Hello Thomas, I build industrial automation as my daily job. This automations have a lot of rails, and the right way to calibrate them is a bit different as you did. Ill try to make an accurate step by step guide. 1- First you put the rails in and the blocks (as you did) 2- Tighten the rail with 2 blocks making sure it is square to the X axis 3- Tighten the blocks on the fixed rail. 4- Place the carriage to one limit, and tighten the othe rail, but just on this limit side. 5- Slide the carriage to the othe side and tighten this side of the rail. 6- Tighten the remainig block. 7- Now you have to loosen the blocks one by one and tighten them again. Ma ke sure to have only one block loose at the time. 8- Check for unwanted friction, if there is any start all over again. Protips: Never use power tools for this applications, they can shift parts around (even blocks that weigh 1kg) You don't have to be super gentle with this stuff, being industrial they are pretty hardcore (but don't drop them) and for last, if you follow all my steps and you can't go away some little noise remember that the rails can bent some times. The only way to make them straight is to fix every hole on a machined plane. Hope this was helpfull :D
For newbies like myself, there are UA-cam videos from bearing and motion-control companies that are demos of the described "proper" procedures for installing and aligning bearings. Thanks for the instructions!
@@JeffDM I agree but my main concern is not over torquing. You lose a lot of feeling for how things are falling in place. Tighten a bolt a little bit, check, align, tighten another one, adjust until everything falls in to place slowly and smoothly. Thomas went for the AvE approach which might not be the best with precision machinery.
@@MadeWithLayers But it is not, despite it being a torque limiter clutch, it still has enough mass to be equivalent to a mini impact wrench for a precision component that can throw off alignment. Would you like someone to build your 5axis CNC machine with an impact wrench?
I was pretty sure that Josef Prusa said in an interview that they found that using other methods such as linear rails had no affect on print quality, so I guess it's good to know that he was correct
I think that the way the printer is designed to use the guides is very clever. As Tom noticed, the original design is just constrained, and well constrained. With rails, it becomes overconstrained. That is unlikely to improve precision, but is very likely to increase friction, which can lead to decreased precision.
the primary reason hiwin or other square profile linear guides are more rigid is because it can be mounted directly to a subframe and not just at each end (also things like preload but you probably want to avoid that in a printer). By mounting it only at each end it acts like a leaf spring when loaded in the middle of the guide.
Hi Tom, love your videos. When you're talking about difficult to see artifacts on the prints it would be immensely helpful to have some b-roll with something like a macro lense up close on the part so we can see too. In this video, I feel like I'm mostly just talking your word on all of the artifacting instead of seeing it myself. Still a good video though, looking forward to more.
Yep, macro cameras work well. I still have a Konica Minolta dImage Z5 that I use for macro work down to 2mm - it's perfect for examining individual pixels of LCD screens. The Ipevo V4K USB document camera is surprisingly good, too, though I haven't explored it completely yet.
Using those Linear rails, in an industrial manufacturing plant setting, we used to very tightly rubber shockmount those carriages [Just a thin rubber gasket really]. In our cases, over 1 metre length rails, without the rubber mount the rails would bind, and resonances would be setup through the rigid chassis and moving tool head. With them the rails would more or less self align, and run extremely smoothly.
I build machinery as part of my job. When we are using a pair of linear rails, we align and lock down one rail. The lock all bearings to the platen. This will align the rails to each other. Then we tighten down the other rail.
Hi Thomas, I study High Tech engineering, which is a continuation on Mechanical Engineering. Within this study, my specialization is Mechatronic system design. I'm currently doing my Master's thesis and am working on a properly constrained gantry system, which uses state of the art technology. I can not say much about it, since I intend to start a company with it, selling 3D printers and laser cutting machines. Basically every machine on the market nowadays uses a heavily overconstraint set-up. Having 2 cylindrical bearings perfectly aligned (functioning as a single bearing, theoretically) and having another one to cancel the rotation, already overconstraints your system by 3 (1 free translation means 5 constraints, each set has 4 constraints, totalling 8 constraints. 8-5=3). Now, doing the same for hiwin linear rails, we get a sum of 5 overconstraints (5+5=10 constraints, thus 5 overconstraints). What's the result of overconstraining? Not only does it introduce higher order harmonics, your system is no longer statically determined. Binding becomes more of an issue, which might not be noticeable at first or by hand, however at the microscale it's noticeable. Often relaxing a certain bearing helps, since you reduce the amount of overconstraints and thus the harmonics. One thing that confused me in this video is... Why do you wiggle the bed back and forth sideways, while the direction of actuation is through the centre of mass of the bed in the same direction as the bearings slide... The vibrations from the x-axis shouldn't cause much vibrations on the bed, since the frame is very stiff and its amplitude is very much degraded. There's one side question I have. Once I finish my product, I intend to start-up and kickstart it. What conventions do you usually go to in Europe? Keep up the good work!
For this application linear rails from your favorite chinese store for around 15 Euros for a 500mm rail are just as fine. Most of the time they are perfectly usable out of the box. Should they not run perfectly smooth just tear down the block and wash it and the bearing balls in solvent to clean out the dirt from manufacturing. I have been happily running banggood rails on my corexy for a long time.
Pro tip: Use a guitar tuner (or app) on the belt before and after to make sure you get the same tension. Also, use silk PLA or shiny PLA to make ringing artefacts more visible. Black PLA printed hot also shows artefacts more.
For a while I was having all sorts of similar Z banding issues. But if I ran my finger along it I couldn't feel any sort of banding. The issue ended up being that the eSUN grey I was using (that I buy in 3kg rolls) had a large amount of pigment variation. So it wasn't actually a layer or printer issue at all. Changing to a different vendor's filament made it disappear all together.
If you trace the history of the short-lived E3D BigBox project, you find the banding issue there as well. It seems that their primarily used stepper controllers caused some of it by introducing rounding errors in some zones. You always have to look at the whole system.
@Jim Bob I think it was just the bulk rolls - printing out some vase mode models made it visible and obvious when you looked at it. I'd not had that issue with any other eSun filaments, and I've been through probably 50kgs of it.
8:00 i think the y-carriage plate is symmetrical so you can have 2 bearing right or left, you can just turn/flip the plate. A bit upsetting that the rails are not sitting on the aluminium profiles but instead sitting on wobbly plastic pieces.
As a physics major, I'm reminded of the damped harmonic oscillator (mass on a spring with friction) and the classic analysis about over-damped, critically-damped, and under-damped conditions. After increasing stiffness, maybe you can also, somehow, increase dampening and/or modify the mass of the system? It would be interesting to analyze a Prusa i3 with something like a vibration sensor, and then do a "proper" analysis of the ringing and then optimization to reduce it. Maybe a wider belt for the y-axis would increase dampening?
@@evlsc400 Changing the tension of the belt will only change the frequency of the belt oscillation. If you want to dampen its resonance, you need to add some spring made of a different material. For example a belt tensioner with a metal spring. Maybe not one on the belt, but as a holder of the belt wheel. Still this only reduces any resonance frequency, but it will not remove belt vibration completely.
Some recent releases from Chinese market participants might match a better price/quality balance. You simply don't need the HiWin precision when other parts of your gear don't meet the same precision level. Still have to see a 3D printer with a cast mineral concrete bed... :)
We see this on our clients that buy Raise3D units, particularly engineering clients, that have an instinct to tighten every bolt. This is a mistake on the Z axis where the threaded bearings are intentionally loose to deter Z axis artifacts. They get binding as soon as it's over tightened so agreed the first one looks better for similar reasons!
"There is no room for inaccuracy" looks like you have loose hiwinas so there is still room for inaccuracy. Hiwin carriages are divided into several types in this respect, one has a clearance (smaller balls), the other type does not have this clearance but requires more powerful motors. Look on the box ZO, ZA, ZB is this parameter.
Yes! And they should be mounted to honed or finished surfaces, meaning cnc milled, precisions fitting edge! Even a aluminum extrusion frame has a large deviation compared to the tolerances required for these rails to funtion to full specs!
