhhh don't press on scales like that! If it says you went over the limit which you did at least once you risk damaging the scale and getting bad readings, you should re-call it and becareful not to do that :D
i bought the 10mm CF tubes from AliExpress and the ridiculously expensive (for what they are) plastic bushes from RS components and was expecting a snug, maybe slightly overly tight fit. They're as loose as fuck!
This is very puzzling, because inertia only applies to moving parts. The only time when those rods are actually moving, is when the gantry moves up, so that would be the Z direction, I suppose that the ghosting that is found is caused by the poor fit of the bearings to the steel rods. Another issue with aluminium or carbonfiber rods is the poor hardening, causing the bearings to wear out channels on the rods, and causing ghosting free a while on the Y axis. The solution would be using (expensive) hardened aluminium rods by Igus or drylin bearings on softer rods, but because those drylin bearings need to be perfectly parallel, this willing also introduce wearing and ghosting after a while,.so the solution for that would be auto-alighning drylin bearings,.which are hard to come by,, and expensive.
Agreed. Here they were fixed at 10mm/s. Any slower and the bulging corners appear no matter what speed you choose. Too fast and I get ripples with CF tubes ;)
Excellent vid, thank you. I have the same affects you mention on prints on my Wanhao Select Maker Plus. I though these issues were common quality issues because of the type of printer we have to live with. Do you have any opinions on the Carbon Fiber tubes as far as thickness vs rigidity and quality. Seems like a very inexpensive upgrade. Thanks.
really interesting. Out of interest can you give any example of the print time difference for the CF vs Steel rods with the accel and speed on the steel set to levels to give similar quality?
Printing the ripple test cubes at 60mm/s with 1500mm/s2 accel, print time was 13:21 for all rails. At 1000mm/s2 and 500mm/s2 accel, print time increased to 14:04 and 15:34 respectively.
Very interesting test. I didn't know you could get Carbon Fibre rods for printers. I have the Anet A8 and am currently modding it to improve it a bit. Next mods will be changing from direct drive to bowden, and an e3d v6 head. Maybe some dry bearings as well. Not sure what else.
I'm thinking of getting a 3D printer and i was wondering if you're supposed to sand finished products? I've been looking at some videos about issues to see if im willing to go through the trouble of fixing them but most issues i see seem to be fixable with a bit of sandpaper.
![Jack Harlow - Hello Miss Johnson [Official Music Video]](http://i.ytimg.com/vi/Q86_nlRoIGw/mqdefault.jpg)
Hi, author of the Vibration / Ripple / Shadow / Ghosting test thing here. Funny thing, i accidentally came across this vid on youtube and immediately spotted one thing - the changed orientation.
The initial orientation it loads in is actually correct, just not intuitive :)
The idea is that the letter with the ghosting tells you which *axis is exhibiting vibration* on direction change, not which axis the face of the "cube" is parallel to. So X is actually on the Y side, but if it's exhibiting ghosting, that means the X axis is responsible (ripples are oscillations along the X axis in that orientation).
exactly what I said
So what he is telling me in this video is that the ghosting is not from sharp turns like a car fishtailing. It is from acceleration like it's backfiring as it goes in a straight line less and less as it gets further away from an object. But it still doesn't make sense to me, why would that only happen after a sharp corner and not all over the object. I still think I agree with you that it is in the wrong position and is from x axis motion while the y axis should be the only side moving.
@@chromeaviation You don't seem to know what acceleration is. When there's a direction change (any acute corner) then the printer needs to decelerate and accelerate. The more open the corner is, the less the printer needs to slow down, thus the effects of acceleration are less visible.
@@alejandroperez5368 ya i kindof have it figured out now. Still a little confused why you don't see ringing during decelerations. But I get the acceleration ringing.
@@chromeaviation you get ringing when there's a direction change, and that involves 2 axis, think about the nozzle wanting to keep it's previous trajectory and it oscillates until it stops, that's when the ringing ripples cease.
Ringing on only 1 axis movement by accelerating doesn't just happen, or at least I'm not aware of it. But you do get other problems related to nozzle pressure delay (check Linear Advance).
You Can also turn the object a little in cura so it does not move only along one axis but always along both axis at the same time. That also improves your printquality in the corners.
you referring to s^2 as per second per second threw me of at first, but because of that i thought about it and understand now why s^2 is being used for acceleration amounts. And yes great vid as always
11:28 Well hello there Dave Jones
Lol, yep, EEVblog dude here
thank you for this design! I am working on mine right now. just need to get the software up and running.
