I hang a 20g scale weight from the center point of my X axis belts, then use the measurement between the 2 sides of the belt as a reference point to start with. For my Y axis belt, which is a vertical mount, I use a DIY 3D printed - open ended - wedge that fits between the belt and the rail - and use the deflection of the 2 sides of the wedge to get a reference point. As I run through prints, I use these reference points to adjust belt tensions. It takes time, but now I have a stable belt tension setup and my prints get great results.
Exactly! This bothered me right off the bat, for example, on AnyCubic printers, and it indeed has a visible effect. I haven't tested it (yet), but I also believe that different tooth profile shape also causes different visoble effects. For example a HTD-3M semi-circle profile belt engages more softly with the pully than a trapezoidal XL, MXL or GT2 belt. I haven't tested it yet this professionally on a 3D printer, but in other applications I hear them less noisy.
@@schizophrenicgaming365 I recently returned a Creality K1 because of that, not only where tooth sides running on small smooth idlers but the belts also didn't line up and caused a 0.8-1.6% deformation on parts near the edges and sides
Extraordinary, well constructed and executed experiment and video script. Thanks for the contribution to the community. This video deserves more views and attention.
One of the most educational videos I've seen on 3d printers! Has anybody tried the ender 3 v2 for an optimal tension, couldn't find data for that in the given link
For ender 3v2 we do not have an official value, but in the tension guide, there is a tension table calculator to help you find the optimum tension on your own.
this is interesting but you have to also consider the radial force that the motor shaft can actually sustain (which depending on the pulley position might not be that much).
This makes me think that it would be interesting to add a load cell "weight" sensor to a printer's axis, and with something like HX711 you could measure the weight created by the belt during the actual ongoing printing on the printer itself. Much cheaper than an optical sensor like used in this video, and can be cheaply added for each of the axis on a budget 3D printer.
Interesting but, they weren't measuring the tension on the belt with the optical sensor. What benefits would measuring the belt tension on the fly provide?
@@JeanSolagnier my guess is that it would be possible to find when tension changes because of velocity/acceleration fluctuations during movement. And then an algorithm could be used to reduce the fluctuations in tension, leading to a more stable tension for the duration of the print. The hypothesis being that a more stable tension during extrusions will lead to more even extrusions and thus less surface artifacts.
Using a load cell snsor is a different approach to be researched, but obtaining the "force" signal does not directly indicate the change in position, and the resolution and accuracy of this approach is unclear. In our research, we have used an optical linear rail that has 1um resolution.
What surprises me is that nobody comes up with automatic belt tensioners. One can simply use idlers on hydraulic or pneumatic pistons to tension the belt. A pump and pressure sensor can then be used to maintain a constant oil/air pressure so that any wear, stretching or other deviation will be automatically compensated.
I can’t remember exactly where I saw / read it, but a while ago someone did test that hypothesis and could not find any difference between smooth and teethed idlers…
@@UloPe Well i have an opposite opinion from my first hand experience, i don't have rigorous test results, but at higher tensions it does feel a lot more consistent by hand feel when hand jogging. However it took me a few tries to get toothed idlers from China which were actually machined deeply enough and had bearings good enough to actually work, and the good ones weren't all that cheap. At lower tensions it's probably less relevant but at lower tensions you have more belt compliance and vibration, you're supposed to have more preload. Did you know the belt is undersized and reaches the nominal 2mm/tooth pitch only at pretty high tensions? Unfortunately stretching the belt to spec is also a little borderline and ruins the printer stepper bearings, since you're supposed to not have the shaft just stick out unsupported with that much of a side load... Anyway i'm the sort of person to add more belt tension, add more stepper current, add more PCB cooling, and then juggle the consequences with pleasure. If the printer can't break your arm, is it even a real printer?