I guess it prevents them from bending if you do that, but I'm not sure it's strictly needed. Also if you fix it to a different material technically it could bend depending on temperature.
@@StrathpefferJunction Then you can add supports and shims for each hole. I would have expected to see Tom using a dial gauge to show any before / after effect. They cost about 15 Euros in an electronic version, the purely mechanical fashion less than 9 Euros, when I looked last week. And you can use them the rest of your life.
@@emmettmartin2639 Because he said he preferred them before it was changed? So wouldn't it make sense he might consider switching them back? Come on this isn't rocket science.
I have a DYI Prusa with several modifications. It has considerable amunts of play in the y and x axis, but always in directions perpendicular to the axis movement. So far, I am happy with the amount of ringing in the prints, nothing out of the ordinary.
The best way to fix the inconsistent extrusion problem (which i assume is the cause of your issues based on your description) I’d suggest getting the bondtech upgrade for the mk3, which uses the original bmg geared style extruder. A lot of users had this issue, and fixed it with this upgrade.
Hi Tom, one of the thing we saw in the early days of DIY 3D printing was similar artifacting. Ours was caused by really tight belts. The 2mm teeth were giving us that slight imperfection. We wanted tight belts for precision. We gave our belts 1/2 a twist so they had the smooth side rolling against the pulley/bearing. That simple fix made our prints super smooth.
I noticed on my Prusa i3Mk3S was producing Z axis banding as yours also does. I also noticed the extruder was sometimes pulling hard enough on the filament to slightly deflect the print head. This was with the stock Prusa filament holder. My fix was a completely new filament reel and axle running on sealed bearings that puts nearly zero stress on the print head while feeding filament. This change made a noticeable improvement in the layer artifacts. My latest prints are about 50mm tall and are very close to perfect exhibiting almost no layer issues. And yes high quality filament is a must... great video as always Thomas.
Those Z axis artifacts you're talking about may be related to the bed's temperature control. The bed expands and contracts while is trying to keep the temperature constant, thus changing the top surface height. You should check your BED PID values although it may not improve your results.
I am a CNC professional of 30 years. Hiwin is fine brand of linear guide. To really take advantage of the Hiwin linear rails, you should do *exactly* what every CNC machine tool builder does: 1. Machine both surfaces of the Y axis extrusions, generating mounting surfaces with both flatness and straightness. 2. Install one of them, tight against the straight edge of the pocket you machined. 3. Install the other side loosely, and let it float. 4. Put an indicator base on the tight side's carriage, with the arm and indicator reaching over to the other side. 5. Slowly work the indicator back and forth while tightening the loose rail side one screw at a time, so that the total indicator reading is within 1 micron parallelism between both rails. This is the *only* way to enjoy the benefits of the Hiwin rails. If you are going to mount them to unqualified extrusion surfaces, you should use cheaper rails and carriages - they exist. There really is no point in using 200 euro rails only to mount them to unqualified, unflat, unparallel surfaces. You are not showing off the rails, you are *showcasing* the out-of-flatness and out-of-parallelism of the Y axis extrusions.
That's because it's a timelaspe recording put in a picture in picture placed in after just to show you it printing. Does not mean it was not the actual print :D
@@brianwild4640 But the print wasn't going when he said it was. That was the dry humour Pouya was going for. Do you always come in when someone makes a joke and tell everyone "well technically...."?
From my experience, the belt tension is the biggest contributor to Y axis problems on MK3. Cranking it up helps constrain the bed more than fiddling with the rails. What I'm thinking about doing, at some point in the future, is to swap the X axis rods for a rail. The print head is rather heavy and the entire assembly is plastic and thin rods, so there might be some play in there.
Yes. From years of experience with home-brew plotters , engravers, cnc and a 3D printer , keeping the belt resonance under control is vital (I built a wire drive plotter once , it rang like a guitar string , wild resonance!) adding a couple of idler wheels (or even long slides) on the long return run can help . Too much tension is as bad as too little - it really a case of observing and making adjustment s to kill any bad resonance.
Back in July when I did upgrade my Prusa MK3 to linear rails, it cost me under $40 bucks not 200 Euros. Print quality stayed great but the lower noise level of the heat bed was well worth doing it. Also you can gain up to 15mm or Y travel if you want it and more Z if you want it but that will stop the Self tests from working so I kept my travel the same.
I've been trying to understand the inconsistency issue(s) of the Mk3 for months now, read countless forum posts and asked the 602 community in discord (which I can recommend btw). First of all I installed a reverse bowden as suggested in the video too, and it really helped a lot. Mind, that you don't need to minimize slop here, so a 3mm inner diameter/4mm outer diameter PTFE tube works wonderfully! Everything else probably comes down to the extruder itself. Many people reported significant improvement after installing a geared extruder like the Bondtech upgrade. Although a bit pricey it also reduces weight and extruder motor temperature, which on top of that can't be transferred to the actual gears and with that the filament anymore, so it won't soften. Prusa actually uses this method in their Pursa Mini printers now, which I find somewhat ridiculous... Another thing of course is heat creap which causes the filament to stick slightly on its way to the nozzle and therefore causes layer inconsistencies. Swapping the silent but weak noctua fan for one with higher static pressure like models by Sunon seemed to help a lot by keeping the cold end, well... colder. So these are the most important upgrades I'm planning to install next. Hope that helps some people =)
I expect the difference to be highlighted by larger prints where the mass creates more kinetic issues. Perhaps loading the linear rail with a 1lb weight while printing smaller test parts like you did would be more illustrative of the linear rail performance.
The benefit of profile linear rails is they can be mounted on a strong support beam! They ARE NOT designed to be used floating like this. Those circular bearing rails actually have very tight tolerance, fully enclosing the rail, the downside is just they are unsupported.
Some of that ringing could be related to the mount arrangements for the Y axis belt drive system. Both ends are at the center of a flexible beam. There is no solid support below the belt from the back to the front of the printer. I had the vibration on my Anet A8 which has similar but plastic construction. I mounted a threaded rod below the belt tied in at the front middle and back. No more vibration on in the print.
Despite there being no apparent quality benefits I’ll be doing this upgrade soon in the hope that it will get it back to somewhat quiet operation. The Y bearings on my MK3 have become increasingly loud in the last couple of months (after only ~150 h of printing) to the point where I’m afraid of keeping the neighbors up.
I floors me that time and effort was taken to reduce 'wobble' in the wrong place. No matter how rigid the frame, rails, table, floor, building, Earth, there is a weak link right under to the model! Springs. They are a safety device to prevent damage, until you have learned enough to prevent nozzle-strikes! Put the leveling knobs where the springs lived. Install lock-nuts where the leveling knobs came from. Mechanical Leveling becomes a truly 'one-time' event. Level opposite corners with the knobs, then lock them down with the lock-nuts. Never adjust one knob by itself! Once you have 'A' and 'C' locked down, 'B' and 'D' are next. Never ever ever ever ever ever 'level' just one corner, you are warping the plate! Three points define a plane, and there are four points attached. What happens to the shape of that plane if only ONE point is moved? Warping. And we don't like warping. Doesn't matter if it makes that corner 'right', it is inducing mechanical stress for the other three corners. Adjust opposite corners as a pair! One goes up, one goes down, until they are both 'perfect'. Only then can the other two be adjusted in the same way. Anycubic MEGA X, Mingda Rock 3 Pro. No springs, no adjustment to bed in 100's of hours of printing. Take the printer apart, put it back together...don't even have to check leveling! It just works. Robert is you parents sibling.
Willyminime covered every mechanical issue that should have been addressed. One thing you highlighted was using the standard settings for all three tests. This in itself, is a testing flaw if the goal is to improve the precision of the model. In order to take advantage of a fully constrained and rigid environment, the settings need to be adjusted to account for the increased accuracy and coordinate consistency. The print settings of the model should now be able to leverage maximum performance - if the coordinates of both the X and Y axis are perfectly aligned during the entire process. The linear rail upgrade to the platform, using the exact same settings for all three models, *proves* how badly the printer is when using *default* settings after the linear upgrade. Perhaps you should run another test and see if your results improve. I'm betting they will.