You are a life saver. I've been scratching my head at this exact problem for days. Going to adjust my acceleration and print speed now!
Had a scroll down the comments to see if this was mentioned, but it appears that nobody had considered the bed. As a cantilever component it's far more likely to be moving due to vibration than the gantry. While your testing is valid, the steel Rods result in more ghosting, it's not the Rods that cause it. It's because the bed is an supported cantilever and the heavy moving mass is causing that to wobble. That'd be my theory anyway. Thanks for the video
Adam Meadows It's the rods, or the CF wouldn't have fixed it! If it was the bed, the rippling would still be there.
No More BS Please removing the weight that was causing the inertia will obvisly stop the ripples that doesn't mean it's the problem, that's why slowing the acceleration down also worked. Bolt the whole machine to a concrete floor and use a fixed bed instead of a floating one will also stop it.
if all the motion control components are solid and have no play then the wobble has to come from somewhere else. Simple to test wobble the table when its printing and see what happens.
Exactly the same issue with cnc routers and lasers have if the bed or the machine tabke isn't solid.
@@NoMoreBsPlease Correlation vs. causation (Freakonomics)
Gantry mass *induces* bed vibration, in turn, resulting in ghosting, etc.
Correlation: Cantilever bed vibration.
Resolution: Modify bed design.
Don't stop at the first apparent cause. Sometimes you'll find that the tail does, indeed, wag the dog.
Going further, as the clip of the water glass shows, the mass of the steel rods induces vibration. These vibrations are transmitted to the build plate and manifested as ghosting in the test object.
Cause: Gantry mass.
Resolution: Reduce gantry mass.
Result: Fast print speed and rapid acceleration with high precision.
This could be tested by printing each test at different Y positions. Prints further away from the Y support should have more ringing.
Oh, and you've absolutely nailed it w/ respect to acceleration / jerk being the key configuration variable to affect this. I once asked Richard Horne to test a printer using this, and he didn't, berating me (in a vid, nonetheless) for suggesting printing such a small object as a test, as it's "clearly too small for the printer to reach the speeds where that would matter" (or something to that accord).
Which is clearly incorrect, as it's the speed vector *change* that's responsible, not the speed (i.e. acceleration and jerk).
Neat video! I like your content.
Also, damping not dampening. No one is getting wet here. And underdamped oscillation for that matter.
Harmonic oscillations relative to rigidity of the gantry design and induced harmonics from the printer movement. Try a steel tube as opposed to a steel rod as well...
You could also get rid of the ripples by lowering your jerk setting to something like 0.5 instead of 10, but you also wouldn't be able to achieve the same top speeds during a small print like that.
Print that same cube at 200x200mm and lower jerk to 1 on both x and y so the print has time to reach max speed on each line and those ripples probably won't show on either axis.
Fantastic. I've been trying to eliminate my corner bulge for ages and found that decreasing the acceleration made it worse. I thought i was going crazy because everyone was telling me accel was too high. This fits perfectly with my experience as well.
Also how do you stop non steal rods from wearing. I can imagine that carbon tubes have only a resin surface which would wear very quickly?
The load on these axis is very little, if it is properly setupand high quality carbon rods it shouldnt wear that fast at all. Epoxy is actually quite hard and, like i said, the load that these bearings have to transmit is very low.
Thank you for this video. Your explanations are clear and concise and still helping your fellow makers 4 years after publishing.
Slowing down accelleration requires retuning "advance" or pressure compensation as well or you get those bulges on the corners...
There is no reason to get those with correct setup where extrusion will compensate for pressure in melt chamber.
Nice video. It was cool to see the differences and carbon fiber tubes would really pimp up my printer. Added to my long todo list.
FWIW - I have a stock Anet A8 which has steel rods and a direct drive extruder so there's a bit of mass moving around. I didn't look at weight factors but I did observe a substantial reduction in ghosting, similar to your qualitative assessment, by tightening the X and Y drive belts. I printed tensioners for X and Y and now they are taught. When I had ghosting I could literally see a wave propagate along the belts with each direction change. Now the belts just move back and forth with no judder at all.
Yes belt tension helps reduce, or at least change the frequency/repetition of the ripples. On the A8 the rails aren't moving so no need for CF.