@@UloPe I do think quite a few people concluded that toothed idlers aren't needed or all that useful, from tests or first hand experience; but you'll also find quite a few people who swear by them. I think ultimately that's the thing, if a toothed idler binds, even momentarily, you're in for a rude surprise. If a smooth one binds due to bearing fault, you're not even going to notice it most likely, and you can get much bigger bearings for a given idler size. It's kind of a design meme to just use flanged bearings as iders, these are super robust because they have big balls and spacious cages, so even if one is a little worse off, it's still going to be fine. So for a manufacturer, regardless of possible advantages of a toothed idler, it's just much safer to use smooth ones, less complaints less warranty less service to do, quality be secondary. Also like on something like Prusa earlier generations there's an extruder stepper and gears imprinting on the extrusion bead, there's a high frequency pattern there that will cover up any pattern from the belt chatter. Now that MK4 has an actually amazingly well designed extruder, other sources of disturbance should be paid more attention to.
Very detailed, a lot of work went into this thankyou. Next please test how the tension affects accuracy with increasing speed and acceleration. Also what about a genuine gates Vs copy as genuine gates belts appear a lot stiffer
Can you run pulleys with gears vs the "flat" bearings? Most of my ripple was removed by changing idlers to toothed bearings. Not sure who thought belt teeth bouncing over a hump would be a good idea?
@@jordandotan Nero made a comment today suggesting if it was a problem it was more due to belt tension. Which I can agree with in part especially if really tight. But I removed the extra idler entirely on the ender 3, at the back and used a toothed pulley up front. The belt teeth bumping up onto the back tensioner idler was a big source. You could feel it moving the bed. bump-bump-bump
i mean 2 - 4 seems like a winner here, but it seems to be tested on snail printing speed. Everyone is pushing the machines to the limit where accelerations and deceleration are much bigger than what was shown in the axis simulator. In those cases your belt will start to slip on low tension.
The best tension i found is to hand pull and tighten. The knobs make you lose that leveling or feeling of tightness. Extremly hard to find the right tension with the fancy belt tensioners
Excellent video thanks!! I may have missed it but how do I actually set my belt tension to a specific lbs? Is there a chart that converts lbs to frequency?
I tried this motor QSH4218-47-28-040 and it had way less motor vibrations than the other three types I had at home. Then I tried DM542 and it made it vibrate even less than the Tmc2209. Have not tested ny more. However Gecko driver makers said that the low inductance (600 uh) of the motor above will give too high ripple current with switching drivers.
Loots of variables to take in to consideration for making a conclusion to mention a few: Speed, deceleration, acceleration, moving mass, structural resonance and rigidity, temperature, belt compliance.................... not easy for sure ;)
Do you plan on developing a 0.9 degree motor at all? I know 0.9 degree motors are usually used because the MMR that they produce is higher frequency and (usually) lower amplitude, but they also have the benefit of additional resolution, especially when used on the Z Axis.
For Voron 2.4 we do not have an official value yet, but in the tension guide there is a tension table calculator and instructions to help you find the optimum tension on your own!
I wonder if my belt tension is causing my extrusions on top layers to be looking good one direction but bad when it travels in the opposite direction? I notice it when I print a brim.
Great data and summary, but you've gotta change that narrator. Nobody wants to hear a robot talk. Get a human to present and you'll have thousands of subs in no time.
This video explains a lot, thanks. Marlin now has input shaping. It's still experimental, but was released in 2.1.2. Once refined in Marlin, there may be no need for an external Pi to do input shaping. Not saying that there are not other benefits to Klipper though...
Correction, the Klipper firmware does input shaping and it won't help with belt tension issues beyond fixing the ghosting. Input shaper is about accel/decel
@@saltwaterrook4638 The X1 can detect belt tension with it's short vibration test it does before each print. I found out when it suggested I tighten the xy belts before my print started, after loosening and re-tightening 4 screws for the tensioners I ran the calibration for Resonance Compensation and it cleared the message. So it does use that to check belt tension.