That the liniar guider are not the best upgrade for the printer. The issue, I think, is the belt teansmision, a ballscrew will offer better results. But the upgrade is a really hard one. Love to see you do it. Also when mounting linial giudes, there a master 1 and a slace 1. You put the master and then check with a gouge that the other raile is paralel. I have done this on machine tools.
Add some cork feet or a thick rubber mat under the printer to reduce resonance 'ringing' on the prints. you can also add them between the stepper motors and the stepper mounts to reduce it as well.
You should start by acknowledging that the stock Prusa extruder and Noctua fan provide insufficient cooling for the v6 heatsink, and swap out that fan for a more powerful one, ie one that has a better static pressure rating. Prusa has also published newer revisions of the extruder which you should definitely upgrade too, or even look into other extruder designs provided by the community, if you dare try them. This has been recognized by many people as one of the main cause for the layer misalignment artifacts that you mention at the beginning (but were not inclined to acknowledge when the MK3 first came out) I also think it would have given one less argument against this experiment if the exact same model had been used in the test prints. I understand the feature difference is minimal, but people will always get you on the fine detail, and in the spirit of scientific process I think this has value.
Looking at how you are printing the model, aren't the nose ringing artifacts along the x axis ?? why are you judjing the improvment of installing linear rails on Y axis by artifact occured in the x Axis? Am i missing something?
Yep; there are a *lot* of issues with this video. It's kind of sad; I really like Thomas' content, and he really does put in an honest effort to teach people and clear up misinformation, but he produces so much misinformation himself and seems completely unaware of it. His recent Solar Battery video was a *really* bad case of this, and so is this video.
I would test a better spool holder. The resistance of the spool has negative impact and x gantry wobble. I have found that on a direct drive it is noticeable and you touched on it.
I think the Z-banding is mostly from the Z axis. I have an Anycubic i3 Mega and I printed a pair of flexible Z-couplings from thingiverse and eliminated most of the Z banding. The author explained that the Z leadscrews are not completely concentric with the steppers and will cause Z-axis to wobble when the leadscrews turn. The Z-axis smooth rods and bearings are not exactly rigid, either plus there is no load on them as on the X and Y axis so it's easier for them to move by the lead screws.
Some of us printing visor headbands for health services are seeing splits down the middle of prints due to backlash. They occur on curved narrow strips sliced with lots of walls so they are 100% filled without using infill. Where the walls meet in the middle we get splits as each side is printed in opposite directions. Would be nice to know if linear rails have less backlash than other designs.
Thanks for the info... I have been waiting for a linear rail version of the Prusa printer for a long time. I think it may have been beneficial to print a simple test cube. That way you can see how much effect you are having on the Y axis. Cheers, Greg
I installed linear rails (Not the kit) and I can see an improvement on the Y axis artifacts, some ringing on the print has improved. I compared the same prints before and after the change, you can visually see the improvement and also can feel it, with the linear rails the print is smoother. Also the play on the printing bed disappeared now is totally rigid.
I'd actually like to see how well this prints with ballscrews instead of the the belts and pulleys. In my opinion the belts are always the sloppiest part of most consumer 3D printers.
@@AudreyRobinel you would have to spin the motor faster for the same amount of lateral speed, but generally I'd say most 3D printers are printing at a rate much slower than what the motors are capable of. A quick check, it seems like most belt systems should go about 40mm per rotation, whereas a 1204 ballscrew (the size that I would recommend for a 3D printer like this) travels 4mm per rotation. You'd need to spin the X/Y motors 10x as fast to have the same print speed, but conversely you'd have 10x the resolution on positioning.
@@cavemaneca And that 10x speed requirement instantly puts stepper motors inadequate. Don't forget that 10x speed means 100x more energy in the rotor, and you are adding a steel rod to it too. But I am curious what a properly engineered ODrive + servo + ballscrew + linear rail printer would do. 😉
The H high precision designation on the rails is probably unnecessary but definitely doesn't hurt. What you're probably thinking about is the preload on the bearings. Less preload is better if your mounting parts aren't super accurate. But preload is a completely separate variable than accuracy.
Hey Tom, thanks for this tryout! I'm now glad with my original bearings and the little space they have. So my Mk3s is fine as it is :). Schönes Video und Grüße aus Süddeutschland!
Yeah, the layer lines not perfectly aligning and vertical lines (ringing?) have been driving me crazy. Belt tension does seem to affect the vertical lines. I think it may have something to do with the printer not micro-stepping properly. The printer prints mechanical components just fine, I can reliably get to 0.1mm on Makers Muses' tolerance test print, I don't even need to force it free. It just makes it impossible to print decoration prints at anything faster than snails' pace
Tom, this experiment makes me wonder if you should replace the u-bolts of the linear bushings with zip-ties, and compare the quality of the print again.
I used belleville washers and distorted-thread locknuts on the U-bolts. The idea is to get some tension force in the joint that will stay relatively consistent over time. I think the Prusa instructions have you tighten them until they touch and then another 90 degrees. This is also a valid technique but I don't know if everyone will interpret 'touching' the same way.
With regards to the loss of height that 3mm is actually within the design tolerances of th MK3. You could combine this upgrade with an upgrade to the MK3S which would get you the height back. Another way to gain height back is to use the full volume volcano modification with which when fitted with the standard ED3 heaterblock with out the standard height adjusters would gain you 8.6mm.
in my opinion, mounting the rails like this(not supported/screwed in along the entire length of the rail) takes away one of the major advantages of linear rails in comparison to smooth rods and linear bearings. now the rail itself can flex/rotate
Try testing with a very tall cone or something near the Z limit. The further the print head is from the plate, the more that slop will translate into wobble distance. These 7cm high squat chunky shapes will not reveal that problem. Still, if you don't have a problem where that kind of thing is concerned, doesn't need to be fixed.
Hi, by experience, if you tight the U-brackets bolts too much, the barrel of the bearing will deform and the bearing will slowly carve a line into the rod. Luckily it is only a line and there will be not misalignment
I love this, I learned a lot. I wish you would still show us the more technical side of 3D printing and not so much of the here's where I can go and here's what I can do side of it all. I am not trying to bring your channel down, I have learned a lot but I love the old stuff and wish you would do more in that aspect.
Since the bed is over-constrained and causes ringing in the system do you think adding rubber spacers underneath the u-bolt nuts would help to alleviate 1. some of the excess ringing, and 2. the wear of the textured surface over time? Of course you would have to use the same torque value over each nut to ensure consistent tightening. I would be curious so see the results. These would be similar to rubber motor mounts to reduce vibration in the entire system. Thoughts?
Good video!! You will want a cnc plate with those linear rails. The rail system Prusa uses is already stiff enough IMO. I think you need to make X and Y stiffer and some adjustments to the printer profile. What you will get more consistency between prints. You didn't show that in this video.
Ringing has nothing to do with the linear rails, but the linear rails expose the weakness and compromise with most 3D printer designs. There are a couple things that will affect the resonant frequency of a machine - its rigidity, and the inertia ratio (the ratio of the inertia of the load and inertia of the motor.) As to rigidity, probably the polar opposite of a Prusa frame would be an epoxy-granite frame with linear rails and ballscrews, or ACME or trapezoidal leadscrews and anti-backlash nuts. I also believe the belts can act as shock absorbers of sorts, but they only mask some of the underlying problems with designs like these. As to inertia ratio, having a too high inertia ratio can also lower the resonant frequency of the machine. One way of lowering the inertia ratio is by gearing down, which lowers the inertia of the load by the square of the reduction ratio. The only problem with that potentially is that the motor speed has to be increased by the gear ratio to reach the same speed. The big problem with direct-drive belt designs is that the pinion acts as a gear increaser, since one turn of the stepper will move the load the diameter of the pinion times pi. As just as the inertia of the load decreases with gear reduction, it increases with a gear multiplier. The other way of lowering the inertia ratio is to use a stepper with a higher inertia - a larger stepper. Changing to a less aggressive acceleration/deceleration setting can help as well.