Thanks for taking one for the team and doing this invaluable research!
i have never owned a 3d printer but i am pretty sure its about the resonance and harmonics of the thing. putting something floppy on the nozzle would proly help alot
On my P3Steel Toolson Mk2 with steel rods I do not get the letter's ghosting at 60mm/2.
I use sinter bronze bearing.
But those corner bulges. They also occur with your carbon fire tubes. They can be reduced by printing the exterior perimeter with around 25mm/s while maintaining a speed of 60mm/s for the inner perimeters.
I've learned more watching this channel than any of the others on the subject. Thank you!
Why not use steel tubes? Thin wall precision tubes are available, and a tube is always going to be stiffer than a rod of the same diameter and material, as long as the walls aren't too thin.
Also, one interesting tidbit from my own battle with ripple back in the day... I had a Y axis with a movable bed, and an X gantry. You'd expect the Y axis to exhibit the effect more, right? The bed being heavier and all (at least mine was, it was a solid aluminium plate with another plate on that, then glass, plus a bunch of V wheels for the Makerslide to keep the thing solid).
Well, it turned out the X axis was exhibiting way more ripple than Y. It's not just the mass, it's actually a combination of mass and rigidness. If you're familiar with the spring oscillations in physics / mechanics, there are three components to it - the dampening factor (friction), the mass, and the spring constant. The spring constant, in this case, is the rigidness of the whole system, and the belt tension. You can reduce the frequency of the oscillations (making them invisible, or less noticeable) by reducing belt tension, as long as that doesn't induce slop. That was the case with my Y/bed axis.
You can also increase the dampening in some cases, by, for example, tightening the V wheels to the rail (or whatever the equivalent for your motion system of choice). This makes the steppers work harder, but there's always some margin left if they're not overheating to begin with.
You sir have quite exactly Bob Odenkirk's voice and accent. It was like Saul Goodman was teaching me ho to cut corners... on a 3D print!
This analysis of the ringing of your 3D printer is really cool. Reducing the moving mass is always a good idea. However, I am puzzled if it was possible to counteract the ringing in software:
It should be possible to measure the ringing amplitudes at different positions of your print bed.
In the next step, one can generate a model for the ringing amplitude.
Finally, the G-code generator can be modified to add a ringing with inverted amplitude to the G code.
As a result, the ringing should be canceled out.
This should allow significantly higher print speeds without changing the hardware. So it would allow for higher printing speeds for cheap printers but could also further speed up fast printers.
I am new to 3D printing, and I am not sure if this is very difficult to implement, but as far as I see the G-code is very modular.
Its is oriented the right way X is supposed to be on the Y axis, and Y on the x axis. to make it easier to see witch axis to work whit.
This is because movements on the X axis is making the rippel on y axis and movement on y axis is causing the rippel on the x side. It's been my experience anyways but i might be wrong :-D
I've watched all your videos and am very impressed with the science you apply to all your testing.
Indeed, weight is the enemy of all 3D printers and lugging around steel is a very bad thing to do.
On Prusa i3 type printers, I've never understood why people put glass on the bed because this can weigh around 300g in itself.
I've printed your new X carriage design for an E3D hotend and been very impressed with the results.
What kind of time improvement do you get with the carbon fibre tubes whilst still maintaining an equivalent print quality comparable to the steel rods?
Interesting, didn't consider a glass bed would have that drawback. Thanks for pointing that out.
Thanks for the video.. Very interesting.
I'm building a hypercube based machine, but have started making changes to the design and so it'll barely resemble yours. Because of this I've decided to build two! One much like yours: lightweight, open and good for PLA/PETG and my own machine that will be slower, hotter and enclosed with a direct drive head. Hopefully better for difficult materials.
I had no luck with the carbon tubes I bought. They were poor in both linearity and diameter all the way along. I've since found a better source so will get some more after this video. I also bought a Titan Aero because I prefer direct drive; but this video has convinced me that I should also make the lighter machine a bowden type and give it a go.
More videos like this one please! Thanks again.
Haha very good. You can still choose direct drive, just slow the print down is all ;)
You might want to experiment with linear advance. It is designed to fix the extrusion pressure of printing fast with slower Accel.
Okay, Tech2C. You've sold me.
I've been - theoretically - building a 3d printer with a 500mm X 500mm bed and a 450mm lift. But it's taken so long for parts to arrive from China to Canada - about 14 weeks, on average - that I'd kind of lost interest. I had been considering 8mm hardened steel rods for the 500mm span, but I saw your earlier video on the CF tubes you installed, so I held off. The trouble with a 500mm bed is that you need a 600mm tube; and you can get 600mm hardened steel tubes, but you can't get 600mm carbon fiber tubes. The best deal I could find was 5 - 1000mm tubes... delivered, to Canada, for $75.00. Ouch.