Why is it that you can have perfectly scientific data presented but when a computer reads the it, the credibility of said data seems to drop a few points...
If firmware can access motor resistance or whatever you said, then why isnt there a belt tension ui under a settings menu? Someone should do that. Just like you can jog a machines x,y,z … u could jog the belt tension.
The Bambu Lab X1 Carbon can detect loose belts. It checks before each print by doing some vibrations for each axis. I thought it was just part of the input shaping, but it gave me a message to tighten the xy belts one day just as my print started. I let it continue and the print looked good, so maybe it's pretty sensitive at detecting a small change in tension. After I loosened and re-tightened the 4 screws in back for the xy tensioners, I ran the calibration test and the message went away. Still printing great! I was surprised though because I never heard of any consumer printers that detect loose belts and I never heard Bambu Lab mention it as one of its features. I wonder how many other features it has hidden away. It would make another great selling point in comparison to other printers. They 're definitely on a mission to change the game.👍
It would be difficult to find the _ideal_ tension, but you can test for some edge cases. If the belts are too loose, then the first few steps will not meet any resistance - they will just be picking up the slack. After a few steps, the slack will be taken up and the resistance will increase. That would be a clear indication that the belt is too loose. If the belt is too tight, there will be excessive amounts of resistance due to friction on all the rotating parts. If there is an expected level of resistance known for a particular printer, then you could set an upper bound on how much would be excessive. That could be for more reasons than just the belts being too tight (dirty, in need of lubrication, misaligned, etc.), but it would still be a valuable test. However, you couldn't tell from resistance alone what the positional accuracy is, since the very cause of this error is a difference in the motor's position and the head's position due to linkage in between (the belt). I'm not sure what the resolution is on the measurement of the motor current, but I doubt you could test for the amplitude or frequency of ringing in the belt based on those measurements. Not that it is impossible (a similar technique is used to control brushless motors), but I don't think the usual TMC stepper drivers provide this level of resolution.
Mind you that your belt setup on right side is not parallel to the belt below. That creates also error in position on your setup. Many printers has this issue having the belt angled and not parallel!
There is nothing special about the situation on the CoreXY printer, you can refer to the voron guide for more details:docs.vorondesign.com/tuning/secondary_printer_tuning.html
Threaded rods are awful, but leadscrews can be very accurate. That said, belts still have less backlash. Belts are great for 3D printers and laser cutters where you are applying a relatively low force. Screws work better for higher forces like routers and mills, whose high forces would cause consider change in size of the belt (it's a spring).
This is great content, but you probably see it in the retention rating that the text to speech voice over isn't going to do you any favors, I was basically out the door upon hearing it.
How do you accurately determine tension by pitch? I find that plucking a shorter section makes a lower sound for example. Also don't use ai voice pls. Sounds really cheap. We don't care if you have an accent or whatever just record with your own voice :)
So Chep on the Filament Friday RU-vid channel says just the opposite and his test prints proved it. The tighter the belt the better the quality of the prints. So I tried it and my print quality improved dramatically by making the belts tighter. All I see here is data, no actual proof with a 3D printed object. He's been in the 3D printing world a long time and I'll trust him over your data. The prints at the end are only first layer prints. Go watch his video and try again. Looking forward to your updated video.
Man, if I can give you a tip... these videos are WAYYY too dry and technical, and not nearly "entertaining" enough. You are putting in SO much interesting and valuable work, but your videos aren't going as far (and not as many people are subscribing) because this is tooooo much in the weeds and not "entertaining" enough.
Not straight to the point!! It should be a 50 seconds video, you're very good, you should have two channels, one straight to the point, and this one for those who like expert settings
No, they clearly show that both too tight and too loose induce errors. Tighter belts reduced belt ripple in most cases, but increased motor resonance ripple.
It's interesting but the annoying voice made me quit after 2 min. I just can't live with the artifacts created by text-to-speech. Is there a transcript or something I can follow along with?