Interesting video. I wonder if the results would have been different or more clear if the test model were more challenging - those planters are cute but look pretty easy to print.
The plastic roll needs to be in an actual bearing containing hub to prevent the jerking as it feeds down. It must be pulled which results in microscopic jumps as it unloads material
@Bubz of Steel That's a valid point. However, from my own experience I can say that sometimes it's difficult to take a camera picture of a print irregularity. Especially with light-colored filaments. I'd say you don't need to take his word - you can test it yourself and see if you are going to get the same results. So far, from my experience, when I've tested Tom's recommendations I had success. So for me, he is a benchmark. Feel free to put your trust into any other person.
As someone who has used them a lot for industrial apps, it's pronounced High-Win. Not "he-win". (I have spoken to their reps at length on multiple occasions...)
So the uptake in resin printers has me thinking: Dude! I'd like to, for science, challenge/request/suggest (whatever works!) you to see how strong a part can be made by printing (either method to be honest, UV it FDM) a single wall outline and then filling it with epoxy resin. The idea being that a complicated part can be printed and then have the strength of a single object, with epoxy being difficult at best to shape into complicated parts. Will post this comment on a couple of other channels if you can't be bothered, but it sounds to me like a great for science episode!
I maintain and run two MK3s for work. I do about 100 hours a week on each printer. I’ve had more issues with the LMU8s. I might have gotten a bad batch, I went with upgrades from Misumi and have had good results. This kit would have been worth it for me just to get rid of the variables that the cheap 8 mm bearings introduced.
I like that it has 2 linear rails, however, I don't like that the rails are "floating" and only connected to the ends. The linear rail kit for the Ender3 from 3dFused has only one rail on the Y axis, however, it's bolted down along the whole rail. I do love the rails on the Ender3, and I'm sure they would be great on the Prusa as well.
I have a Prusa i3 replica, the reason being its cheaper. I love the way people are upgrading and going further into the premium. category. my concern is that we are now outdoing these printers and the goal for the fdm printers; cheap and accessible. I love the way Prusa mini has come across, but still far too expensive, would you agree? I would like to see a challenge that uses cheap accessible electronics and print the rest of the build, such as the bearings and brackets. In a way that it's replicable as well as accessible to the wide audience.
True statement. The bearings have seals on them and grease in them. This will take the energy. As with the stock system there will also be energy taken by bending the wire harness to the heated bed and in the belt.
With some printers, the z wobble is caused by the heated bed turning on, and slightly lowering the 24 V rails. To test this, you can turn off the heated bed during printing. If the extrusion is smooth with the heater off, you have found the culprit!
Could be thermal expansion due to one side of a bed being hotter than the other side. I saw a horrible instance of this on another brand of printer when I installed a poorly-tuned retrofit. The one-sided thermal expansion curls the bed and depending on the method of constraint this can move the Z up in the middle.
It is the filament wind probably most are not perfect and make the head pull harder some times and softer other times. I think the multimaterial add on helps with that cause it is pulled 2 times not 1.
Maybe now the frame needs to be more rigid as well because the forces from Y go directly into the frame without wobble? Would be interesting to see if the print quality does improve with a bear frame and linear rails.
I have the MK2s (not even done the 2.5s upgrade yet, but that bed is tempting) but the first (and only) upgrade I did was to swap those u bolts for a printed block (see thingiverse) so it did not pinch the bearings and I could tighten them up securely.
Why not try 1 HIWIN and 1 standard. To see if your right about the over constraint? I admit it would look"Jankey" but it would prove your point and still give you a HIWIN for the X axis. Great video. Keep up the good work.
Could it be that the backlash in the y axis was previously obscuring artifacts from other areas? Perhaps the vertical artifacts are a more accurate representation of the print that the whole system was capable of producing which the backlash was "low pass filtering" away, meaning that the rails actually technically were an "improvement" in that area despite looking worse.
At maker faire I saw a screw-drived printer (used instead of belts). I guess a screw drived XYaxis is quite slower, but could it be an improvement in terms of quality?
Hello Thomas, Do you think using a "thread locking product/glue " may help in these circumstances. As they are often used in high vibration machinesry to prevent loosening. Apologies if this has been asked before.
Hi Thomas , if I want to modify the original Prusa i3 mk3 printer and change my heated bed for a bigger size bed or extend z axis , does the printer recognise the extension in the axis and would auto bed levelling works
Im glad you spent the 200 euro's and not me lol, I stick by what I said before, the issue is not that the Y isn't stiff enough! its that damm 3 bearing slop setup! I solved that by using a 3/4 inch PVC tube, that the lm8uu bearing slips in tightly and anchored it to the bed, the total cost was a whole 10 dollars for the pipe and all the bed artifacts have been solved.
@@MadeWithLayers thanks for the transparency, Either way, you saved a lot of people some money by showing it may not be worth the amount it would cost to retrofit with the hiwin's. Two thumbs up as always!
I’m really glad I watched this - I just put my order in this morning and was trying to find any upgrades I could install while building up the printer. As someone who works with industrial CNC machinery, it’s hard to look at 3D printed parts supporting the main axis, but clearly that’s not a bottleneck in performance! Have you played around with any of the complicated analysis tools that use a print head accelerometer to tune motion on the fly to negate ringing and ghosting?
Thermal expansion of the y carriage made impossible to set the rails for me. Its fine when cold but binding when heated to petg temp. Changed them to 10mm rods.
Good video Thomas, I think the main reason of the linear rail not just the print quality, but less noisy than stock one, I hate the of the prusa mk3 when it's print with speed it't not quiet anymore, and start see some Grove on the rods, so it's worth upgrade imp 😁
Hello Thomas,
I build industrial automation as my daily job.
This automations have a lot of rails, and the right way to calibrate them is a bit different as you did.
Ill try to make an accurate step by step guide.
1- First you put the rails in and the blocks (as you did)
2- Tighten the rail with 2 blocks making sure it is square to the X axis
3- Tighten the blocks on the fixed rail.
4- Place the carriage to one limit, and tighten the othe rail, but just on this limit side.
5- Slide the carriage to the othe side and tighten this side of the rail.
6- Tighten the remainig block.
7- Now you have to loosen the blocks one by one and tighten them again. Ma ke sure to have only one block loose at the time.
8- Check for unwanted friction, if there is any start all over again.
Protips: Never use power tools for this applications, they can shift parts around (even blocks that weigh 1kg)
You don't have to be super gentle with this stuff, being industrial they are pretty hardcore (but don't drop them)
and for last, if you follow all my steps and you can't go away some little noise remember that the rails can bent some times. The only way to make them straight is to fix every hole on a machined plane.
Hope this was helpfull :D
willyminime good post 😁
Great advise!
For newbies like myself, there are UA-cam videos from bearing and motion-control companies that are demos of the described "proper" procedures for installing and aligning bearings. Thanks for the instructions!
Can confirm, this is similar to what we do at DMG MORI to our machines
@Thomas Sanladerer Pin this! This is very important!!!
"pulls out high precision rails"
"tightens with a power tool"
oh boy..
I seem to hear the clutch mechanism clicking, which is pretty decent on the Bosch drivers.
@@JeffDM I agree but my main concern is not over torquing. You lose a lot of feeling for how things are falling in place.
Tighten a bolt a little bit, check, align, tighten another one, adjust until everything falls in to place slowly and smoothly.
Thomas went for the AvE approach which might not be the best with precision machinery.
I've had the clutch on setting "3" almost the entire time - which I fund to be extremely consistent for mid/light tightening of M3.
@@MadeWithLayers But it is not, despite it being a torque limiter clutch, it still has enough mass to be equivalent to a mini impact wrench for a precision component that can throw off alignment. Would you like someone to build your 5axis CNC machine with an impact wrench?
Spent the whole video just wishing he grabed a plain old screwdriver! After I heard the price for these linear rails, it made me even more unease!
I was pretty sure that Josef Prusa said in an interview that they found that using other methods such as linear rails had no affect on print quality, so I guess it's good to know that he was correct
@Jazz Josef also said "we dont need 32 bit boards" :D
@@Arnipll Bill Gates said: "640K is more memory than anyone will ever need on a computer."