So, I ordered 5 4mm X 8mm X 1000mm pultruded tubes. Pultruded has less crush and torsional resistance than woven, but it is stronger at 90 degrees - if you try to bend it, for example. And the walls are 2mm thick, so it ought to work.
Geez. You have been absolutely the worst thing for my bank account. But now I have enough carbon fiber tubing to build a bunch of printers.
Sigh.
haha I thought your printer would have been finished ages ago! Oh well the fun is yet to begin when you actually start printing...
For such a large span of the CF tubes I'm glad you chose a thicker version. I haven't had shipping delays from China to Oz in a awhile, but 14 weeks sounds ridiculous!
Yep. Used to be about 6 weeks, but I ordered some stuff around December 5... and that's when the delays started. Canada Post used to provide tracking support for China registered air mail, but apparently they've gotten into a spat. CanPost now prominently states on their website that they no longer provide tracking support for China registered mail... and it appears that they're also 'storing' China shipments for a couple months in Customs. Or so I suspect. I had also finished the design of the printer, right down to every single part, bearing and bushing... and through sheer, unforgivable, utter stupidity I formatted the drive and lost it all. Tried recovery software, but got nothing. And that was just about the time that Autodesk announced that they were pulling the plug on 123Design, and that I'd have to shift over to Fusion360. Sigh. Just got a little bummed out for a while. But you got me back on track, so I've ordered the carbon rods - and the 8mm bronze bushings to go with them. So, lets see. It's the middle of May...I'm sure they'll be here by late August, early September. Sigh. At least it gives me a bunch of time to learn Fusion 360 and redesign everything!
I am in that same situation myself and worse. My print area will be 300x300x360ish which is already huge enough to require 450 watt mains powered heated bed. I don't like having to do the mains powered heated bed at all.
If you can find a discarded desktop computer somewhere, you can cannibalize the power supply. Failing that, you can buy one for maybe $50 or so. They usually can supply 400 - 500 watts on the 12 volt rail, so they make a great cheap power supply. No need to go with a mains supply, a cheap fiberglass mat heater (usually red in color) is typically around $10 - $20 and has resistances designed for 12 volts. You can use them with a plate of glass, or bond them to the underside of a sheet of aluminum. I don't like having mains voltage in a system, either. A decent 500 watt computer power supply should get you rockin pretty quickly. Use a solid state relay triggered from the 12V heater output on your motherboard to switch the power supply in and out. Good luck!
charles worton you don't need it to be mains powered for that size, i am using a 300x300 24v silicone heater that is rated at 360w but i have been running it at 400w+ using a 24v 20A supply and it got hot enough to crack piece of ordinary glass that i was using originally because i had my thermistor in the wrong place so it was seeing about 70 degrees when it was really at 140°c, i have since moved the thermistor to the center of the bed and I'm now using a tempered piece of glass and it will go over 100°c easily.
"Carban Fibah"
Great video, thanks for clarifying this “issue” for me......
its not unusual that a caliper is plus/minus 0,1mm, for a more exact measurement you'll need a micrometer.
Even if the retailer said that the rod is "exakt" 10mm, axels are always plus/minus a few 1/100mm.
Smear a tiny bit of silicon on the bottom of your heatbed and bond that cork so it won't sag loose and not insulate properly
next steps: external cooling, smaller probe... titanium screws?
now what is oscillating? the belt? the motors? the frame? i think the MK2 has defaults at 1500 acceleration, does not seem to ring as bad as this.... perhaps the frame just isn't solid. some aluminum panels covers might help a bunch. maybe just wood or acrylic would make it stronger even.
I assume the head glides via roller bearings.? Is there some kind of adjustment so to remove clearance? Was that done when swapping the various rods?
brilliant.. i do with you would check more how jerk affects the quality (it is the the speed to which it decelerates when changing direction) - low acceleration, high jerk?
I am a bit confused. The rippling should occur due to the sharp cornering while newton told the weight to go straight on, resulting in an oscillation on the face after the corner. This means that weight moving in the Y direction should affect the faces perpendicular to it. The heavy Y-gantry should affect the face that is marked X in your print because you rotated it. Then Y oh Y is there still an obvious difference in ghosting where it shouldn't be?