Oh boi, suck a spoiler there
@@RobNisters That has never actually been proven.
I think that the way the printer is designed to use the guides is very clever. As Tom noticed, the original design is just constrained, and well constrained. With rails, it becomes overconstrained. That is unlikely to improve precision, but is very likely to increase friction, which can lead to decreased precision.
the primary reason hiwin or other square profile linear guides are more rigid is because it can be mounted directly to a subframe and not just at each end (also things like preload but you probably want to avoid that in a printer). By mounting it only at each end it acts like a leaf spring when loaded in the middle of the guide.
Hi Tom, love your videos. When you're talking about difficult to see artifacts on the prints it would be immensely helpful to have some b-roll with something like a macro lense up close on the part so we can see too. In this video, I feel like I'm mostly just talking your word on all of the artifacting instead of seeing it myself. Still a good video though, looking forward to more.
1:12 If it's hard to show on camera just take a picture.
Yep, macro cameras work well. I still have a Konica Minolta dImage Z5 that I use for macro work down to 2mm - it's perfect for examining individual pixels of LCD screens. The Ipevo V4K USB document camera is surprisingly good, too, though I haven't explored it completely yet.
Seriously. Replace the camera man on camera two with a still camera, and the video would become much more informative.
@@darrinbrunner6429 i think there is only some one on b cam
Using those Linear rails, in an industrial manufacturing plant setting, we used to very tightly rubber shockmount those carriages [Just a thin rubber gasket really]. In our cases, over 1 metre length rails, without the rubber mount the rails would bind, and resonances would be setup through the rigid chassis and moving tool head. With them the rails would more or less self align, and run extremely smoothly.
I build machinery as part of my job. When we are using a pair of linear rails, we align and lock down one rail. The lock all bearings to the platen. This will align the rails to each other. Then we tighten down the other rail.
Hi Thomas,
I study High Tech engineering, which is a continuation on Mechanical Engineering. Within this study, my specialization is Mechatronic system design. I'm currently doing my Master's thesis and am working on a properly constrained gantry system, which uses state of the art technology. I can not say much about it, since I intend to start a company with it, selling 3D printers and laser cutting machines.
Basically every machine on the market nowadays uses a heavily overconstraint set-up. Having 2 cylindrical bearings perfectly aligned (functioning as a single bearing, theoretically) and having another one to cancel the rotation, already overconstraints your system by 3 (1 free translation means 5 constraints, each set has 4 constraints, totalling 8 constraints. 8-5=3). Now, doing the same for hiwin linear rails, we get a sum of 5 overconstraints (5+5=10 constraints, thus 5 overconstraints).
What's the result of overconstraining? Not only does it introduce higher order harmonics, your system is no longer statically determined. Binding becomes more of an issue, which might not be noticeable at first or by hand, however at the microscale it's noticeable. Often relaxing a certain bearing helps, since you reduce the amount of overconstraints and thus the harmonics.
One thing that confused me in this video is... Why do you wiggle the bed back and forth sideways, while the direction of actuation is through the centre of mass of the bed in the same direction as the bearings slide... The vibrations from the x-axis shouldn't cause much vibrations on the bed, since the frame is very stiff and its amplitude is very much degraded.
There's one side question I have. Once I finish my product, I intend to start-up and kickstart it. What conventions do you usually go to in Europe?
Keep up the good work!
For this application linear rails from your favorite chinese store for around 15 Euros for a 500mm rail are just as fine. Most of the time they are perfectly usable out of the box. Should they not run perfectly smooth just tear down the block and wash it and the bearing balls in solvent to clean out the dirt from manufacturing. I have been happily running banggood rails on my corexy for a long time.
Is there a specific vendor you would recommend?
The ones I got have play in the block ;-;
Pro tip:
Use a guitar tuner (or app) on the belt before and after to make sure you get the same tension.
Also, use silk PLA or shiny PLA to make ringing artefacts more visible. Black PLA printed hot also shows artefacts more.
For a while I was having all sorts of similar Z banding issues. But if I ran my finger along it I couldn't feel any sort of banding. The issue ended up being that the eSUN grey I was using (that I buy in 3kg rolls) had a large amount of pigment variation. So it wasn't actually a layer or printer issue at all. Changing to a different vendor's filament made it disappear all together.
If you trace the history of the short-lived E3D BigBox project, you find the banding issue there as well. It seems that their primarily used stepper controllers caused some of it by introducing rounding errors in some zones. You always have to look at the whole system.
@Jim Bob I think it was just the bulk rolls - printing out some vase mode models made it visible and obvious when you looked at it. I'd not had that issue with any other eSun filaments, and I've been through probably 50kgs of it.
8:00 i think the y-carriage plate is symmetrical so you can have 2 bearing right or left, you can just turn/flip the plate.
A bit upsetting that the rails are not sitting on the aluminium profiles but instead sitting on wobbly plastic pieces.
They specifically point out the orientation you should be using in the guide.
As a physics major, I'm reminded of the damped harmonic oscillator (mass on a spring with friction) and the classic analysis about over-damped, critically-damped, and under-damped conditions. After increasing stiffness, maybe you can also, somehow, increase dampening and/or modify the mass of the system?
It would be interesting to analyze a Prusa i3 with something like a vibration sensor, and then do a "proper" analysis of the ringing and then optimization to reduce it.
Maybe a wider belt for the y-axis would increase dampening?
perhaps giving enough 'slack' in the tension of the Y belt might help some what? I like the widening idea.
Stefan has got you covered! ua-cam.com/video/Ws1JfHl3Y0o/v-deo.html
Thanks Tom! Will definitely go watch Stefan's video.
@@evlsc400 Changing the tension of the belt will only change the frequency of the belt oscillation. If you want to dampen its resonance, you need to add some spring made of a different material. For example a belt tensioner with a metal spring. Maybe not one on the belt, but as a holder of the belt wheel. Still this only reduces any resonance frequency, but it will not remove belt vibration completely.
Gijs Noorlander so a spring loaded belt tensioner? If those exist I’ll test them this weekend.
7:12 They likely had the thought, but ultimately decided against it to keep the price a bit lower. Those rails are pricey.
Some recent releases from Chinese market participants might match a better price/quality balance. You simply don't need the HiWin precision when other parts of your gear don't meet the same precision level. Still have to see a 3D printer with a cast mineral concrete bed... :)
We see this on our clients that buy Raise3D units, particularly engineering clients, that have an instinct to tighten every bolt. This is a mistake on the Z axis where the threaded bearings are intentionally loose to deter Z axis artifacts. They get binding as soon as it's over tightened so agreed the first one looks better for similar reasons!
I got the best improvements when I upgraded the Z axis bearings to PTFE bushes
I would love to hear more about that !
Where could we get those? I did a search but couldnt find that. Also would that work for mk3s?
@@Gefionius IGUS bushings, I printed my own to the tolerances of my printer in TPU.
"There is no room for inaccuracy" looks like you have loose hiwinas so there is still room for inaccuracy.
Hiwin carriages are divided into several types in this respect, one has a clearance (smaller balls), the other type does not have this clearance but requires more powerful motors.
Look on the box ZO, ZA, ZB is this parameter.
I always thought these HIWIN rails need to be supported over their whole length...🤔
Yes!
And they should be mounted to honed or finished surfaces, meaning cnc milled, precisions fitting edge!
Even a aluminum extrusion frame has a large deviation compared to the tolerances required for these rails to funtion to full specs!
I guess it prevents them from bending if you do that, but I'm not sure it's strictly needed. Also if you fix it to a different material technically it could bend depending on temperature.
Totally. They flex in the middle. The kit should include a shim for the middle sections.
@@StrathpefferJunction Then you can add supports and shims for each hole. I would have expected to see Tom using a dial gauge to show any before / after effect. They cost about 15 Euros in an electronic version, the purely mechanical fashion less than 9 Euros, when I looked last week. And you can use them the rest of your life.
Ulrich Kliegis Tom has multiple gauges as you’ve described which he has used in other videos.
Now the question is: Are you leaving them on, or switching back to the original rod setup?