Because theory and practice sometimes don't match, well at least in this example. The ripples you see originate from stationary to acceleration moves only. Hence why you don't see ripples on both sides of the Y letter.
Thanks for this test, very useful to see the differences in quality control settings.
Would you ever consider testing/reviewing the various changes that people have made to the Hypercube design? Perhaps even include some of them into your designs if they turn out to improve the build or print quality?
I note that you did rotate the object (cube) to make it 'flat' to the y axis. What would happen if you only rotated it 45 degrees? so it now sits diagonal on the build plate?
Tell me Y!
Ain't nothing but a heartache...
Tell me Y!
Ain't nothing but a mistake...
.
.
.
.
.
.
.
.
.
.
.
yeah. backstreet boys ref
Great explanation, thank you.
Would you mind watching the MakersMuse video and making the tolerance test that he created?
I have printed a few with my HyperCube and found a 0.20 tolerance to be the best I can get at this moment.
I can imagine that printing it with Carbon Fiber rods will give better results, but I only have anodized Aluminum here.
Measure the shafts with a micrometer for proper accuracy.
Calipers are good, but for true precision mic em....
Ray Jones thay are "only " 70 € I got one that will do 0.001 mm but I use it for school so if you don't have one 0.05 mm is good enough for your purpose
I was thinking the same, its probably those $10 calipers that aren't accurate.
AussieMatt That they are. The funny thing is it probably is 10mm, just an inaccurate measuring instrument.
Another factor here you could experiment with for the bulging corners is called linear advance. Thats its name in marlin firmware but it's also been implemented in some others like smoothie. Basically it gradually lowers the pressure in the nozzle when you approach a point of zero extrusion so that none will ooze out. Might be worth looking into or even doing a video on!
I wasn't aware of that feature - thanks for sharing! I'll look into it.
No worries! marlinfw.org/docs/features/lin_advance.html the documentation here is really useful
Hey mate, I tried lin_advance, but I didn't notice an improvement at any speed or acceleration. I'd set it up in configuration_adv.h and also using M900 commands, and could hear the extruder motor skipping if the K factor was over 200, but it just didn't improve the bulging corners.
I did read that this feature is best suited to direct drive, and that Bowden may be less useful? Bit of a let down. Have you used it?
Hmm.. did you follow the calibration instructions on the documentation website? I mean doing the test cube with the specific values it specifies and making sure you have wipe and coasting and all of those other features turned off. I haven't personally used it yet because prusa marlin hasn't implemented it. However, I have seen really really nice results from prints done on a tevo tarantula using lin_advance which is bowden
Linear advance should be more useful with bowden setups. You might get better results in direct drive though only because you're already 90% of the way there as far as having better pressure control due to having a short filament path. The whole purpose of linear advance is to model the actual pressure in the hotend to provide consistent extrusion due to the springy nature of the filament as it's pushed and compressed into the hotend by the extruder. The longer the path is from the extruder to the hotend, the more compression force will have to build up in order to actually push the filament through the hotend, so then when you stop (or slow down) extruding, you get extra extrusion at the end as that compression force is then released out the nozzle. Linear advance attempts to regulate this by slowing down the extruder or even reversing it if needed, essentially decoupling it from the X and Y axes.
Very interesting demo -- quick question though. Have you ever calculated the total weight of the gantry? i.e. sum up the carriage, extruder, fan, bushings, Y-rail LM8 bearings and connectors, etc.
Would be interesting to see the difference in power consumption due to the higher forces required.
Can you use polished wood dowels, or acrylic, or perhaps fiber glass rods?
Thanks for this video, it provided a lot of really helpful information for me to address my bulging corner issue. It's surprising to me that slicers still don't take the corner slowdown into account and change the extruder feed rate to compensate. I imagine that someone could create a slicer that would adjust the speed, acceleration and extruder feed rate dynamically during the print to optimize speed without causing any artifacts. The slicer has all of the information it would need about the geometry of each line or curve it will have to draw and should be able to use physics to determine where to slow down. When I'm driving my car I would never use cruise control on curvy country roads because I'd fly off the road on the corners. Instead I adjust my speed and acceleration on the fly. 3D printers should be able to apply the same logic.
I noticed that your cork insulation appears to have an air gap from the heated plate. If you use 100% silicone sealant between the cork and heated bed surface it works excellent as an adhesive. I put about 10-12 small globs and press the cork against the globs to flatten them out. Let it sit overnight and you are golden!