Why would he switch back?
@@emmettmartin2639 Because he said he preferred them before it was changed? So wouldn't it make sense he might consider switching them back? Come on this isn't rocket science.
I have a DYI Prusa with several modifications. It has considerable amunts of play in the y and x axis, but always in directions perpendicular to the axis movement. So far, I am happy with the amount of ringing in the prints, nothing out of the ordinary.
The best way to fix the inconsistent extrusion problem (which i assume is the cause of your issues based on your description) I’d suggest getting the bondtech upgrade for the mk3, which uses the original bmg geared style extruder. A lot of users had this issue, and fixed it with this upgrade.
Hi Tom, one of the thing we saw in the early days of DIY 3D printing was similar artifacting. Ours was caused by really tight belts. The 2mm teeth were giving us that slight imperfection. We wanted tight belts for precision. We gave our belts 1/2 a twist so they had the smooth side rolling against the pulley/bearing. That simple fix made our prints super smooth.
Very interesting, I actually thought of that too. Don't they skip occasionally tho?
I noticed on my Prusa i3Mk3S was producing Z axis banding as yours also does. I also noticed the extruder was sometimes pulling hard enough on the filament to slightly deflect the print head. This was with the stock Prusa filament holder.
My fix was a completely new filament reel and axle running on sealed bearings that puts nearly zero stress on the print head while feeding filament. This change made a noticeable improvement in the layer artifacts. My latest prints are about 50mm tall and are very close to perfect exhibiting almost no layer issues. And yes high quality filament is a must... great video as always Thomas.
Those Z axis artifacts you're talking about may be related to the bed's temperature control. The bed expands and contracts while is trying to keep the temperature constant, thus changing the top surface height. You should check your BED PID values although it may not improve your results.
My heart is breaking with all that power drill tightening
M3 screws, through a 3D printed plastic adapter and the rails are supported only on the ends, so...it doesn't matter anyway :D
I am a CNC professional of 30 years. Hiwin is fine brand of linear guide. To really take advantage of the Hiwin linear rails, you should do *exactly* what every CNC machine tool builder does:
1. Machine both surfaces of the Y axis extrusions, generating mounting surfaces with both flatness and straightness.
2. Install one of them, tight against the straight edge of the pocket you machined.
3. Install the other side loosely, and let it float.
4. Put an indicator base on the tight side's carriage, with the arm and indicator reaching over to the other side.
5. Slowly work the indicator back and forth while tightening the loose rail side one screw at a time, so that the total indicator reading is within 1 micron parallelism between both rails.
This is the *only* way to enjoy the benefits of the Hiwin rails. If you are going to mount them to unqualified extrusion surfaces, you should use cheaper rails and carriages - they exist. There really is no point in using 200 euro rails only to mount them to unqualified, unflat, unparallel surfaces. You are not showing off the rails, you are *showcasing* the out-of-flatness and out-of-parallelism of the Y axis extrusions.
Liking your honest review - no manufacturer favoritism here - and I agree with your findings - makes total sense 👍
Getting rid of the z axis artifacts fully is near impossible tbh
13:57 "while this last print is going..." ehm... Tom we can see the finished third print in front of you :D
That's because it's a timelaspe recording put in a picture in picture placed in after just to show you it printing. Does not mean it was not the actual print :D
@@brianwild4640 But the print wasn't going when he said it was. That was the dry humour Pouya was going for. Do you always come in when someone makes a joke and tell everyone "well technically...."?
Krytern yes I do
From my experience, the belt tension is the biggest contributor to Y axis problems on MK3. Cranking it up helps constrain the bed more than fiddling with the rails.
What I'm thinking about doing, at some point in the future, is to swap the X axis rods for a rail. The print head is rather heavy and the entire assembly is plastic and thin rods, so there might be some play in there.
Yes. From years of experience with home-brew plotters , engravers, cnc and a 3D printer , keeping the belt resonance under control is vital (I built a wire drive plotter once , it rang like a guitar string , wild resonance!) adding a couple of idler wheels (or even long slides) on the long return run can help . Too much tension is as bad as too little - it really a case of observing and making adjustment s to kill any bad resonance.
Back in July when I did upgrade my Prusa MK3 to linear rails, it cost me under $40 bucks not 200 Euros. Print quality stayed great but the lower noise level of the heat bed was well worth doing it. Also you can gain up to 15mm or Y travel if you want it and more Z if you want it but that will stop the Self tests from working so I kept my travel the same.
I've been trying to understand the inconsistency issue(s) of the Mk3 for months now, read countless forum posts and asked the 602 community in discord (which I can recommend btw).
First of all I installed a reverse bowden as suggested in the video too, and it really helped a lot. Mind, that you don't need to minimize slop here, so a 3mm inner diameter/4mm outer diameter PTFE tube works wonderfully!
Everything else probably comes down to the extruder itself. Many people reported significant improvement after installing a geared extruder like the Bondtech upgrade. Although a bit pricey it also reduces weight and extruder motor temperature, which on top of that can't be transferred to the actual gears and with that the filament anymore, so it won't soften. Prusa actually uses this method in their Pursa Mini printers now, which I find somewhat ridiculous...
Another thing of course is heat creap which causes the filament to stick slightly on its way to the nozzle and therefore causes layer inconsistencies. Swapping the silent but weak noctua fan for one with higher static pressure like models by Sunon seemed to help a lot by keeping the cold end, well... colder. So these are the most important upgrades I'm planning to install next.
Hope that helps some people =)
Nice to see upgrade videos still coming for MK3 with the mini coming soon !
I expect the difference to be highlighted by larger prints where the mass creates more kinetic issues. Perhaps loading the linear rail with a 1lb weight while printing smaller test parts like you did would be more illustrative of the linear rail performance.
The benefit of profile linear rails is they can be mounted on a strong support beam! They ARE NOT designed to be used floating like this.
Those circular bearing rails actually have very tight tolerance, fully enclosing the rail, the downside is just they are unsupported.
Some of that ringing could be related to the mount arrangements for the Y axis belt drive system. Both ends are at the center of a flexible beam. There is no solid support below the belt from the back to the front of the printer. I had the vibration on my Anet A8 which has similar but plastic construction. I mounted a threaded rod below the belt tied in at the front middle and back. No more vibration on in the print.
Despite there being no apparent quality benefits I’ll be doing this upgrade soon in the hope that it will get it back to somewhat quiet operation. The Y bearings on my MK3 have become increasingly loud in the last couple of months (after only ~150 h of printing) to the point where I’m afraid of keeping the neighbors up.
I'm curious about linear rails for this same reason, and also have insanely loud Y bearings after only a couple of months
Moving belt mounting point 3mm higher also makes Y movement just slightly less uniform
I floors me that time and effort was taken to reduce 'wobble' in the wrong place.
No matter how rigid the frame, rails, table, floor, building, Earth, there is a weak link right under to the model!
Springs.
They are a safety device to prevent damage, until you have learned enough to prevent nozzle-strikes!
Put the leveling knobs where the springs lived.
Install lock-nuts where the leveling knobs came from.
Mechanical Leveling becomes a truly 'one-time' event.
Level opposite corners with the knobs, then lock them down with the lock-nuts.
Never adjust one knob by itself!
Once you have 'A' and 'C' locked down, 'B' and 'D' are next.
Never ever ever ever ever ever 'level' just one corner, you are warping the plate!
Three points define a plane, and there are four points attached.
What happens to the shape of that plane if only ONE point is moved?
Warping.
And we don't like warping.
Doesn't matter if it makes that corner 'right', it is inducing mechanical stress for the other three corners.
Adjust opposite corners as a pair!
One goes up, one goes down, until they are both 'perfect'.
Only then can the other two be adjusted in the same way.
Anycubic MEGA X, Mingda Rock 3 Pro.
No springs, no adjustment to bed in 100's of hours of printing.
Take the printer apart, put it back together...don't even have to check leveling!
It just works.
Robert is you parents sibling.
I heard both "He-win" and "High-win" in this video. After a little googling I believe "high-win" is the correct pronunciation.