Great idea, I have the same problem.
Pressure advance FTW.... low acceleration, sharp corners, even the 1500mm/2 accel can benefit (usually does). Though it's another dimensions to tweak per filament. unfortunately it ain't so good on a bowden.
Gave a like at the beginning for the TL;DW. You sir are a scholar and a gentleman.
6:58 -> you can see much more artifacts with steel rods than aluminum or carbon fiber. It really shows a lot in the light.
Replacing steel rod is good for reducing vibration. But I think it is bad for accuracy.
Carbon or Aluminum pipe is no precision.
It is necessary to think about a structure to suppress vibration of the bed.
Although true 3D printers are not there to the point that it actually matters.
Great info. Where did you get your carbon fiber rods/tubes? Are there any issues over time with the linear bearings on the carbon fiber?
Would larger steppers solve the problem of the extra weight?
I have 8mm steel rods for the gantry and I am printing with 220c, 150mms speed, and acceleration 3000mms and I am having a little rippling effect! and btw, my printer is literally dancing when printing with those speeds, not just shaking and my neighbor told me that he is hearing my printer's vibrations in his room :lol
Heyya, great video, thanks :D
I'd just like to clarify, that it's just your printer interprets the X axis as Y. In cura, and most other software, X is on the Red axis. you can verify this by clicking the move tool.
So why wouldn't you replace all steel with carbon? so also Y for a delta and YZ for a corexy printer?
The letter X should be facing the Y axis, since each face is measuring the motion perpendicular to that face. You should have seen the difference on the Y face.
oof. I'll be using (if I actually make one) steel chrome plated ones cause they're the only ones that work with a ball bearing from what i've heard 0_0 brass bushings and everything are just too much effort to get working properly
...Ive been using solid 10mm bearing steel for months, ive gotten none of these problems. Idk what you did, but im using the same firmware you are.
Thank you very much for your tests.
This video is now outdated cura is using M204 Sxxx.xx for acceleration in later versions but marlin is using M204 Pxxx.xx Rxxx.xx Txxx.xx (print, retraction, travel) so you can set the values in cura but they wont affect the acceleration in actuall printing.
To get this to work with cura modify the start Gcode to include:
M204 P1500 T1500; Set acceleration
Keep the settings in cura set as this will keep your print time estimates more accurate and only increases gcode file size which is not a big deal.
As for corners becoming out of shape you need to enable linear advance in marlin then you can use a K value to fix the issue.
Looking forward to a update video.
What happened mate ? Your videos were awesome like this one, perfectly summed up the problem with a brief explanation and solution BOOM hats off.
Iv thought about it before with the directions of the x y it’s looks wrong in the slicer but I think you should put the x on the y axis because because the face on the y axis will only show ghosting in the x axis as in side to side in relation to the y axis. Damm Iv not said that well
Thank you. This video was very informative and also fixed the issues I had. Please keep up the good work
Good vid.
Have you tested this the other way where you use the lightest of the rods, then turn up the speed to see the limitations?
My opinion because I'm impatient like most people that have a backload of stuff to print and just want to go faster.
Perhaps the print head should have an accelerometer on it and the system go through a setup phase so that it can detect those issues and subsequently negate them through better print head control, i..e through dynamic software.
Great idea this is exactly what Klipper does nowadays. You were spot on mate.
What about a aluminum and steel tubes with the same wall thickness as the carbon fiber tubes?
12:01 There is most definitely ghosting on the print on the right...You can see it to the right of the Y. Watch it on a desktop, not a phone. YES the print looks better, but ghosting (The repeating pattern of the Y) is clearly visible.
It's clearly visible with a 12-megapixel photo zoomed in 10x, but to the naked eye it isn't clear at all... ;)
Normally I would tend to agree...but I saw it on the video on a 1920x1080 video. Apparently, with 12 megapixel reverse-engineered vision.. ;-)
there's a lot to be said about the way cameras see the world vs the human eye. I'd wager that the difference is imperceptible even with the tongue test and that what you see is most likely an effect of the contrast and slight difference in angles to the camera.
Lol, yeah...clearly I know nothing about cameras.
Finding the bushings for the carbon tubes that's very dificult.
Video after Video, your content is teaching me new information! Before I use my new printer I have been learning as much as I can to improve the quality of my prints and printer. You have much experience and I appreciate your channel. Thank you, Subscribed
I see the video again, I find it excellent. Congratulations!!