I heard both "proosha" and "proosa" in this video. I love the pronunciation variation.
Willyminime covered every mechanical issue that should have been addressed. One thing you highlighted was using the standard settings for all three tests. This in itself, is a testing flaw if the goal is to improve the precision of the model. In order to take advantage of a fully constrained and rigid environment, the settings need to be adjusted to account for the increased accuracy and coordinate consistency. The print settings of the model should now be able to leverage maximum performance - if the coordinates of both the X and Y axis are perfectly aligned during the entire process.
The linear rail upgrade to the platform, using the exact same settings for all three models, *proves* how badly the printer is when using *default* settings after the linear upgrade. Perhaps you should run another test and see if your results improve. I'm betting they will.
Ah yes, the tool to fasten RC Car wheels at 3:23 . I also use mine as default hex nut fastener for small ones. :D
That the liniar guider are not the best upgrade for the printer. The issue, I think, is the belt teansmision, a ballscrew will offer better results. But the upgrade is a really hard one. Love to see you do it.
Also when mounting linial giudes, there a master 1 and a slace 1. You put the master and then check with a gouge that the other raile is paralel. I have done this on machine tools.
Why can't I see the horizontal print lines? Is it the camera?
Add some cork feet or a thick rubber mat under the printer to reduce resonance 'ringing' on the prints. you can also add them between the stepper motors and the stepper mounts to reduce it as well.
You should start by acknowledging that the stock Prusa extruder and Noctua fan provide insufficient cooling for the v6 heatsink, and swap out that fan for a more powerful one, ie one that has a better static pressure rating. Prusa has also published newer revisions of the extruder which you should definitely upgrade too, or even look into other extruder designs provided by the community, if you dare try them. This has been recognized by many people as one of the main cause for the layer misalignment artifacts that you mention at the beginning (but were not inclined to acknowledge when the MK3 first came out)
I also think it would have given one less argument against this experiment if the exact same model had been used in the test prints. I understand the feature difference is minimal, but people will always get you on the fine detail, and in the spirit of scientific process I think this has value.
Looking at how you are printing the model, aren't the nose ringing artifacts along the x axis ?? why are you judjing the improvment of installing linear rails on Y axis by artifact occured in the x Axis? Am i missing something?
Yep; there are a *lot* of issues with this video. It's kind of sad; I really like Thomas' content, and he really does put in an honest effort to teach people and clear up misinformation, but he produces so much misinformation himself and seems completely unaware of it. His recent Solar Battery video was a *really* bad case of this, and so is this video.
I would test a better spool holder. The resistance of the spool has negative impact and x gantry wobble. I have found that on a direct drive it is noticeable and you touched on it.
I think the Z-banding is mostly from the Z axis.
I have an Anycubic i3 Mega and I printed a pair of flexible Z-couplings from thingiverse and eliminated most of the Z banding. The author explained that the Z leadscrews are not completely concentric with the steppers and will cause Z-axis to wobble when the leadscrews turn. The Z-axis smooth rods and bearings are not exactly rigid, either plus there is no load on them as on the X and Y axis so it's easier for them to move by the lead screws.
thank you for this review thomas
how is the noise level of the hiwin compared to stock on travel speeds > 100ms/s ?
tia
I printed PETG bearing blocks for my bed and it's near rock solid, maybe 0.2mm wobble up and down under force. And prints look better than before.
Some of us printing visor headbands for health services are seeing splits down the middle of prints due to backlash. They occur on curved narrow strips sliced with lots of walls so they are 100% filled without using infill. Where the walls meet in the middle we get splits as each side is printed in opposite directions. Would be nice to know if linear rails have less backlash than other designs.
Thanks for the info... I have been waiting for a linear rail version of the Prusa printer for a long time. I think it may have been beneficial to print a simple test cube. That way you can see how much effect you are having on the Y axis.
Cheers,
Greg
Im slowly building the theory that for the best print quality overall the whole printer actually needs to be set up in a way that it can float a bit.
I installed linear rails (Not the kit) and I can see an improvement on the Y axis artifacts, some ringing on the print has improved. I compared the same prints before and after the change, you can visually see the improvement and also can feel it, with the linear rails the print is smoother. Also the play on the printing bed disappeared now is totally rigid.
I'd actually like to see how well this prints with ballscrews instead of the the belts and pulleys. In my opinion the belts are always the sloppiest part of most consumer 3D printers.
but ballscrew would be much slower than the belt, no? for Z ok, but for X and Y i'm not convinced.
@@AudreyRobinel you would have to spin the motor faster for the same amount of lateral speed, but generally I'd say most 3D printers are printing at a rate much slower than what the motors are capable of. A quick check, it seems like most belt systems should go about 40mm per rotation, whereas a 1204 ballscrew (the size that I would recommend for a 3D printer like this) travels 4mm per rotation. You'd need to spin the X/Y motors 10x as fast to have the same print speed, but conversely you'd have 10x the resolution on positioning.
@@cavemaneca And that 10x speed requirement instantly puts stepper motors inadequate. Don't forget that 10x speed means 100x more energy in the rotor, and you are adding a steel rod to it too. But I am curious what a properly engineered ODrive + servo + ballscrew + linear rail printer would do. 😉
The H high precision designation on the rails is probably unnecessary but definitely doesn't hurt. What you're probably thinking about is the preload on the bearings. Less preload is better if your mounting parts aren't super accurate. But preload is a completely separate variable than accuracy.
Hey Tom, thanks for this tryout! I'm now glad with my original bearings and the little space they have. So my Mk3s is fine as it is :).
Schönes Video und Grüße aus Süddeutschland!
Yeah, the layer lines not perfectly aligning and vertical lines (ringing?) have been driving me crazy. Belt tension does seem to affect the vertical lines.
I think it may have something to do with the printer not micro-stepping properly. The printer prints mechanical components just fine, I can reliably get to 0.1mm on Makers Muses' tolerance test print, I don't even need to force it free. It just makes it impossible to print decoration prints at anything faster than snails' pace
Tom, this experiment makes me wonder if you should replace the u-bolts of the linear bushings with zip-ties, and compare the quality of the print again.
I used belleville washers and distorted-thread locknuts on the U-bolts. The idea is to get some tension force in the joint that will stay relatively consistent over time. I think the Prusa instructions have you tighten them until they touch and then another 90 degrees. This is also a valid technique but I don't know if everyone will interpret 'touching' the same way.
With regards to the loss of height that 3mm is actually within the design tolerances of th MK3. You could combine this upgrade with an upgrade to the MK3S which would get you the height back. Another way to gain height back is to use the full volume volcano modification with which when fitted with the standard ED3 heaterblock with out the standard height adjusters would gain you 8.6mm.
in my opinion, mounting the rails like this(not supported/screwed in along the entire length of the rail) takes away one of the major advantages of linear rails in comparison to smooth rods and linear bearings. now the rail itself can flex/rotate
Try testing with a very tall cone or something near the Z limit. The further the print head is from the plate, the more that slop will translate into wobble distance. These 7cm high squat chunky shapes will not reveal that problem. Still, if you don't have a problem where that kind of thing is concerned, doesn't need to be fixed.
Hi, by experience, if you tight the U-brackets bolts too much, the barrel of the bearing will deform and the bearing will slowly carve a line into the rod. Luckily it is only a line and there will be not misalignment
I love this, I learned a lot. I wish you would still show us the more technical side of 3D printing and not so much of the here's where I can go and here's what I can do side of it all. I am not trying to bring your channel down, I have learned a lot but I love the old stuff and wish you would do more in that aspect.
A great video, I like the end result that spending a lot of money may not make much or any difference, but , many of us enjoy the hunt
The mk2.5 upgrade kit come with a drill bit and a holder so you can drill out the holes for the zip ties so you can use u bolts.
Since the bed is over-constrained and causes ringing in the system do you think adding rubber spacers underneath the u-bolt nuts would help to alleviate 1. some of the excess ringing, and 2. the wear of the textured surface over time? Of course you would have to use the same torque value over each nut to ensure consistent tightening.