What brand of filament is the one you use for the test? Is ABS?
Interesting video. I bet the steel rods are close to advertised size. Digital calipers cannot replace high quality micrometer for really fussy measurement. Even then it takes some practise to get the feel of using a micrometer. Lots of beginners try to use them like a C clamp : )
Did you check your vernier for accuracy against a calibration machinist square ,block before concluding 10mm rod ?
Great test.
Excellent video. Now this means I will tinker with my settings, a bit more. Before, was not sure what settings affected what outcome. Now a little less confusing. I also want to make parts that fit together, and this does require much more knowledge. Please, please continue on this theme, I am sure you will make us all pro, very soon, regards
How do the stiffness of the rods compare ? Eg if you hang a mass in the middle of the rods and measure the deflection
Amazing!!! Thanks for the video.
Did you try to change jerk from 10 to 50? what will happen with the same rods?
I appreciate all the effort in the testing, but the whole time I was looking at the edges of all the 60mm/s prints like, is he going to address what a mess these edges are? Only the 30mm/s had decent edges. It may be a solvable problem, but until it's solved there's no point cutting weight to go high speed when all the high speed prints look like that.
the steel print looks like an old bad tv signal :)
Ahhhhhhh don't press on scales like that! If it says you went over the limit which you did at least once you risk damaging the scale and getting bad readings, you should re-call it and becareful not to do that :D
could the stiffness if the belts have a rubber band effect
my table wobbles allot, thats maybe why i get the ghosting?
i bought the 10mm CF tubes from AliExpress and the ridiculously expensive (for what they are) plastic bushes from RS components and was expecting a snug, maybe slightly overly tight fit.
They're as loose as fuck!
lel comming from the delta world your "fast" speed kinda made me giggle
Is it even possible to get these anodized aluminum tubes anywhere still? Anyone tried 8mm steel shafts instead of 10mm?
Thank you for this video. Very good info and content.
This is very puzzling, because inertia only applies to moving parts.
The only time when those rods are actually moving, is when the gantry moves up, so that would be the Z direction,
I suppose that the ghosting that is found is caused by the poor fit of the bearings to the steel rods.
Another issue with aluminium or carbonfiber rods is the poor hardening, causing the bearings to wear out channels on the rods, and causing ghosting free a while on the Y axis.
The solution would be using (expensive) hardened aluminium rods by Igus or drylin bearings on softer rods, but because those drylin bearings need to be perfectly parallel, this willing also introduce wearing and ghosting after a while,.so the solution for that would be auto-alighning drylin bearings,.which are hard to come by,, and expensive.
What about "jerk" settings? those will also have effect.
Agreed. Here they were fixed at 10mm/s. Any slower and the bulging corners appear no matter what speed you choose. Too fast and I get ripples with CF tubes ;)
Excellent vid, thank you. I have the same affects you mention on prints on my Wanhao Select Maker Plus. I though these issues were common quality issues because of the type of printer we have to live with. Do you have any opinions on the Carbon Fiber tubes as far as thickness vs rigidity and quality. Seems like a very inexpensive upgrade. Thanks.
The units for jerk ought to be mm/s^3.
really interesting. Out of interest can you give any example of the print time difference for the CF vs Steel rods with the accel and speed on the steel set to levels to give similar quality?
Printing the ripple test cubes at 60mm/s with 1500mm/s2 accel, print time was 13:21 for all rails. At 1000mm/s2 and 500mm/s2 accel, print time increased to 14:04 and 15:34 respectively.
I don't know if I'd even trust those crappy calipers to be accurate within .05mm. Great and quite informative video nevertheless.
Where can one buy the carbon shafts from ?
Very interesting test. I didn't know you could get Carbon Fibre rods for printers.
I have the Anet A8 and am currently modding it to improve it a bit. Next mods will be changing from direct drive to bowden, and an e3d v6 head. Maybe some dry bearings as well. Not sure what else.
I'm thinking of getting a 3D printer and i was wondering if you're supposed to sand finished products? I've been looking at some videos about issues to see if im willing to go through the trouble of fixing them but most issues i see seem to be fixable with a bit of sandpaper.
Where did you get the carbon fiber rods?
Where can i get the aluminum rods? The ones I've found so far are either very short or not smooth for use in a 3d printer.
What ID and OD carbon fibre tube did you use/ what do you recommend using