I would be curious so see the results. These would be similar to rubber motor mounts to reduce vibration in the entire system. Thoughts?
Good video!!
You will want a cnc plate with those linear rails. The rail system Prusa uses is already stiff enough IMO.
I think you need to make X and Y stiffer and some adjustments to the printer profile. What you will get more consistency between prints. You didn't show that in this video.
Ringing has nothing to do with the linear rails, but the linear rails expose the weakness and compromise with most 3D printer designs. There are a couple things that will affect the resonant frequency of a machine - its rigidity, and the inertia ratio (the ratio of the inertia of the load and inertia of the motor.) As to rigidity, probably the polar opposite of a Prusa frame would be an epoxy-granite frame with linear rails and ballscrews, or ACME or trapezoidal leadscrews and anti-backlash nuts. I also believe the belts can act as shock absorbers of sorts, but they only mask some of the underlying problems with designs like these. As to inertia ratio, having a too high inertia ratio can also lower the resonant frequency of the machine. One way of lowering the inertia ratio is by gearing down, which lowers the inertia of the load by the square of the reduction ratio. The only problem with that potentially is that the motor speed has to be increased by the gear ratio to reach the same speed. The big problem with direct-drive belt designs is that the pinion acts as a gear increaser, since one turn of the stepper will move the load the diameter of the pinion times pi. As just as the inertia of the load decreases with gear reduction, it increases with a gear multiplier. The other way of lowering the inertia ratio is to use a stepper with a higher inertia - a larger stepper. Changing to a less aggressive acceleration/deceleration setting can help as well.
Interesting video. I wonder if the results would have been different or more clear if the test model were more challenging - those planters are cute but look pretty easy to print.
The plastic roll needs to be in an actual bearing containing hub to prevent the jerking as it feeds down. It must be pulled which results in microscopic jumps as it unloads material
Tom - the benchmark of honest reviews. Thanks for doing what you're doing.
@Bubz of Steel That's a valid point. However, from my own experience I can say that sometimes it's difficult to take a camera picture of a print irregularity. Especially with light-colored filaments.
I'd say you don't need to take his word - you can test it yourself and see if you are going to get the same results.
So far, from my experience, when I've tested Tom's recommendations I had success.
So for me, he is a benchmark. Feel free to put your trust into any other person.
As someone who has used them a lot for industrial apps, it's pronounced High-Win. Not "he-win". (I have spoken to their reps at length on multiple occasions...)
He said it correctly once, but yeah...
So the uptake in resin printers has me thinking:
Dude! I'd like to, for science, challenge/request/suggest (whatever works!) you to see how strong a part can be made by printing (either method to be honest, UV it FDM) a single wall outline and then filling it with epoxy resin. The idea being that a complicated part can be printed and then have the strength of a single object, with epoxy being difficult at best to shape into complicated parts.
Will post this comment on a couple of other channels if you can't be bothered, but it sounds to me like a great for science episode!
I maintain and run two MK3s for work. I do about 100 hours a week on each printer. I’ve had more issues with the LMU8s. I might have gotten a bad batch, I went with upgrades from Misumi and have had good results. This kit would have been worth it for me just to get rid of the variables that the cheap 8 mm bearings introduced.
I like that it has 2 linear rails, however, I don't like that the rails are "floating" and only connected to the ends. The linear rail kit for the Ender3 from 3dFused has only one rail on the Y axis, however, it's bolted down along the whole rail. I do love the rails on the Ender3, and I'm sure they would be great on the Prusa as well.
I have a Prusa i3 replica, the reason being its cheaper. I love the way people are upgrading and going further into the premium. category. my concern is that we are now outdoing these printers and the goal for the fdm printers; cheap and accessible. I love the way Prusa mini has come across, but still far too expensive, would you agree? I would like to see a challenge that uses cheap accessible electronics and print the rest of the build, such as the bearings and brackets. In a way that it's replicable as well as accessible to the wide audience.
The tighter/rigid everything is, the more accuracy you get, but also the more ringing, the energy has to go somewhere.
True statement. The bearings have seals on them and grease in them. This will take the energy. As with the stock system there will also be energy taken by bending the wire harness to the heated bed and in the belt.
How about adding rubber mounts b/w bearing and build plate. Maybe that can damper the slight vibrations
It is a pity that the kit does not have 2 milled aluminium plates to fully support those beautiful rails.
With some printers, the z wobble is caused by the heated bed turning on, and slightly lowering the 24 V rails. To test this, you can turn off the heated bed during printing. If the extrusion is smooth with the heater off, you have found the culprit!
Could be thermal expansion due to one side of a bed being hotter than the other side. I saw a horrible instance of this on another brand of printer when I installed a poorly-tuned retrofit. The one-sided thermal expansion curls the bed and depending on the method of constraint this can move the Z up in the middle.
Would it be a good idea to use the hiwin rails only on one side to not "overconstrain" the system ?
It is the filament wind probably most are not perfect and make the head pull harder some times and softer other times. I think the multimaterial add on helps with that cause it is pulled 2 times not 1.
Maybe now the frame needs to be more rigid as well because the forces from Y go directly into the frame without wobble? Would be interesting to see if the print quality does improve with a bear frame and linear rails.
I have the MK2s (not even done the 2.5s upgrade yet, but that bed is tempting) but the first (and only) upgrade I did was to swap those u bolts for a printed block (see thingiverse) so it did not pinch the bearings and I could tighten them up securely.
Why not try 1 HIWIN and 1 standard. To see if your right about the over constraint?
I admit it would look"Jankey" but it would prove your point and still give you a HIWIN for the X axis.
Great video. Keep up the good work.
The rails are a different height than the rods.
Would work with a printed spacer
Could it be that the backlash in the y axis was previously obscuring artifacts from other areas? Perhaps the vertical artifacts are a more accurate representation of the print that the whole system was capable of producing which the backlash was "low pass filtering" away, meaning that the rails actually technically were an "improvement" in that area despite looking worse.
At maker faire I saw a screw-drived printer (used instead of belts). I guess a screw drived XYaxis is quite slower, but could it be an improvement in terms of quality?
This is why I am subscribed to this channel.
Hello Thomas,
Do you think using a "thread locking product/glue " may help in these circumstances. As they are often used in high vibration machinesry to prevent loosening. Apologies if this has been asked before.
Hi Thomas ,
if I want to modify the original Prusa i3 mk3 printer and change my
heated bed for a bigger size bed or extend z axis , does the printer
recognise the extension in the axis and would auto bed levelling works
Any chance that the wobble is applying a naturally reduced jerk, and that by reducing jerk on y-axis you'll get better results?
The prints looked flawless....before and after the hiwin rails...
How many takes did it take for you to get through that first 5 second mouthful?
Im glad you spent the 200 euro's and not me lol, I stick by what I said before, the issue is not that the Y isn't stiff enough! its that damm 3 bearing slop setup! I solved that by using a 3/4 inch PVC tube, that the lm8uu bearing slips in tightly and anchored it to the bed, the total cost was a whole 10 dollars for the pipe and all the bed artifacts have been solved.
To be fully transparent: Watterott sent me the upgrade kit along with a set of TMC2209 drivers to test out (at no charge)
@@MadeWithLayers
thanks for the transparency, Either way, you saved a lot of people some money by showing it may not be worth the amount it would cost to retrofit with the hiwin's. Two thumbs up as always!
I’m really glad I watched this - I just put my order in this morning and was trying to find any upgrades I could install while building up the printer. As someone who works with industrial CNC machinery, it’s hard to look at 3D printed parts supporting the main axis, but clearly that’s not a bottleneck in performance!
Have you played around with any of the complicated analysis tools that use a print head accelerometer to tune motion on the fly to negate ringing and ghosting?
Thermal expansion of the y carriage made impossible to set the rails for me. Its fine when cold but binding when heated to petg temp. Changed them to 10mm rods.
Good video Thomas, I think the main reason of the linear rail not just the print quality, but less noisy than stock one, I hate the of the prusa mk3 when it's print with speed it't not quiet anymore, and start see some Grove on the rods, so it's worth upgrade imp 😁